Water use characteristics of different pioneer shrubs at different ages in western Chinese Loess Plateau:Evidence from δ2H offset correction

ZHANG Yu, ZHANG Mingjun*, QU Deye, WANG Shengjie,Athanassios A ARGIRIOU, WANG Jiaxin, YANG Ye

1 College of Geography and Environmental Science, Northwest Normal University, Lanzhou 730070, China;

2 Key Laboratory of Resource Environment and Sustainable Development of Oasis, Gansu Province, Northwest Normal University, Lanzhou 730070, China;

3 Laboratory of Atmospheric Physics, Department of Physics, University of Patras, GR-26500 Patras, Greece

Abstract: Caragana korshinskii Kom. and Tamarix ramosissima Ledeb. are pioneer shrubs for water and soil conservation, and for windbreak and sand fixation in arid and semi-arid areas. Understanding the water use characteristics of different pioneer shrubs at different ages is of great importance for their survival when extreme rainfall occurs. In recent years, the stable isotope tracing technique has been used in exploring the water use strategies of plants. However, the widespread δ2H offsets of stem water from its potential sources result in conflicting interpretations of water utilization of plants in arid and semi-arid areas. In this study, we used three sets of hydrogen and oxygen stable isotope data (δ2H and δ18O,corrected δ2H_c1 based on SW-excess and δ18O, and corrected δ2H_c2 based on -8.1‰ and δ18O) as inputs for the MixSIAR model to explore the water use characteristics of C. korshinskii and T. ramosissima at different ages and in response to rainfall. The results showed that δ2H_c1 and δ18O have the best performance, and the contribution rate of deep soil water was underestimated because of δ2H offset.During the dry periods, C. korshinskii and T. ramosissima at different ages both obtained mostly water from deeper soil layers. After rainfall, the proportions of surface (0-10 cm) and shallow (10-40 cm) soil water for C. korshinskii and T. ramosissima at different ages both increased. Nevertheless, there were different response mechanisms of these two plants for rainfall. In addition, C. korshinskii absorbed various potential water sources, while T. ramosissima only used deep water. These flexible water use characteristics of C.korshinskii and T. ramosissima might facilitate the coexistence of plants once extreme rainfall occurs. Thus,reasonable allocation of different plants may be a good vegetation restoration program in western Chinese Loess Plateau.

Keywords: stable isotope; Caragana korshinskii; Tamarix ramosissima; water uptake pattern; isotope depletion

1 Introduction

Water is an important factor for vegetation growth, which determines the ecosystem functions in arid and semi-arid areas (Porporato et al., 2004; Wang et al., 2017). It restricts vegetation coverage, species richness, vegetation growth, biomass, and diversity in ecologically vulnerable areas (Bai et al., 2004; Chang et al., 2019; Gao et al., 2011; Heras et al., 2011). It also influences the sustainability of the reestablishment of vegetation in water-scarce areas (Jia et al., 2012; Yang et al., 2014; Huo et al., 2018). Plants satisfy their water needs from deeper soil water and groundwater during the dry season, while they uptake water from shallow soil layers during the wet season (Dawson and Pate, 1996; Asbjornsen et al., 2008; Wang et al., 2017). In addition,plants with dimorphic root systems can absorb water from both shallow and deep soil layers(Dawson and Pate, 1996; Nie et al., 2011; Yang et al., 2015). These flexible water use characteristics reflect ecological plasticity, and are very beneficial to the growth, reproduction,and competition of plants (Eggemeyer et al., 2009; Moreno-Gutiérrez et al., 2012). Thus, water use characteristics of different plants play a significant role in understanding the interspecies competition and simulating the hydrological process at the soil-plant-atmosphere continuum(Sprenger et al., 2016; Chen et al., 2017; Vargas et al., 2017).

Stable isotopes of hydrogen and oxygen have been widely used in determining plant water sources (Rothfuss and Javaux, 2017; Wang et al., 2017; Allen et al., 2019; Chen et al., 2021)based on the assumption that no isotopic fraction occurs during root water uptake (Dawson and Ehleringer, 1991). The isotopic ratios of plant xylem water can be regarded as the weighted average of potential water sources and their contribution rates (Ehleringer and Dawson, 1992).However, Lin and Sternberg (1993) found that the δ2H of vacuum-extracted xylem water from mangroves was more depleted than that of source water. The reason can be attributed to the hydrogen isotope fractionation during the suction of water by the plant. Ellsworth and Williams(2007) studied sixteen xerophytic and semi-xerophytic trees and shrubs under controlled conditions, and found a 3‰-9‰ depletion in the δ2H of plant xylem water. This hydrogen isotope depletion is not just a special case, but also has been found in other tree species, including semi-arid shrub species (Wang et al., 2017), coniferous and broadleaf forests (Brooks et al., 2010;Bowling et al., 2017; Geris et al., 2017; Barbeta et al., 2019; Goldsmith et al., 2019), and tropical rainforests (Hannes et al., 2018; Brum et al., 2019). The isotope tracing technique has contradictory interpretations in water utilization of plants if δ2H offsets are not considered(Barbeta et al., 2019; Barbeta et al., 2022).

