Strongly Active Responses of Pinus tabuliformis Carr. and Sophora viciifolia Hance to CO₂ Enrichment and Drought Revealed by Tree-Ring Isotopes on the Central China Loess Plateau

Abstract

Understanding the water-use strategy of human-planted species used in response to climate change is essential to optimize afforestation programs in dry regions. Since 2000, trees on the central Loess Plateau have experienced a shift from strengthening drought to weakening drought. In this study, we combined tree-ring δ¹³C and δ¹⁸O records from Pinus tabuliformis (syn. tabulaeformis) Carr. (a tree) and Sophora viciifolia Hance (a shrub) on the central Loess Plateau to investigate species-specific responses to rising atmospheric CO₂ (C_a) and drought. We found summer relative humidity controlled the fractionation of tree-ring δ¹⁸O, but the magnitude of the climate influence on δ¹³C differed between the species. The intrinsic water-use efficiency (iWUE) trends of both species suggested a strongly active response to maintain constant intercellular CO₂ concentrations as C_a rose. The tree-ring δ¹³C and δ¹⁸O of both species using first-difference data were significantly and positively correlated, with stronger relationships for the shrub. This indicated the dominant regulation of iWUE by stomatal conductance in both species, but with greater stomatal control for the shrub. Moreover, the higher mean iWUE value of S. viciifolia indicated a more conservative water-use strategy than P. tabuliformis. Based on our commonality analysis, the main driver of the increased iWUE was the joint effect of C_a and vapor-pressure deficit (25.51%) for the tree, while it was the joint effect of C_a and the self-calibrated Palmer drought severity index (39.13%) for the shrub. These results suggest S. viciifolia will be more drought-tolerant than P. tabuliformis and as C_a continually rises, we should focus more on the effects of soil drought than atmospheric drought on the water-use strategy of S. viciifolia.

1. Introduction

The atmospheric carbon dioxide concentration (C_a) has increased rapidly from 303 ppm in 1920 to 410 ppm in 2020 [1]. During the same period, extreme drought events have occurred frequently in parallel with global warming [2]. Rising C_a and global warming have increased tree water-use efficiency around the world and stimulated tree growth and forest productivity [3,4,5]; on the other hand, gradual warming has exacerbated both atmospheric vapor and soil water deficits, which restricts forest growth and endangers ecosystem health [6,7]. The trade-off between these two changes is a crucial determinant of changes in the terrestrial carbon cycle and in the sustainable development of forests [8], especially on dryland vegetation cover [9].

China’s Loess Plateau is a typically fragile ecotone located in the transition zone between China’s arid and humid regions [10]. Since 1999, the Grain for Green Project (GFGP) has been implemented across China, including on the Loess Plateau, to conserve soil and sequester carbon through afforestation and grassland restoration [11]. Although the vegetation cover increased greatly, evidence provided the warning that the regional revegetation was approaching the sustainable limits imposed by the regional water resource [11,12]. Again, the frequently occurring drought hazards on the Loess Plateau have threatened the sustainability of plantations [13,14]. One way to deal with this risk is to study the water-use characteristics of human-planted species, which can identify suitable species for improving next-stage afforestation.

In contrast to field control experiments and eddy-flux measurements with sparse distributions, the stable carbon isotope (δ¹³C) in tree rings provides an opportunity to investigate long-term plant water-use strategies. Discrimination against ¹³C (Δ¹³C) during photosynthesis in C₃ plants reveals a specific response to the ratio of intercellular CO₂ concentration (C_i) to C_a (i.e., C_i/C_a). In turn, the C_i/C_a level is determined by the relationship between photosynthesis (A) and stomatal conductance (g_s) [15], which is referred to as intrinsic water-use efficiency (iWUE = A/g_s), at the individual tree level. iWUE indicates the balance between water loss and photosynthetic carbon assimilation [16]. Different from δ¹³C, the stable oxygen isotope (δ¹⁸O) in tree rings is inversely linked to g_s but is independent of A and can, therefore, be assumed to be a proxy for g_s in certain environments [17]. Variation in g_s can affect δ¹⁸O in leaf water and derived photosynthate through: (1) evaporative cooling of the leaf; (2) the diffusion of water vapor; and (3) the Péclet effect [18]. Based on these mechanisms, a semiquantitative model was developed to link A and g_s with δ¹³C and δ¹⁸O [19], and this has been used to determine whether variation in iWUE is caused by changes in A, g_s, or both [20,21]. Thus, the dual-isotope records in annual growth rings can reveal the water-use strategy of human-planted species used to survive under environmental constraints.

