The study area was on the southern border of the Gurbantunggut Desert located in the Central Asian continental desert, which was near the Fukang Station of Desert Ecology, Chinese Academy of Sciences (44°22′N, 87°55′E). The research area has a typical temperate continental arid climate, with a hot, dry summer lasting from June to the end of August, and a cold, snow-covered winter lasting from December to March. The mean annual temperature is 6.6°C (1997–2016), the annual potential evaporation is approximately 1,000 mm, and the annual precipitation ranges from 70 to 180 mm (Fan et al., 2014; Xu et al., 2016). The C₃ and C₄ shrubs are the dominant vegetation types distributed across the area from the fringe of the alluvial plain to the interior of the desert.

Three shrub communities with distinct vegetation structures were selected along the groundwater gradient from the southern border to the interior of the desert to identify the divergent eco-physiological performances of two different C₃ and C₄ shrubs to deteriorating soil water conditions. The C₄ shrub community (C₄ site hereafter) was chosen from the interior of the sandy desert, where the groundwater depth was approximately 10 m. The C₃ and C₄ mixed shrub community (C₃/C₄ site hereafter) and the C₃ shrub community (C₃ site hereafter) were chosen from the fringe of the alluvial plain, where the groundwater depths were 3.5 m and 6.5 m, respectively ( Figure 1 ). The selected three sites had comparable climatic characteristics and weather patterns and were 5–8 km apart. The C₄ shrub H. ammodendron (C. A. Mey.) Bunge was chosen in the C₄ site, and the C₃ shrub T. ramosissima Ledeb. Fl. Alt. was selected in the C₃ site, and both shrubs were chosen from the C₃/C₄ site for comparison. The characteristics of these shrubs are presented in Supplementary Table 1 . The study was conducted for two consecutive years, 2015 and 2016, with similar annual precipitation amounts (approximately 170 mm) and mean annual temperatures ( Figure 1 ). The soil moisture differed with the groundwater depth in the selected three shrub communities, with the C₃/C₄ site having the highest soil moisture and the C₄ site having the lowest soil moisture ( Figure 1 ).

Precipitation, soil, plant stems, and groundwater samples were collected to determine the plant water sources. Soil, stem, and groundwater samples were collected in April, June, August, and October 2015 and in April, July, and September 2016. Precipitation samples were collected after each rainfall (or snowfall) event using a collector installed in the ground of each site. The meteorological data were collected from the weather station located in the study area. Detailed information on the soil and stem sample collection method and soil water content measurement were described by Tiemuerbieke et al. (2018). Soil and xylem water were extracted using a cryogenic vacuum distillation system, and the oxygen and hydrogen isotope compositions of all the water samples were analyzed using a liquid water isotope analyzer (LWIA, DLT-100, Los Gatos Research Inc., Mountain View, CA, USA), with precision levels of 0.25‰ and 0.1‰ for δ¹⁸O and δ²H, respectively. The specific procedures were described by Tiemuerbieke et al. (2018) and Wu et al. (2014).

A portable gas exchange system (Li-6400; Li-Cor, Lincoln, NE, USA) was used to measure the leaf gas exchange parameters of C₃ and C₄ shrubs monthly from June to September. This study detected the responses of leaf gas exchange to photosynthetically active radiation and intercellular CO₂ concentration from 08:00 to 14:00 on clear sunny days. To measure the light response of plants, the light source leaf chamber (20 × 30 mm²) was used to measure the light response of three to four mature, fully extended sunlit leaves from each tree, and the light intensities were set at 0 μmol m⁻² s⁻¹, 20 μmol m⁻² s⁻¹, 50 μmol m⁻² s⁻¹, 100 μmol m⁻² s⁻¹, 150 μmol m⁻² s⁻¹, 200 μmol m⁻² s⁻¹, 400 μmol m⁻² s⁻¹, 600 μmol m⁻² s⁻¹, 800 μmol m⁻² s⁻¹, 1,200 μmol m⁻² s⁻¹, 1,600 μmol m⁻² s⁻¹, 1,800 μmol m⁻² s⁻¹, 2,000 μmol m⁻² s⁻¹, and 2,200 μmol m⁻² s⁻¹. The chamber temperature was adjusted to 30°C to keep the ambient air temperature and relative humidity, and the gas flow rate was set at 400 μmol/s with a 400 μmol/mol reference CO₂ concentration. A modified rectangular hyperbola model (Ye and Yu, 2008) was used to derive the light saturation point (LSP) maximum net photosynthetic rate (P_nmax ):

where P_nmax represents the maximum net photosynthetic rate; I represents the PPFD_i; α represents the initial slope of the photosynthetic light-response curve when PPFD_i approaches zero, namely, the apparent quantum efficiency; β and γ are coefficients that are independent of I and obtained by curve fitting (Ye & Yu, 2008); and Rd represents the dark respiration. The stomatal conductance (g_s ) and transpiration rate (T_r ) were measured at PPFD_i = 1,800 μmol m⁻² s⁻¹. Nine measurements were performed for each species, with three duplicates for each species at each site.