Caragana korshinskiiKom. is a plant of the genus LeguminousCaraganawith a relatively developed root system and strong adaptability. It is the preferred shrub for soil and water conservation and ecological restoration in water-scarce areas (Fang et al., 2013; Chen et al., 2021).Tamarix ramosissimaLedeb. is a shrub with a well-developed root system, which is droughttolerant and salt-alkali-resistant. It plays an important role in wind prevention, which is used for sand fixation and ecological restoration on the Chinese Loess Plateau (Li et al., 2015). On the Loess Plateau, the groundwater is buried too deeply to be a source of water for vegetation, and plants can only survive on soil moisture replenished by rainfall (Wang et al., 2017; Chen et al.,2021). However, the climate of northwestern China has undergone a transition from warm-drying to warm-wetting in recent years (Yao et al., 2020; Zhang et al., 2021), which has inevitably lead to variations in plant water use patterns. Thus, whether these two pioneer shrubs can adequately adjust water use characteristics in response to highly variable rainfall patterns is crucial for their survival. Previous research has shown thatT. ramosissimamainly uses deep soil water (Zhou et al., 2017; Su et al., 2020), andC. korshinskiiexhibits significantly seasonal patterns in water source uptake (Gao et al., 2018; Zhang et al., 2020). However, these researches did not consider δ2H offsets of plant xylem water. Barbeta et al. (2019) found that monoisotopic tracers are insufficient to be identified as plant water sources when the stem water isotopic composition is matched to multiple water sources. Therefore, it is necessary to correct the δ2H offsets when quantifying sources of root water absorption (Li et al., 2021). Thus, this paper researched the water use characteristics ofC. korshinskiiandT. ramosissimaat different ages after rainfall in the Chinese Loess Plateau based on corrected isotope data. The objectives of this study were to: (1)analyze the performance of these three sets of data (δ2H and δ18O, corrected δ2H_c1based on SW-excess and δ18O, and corrected δ2H_c2based on -8.1‰ and δ18O) input into the MixSIAR model; and (2) investigate the water use characteristics ofC. korshinskiiandT. ramosissimaat different ages after rainfall and their responses.

2 Materials and methods

2.1 Study area

This study was conducted in the western Chinese Loess Plateau, Lanzhou City, Gansu Province(36°07′N, 103°44′E; Fig. 1). The study area has a mid-temperate continental climate with mean annual precipitation ranging from 270 to 320 mm. The precipitation is unevenly distributed throughout the year (mainly from June to September). The annual mean air temperature is 10.0°C,with a minimum value of -9.0°C in January and the highest value of 29.9°C in July. The terrain is higher in the north and lower in the south, the altitude ranges between 1560 and 2067 m a.s.l., and the slope is generally steeper than 30° (Wu et al., 2006). The soil is mainly light sierozem, with a pH value of 8.0-9.0. The vegetation coverage is low, mainly consisting of shrubs such asTamarix ramosissima,Caragana korshinskiiandReaumuria soongorica(Pall.) Maxim., and herbs such asAgropyron cristatum(L.) Gaertn. andPeganum multisectum(Maxiam.) Bobr.(Zhang et al.,2020).

Fig. 1 Sample location on the western Chinese Loess Plateau (a) and sample sites of soil, xylem, and rainfall (b)

2.2 Experimental design and sample collection

The twig xylem of juvenile, intermediate, and adultC. korshinskiiandT. ramosissimawere sampled from July to October 2020 (Table 1). Sampling were conducted on days 1, 3, and 5 after rainfall events on 17-18 July (12.9 mm) and 23 August (15.0 mm). Each sampling was performed between 08:00 and 10:00 (LST). Three replicate samples were taken from the shrubs. Twig xylem samples were cut-off from well-grown plants; and the length of each sample ranged between 3 and 5 cm, and its diameter was approximately 0.5 cm. The bark was quickly removed to retain only the xylem, placed into a glass bottle with screw caps, sealed with parafilm, and kept frozen in a refrigerator (-20°C). Simultaneously, we excavated a soil core at a depth of 200 cm near the sampled shrubs, and soil samples were taken at depths of 5, 10, 15, 20, 30, 40, 50, 60, 70, 80, 90,100, 120, 140, 160, 180, and 200 cm, with two replicates per individual. Each soil sample was divided into two parts: one part was placed into glass vials, sealed with parafilm, and stored in a refrigerator (-20°C) until isotopic analysis, and the other was used to measure its gravimetric water content (SWC, %) by drying it at 105°C for 24 h. The rainfall samples in 2020 were collected from a rain collector comprised of a polyethylene tank and a funnel fitted with a ping-pong ball, which was set in the meteorological field of the New Campus of Northwest Normal University, China, at approximately 2.5 km from the soil and xylem sample site. The temperature and precipitation data in 2020 were collected from the nearest Gaolan station (China Meteorological Data Service Centre) to the study site.

Table 1 Morphological traits of C. korshinskii and T. ramosissima at different ages

2.3 Isotopic analyses

All xylem and soil water samples were extracted using an automatic cryogenic extraction system(LI-2200, BJJL, Beijing, China) in the Stable Isotope Laboratory, College of Geography and Environmental Science, Northwest Normal University, China. The vacuum threshold was controlled below 1 Pa/s, the heating temperature was 105°C, and the extraction time was 3 h. We randomly selected portion of those samples, weighed them after the extraction, and dried them at 105°C for 24 h to ensure that the extraction efficiency exceeded 98% (Yang et al., 2015).

The isotopic compositions of the extracted soil water, xylem water, and rainfall samples were measured using an isotopic ratio infrared spectroscopy (IRIS) system (T-LWIA-45-EP, ABB-Los Gatos Research, CA, USA). The analytical accuracies were ±1‰ for δ2H and ±0.3‰ for δ18O.The composition of δ2H and δ18O is expressed in per mil relative to the Vienna Standard Mean Ocean Water (VSMOW):

whereRsampleandRstandardare the ratios of2H/1H or18O/16O of the sample and the standard(VSMOW), respectively.

Organic contaminants (such as methanol, ethanol, and other biological volatile substances) in the water extracted from the twig xylem and soil used cryogenic vacuum distillation would affect the isotopic measurements by the IRIS method (Martín-Gómez et al., 2017). Therefore, we used the spectral analysis software of Los Gatos Research (ABB-Los Gatos Research, CA, USA) to correct the soil and twig xylem water isotope data based on the index of the measured absorption spectrum to eliminate organic contaminants (Schultz et al., 2011; Leen et al., 2012).