Previous studies of tree-ring isotopic records on the Loess Plateau mainly focused on paleoclimate reconstructions, such as relative humidity (RH) [22,23] and drought history [24,25]. Yet, relatively few researchers examined plant physiology using δ¹³C or δ¹⁸O data. For example, Pinus tabuliformis primarily utilized shallow water based on an analysis of δ¹⁸O [26], and the higher iWUE derived from δ¹³C indicated that this species had more conservative water use than Robinia pseudoacacia [27] on the Loess Plateau. However, to our knowledge, few studies have been concentrated on the human-planted species’ medium for long-term ecophysiological responses to regional climate changes using both carbon and oxygen isotopes.

Pinus tabuliformis Carr. (a tree) and Sophora viciifolia Hance (a shrub) are both native pioneer species with high drought tolerance and a high survival rate, and they were widely distributed and planted to halt soil erosion on the Loess Plateau [28,29]. In this study, we combined tree-ring α-cellulose δ¹³C and δ¹⁸O of P. tabuliformis and S. viciifolia growing on the central Loess Plateau to explore species-specific responses to rising C_a and drought. Our main goals were to (1) identify the dominant climate drivers of tree-ring δ¹³C and δ¹⁸O responses; (2) explore the mechanisms of iWUE changes in the two species; and (3) quantify the contribution of atmospheric CO₂ concentrations and drought to the iWUE of the two species. We hypothesized that S. viciifolia is more sensitive to rising C_a and frequent drought than P. tabuliformis.

2. Materials and Methods

2.1. Study Area and Climate

Ravines and gullies cross the central Loess Plateau (Figure 1), and most areas are covered with a loess soil, which is highly vulnerable to erosion. Groundwater is buried deeply, and soil water is the main water source for plants [30]. Because of long-term climate change and human activities, the region’s primary forest vegetation has been severely damaged and replaced by secondary and artificial vegetation (e.g., GFGP).

The annual average temperature at the Yan’an meteorological station (36.36° N, 109.30° E, Figure 1) was 10.5 °C, with mean monthly temperature ranging from −4.8 °C in January to 23.5 °C in July, for the period from 1986 to 2017 (Figure S1a,b). Annual total precipitation was 516 mm from 1986 to 2017, and up to 90% of the precipitation fell during the growing season (April to October) (Figure S1a,c). Extreme precipitation event happened in July 2013. Vapor-pressure deficit (VPD) reflects the atmospheric vapor demand (i.e., atmospheric drought), while self-calibrated Palmer drought severity index (scPDSI) denotes the balance between evaporation and precipitation (i.e., soil drought), which are both common and effective indictors of drought on the Loess Plateau [31]. We estimated the monthly mean VPD based on the monthly air temperature and RH [32]; VPD ranged from 0.21 kPa in January to 1.13 kPa in June (Figure S1a). We also extracted scPDSI datasets based on 0.5° gridded datasets from the Climate Explorer database (https://climexp.knmi.nl/, accessed on 13 February 2021) as shown in Figure 1.