A pressure chamber (Model 3005, PMS Instrument Company, Albany, NY, USA) was used to test the stem water potential for three duplicates of shrubs at each site (a total of nine measurements were taken for each species) from June to September of 2015 and in June, August, and September of 2016. The specific procedures were described by Tiemuerbieke et al. (2018).

The leaf sample collection for leaf non-structural carbohydrate (NSC) analysis was conducted simultaneously with gas exchange measurements. After being collected, the plant leaves were frozen immediately in liquid nitrogen to halt any enzymatic activity. Then, samples were dried in the laboratory at 65°C for 48 h before being ground to a fine powder to measure NSC concentrations, including leaf solute sugars and starch. This study used a modified microplate method (Zhao et al., 2010) to analyze foliar NSC, including foliar solute sucrose and starch content. To extract ethanol-soluble sucrose, the samples were heated in an 80°C water bath for 15 minutes. A mixture of 70 mg of ground tissue and 2 mL of 80% (v/v) ethanol was utilized for the extraction process. The starch concentration was determined by measuring the glucose in an aliquot of the supernatant following the hydrolysis of starch in the sample residue remaining after the EtOH extraction of soluble NSC fractions. The specific procedures were described by Zhao et al. (2010).

The inter- and intraspecific differences, seasonal (or monthly) differences in water potential, gas exchange parameters, and foliar non-structural carbohydrate content between the two plants were tested using univariate analysis based on a general linear model (GLM) and one-way analysis of variance (ANOVA), in which the species, season (or month), and sites were set as fixed effects. Prior to analysis, normality and homoscedasticity tests were conducted on all data, and transformations (logarithmic or square root) were applied as needed. For post hoc tests, Tukey’s honestly significant difference (HSD) was utilized. Pearson’s correlation coefficient was employed for correlation analysis. All data analyses were carried out using SPSS 19.0 (SPSS Inc., Chicago, IL, USA), while linear and non-linear regression analyses and graphing were performed using Origin 8.5 (Origin Lab Corp., Northampton, MA, USA).

The relative reduction ratios (RRRs) for some of the parameters were calculated to evaluate the impact of groundwater decline on the plant’s physiological performance:

where B is the mean value of the parameters at the shallower groundwater depth site and A is the mean value of the parameters at the deeper groundwater depth site. Therefore, this method evaluates the relatively reduction ratio of given parameters at the site with deeper groundwater table depth compared to those of the site with shallower groundwater table depth.

The water sources of the H. ammodendron differed at the two sites with different groundwater table depths ( Figure 2 ). The percentage of groundwater use was higher for the H. ammodendron at the C₄ site, whereas the percentage of soil water was higher for the H. ammodendron at the C₃/C₄ site ( Figure 2 ). However, the water sources of T. ramosissima did not differ between the two sites with different groundwater table depths, and the shrub mainly used soil water over groundwater at both sites ( Figure 2 ).

Significant differences were detected in the gas exchange parameters between the shrubs at different sites for both H. ammodendron and T. ramosissima ( Figure 3 ). The significant differences in the Pnmax (Equation 1) and gs of H. ammodendron in the two sites varied with months, with the Pnmax and gs of H. ammodendron in the C4 site reduced significantly during August and September compared with the H. ammodendron in the C3/C4 site (Figure 3). The LSP and T_r of H. ammodendron at the C₄ site exhibited a decreasing trend from June to September, but the differences were not statistically significant ( Figure 3 ). For the T. ramosissima, decreases in the P_nmax , g_s , T_r , and LSP of T. ramosissima in the drier site (C₃ site) were detected during the entire study period of both years, and the reductions in P_nmax , g_s , and LSP went greater during the drier months of both years ( Figure 2 ).