2.4 Data analysis

Landwehr and Coplen (2006) proposed the line-conditioned excess (lc-excess) to evaluate whether there is a δ2H offset between soil water (river water and groundwater) and precipitation:

whereaandbare the slope and intercept of local meteoric water line (LMWL), respectively. The lc-excess describes the non-equilibrium dynamic fractionation caused by evaporation (Landwehr et al., 2014). The average value of lc-excess of precepitation is 0‰. The lc-excess for other water bodies (e.g., soil water and river water) affected by evaporation, is usually less than 0‰. However,since the water inside a plant is more likely to originate from soil water than directly from precipitation, Barbeta et al (2019) revised Equation 2 and proposed SW-excess to describe the deviation of plant xylem water concerning soil water line (SWL):

whereasandbsare the slope and intercept of SWL at the same sampling point in a given period,respectively. Positive SW-excess indicates that plant xylem samples are more enriched in δ2H than SWL, and vice-versa.

The corrected hydrogen isotope of xylem water (δ2H_c1) is:

In addition, Chen et al (2020) found that plant xylem water cryogenic extraction bias that could originate from a dynamic exchange between organic combination deuterium was the key to explaining the hydrogen isotope depletion. They proposed a stem water hydrogen isotope correction method based on the stem relative water content:

where δ2H_c2is the corrected hydrogen isotope of stem water;εis the deuterium offset between cryogenically extracted stem water and true xylem water, which is the regression function of the relative water content of the stem; andSis the measurement uncertainty (Evaristo et al., 2015):

Since we did not measure the stem relative water content, we treatedεas a fixed value of-8.1‰, which corresponds to the species-averaged value for the offset of δ2Hstem_CVD(the hydrogen isotope ratio of water cryogenically extracted from plant stem samples) from δ2Hxylem(δ2H of plant source water) as obtained from the study of Chen et al. (2020).

At present, there are mainly methods such as the graphical inference method, two-source or three-source linear mixing models, multiple linear mixing model (IsoSource), Bayesian mixing models (MixSIR, SIAR, and MixSIAR), and other methods based on the stable isotopes to trace water sources of plants. Among them, MixSIAR not only considered the uncertainty of root water absorption but also provided an optimal solution rather than a series of feasible solutions (Stock and Semmens, 2013; Rothfuss and Javaux, 2017; Antunes et al., 2018; Wang et al., 2019).Therefore, the Bayesian mixing model MixSIAR was used to estimate the contribution of different potential water sources to the plants. Since isotope fractionation does not occur during root water uptake, the discriminant value was set to 0 (Brunel et al., 1995; Oerter et al., 2019).The run length of the Markov Chain Monte Carlo (MCMC) was set to 'long' (chain length=300,000; burn-in=200,000; thin=100; and chains=3) to ensure that the model converged,which was tested by applying the Gelman-Rubin and the Geweke diagnostic. The mean value was presented as the output of the MixSIAR. To evaluate the effect of δ2H offset in xylem water on quantifying root uptake water, we used three sets of data: δ2H and δ18O, corrected δ2H_c1based on SW-excess and δ18O, and corrected δ2H_c2based on -8.1‰ together with the δ18O input MixSIAR model. The performance of three sets of data was assessed by the Akaike information criterion (AIC), Bayesian information criterion (BIC), and root mean square error (RMSE). The input dataset with the smallest AIC, BIC, and RMSE values was the best type.

wherenis the number of validation samples; andpiandoiare the predicted and observed values of xylem water isotope values, respectively.

wherejis thejthwater source used by plants;kis the number of water sources (k=4 in this study);fis the proportion of water sources calculated by the MixSIAR model; and δAis the isotopic composition of the water sources.

One-way analysis of variance (ANOVA) with the least significant difference (LSD) method(P＜0.05) was used to examine isotope differences of different water bodies; ifP≥0.05, it was considered that there was no difference. In addition, the relationship between the contribution rate of potential soil water sources and environmental factors was analyzed using Pearson's correlation coefficient test. We divided the whole sampling profile into four potential soil water sources based on the variability in δ2H, δ18O, and SWC, i.e., surface (0-10 cm), shallow (10-40 cm),middle (40-100 cm), and deep (100-160 cm) soil water.

3 Results

3.1 Soil water content

The vertical distribution of soil water content forC. korshinskiiandT. ramosissimaat different ages is shown in Figure 2. The 0-40 cm soil layers ofC. korshinskiiandT. ramosissimaat different ages were found to have relatively higher soil water contents, and the temporal variation of soil water content after rainfall events in the 0-40 cm was more obvious than that in the 40-100 cm, suggesting 0-40 cm soil layers are more susceptible to rainfall infiltration and evaporation. In addition, the 40-100 cm soil water content had relatively small seasonal differences with gentle fluctuation, and the soil water content below 100 cm was the lowest with smooth and low fluctuation. The soil water content of the whole profile was the smallest before rainfall (9 July and 16 August), then increased rapidly at 1 d after rainfall (19 July and 24 August),and gradually decreased over time.

Fig. 2 Variations in the soil water content of juvenile C. korshinskii (a1-a3), intermediate C. korshinskii (b1-b3),adult C. korshinskii (c1-c3), juvenile T. ramosissima (d1-d3), intermediate T. ramosissima (e1-e3), and adult T.ramosissima (f1-f3)

3.2 Isotopic composition of different water bodies

The δ18O of precipitation ranged from -18.06‰ to 1.58‰, and the annual precipitation weighted average values were -8.21‰. The δ2H ranged from -126.82‰ to -28.73‰, with the annually weighted average value of -49.91‰. The maximum value of the weighted monthly precipitation was observed in June (δ2H and δ18O being -12.16‰ and -2.91‰, respectively), and the minimum value was observed in January (-96.41‰ for δ2H and -13.72‰ for δ18O) (Fig. S1). The LMWL is δ2H=6.95(±0.26)δ18O+3.25(±2.31),R2=0.95, andP＜0.01; its slope and intercept were less than those of the GMWL (Craig, 1961), suggesting the climatic characteristics of intense evaporation in arid and semi-arid areas (Fig. 3).