The scPDSI value in 2000 was lowest (−3.89) during the whole investigated period, which indicated the drought reached the severest level [33] (Figure 2a). An abrupt test using the ‘changepoint.np’ package for R software showed that the year 2000 was the inflection point year. From 1986 to 1999, scPDSI decreased significantly (R² = 0.43, p < 0.05, Figure 2a) while VPD and temperature increased significantly (R² = 0.34, p < 0.05 and R² = 0.53, p < 0.005, respectively; Figure 2b and Figure S1b). Total precipitation decreased but not significantly (Figure S1c). After 2000, the scPDSI (temperature) showed significantly increasing (decreasing) trends (Figure 2a and Figure S1b), while VPD and total precipitation showed no significant change (Figure 2b and Figure S1c). Based on these results, we defined the period from 1986 to 1999 as a period when drought stress strengthened (D_s+) and the period from 2000 to 2017 as a period when drought stress weakened (D_s−).

2.2. Sample Preparation and Isotopic Measurements

The sampling site (36.08° N, 109.17° E, 1146 to 1264 m a.s.l.) was located at the Renjiatai forest plantation on the central Loess Plateau (Figure 1). We collected wood cores from P. tabuliformis and discs from S. viciifolia in June 2018. Following standard dendrochronological techniques [34], we air-dried the samples and then sanded and cross-dated them. Tree-ring widths were measured using version 6.0 of the LINTAB system (RINNTECH, Heidelberg, Germany) with a resolution of 0.01 mm and checked with the COFECHA quality control software [35]. Details of the cross-dating and growth analysis are provided by Li et al. [36].

We selected five cores (five trees) and six discs (six shrubs) from each species for isotope analysis according to our previous suggestions [37]. The selected tree cores had a high intercorrelation with the master chronology (the Pearson’s correlation coefficient ranged from 0.50 to 0.87 (p < 0.005) for P. tabuliformis and 0.61 to 0.77 (p < 0.005) for S. viciifolia). We eliminated the first 5 years of each core or disc to avoid possible juvenile effects and then separated the annual rings using a scalpel under a binocular microscope. The samples were dried in a vacuum-drying oven and then ground to pass through a 60-mesh (300 μm) sieve in an automatic grinder.

We extracted α-cellulose using the method of Liu et al. [37] and Leavitt and Danzer [38]. Briefly, the sample was placed in a filter bag and the α-cellulose was extracted by means of organic extraction, followed by bleaching and alkalization. Finally, the α-cellulose was homogenized using an ultrasonic cell disruptor (JY92-2D, Scents Industry, Ningbo, China) and then freeze-dried for 72 h using a vacuum freeze dryer (LGJ-10c, Foring Technology, Beijing, China) prior to the stable isotope analysis.

The isotopic measurements were conducted at the Laboratory of Stable Isotopes and Global Change, Shaanxi Normal University. We packed 0.14 to 0.16 mg of α-cellulose into silver capsules for the δ¹⁸O analysis and determined the ratio using a high-temperature conversion element analyzer (Flash IRMS EA, Isolink, Germany) coupled through a ConFlo VI interface to a gas isotope ratio mass spectrometer (Delta V Advantage, Thermo Fisher Science, Bremen, Germany). To determine the δ¹³C values, we added 0.10 to 0.12 mg of α-cellulose into tin foil capsules; then, we used the same equipment that we used for the δ¹⁸O analysis with silver capsules. The stable isotope measurements were calibrated following a two-point calibration method [39] using Sigma-Aldrich α-cellulose and IAEA-CH₃. The δ¹⁸O and δ¹³C values were expressed with reference to the corresponding standards, namely, the Vienna Standard Mean Ocean Water (VSMOW) and Vienna Pee Dee Belemnite (VPDB), respectively. The isotope measurements were performed three times for δ¹⁸O and one time for δ¹³C for each sample, and the precision was better than 0.3‰ for δ¹⁸O and 0.05‰ for δ¹³C, respectively.