The Ψ _pd and Ψ _md of H. ammodendron differed significantly at the two sites. The H. ammodendron at the drier site (C₄ site) had more negative water potential than the H. ammodendron at the wetter site from June to September. Particularly, the Ψ _pd of H. ammodendron at the drier site dropped significantly during the drier months ( Figure 4 ). However, the water potential of T. ramosissima did not differ significantly between the two sites during the study period, except for in June and August of 2016 ( Figure 4 ).

The RRR (Equation 2) values of the P_nmax , g_s , and LSP of both shrubs increased gradually with increasing monthly cumulative non-precipitation days (MCND) and monthly cumulative high temperature days before sampling (MCHT), as well as with the decreasing monthly total precipitation amount (MTPA) ( Figure 5 ). The RRR values of the parameters were higher for T. ramosissima than for H. ammodendron. The RRR of Ψ _pd was higher in August and September than in June for H. ammodendron, whereas there was no such trend in the Ψ _pd for T. ramosissima ( Figure 5 ).

The leaf starch content and leaf soluble sugar content of H. ammodendron did not vary significantly with sites ( Figure 6 ). However, there were significant monthly differences observed in these two parameters at both sites. Specifically, the leaf starch content decreased significantly during drier months, whereas the leaf soluble sugar content tended to increase significantly during the drier period (p< 0.05, Figure 6 ). For T. ramosissima, the leaf starch content tended to increase during the drier months of both years and increased significantly during August 2016 ( Figure 6 ). However, the leaf-soluble sugar content of T. ramosissima showed a decreasing trend at the drier site, but a statistically significant decrease was only detected in August 2015 ( Figure 6 ). The NSCs of T. ramosissima did not differ by month in both sites during both years ( Figure 6 ).

The diminishing groundwater had a great impact on the water use, gas exchange, and stem water status of the Central Asian desert shrubs. The water sources of H. ammodendron reverted from the soil water (at the C₃/C₄ site) to the groundwater (at the C₄ site), and the percentage of groundwater use increased with the month in response to the declining groundwater depth ( Figure 2 ). These results were in line with the findings of earlier reports conducted on the same species in the same study area (Dai et al., 2015; Wu et al., 2019), which suggest that this shrub could adjust its rooting depth flexibly in space and time to cope with the water stress caused by fluctuating water conditions (Tiemuerbieke et al., 2018). In addition, the capacity to extract water from relatively stable groundwater via a strong root system would account for the ability of H. ammodendron to maintain relatively constant eco-physiological performance under changing water conditions (Drake and Franks, 2003; Xu and Li, 2006).

The responses of the eco-physiological parameters of H. ammodendron to the declining groundwater depth were manifested during the prolonged summer drought, indicating that the prolonged summer drought further intensified the effect of declining groundwater depth. At the start of the summer drought (during June), the response of H. ammodendron to the declining groundwater depth in terms of gas exchange was not pronounced. The RRR values of the P_nmax and g_s of the shrubs at the drier site were the lowest during this period ( Figure 5 ). However, with the progress of the summer drought, with elevated high temperatures and an increasing number of non-precipitation days, the RRRs of P_nmax , g_s , and LSP increased substantially during the following months, and the highest values were observed in September ( Figure 5 ). The same trends were detected in the predawn water potential of H. ammodendron, which tended to be more negative at the C₄ site during August and September ( Figure 4 ). The RRR of the predawn water potential was greater during the chronic summer drought ( Figure 5 ), suggesting that the ability of H. ammodendron to recover daytime water status was impaired by the prolonged summer drought at the site with deeper groundwater depth ( Figures 4 , 5 ). Similar results were reported in previous studies conducted on the same species (Wu et al., 2019), indicating the pronounced impacts of the summer drought on the plant responses under changing water conditions. Moreover, the ΔΨ reflects the hydraulic balance between water loss from the leaves and supply from the roots and soil, with smaller ΔΨ at the C₄ site ( Supplementary Figure 2 ), indicating that nighttime rehydration under severe drought was insufficient to bring the Ψ _pd into equilibrium with the soil water potential (Taylor et al., 2014). The present study did not detect a significant reduction in the transpiration rate of H. ammodendron at the site with lower groundwater depth. It can be inferred that H. ammodendron sustains the transpiration rate and stomatal conductance essential for growth by relying on stable groundwater. This might be a water-saving strategy for this shrub, as it allows H. ammodendron to respond quickly to changes in the ambient vapor pressure deficit (VPD) to prevent excessive dehydration (Chen et al., 2015).