Fig. 3 Relationships between δ2H and δ18O of rainfall, soil water, and plant xylem water for juvenile C.korshinskii (a), intermediate C. korshinskii (b), adult C. korshinskii (c), juvenile T. ramosissima (d), intermediate T. ramosissima (e), and adult T. ramosissima (f). SWL is the soil water line based on soil water isotope values,and LMWL is the local meteoric water line. GMWL (δ2H=8δ18O+10) is plotted for reference. Insets show the linear regression relationship between δ2H (δ2H, δ2H_c1, and δ2H_c2) and δ18O in plant xylem water.

Hydrogen and oxygen isotopesin soil water showed similar vertical distributions and temporal changes; thus, we used δ18O to analyze the characteristics of soil water isotopic compositions(Figs. S2 and S3). The δ18O of soil water varied with soil depth and plant species (Fig. S3; Table S1). The δ18O of the 0-40 cm soil water exhibited greater temporal variability than that of the 40-100 cm soil water, and then stabilized with increasing of soil depths. In addition, the soil water δ18O of the two shrubs at different ages showed obvious differences (P＜0.05). The δ18O values of soil water were -4.69‰ (±3.39‰) for juvenileC. korshinskii, -5.27‰ (±3.14‰) for intermediateC. korshinskii, -7.01‰ (±3.47‰) for adultC. korshinskii, -4.27‰ (±3.61‰) for juvenileT. ramosissima, -1.06‰ (±4.38‰) for intermediateT. ramosissima, and -0.38‰(±3.80‰) for adultT. ramosissima. Prior to rainfall (9 July and 16 August), the δ18O was the most enriched, then decreased rapidly at 1 d after rainfall (19 July and 24 August), and gradually increased over time (Fig. S3). The relationships between δ2H and δ18O in soil water for juvenile,intermediate, and adultC. korshinskiiandT. ramosissimaare shown in Figure 3. The slope of the soil water lines (SWL) increased with age forC. korshinskii, while the inverse trend was observed forT. ramosissima. All the SWLs were plotted on the lower right of the LMWL and intersected with it, indicating that soil water originated from rainfall and was affected by different degrees of evaporation.

Significant differences in xylem water isotope values between different species were observed(P＜0.05), indicating thatC. korshinskiiandT. ramosissimamay have distinct water use characteristics. Significant differences were observed in the xylem water isotopes ofC.korshinskiiat different ages (P＜0.05), whileT. ramosissimadid not present such a discrepancy with age (P≥0.05). The isotopic composition ofC. korshinskiiandT. ramosissimaat different ages was found to be near that of soil water, suggesting that the plants obtain water primarily from soil horizons (Fig. 3). After the rainfall events of 17-18 July and 23 August, there was no significant variation in the isotopic composition of plant xylem water (Table S1), which was attributed to the fact that the isotope signal of rainfall needed a certain time to transport to xylem of plants.

3.3 δ2H offset of xylem water

The δ2H offset between xylem water and SWL was calculated by SW-excess. The average SW-excess values were 2.43‰, 0.42‰, and 3.48‰ for juvenile, intermediate, and adultC.korshinskii, respectively, while the average SW-excess values were -1.03‰, 3.16‰, and 1.85‰for juvenile, intermediate, and adultT. ramosissima, respectively (Fig. 4). This finding was distinct from that of previous studies. The SW-excess has positive and negative values, which may be related to the occurrence of rainfall events. The average SW-excess after rainfall on 17-18 July (SW-excess＞0) was larger than that on 23 August (SW-excess＜0), which was attributed to the slope of SWL after rainfall on 17-18 July being less than that on 23 August (except for intermediateT. ramosissima) (Fig. S4). It all boils down to the significantly different rainfall isotopic compositions on 17-18 July (δ2H= -33.26‰,δ18O= -5.64‰) and 23 August (δ2H=-68.76‰,δ18O= -11.43‰). Compared with the uncorrected δ2H, the corrected δ2H_c1and δ2H_c2values from both methods were closer to the range of potential water sources (Fig. 3), indicating that these two correction methods were reasonable.

Fig. 4 Temporal variation (a) and plant species difference (b) in SW-excess. The extents of the boxes show the 25th and 75th percentiles, whiskers show the range within 1.5IQR (interquartile range), and the black rhombus denote the outliers.

3.4 Water source apportionment by three input datasets

When the isotopic compositions of potential water sources were similar, it is not sufficient to use a single isotope to analyse the water use strategies of plants to distinguish the contribution rates of different water sources. Thus, we used three sets of isotopic datasets as inputs to the MixSIAR model: δ2H and δ18O, δ2H_c1and δ18O, and δ2H_c2and δ18O, and the obtained results were different (Figs. 5 and 6). The AIC, BIC, and RMSE values showed that the δ2H_c1and δ18O pairs are the best performing types in identifying water use characteristics (Table 2). In contrast, the performance of δ2H_c2and δ18O was not better than that of δ2H and δ18O because we did not directly measure the water content of the plant stems but used the constantεof -8.1‰. This result also showed that the value of -8.1‰ for δ2H bias cannot be used as a general correction factor.Moreover, the contribution of deep soil water was underestimated because of hydrogen isotope offset (Table S2). Therefore, the δ2H_c1and δ18O datasets were used to analyse the root water uptake patterns of pioneer shrubs of different ages.