2.3. Definitions and Basic Equations

The Δ¹³C can be calculated as follows [16]:

$$\mathsf{\Delta }{}_{}{}^{13}\mathrm{C}=\frac{\mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{air}}-\mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{plant}}}{1+ \mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{plant}}∕1000}$$

(1)

where δ¹³C_air and δ¹³C_plant are the δ¹³C values of ambient CO₂ and plant cellulose, respectively. For C₃ plants, Δ¹³C is also a function of the difference between C_i and C_a, and it can be converted to C_i/C_a ratios as follows:

$$\mathsf{\Delta }{}_{}{}^{13}\mathrm{C}=a+\left(b-a\right)\times \frac{C_i}{C_a}$$

(2)

where a (~4.4‰) represents the isotopic discrimination that results from diffusion of CO₂ from the atmosphere into the intercellular space of leaf cells, and b (~27.0‰) is the fractionation associated with carboxylation by Rubisco.

iWUE is the ratio of the net photosynthetic assimilation rate (A) to stomatal conductance (g_s) for water vapor [40]:

$$iWUE=\frac{A}{g_s}=\frac{\left(C_a-C_i\right)}{1.6}$$

(3)

where 1.6 is the ratio of the diffusivities of water and CO₂ in air.

The atmospheric CO₂ concentration rose rapidly since industrialization and increased by 58.3 ppm from 1986 to 2017 (https://gml.noaa.gov/ccgg/trends/global.html accessed on 24 April 2022), as shown in Figure S1d. To identify the climate drivers of the tree-ring δ¹³C, we used a statistical correction for the effects of depleted atmospheric δ¹³C (δ¹³C_air; Figure S1d) due to anthropogenic fossil fuel emissions [15]:

$$\mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{cor}}= \mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{plant}}-\left(\mathsf{\delta }{}_{}{}^{13}\mathrm{C}{}_{\mathrm{air}}+6.4\right)$$

(4)

where δ¹³C_cor represents corrected tree-ring δ¹³C.

If the rise of C_a alone were responsible for changing iWUE, the ensuing measured long-term stomatal regulation of leaf gas exchange would fall within the natural ranges expected for active (constant C_i) or passive (constant C_i/C_a and C_a-C_i) responses [41]. These scenarios differ only in the degree to which the increase in C_i follows the increase in C_a (either not at all, in a proportional way, or at the same rate, respectively) [42]. Mean C_i concentration over the first 3 years of each species was used as the starting point for the scenarios in our study. Therefore, the iWUE of each year for the active response scenario (constant C_i) was calculated using C_a concentration of the current year and the mean C_i concentration of the first 3 years according to equation (3). The iWUE for the negative response scenario was derived following a similar method. More details can be found in Voelker et al. [43] and Saurer et al. [44]; these scenarios denoted different trade-offs between A and g_s, so we applied the scenarios as guidelines to interpret the observed iWUE trends for the tree and shrub, respectively.

2.4. Data Analyses

We computed Pearson’s correlation coefficient (r) for the relationships between the tree-ring δ¹³C_cor and δ¹⁸O of the two species and the relationships between climate variables and dual isotopes. A piecewise regression model was employed to identify the difference in trends of climatic variance and isotopic values since the drought shift happened in 2000. We used the coefficient of determination (R²) to evaluate the goodness of fit of regression models. To avoid juvenile effects and the effects of low-frequency trends on δ¹³C_cor, we also extracted the high-frequency signal using the first-difference data [37]. To compare the difference in isotopic values between the two species, we used an independent samples Student’s t test for the common period (1994 to 2017).

We also used commonality analysis to quantify the contribution of C_a and drought (VPD and scPDSI) to the iWUE of the two species over the investigated period. Commonality analysis can decompose the explained variances into pure and joint effects of the predictors when dealing with variables that exhibit collinearity in regression analysis [45]. We performed this analysis using version 3.6.3 of the ‘yhat’ package for the R software [46].