The carbon assimilation, stomatal conductance, and water transpiration were all suppressed at the site with deeper groundwater depth, and the relative reductions in the average photosynthetic rate and stomatal conductance were greater at the site with deeper groundwater depth and accelerated with drought, indicating that the eco-physiological performance of T. ramosissima was greatly affected by the combined effects of declining groundwater depth and chronic summer drought. However, unlike H. ammodendron, the water sources of T. ramosissima did not shift, and this shrub mainly relied on deep soil water at both sites ( Figure 2 ). The stem water status did not differ significantly with declining groundwater depth but deteriorated with the progression of the summer drought at both sites ( Figure 4 ). These results suggested that the deeper soil water might be the reliable water source that enabled this shrub to maintain comparable water status with the wetter site, but the monthly declining midday water potential indicated that the water status of this shrub would be significantly affected by progressive atmospheric desiccation at both sites.

Both species exhibited a decline in eco-physiological performance due to the reduction in groundwater depth, especially when coupled with prolonged summer drought, and the response patterns differed between the two shrubs. With declining groundwater table depth, the water sources shifted, carbon fixation and stomatal conductance were suppressed, the water status of stems further deteriorated, and the transpiration became less sensitive in H. ammodendron. The water sources and water status responded less sensitively in T. ramosissima, indicating that the two shrubs prioritized water consumption and carbon gain differently when faced with changing water conditions.

The reduced transpiration rate and stomatal conductance and increased instantaneous water use efficiency of H. ammodendron ( Table 1 , Supplementary Figure 1 ) together with its deep root system reflected the adoption of a typical conservative water use mechanism by the C₄ shrub. The deep roots of H. ammodendron at the deeper groundwater site would guarantee a reliable water source to maintain its normal eco-physiological performance. Such a strategy seemed to be beneficial for maintaining its hydraulic functions (Zhu and Cao, 2009; Chen et al., 2015) during the dry season with high temperatures and evaporative demand. The smaller stomatal aperture and higher efficiency of water use in H. ammodendron were consistent with the typical pattern and classical understanding of the traits of C₄ grass species discussed in previous reports and developed models (Kocacinar and Sage, 2003; Taylor et al., 2011; Osborne and Sack, 2012; Taylor et al., 2014). This is also the case for the Central Asian C₄ shrub based on the present study.

As discussed above, strict stomatal regulation of the C₄ shrub could be the key water conservation strategy to limit excessive water loss. H. ammodendron may obtain a great advantage by adopting this water-saving mechanism. On the one hand, the effective regulation of stomata in H. ammodendron is essential for sustaining hydraulic function in hot and dry conditions. Additionally, the transpiration rate in H. ammodendron remained constant at its lowest point across the growing season, which might aid this shrub conserve water in the prolonged dry season and decrease the likelihood of stem embolism (Meinzer and McCulloh, 2013; Chen et al., 2015). On the other hand, the consequence of this avoidance of hydraulic failure is the subsequent limitation of carbon gain that occurs with tight stomatal control (McDowell and Sevanto, 2010; Silvertown et al., 2015). Nevertheless, the CO₂-concentrating mechanism within the C₄ shrub seemed to compensate for the CO₂ limitation caused by tight stomatal control (Ripley et al., 2007; Osborne and Sack, 2012). According to previous studies, rapid reduction in atmospheric CO₂ concentrations and elevated ambient temperatures have selected the evolution of the C₄ pathway (Taylor et al., 2010; Osborne and Sack, 2012), which eliminates the limitation of photorespiration by elevating CO₂ concentrations at the site of the bundle sheath where Rubisco fixes CO₂ (Ripley et al., 2010; Sage, 2014). This allows C₄ plants to maintain the same rate of photosynthetic efficiency at a lower stomatal conductance compared with C₃ ancestors (Taylor et al., 2012; Liu et al., 2019). As revealed in this study, the P_nmax of H. ammodendron did not differ significantly compared with T. ramosissima at the same site during the rapid growth month (June) in both years ( Figure 3 , Table 1 ), demonstrating that H. ammodendron exhibited higher carbon fixation, with stomatal conductance being approximately half that of T. ramosissima. Additionally, it can be inferred from the A–Ci curves that the reduced P_nmax of H. ammodendron might result from metabolic limitations. H. ammodendron seemed to be more susceptible to metabolic limitations under declining groundwater depth than T. ramosissima ( Supplementary Figure 3 ), which aligns with the results of previous reports on C₃ and C₄ grasses (Ripley et al., 2007).