Table 2 Performance of water source contribution using three input datasets to the MixSIAR model

The relative proportions of soil water sources toC. korshinskiiat different ages after rainfall calculated by the MixSIAR model with three input datasets are shown in Figure 5. On 9 July,intermediate and adultC. korshinskiiobtained most of their water from the 100-200 cm soil layers, which was due to the higher soil water content in the 100-200 cm layers (no sampling for juvenileC. korshinskii). JuvenileC. korshinskiiabsorbed water from all four soil water layers at 1 d after rainfall on 17-18 July, and shifted to both use the 10-40 and 40-100 cm soil water at 3 and 5 d after rainfall. On 16 August, juvenileC. korshinskiitended to extract the 10-40 and 40-100 cm soil water, shifted to use the 0-10 cm soil water at 3 and 5 d after rainfall on 23 August, and the proportions of the 0-10 cm gradually increased over time. JuvenileC. korshinskiimainly utilized the 0-10 and 10-40 cm soil water during the late growing season (September and October), and the proportions gradually increased. IntermediateC. korshinskiitended to extract the 0-40 cm soil water at 1 d after rainfall on 17-18 July, and the proportional contributions of the 0-40 cm soil water gradually decreased at 3 and 5 d after rainfall. On 16 August, intermediateC. korshinskiiobtained water mostly from the 0-10 cm soil water, used the 10-40 cm soil water at 1 d after rainfall on 23 August, and then shifted to use the 0-10 cm soil water at 3 and 5 d after rainfall. IntermediateC. korshinskiimainly utilized the 0-10 and 10-40 cm soil water during the late growing season (September and October), and the proportions gradually decreased. AdultC.korshinskiiobtained water from the 40-100 cm soil water at 1 d after rainfall on 17-18 July and then gradually increased the proportions of the 0-40 cm soil water at 5 d after rainfall. On 16 August, adultC. korshinskiiutilized water mostly from the 0-10 cm soil water and used the 0-10 cm (50.7% (±3.7%)) and 100-200 cm (43.1% (±2.1%)) soil water at 1 d after rainfall on 23 August, and the proportions of the 0-40 cm soil water steadily increased over time. AdultC.korshinskiimainly absorbed the 0-10 and 10-40 cm soil water during the late growing season(September and October), and gradually increased the proportional contributions.

Figure 6 shows the relative proportions of water sources toT. ramosissimaof various ages after rainfall, as calculated by the MixSIAR model using three sets of input data. On 9 July, juvenile and intermediateT. ramosissimamainly utilized the 100-200 cm soil water, while adultT.ramosissimaabsorbed the water from the 40-100 and 100-200 cm soil layers. On 17-18 July, a rainfall of 12.9 mm occurred. At 1 d after rainfall, juvenileT. ramosissimamainly used the 0-10 and 100-200 cm soil water and shifted to the 100-200 cm soil water at 3 and 5 d after rainfall.The contribution of the 0-10 cm soil water for intermediateT. ramosissimawas relatively larger(＞70%) at 1 d after rainfall, and the proportion of the 10-40 cm soil water gradually increased at 3 and 5 d after rainfall. On the first day after rainfall, the proportional contributions of the 0-10 cm soil water for adultT. ramosissimawere larger (more than 70%) than those of the other soil layers, and this proportion dramatically increased to a proportion of 98% at 3 d after rainfall.However, the water source for adultT. ramosissimashifted from the 0-10 cm at 3 d to the 100-200 cm soil water at 5 d after rainfall. On 16 August, juvenile and adultT. ramosissimaextracted the 40-100 cm soil layer as the main water source, while intermediateT. ramosissimaused water from the 10-40 cm soil layer. On the first day after rainfall on 23 August (15.0 mm),the proportional contributions of soil water in each soil layer for juvenileT. ramosissimawere approximate, and the 0-10 cm (90%) soil water was used at 3 d after rainfall. Besides, the contribution of the 0-10 and 10-40 cm soil layers for juvenileT. ramosissimawas even higher at 5 d. IntermediateT. ramosissimamainly utilized the 0-10 and the 10-40 cm soil water at 1 d after rainfall, switched to using the 10-40 and 100-200 cm soil water at 3 d, and the 10-40 cm (70%)soil water at 5 d, respectively. AdultT. ramosissimamainly absorbed the 0-10 and 10-40 cm soil water at 1 d after rainfall and shifted to utilize the 0-10 cm soil water at 3 d and 40-100 cm soil layer water (70%) at 5 d, respectively. In September, 100-200 cm soil water was the primary water source for juvenile and adultT. ramosissima, while intermediateT. ramosissimautilized the water in the 40-100 cm soil layer. In October, the proportions of the 40-100 and 100-200 cm,100-200 cm (90%), and 40-100 cm (90%) soil water were higher for juvenile, intermediate, and adultT. ramosissima, respectively.

Fig. 5 Relative proportions of water sources used by juvenile C. korshinskii (a1-a3), intermediate C. korshinskii(b1-b3), and adult C. korshinskii (c1-c3) with the MixSIAR model based on δ2H and δ18O, δ2H_c1 and δ18O, and δ2H_c2 and δ18O

Fig. 6 Relative proportions of water sources used by juvenile T. ramosissima (a1-a3), intermediate T.ramosissima (b1-b3), and adult T. ramosissima (c1-c3) with the MixSIAR model based on δ2H and δ18O, δ2H_c1 and δ18O, and δ2H_c2 and δ18O

4 Discussion

4.1 Variation in stable isotopes of soil water after rainfall

In the Loess Plateau, the groundwater is buried too deeply to be a source of water for plants, and plants can only survive on soil moisture originally derived from rainfall (Wang et al., 2017). In this study, the temporal variation of soil water and isotopic compositions of the 0-40 cm varied dramatically after rainfall than that of deeper soil layers (Figs. 2, S2, and S3), suggesting that the 0-40 cm soil water was more susceptible to rainfall and evaporation (Rossatto et al., 2012). This was consistent with the findings of Sprenger et al. (2017), who proposed subsequent fading of the fractionation effect: the stable isotope in the upper soil water varied dramatically as soon as rainfall infiltrated into the soil, and became weaker with depth. Before rainfall (9 July and 16 August), the isotope values of soil water were most enriched and then decreased dramatically at 1 d after rainfall (19 July and 24 August) because of the infiltration of rainfall with a negative isotopic composition. The isotopic compositions of soil water gradually increased at 3 and 5 d after rainfall due to the stronger evaporation effects (Dai et al., 2015).