3. Results

3.1. Characteristics and Climate Responses of δ¹³C_cor and δ¹⁸O

Table 1. Statistical characteristics for the δ¹³C_cor, δ¹⁸O, and intrinsic water-use efficiency (iWUE) series of Pinus tabulifomis and Sophora viciifolia during the common period from 1994 to 2017. SD represents the standard deviation. Values of a parameter followed by different letter differed significantly (p < 0.05) based on independent samples Student’s t tests during the common period.

Species	Index	Mean	SD
P. tabuliformis	δ13Ccor	−22.8 (‰) b	1.4 (‰)
	δ18O	30.8 (‰) a	1.4 (‰)
	iWUE	107.6 (μmol mol−1) b	14.8 (μmol mol−1)
S. viciifolia	δ13Ccor	−21.9 (‰) a	0.8 (‰)
	δ18O	29.6 (‰) b	1.2 (‰)
	iWUE	117.3 (μmol mol−1) a	11.2 (μmol mol−1)

summarizes the statistical characteristics of the δ¹³C_cor and δ¹⁸O series of P. tabuliformis and S. viciifolia during the common period from 1994 to 2017. The mean δ¹³C_cor value of the tree was lower than that of the shrub, while the mean δ¹⁸O value of the tree was higher than that of the shrub. Figure 3 shows the trends of the δ¹³C_cor and δ¹⁸O series of both species. The δ¹³C_cor series of P. tabuliformis showed an obvious increase (0.373‰ yr⁻¹, p < 0.005, Table S1) in the D_s+ period, with a maximum value of −19.00‰ in 1999 (Figure 3). In the subsequent D_s− period, the series exhibited a nonsignificant decreasing trend (p > 0.05, Table S1). The δ¹³C_cor series of S. viciifolia appeared to increase, but the increase was not statistically significant (p > 0.05) throughout neither the D_s+ period nor the D_s− period (Figure 3; Table S1). The δ¹⁸O series of P. tabuliformis showed slightly but not significantly decreasing trends in both the D_s+ period and the D_s− period (Table S1). The δ¹⁸O series of S. viciifolia exhibited a nonsignificant increasing trend in the D_s+ period, while the series decreased significantly in the D_s− period (−0.117‰ yr⁻¹, p < 0.05; Figure 3; Table S1).

The responses of δ¹³C_cor and δ¹⁸O to changes in the climatic parameters differed in signal strength between the two species (Figure 4). The δ¹³C_cor of P. tabuliformis was significantly and positively correlated with mean temperature in March and from May to July, while it was negatively correlated with RH (except in June with p > 0.05) (Figure 4a). For S. viciifolia, there were no dominant climate variables showing effects on δ¹³C_cor. In the growing season, precipitation in September and VPD in June exhibited significant and positive relationships with δ¹³C_cor of S. viciifolia (Figure 4c). δ¹⁸O of both species exhibited obvious relationships with moisture conditions rather than temperature (Figure 4b,d). δ¹⁸O of both species was significantly and negatively correlated with RH from July to August, while it exhibited no significant relationship (p > 0.05) with precipitation. Furthermore, δ¹⁸O was significantly and positively correlated with VPD in September for P. tabuliformis and with VPD from June to July for S. viciifolia.

3.2. Trends of iWUE and Dual-Isotopes Relationships

The mean iWUE value of S. viciifolia was 9.7 μmol mol⁻¹ higher than for P. tabuliformis during the common period from 1994 to 2017 (Table 1). The iWUE series of both species exhibited increasing trends, but at different rates during the study period (Figure 5). For P. tabuliformis, the iWUE series showed a significant and rapid increase in the D_s+ period (4.57 μmol mol⁻¹ yr⁻¹, p < 0.005, Table S1), and as the drought stress weakened, the iWUE series began to decrease until 2003 and then had a nonsignificant increase (Figure 5; Table S1). However, the iWUE series of S. viciifolia showed a nonsignificant increasing trend in the D_s+ period and then increased significantly at a rate of 1.28 μmol mol⁻¹ yr⁻¹ in the D_s− period (Figure 5; Table S1). Furthermore, the measured iWUE trends of both species were above the theoretical iWUE trend in the active gas exchange scenario: constant C_i (Figure 5). For P. tabuliformis, the difference between the theoretical iWUE in the constant C_i scenario and the measured iWUE gradually increased in the D_s+ period, while the measured iWUE approached the theoretical value in the D_s− period (Figure 5). For S. viciifolia, the measured iWUE trend was nearly in parallel with the theoretical iWUE in the constant C_i scenario during the whole investigated period (Figure 5).