4.2 Water uptake patterns of C. korshinskii and T. ramosissima at different ages

C. korshinskiigradually increased the proportion of deeper (100-200 cm) soil water with age.Juvenile and intermediateC. korshinskiimainly utilized the water from the 0-10 and 10-40 cm soil layers, while adultC. korshinskiimainly absorbed water from the 40-100 and 100-200 cm soil layers. In this study, during the dry period (9 July), intermediate and adultC. korshinskiiobtained water mostly from deeper (100-200 cm) soil water (no sampling for juvenileC.korshinskii).C. korshinskiiat different ages switched their sources from different soil layers after rainfall (Figs. 5 and 6), and the response of water use characteristics ofC. korshinskiiat different ages to rainfall varied (Tables S3 and S4). When antecedent soil water was deficient, the water uptake patterns of juvenileC. korshinskiiafter rainfall tended to be more homogeneous in different soil horizons (after rainfall on 17-18 July). However, when antecedent soil water was sufficient, juvenileC. korshinskiiwater use patterns were more sensitive to rainfall, and the main water sources after rainfall shifted from the 100-200 to the 0-10 cm of soil water (after rainfall on 23 August). This result is in line with the result of Dai et al. (2015), who found that the water use strategies ofHaloxylon ammodendron(C. A. Mey.) Bunge were more sensitive to rainfall in spring with abundant surface soil water than in dry summer. The water use strategies of intermediateC. korshinskiican respond to rainfall immediately, and plants tended to utilize the 0-40 cm soil water after rainfall on 17-18 July and 23 August. AdultC. korshinskiihad delayed responses to water use after rainfall, and the contribution of the 0-40 cm soil water generally increased gradually at 1, 3, and 5 d after rainfall.

T. ramosissimais a deep root plant, and plant roots distributed in surface soil mainly absorbed soil moisture provided by rainfall, while deeper roots absorbed soil water supplied by winter and spring precipitation or groundwater in drought environments (Williams and Ehleringer, 2000;Chimner and Cooper, 2004; Su et al., 2020). The roots in the surface soil may be dormant under drought conditions, and are reactivated by rainfall. Therefore, plants only rely on their deep roots to utilize water from the soil when there is less rainfall and low shallow soil water content(Ehleringer and Dawson, 1992). When the rainfall amount reaches a certain threshold, the shallow soil water content increases, and plant roots begin to form and maintain the function of absorbing surface soil moisture (Duan et al., 2008), which helps to reduce energy consumption as demonstrated in previous studies (Ogle and Reynolds, 2004; Schenk, 2008; Sun et al., 2011). In this study, during the dry period (July 9), juvenile and intermediateT. ramosissimaabsorbed deeper (100-200 cm) soil water, while adultT. ramosissimamainly utilized water from middle(40-100 cm) and deep (100-200 cm) soil layers. After rainfall,T. ramosissimaof different ages tended to absorb surface (0-10 cm) and shallow (10-40 cm) soil water supplied by rainfall, which was the same as the result of Cui et al. (2015) in the Gobi area of Dunhuang, China. The response of the water use characteristics ofT. ramosissimaat different ages to rainfall varied (Tables. S3 and S4). Water use strategies of juvenileT. ramosissimawere sensitive to rainfall when antecedent soil water was sufficient. The water use strategies of intermediate and adultT.ramosissimacan both respond to rainfall immediately. IntermediateT. ramosissimatended to utilize the 0-40 cm soil water after rainfall on 17-18 July and 23 August, and the proportion gradually decreased. AdultT. ramosissimatended to absorb the 100-200 and 40-100 cm soil water at 5 d after rainfall on 17-18 July and 23 August, respectively.

The contribution rates of potential water sources forC. korshinskiiandT. ramosissimaat different ages were also correlated with environmental data (rainfall, vapour pressure deficit, and soil water content) (Table S5). The discrepancy in the water source contribution ofC. korshinskiiwas mainly reflected in the relationship with soil water content. Nevertheless, the difference in contribution rates ofT. ramosissimawas reflected in the correlation with the cumulative rainfall 7 d before sampling, the average vapour pressure deficit 7 d before sampling, and the soil water content.

4.3 Implications

The transition of climate from warm-drying to warm-wetting and extreme precipitation with increasing frequency on the Loess Plateau will impel plants to change their water use patterns(Gao et al., 2018). During the dry period, intermediate and adultC. korshinskiiobtained water mostly from deeper (100-200 cm) soil water and juvenile and intermediateT. ramosissimaabsorbed deeper (100-200 cm) soil water, while adultT. ramosissimaabsorbed mostly water from the middle (40-100 cm) and deeper (100-200 cm) soil layers. After rainfall, the proportions of surface (0-10 cm) and shallow (10-40 cm) soil water forC. korshinskiiandT. ramosissimaat different ages both steadily increased over time. However, there were different responses to rainfall. Compared with the water use strategies of the two pioneer shrubs, we found thatC.korshinskiiabsorbed various potential water sources simultaneously, whileT. ramosissimaonly used deep water. These significantly different water use strategies allow them to achieve a complementary utilization of resources, and promote coexistence between species, which may provide more guidance in ecological rehabilitation and insights for the selection and management of plant species during vegetation restoration on the Loess Plateau. In vegetation restoration,plants with different water use characteristics should be combined to form a good plant configuration, and compensate for the lack of water use by different species. In this study, we attempted to analyze the performance of three datasets: δ2H and δ18O, corrected δ2H_c1based on SW-excess and δ18O, and corrected δ2H_c2based on -8.1‰ and δ18O. The performance of δ2H_c2and δ18O was not the best because we did not directly measure the water content of plant stems but used the constantεof -8.1‰. A higher performance may be obtained if the measured stem water content is used to correct the δ2H offset in future studies. In addition, fine roots determine plant water uptake, and further long-term studies on root biomass, characteristic changes affecting root activity, and changes in stable isotopes over time are needed.

5 Conclusions

This study used three sets of hydrogen and oxygen stable isotope data as inputs for the MixSIAR model to explore the water use strategies and responses to rainfall ofC. korshinskiiandT.ramosissimaat different ages in western Chinese Loess Plateau. We evaluated the performance of the three sets of data input to the MixSIAR model, and found that δ2H_c1and δ18O was the best performance type. In addition, the contribution of deep soil water was underestimated because of the δ2H offset. During the dry periods, deeper (100-200 cm) soil water was the main source for intermediate and adultC. korshinskii, juvenile and intermediateT. ramosissima, while the middle(40-100 cm) and deeper (100-200 cm) soil water was main sources for adultT. ramosissima(no sampling for juvenileC. korshinskii). After rainfall, the proportions of surface (0-10 cm) and shallow (10-40 cm) soil water forC. korshinskiiandT. ramosissimaat different ages both gradually increased. Nevertheless, there were different responses to rainfall. These flexible water use characteristics ofC. korshinskiiandT. ramosissimamight facilitate the coexistence of plants in extreme rainfall. This study will be of great significance for vegetation restoration on the Chinese Loess Plateau.

Acknowledgements

This study was funded by the National Natural Science Foundation of China (41771035, 42071047), the Foundation for Distinguished Young Scholars of Gansu Province (20JR10RA112), the Northwest Normal University (NWNU-LKZD2021-04), and the Department of Education of Gansu Province: "Innovation Star"Program of Excellent Postgraduates (2021CXZX-217).

Table S1 Isotopic composition of different types of water body

Note:*represent annual weighted mean precipitation. Min, minimum; Max, maximum.

Sample δ2H (‰) δ18O (‰)Min Max Mean Min Max Mean Precipitation -126.82 -28.73 -49.91* -18.06 1.58 -8.21*Juvenile C. korshinskii -82.43 -24.28 -50.48 -12.88 5.07 -4.69 Soil water Intermediate C. korshinskii -86.68 -6.93 -48.77 -13.34 7.76 -5.27 Adult C. korshinskii -87.52 -22.63 -56.60 -14.53 4.18 -7.01 Juvenile T. ramosissima -80.11 0.63 -46.60 -12.26 10.34 -4.27 Intermediate T. ramosissima -76.72 -10.00 -43.73 -11.96 12.45 -1.06 Adult T. ramosissima -80.36 -9.70 -39.03 -12.35 9.85 -0.38 Juvenile C. korshinskii -71.40 -27.33 -47.89 -8.64 0.22 -3.83 Plant xylem water Intermediate C. korshinskii -63.97 -27.04 -45.48 -7.67 -2.34 -4.55 Adult C. korshinskii -58.48 -33.58 -47.17 -7.30 -2.44 -5.27 Juvenile T. ramosissima -69.87 -49.46 -62.16 -9.80 -6.00 -7.64 Intermediate T. ramosissima -70.12 -48.69 -57.84 -9.83 -4.94 -6.81 Adult T. ramosissima -70.25 -19.40 -56.61 -9.78 1.50 6.60

Table S2 Contributions of three modes of data input to the MixSIAR model

Note: SD, standard deviation.

Contributions (%)Input data mode 0-10 cm 10-40 cm 40-100 cm 100-200 cm Mean SD Mean SD Mean SD Mean SD δ2H and δ18O 29.67 12.71 22.72 14.33 22.22 12.12 25.38 14.76 δ2H_c1 and δ18O 28.80 12.01 20.78 15.04 23.11 12.47 27.31 14.34 δ2H_c2 and δ18O 26.80 12.07 25.31 14.67 20.65 11.7 27.23 15.00

Table S3 Contribution of the 0-40 cm soil water forC. korshinskiiat different ages after rainfall

Plant Date Contribution of the 0-40 cm soil water (%)δ2H and δ18O δ2H_c1 and δ18O δ2H_c2 and δ18O 19 Jul (1 d after rainfall) 79.8 49.6 86.0 Juvenile C. korshinskii 21 Jul (3 d after rainfall) 51.6 46.4 47.8 23 Jul (5 d after rainfall) 54.5 50.0 43.3 19 Jul (1 d after rainfall) 83.7 83.4 86.8 Intermediate C. korshinskii 21 Jul (3 d after rainfall) 71.3 45.2 91.6 23 Jul (5 d after rainfall) 54.2 48.2 62.9 19 Jul (1 d after rainfall) 53.8 23.7 87.9 Adult C. korshinskii 21 Jul (3 d after rainfall) 32.0 33.9 23.2 23 Jul (5 d after rainfall) 58.1 60.0 50.1 24 Aug (1 d after rainfall) 11.1 10.8 8.2 Juvenile C. korshinskii 26 Aug (3 d after rainfall) 93.6 91.0 86.8 28 Aug (5 d after rainfall) 94.0 93.6 89.5 24 Aug (1 d after rainfall) 61.6 61.1 85.2 Intermediate C. korshinskii 26 Aug (3 d after rainfall) 78.8 52.4 54.4 28 Aug (5 d after rainfall) 87.6 81.9 72.6 24 Aug (1 d after rainfall) 55.8 49.1 54.2 Adult C. korshinskii 26 Aug (3 d after rainfall) 22.9 21.8 17.7 28 Aug (5 d after rainfall) 31.6 30.4 19.5