Figure S2 shows the relationship between δ¹³C_cor and δ¹⁸O of both species either using the raw data or in the first-difference data. For P. tabuliformis, there was no significant relationship between δ¹³C_cor and δ¹⁸O in the D_s+ period, but a significant and positive relationship (r = 0.63; p < 0.01) in the D_s− period based on the raw data (Figure S2a). For S. viciifolia, δ¹³C_cor was significantly and positively correlated with δ¹⁸O in the raw data during both subperiods (Figure S2c). In the first-difference data (Figure S2b,d), we found significant and positive correlations between the two isotopes during both subperiods for both species (r = 0.74 and 0.92 in the D_s+ period and r = 0.56 and 0.70 in the D_s− period for P. tabuliformis and S. viciifolia, respectively). In order to compare the difference in dual-isotope relationships between species at the same sample size, we also calculated the correlations during the common period from 1994 to 2017. We found that the significant and positive relationships between δ¹³C_cor and δ¹⁸O of both species were shown both in the first-difference data (R² = 0.45 and 0.61 for P. tabuliformis and S. viciifolia, respectively; Figure 6) and in the raw data (R² = 0.50 and 0.29 for P. tabuliformis and S. viciifolia, respectively; Figure S3).

3.3. Contribution of CO₂ and Drought to iWUE

In the commonality analysis, we calculated the pure and joint effects of C_a, VPD, and scPDSI on the iWUE of both species. The analysis explained a larger proportion of the iWUE variation of S. viciifolia than that of P. tabuliformis (R² = 65.37% and 53.71%, respectively) (Figure 7). Among the seven components of the explained variance for P. tabuliformis, the combined effects of C_a and VPD accounted for 25.51% of the variance, followed by the combined effects of VPD and scPDSI (18.93%) during the period from 1986 to 2017. For S. viciifolia, the contribution of the combined effects of C_a and scPDSI was greatest (39.13%), followed by the pure effect of C_a (21.07%) during the period from 1994 to 2017.

4. Discussion

4.1. Climate Drivers of δ¹³C_cor and δ¹⁸O

Tree-ring δ¹³C records the balance between A and g_s [16], and the fractionation that affects δ¹³C is influenced by plant physiology, environmental variables, or both simultaneously [15]. The significant relationships between tree-ring δ¹³C and RH, precipitation, and VPD revealed moisture conditions during the growing season, especially in the summer, were generally the main limiting factor for tree growth on the Loess Plateau [47] (Figure 4a,c). These results were in accordance with previous studies in other arid and semiarid regions in China [25,48,49]. In dry conditions, leaves will typically respond by decreasing g_s to conserve water, and reservoirs of CO₂ available for continued photosynthesis are reduced, which increases the gradient of C_i to C_a and leads to higher tree-ring δ¹³C values [15]. The significant relationship between tree-ring δ¹³C and temperature can be explained as high summer temperatures will reduce the C_i concentration if rates of photosynthesis are increased or rates of stomatal conductance are reduced, which weaken carbon fractionation and increase the δ¹³C value [50] (Figure 4a,c).

As suggested in an earlier report, the xylem δ¹⁸O reflects the mixing signal for source water δ¹⁸O and leaf water δ¹⁸O [51]. Our correlation analysis (Figure 4b,d) suggested that RH, rather than precipitation, controlled the tree-ring δ¹⁸O in both species, with the effect becoming significant about 1 month later for S. viciifolia, which suggested strong g_s dominance of oxygen fractionation in the tree rings rather than water source variations during the growing season. RH strongly affects the leaf-to-air vapor pressure ratio, which is one of the factors that directly controls tree-ring δ¹⁸O; a high RH decreases evaporation, resulting in the dilution of leaf δ¹⁸O [52].

4.2. Stomatal Conductance Dominates the Regulation of iWUE

The higher iWUE value of S. viciifolia indicated an overall more conservative water-use strategy and higher drought tolerance than that of P. tabuliformis [27] (Table S1). The difference in water-use strategies between species is related to hydraulic architecture characteristics. For example, the thick cuticle and well-developed palisade tissue of S. viciifolia reduce its evaporation rate and preserve water [53], while the small tracheids and transverse tracheid walls of P. tabuliformis increase the resistance to water transport [54].

In this study, the measured iWUE trends of P. tabuliformis and S. viciifolia were beyond the predicted iWUE trends in the active response scenario (Figure 5). This pattern suggests the intercellular CO₂ concentration of both species keeps constant, no matter how the atmospheric CO₂ changes [44]. Moreover, the actual C_i derived from δ¹³C_plant supported the scenario (Figure S4). Except for the low value during the strong drought stress period from 1997 to 2001 (C_a: 363.8 to 368.5 ppm), the C_i concentration of P. tabuliformis stayed within the level of 221.2 ppm (Figure S4a). Similarly, the C_i concentration of S. viciifolia stayed within the level of 192.5 ppm (Figure S4b). This phenomenon is rare, and we found only one similar report about this phenomenon in spruce forests polluted by copper smelter emissions in Canada’s Abitibi region [41]. If this leaf gas exchange strategy continues, C_i/C_a will gradually decrease as atmospheric CO₂ levels continue to rise, and the water-use efficiency of forests may increase greatly [43]. Conversely, maintaining a constant C_i/C_a with rising C_a is the commonest response in most species and environments [43]. The constant C_i/C_a scenario agrees with the least-cost optimality hypothesis [55], which is described as an approach in which “leaves minimize the summed unit costs of transpiration and carboxylation”, and this can be achieved by a synchronous decrease in both g_s and A [56]. Our results could possibly be explained by the reduced g_s that resulted from the combined effects of rising C_a and drought [57], which was demonstrated in subsequent dual-isotope conceptual models. The short study period may be another potential reason, because recent studies have shown that the iWUE patterns in the first two or three decades of isotope chronologies resemble the constant C_i scenario, and trees will change to the constant C_i/C_a scenario as they age [52,58]. Additionally, iWUE trends in our study were also potentially influenced by overestimated δ¹³C_plant due to juvenile effects, even though we removed the initial 5 years to avoid this potential bias.