Table S4 Contribution of the 0-40 cm soil water forT. ramosissimat different ages after rainfall

Plant Date Contribution of the 0-40 cm soil water (%)δ2H and δ18O δ2H_c1 and δ18O δ2H_c2 and δ18O 19 Jul (1 d after rainfall) 77.6 58.9 54.0 Juvenile T. ramosissim 21 Jul (3 d after rainfall) 21.2 27.3 7.8 23 Jul (5 d after rainfall) 21.0 17.5 34.4 19 Jul (1 d after rainfall) 95.3 93.6 2.9 Intermediate T. ramosissim 21 Jul (3 d after rainfall) 53.8 72.1 35.0 23 Jul (5 d after rainfall) 46.6 42.9 24.6 19 Jul (1 d after rainfall) 90.1 89.0 89.0 Adult T. ramosissim 21 Jul (3 d after rainfall) 99.1 99.0 99.0 23 Jul (5 d after rainfall) 27.7 22.3 35.9 24 Aug (1 d after rainfall) 55.3 47.3 45.6 Juvenile T. ramosissim 26 Aug (3 d after rainfall) 97.1 94.7 94.1 28 Aug (5 d after rainfall) 66.4 99.6 99.4 24 Aug (1 d after rainfall) 60.9 67.5 43.5 Intermediate T. ramosissim 26 Aug (3 d after rainfall) 35.8 39.7 65.3 28 Aug (5 d after rainfall) 82.3 81.2 79.7 24 Aug (1 d after rainfall) 79.1 74.1 73.2 Adult T. ramosissim 26 Aug (3 d after rainfall) 93.7 93.1 91.5 28 Aug (5 d after rainfall) 8.4 12.0 12.1

Table S5 Correlation between contribution calculated byδ2H_c2and δ18O input into the MixSIAR model and cumulative rainfall amount of 7 d before sampling

Note: VPD, vapour pressure deficit; SWC, soil water content;*,P＜0.05 level;**,P＜0.05 level.

Plant Depth(cm)Rainfall(7 d amount)VPD(7d mean)0-10 cm SWC 10-40 cm SWC 40-100 cm SWC 100-200 cm SWC 0-10 -0.365 0.174 -0.375 0.177 -0.039 -0.014 Juvenile C. korshinskii 10-40 -0.118 -0.072 -0.095 -0.686* -0.488 -0.416 40-100 -0.081 0.008 -0.345 -0.615 -0.412 -0.382 100-200 0.639 -0.211 0.826** 0.579 0.626 0.530 0-10 0.332 -0.602 -0.244 -0.342 -0.387 -0.335 Intermediate C. korshinskii 10-40 -0.234 0.210 0.795** 0.659* -0.618 0.587 40-100 -0.242 0.398 -0.067 0.066 -0.068 -0.007 100-200 -0.032 0.351 -0.636* -0.405 -0.217 -0.284 0-10 0.019 -0.198 -0.090 0.035 0.228 0.333 Adult C. korshinskii 10-40 -0.286 -0.106 -0.284 -0.148 0.039 -0.140 40-100 -0.169 0.258 0.460 0.316 -0.013 -0.345 100-200 0.255 -0.046 -0.213 -0.250 -0.197 0.113 0-10 0.183 0.251 0.405 0.414 0.231 0.161 Juvenile T. ramosissim 10-40 0.153 -0.610 -0.046 -0.078 -0.316 -0.412 40-100 0.475 -0.047 -0.410 -0.427 -0.357 -0.174 100-200 -0.688* 0.199 0.008 0.034 0.294 0.263 0-10 0.117 0.288 -0.027 0.343 0.669* 0.718*Intermediate T. ramosissim 10-40 0.598 -0.440 0.049 -0.306 -0.588 -0.580 40-100 -0.876** -0.250 0.688* 0.685* 0.156 0.174 100-200 -0.083 0.246 -0.370 -0.421 -0.214 -0.298 0-10 0.110 0.799* 0.579 0.454 0.383 0.366 Adult T. ramosissim 10-40 -0.712* -0.263 0.339 0.321 0.255 0.514 40-100 0.626 -0.603 -0.589 -0.500 -0.349 -0.584 100-200 -0.773 -0.295 -0.109 -0.046 -0.151 0.151

Fig. S1 Variation of monthly precipitation weighted δ2H and δ18O. No precipitation data in February, March,April, and November.

Fig. S2 Variation of soil water δ2H in juvenileC. korshinskii(a1-a3), intermediateC. korshinskii(b1-b3), adultC. korshinskii(c1-c3), and juvenileT. ramosissima(d1-d3), intermediateT. ramosissima(e1-e3), and adultT.ramosissima(f1-f3)

Fig. S3 Variation of soil water δ18O in juvenileC. korshinskii(a1-a3), intermediateC. korshinskii(b1-b3),adultC. korshinskii(c1-c3), juvenileT. ramosissima(d1-d3), intermediateT. ramosissima(e1-e3), and adultT.ramosissima(f1-f3)

Fig. S4 Linear regression relationship between δ2H and δ18O in soil water after rainfall on 17-18 July and 23 August in juvenile, intermediate, andadultC. korshinskii(a-c), and juvenile, intermediate, andadultT.ramosissima(d-f). LMWL is the local meteoric water line.