Variations in the tree-ring δ¹³C were reflected by A or g_s or both, while the tree-ring δ¹⁸O was mainly determined by the δ¹⁸O in the source water and the evaporative enrichment of ¹⁸O in the leaf water, the latter of which is predominantly controlled by g_s [15,16]. If tree-ring δ¹³C and δ¹⁸O shared a common variability, variations in tree-ring δ¹⁸O should mainly result from changes in g_s [18,19]. In the dual-isotope conceptual model, the significant and positive relationship between δ¹³C_cor and δ¹⁸O could be interpreted as domination of the regulation of iWUE by g_s rather than by A [19] (Figure 6 and Figure S2). iWUE driven by stomatal closure was a common phenomenon in the water-limited regions where water saving would be prioritized over carbon gain [4]. Owing to potential juvenile effects, the dual-isotope relationship in the first-difference data was more reliable. We also found the coefficient of determination in the regression model for S. civiifolia (R² = 0.61) was higher than that for P. tabuliformis (R² = 0.45) when using the first-difference data (Figure 6). Moreover, the higher coefficient of determination indicated that S. viciifolia had greater stomatal control than P. tabuliformis [21] (Figure 6), which was in line with our initial hypothesis. Combined with the higher iWUE value, we suggest S. viciifolia will be more drought-tolerant than P. tabuliformis. Meanwhile, we should realize that species with relatively closed stomata are suggested to have less efficient carbon use with future warmer and drier climates [37,49] and may be at risk of carbon starvation [59]. Future research should be paired with measurements such as growth monitoring and physiology indicators (e.g., xylem embolism resistance and hydraulic conductivity). Previous studies suggested that Scheidegger’s model should be utilized with caution [60]. One of the major challenges for interpreting the model’s output is whether to assume constant δ¹⁸O in the source water and water vapor among investigation periods [20]. In our study, changes in cellulose δ¹⁸O were caused by variations in leaf water δ¹⁸O enrichment due to RH rather than precipitation (Figure 4b,d), and our study covered a small spatial scale with uniform topographic and meteorological conditions. On the other hand, we were aware that no gas exchange measurements were made in this study, and trees may utilize water in deep soil layer by extending their roots in an extreme drought year. Taken together, these factors suggest that our interpretation of the results need to be strengthened with further works.

4.3. The Contribution of CO₂ and Drought to iWUE

Based on the commonality analysis, the main driver of the increased iWUE was the joint effect of C_a and VPD (25.51%) for P. tabuliformis, while it was the joint effect of C_a and scPDSI (39.13%) for S. civiifolia (Figure 7). Rising C_a improves iWUE by increasing A, reducing g_s, or both, which describes the so-called CO₂ fertilization effect [3]. Either increased atmospheric vapor deficit (e.g., VPD) or decreased soil moisture (e.g., scPDSI) causes partial stomatal closure and reductions in g_s. Assuming leaf photosynthetic capacity does not instantly change, reduced g_s leads to increased C_i/C_a ratios and thus increased iWUE [61]. This divergence was possibly caused by differences in plant functional types (e.g., root depth) between the tree and shrub. For instance, shallower root depth limits shrubs from acquiring deep stores of soil moisture, which may buffer the leaf-level responses of trees to fluctuations in aridity [62]. Drought-related limitations on tree growth such as dieback of branches or even whole-tree mortality are non-negligible risks to forest ecosystems, and increased iWUE can be a plant response that aims to promote survival and minimize the growth decrease caused by decreased water availability [20]. Recently, Xiao et al. [63] reported that in areas where shrub species are naturally distributed, populations gradually become concentrated in micro-geomorphic regions with better soil moisture conditions under a regional climate warming trend. Intensive vegetation restoration might lead to increased transpiration that can aggravate water shortages, and a dried soil layer has spread on the Loess Plateau [64]. Thus, we suggest that as C_a continually rises, researchers should focus more on the effects of soil drought than atmospheric drought on the water-use strategy of S. viciifolia.

5. Conclusions

We established tree-ring δ¹³C and δ¹⁸O chronologies for P. tabuliformis and S. viciifolia growing on the central Loess Plateau to investigate species-specific responses to atmospheric CO₂ enrichment and drought stress. Summer RH controlled the fractionation of δ¹⁸O at our study site. The iWUE patterns of both species demonstrated a strongly active response to maintain constant C_i and were dominated by g_s. The main driver of the increased iWUE was the pure and joint effects of C_a and VPD for the tree, while it was the pure and joint effects of C_a and scPDSI for the shrub. We suggest S. viciifolia will be more drought-tolerant due to its more conservative water-use strategy and greater stomatal control compared with P. tabuliformis.

The short period covered by our samples suggests that juvenile effects may have affected our δ¹³C, despite us having partly eliminated these effects. In future research, older trees that provide a longer time series should be studied. In addition, it is unclear whether iWUE stimulates tree growth, which is needed to complete ongoing research. Finally, additional research should also be carried out on the physiological mechanisms that underlie the biochemical phenomenon observed in this study.