Contrary to H. ammodendron, the C₃ shrub T. ramosissima adopted a contrasting water–carbon balance. T. ramosissima maintained higher stomatal conductance during the growing season. Consequently, the transpiration rate and photosynthetic rate were higher ( Table 1 ), indicating that T. ramosissima might guarantee a higher carbon fixation rate at the great cost of higher water loss and lower water use efficiency ( Figure 7 ). This was also consistent with the earlier research on the same shrub in the same study region, which suggested that T. ramosissima has the tendency to assimilate carbon at its maximum level, at the risk of greater water loss (Xu et al., 2007). However, the higher RRR in gas exchange parameters ( Figure 5 ) indicated the downregulation of its eco-physiological performance with declining groundwater depth. Retaining greater stomatal conductance under a declining groundwater table gradient might make this shrub more susceptible to hydraulic failure, which might be the main factor determining its absence in the central areas of the desert, where the groundwater table is even lower. Therefore, H. ammodendron achieved a higher rate of carbon fixation with greater water use efficiency (lower stomatal conductance) than the C₃ shrub T. ramosissima to which a higher rate of carbon assimilation comes with far greater stomatal conductance.

Additionally, the stem water status of the two shrubs was affected differently by declining groundwater depth. The stem water status of T. ramosissima was more sensitive to the progressing summer drought than the declining groundwater table, whereas both the groundwater table and the prolonged summer drought strongly affected the water status of H. ammodendron. It could be inferred that the two shrubs might adopt different stem hydraulic and morphological properties to guarantee the water transport in stems under changing water conditions. H. ammodendron could tolerate greater negative water potential with the decline in groundwater table depth, which might be associated with specific stem morphology and hydraulic properties (such as different xylem morphological traits, stem hydraulic conductance, and stem vulnerability). Previous reports have pointed out that plants with lower water potentials have smaller and thicker xylem conduits with fewer pits specialized in membranes (Lens et al., 2013). These plants are able to handle low water potentials without experiencing embolism, which allows them to safely transport water to the leaves even in extremely dry environments at near-maximum rates (Manzoni et al., 2013). These embolism-resistant stems would guarantee hydraulic safety for these shrubs under unfavorable conditions.

When data were pooled for all individuals of both shrubs at two sites during the entire study period, the greater negative stem water potential of H. ammodendron coupling with lower transpiration rate and stomatal conductance were associated with improved instantaneous water use efficiency and more positive δ ¹³C values. The higher water potential of T. ramosissima coupled with its higher stomatal conductance and transpiration rate was associated with lower instantaneous water use efficiency and more negative δ ¹³C values ( Figure 7 ). Therefore, it could be inferred that the evaluated carbon fixation of T. ramosissima was accompanied by the coordination with more favorable stem water status and higher water conductance from the leaf to the atmosphere. However, the progressive atmospheric desiccation during summer drought might greatly inhibit this process. For H. ammodendron, the water conservation mechanism seemed to rely on the coordination of the stomatal behavior, photosynthetic pathway, and stem hydraulic properties. Tight stomatal regulation was coordinated with an embolism-resistant stem that could tolerate greater negative water potentials. This was in line with previous reports that have proposed that plants’ regulation of stomatal behavior could be in balance with the leaf water potential above a threshold value that limits excessive water loss to avoid hydraulic failure in the xylem to maintain the water-carbon balance (Sperry, 2000; Meinzer et al., 2009; Klein, 2014; Roman et al., 2015). Furthermore, the reduction in stomatal conductance would not necessarily give rise to a marked reduction in photosynthetic assimilation due to the C₄ pathway ( Figure 5 ). Therefore, it is suggested that the evolution of stem hydraulic traits occurred in coordination with the evolution of photosynthetic pathways so that water conductance would keep pace with water consumption. As suggested by Taylor et al. (2014), the diversification in plants’ hydraulic traits may have been crucial in the colonization and adaptation of C₄ grasses to arid and open environments. This was also the case for C₄ and C₃ desert shrubs in the Gurbantunggut Desert, where Haloxylon species were mainly distributed at the fringe of the desert, whereas the C₃ species were distributed at the edge of the desert.

The partitioning pattern of leaf NSCs differed between the two shrubs. The significant monthly variations in the non-structural carbohydrates of the two shrubs demonstrated different carbon allocation processes in the two shrubs. The significant negative association of leaf soluble sugar content with midday water potential in H. ammodendron ( Figure 8 ) revealed that the reduction of leaf starch content and the significant increase in leaf solute sugar content in drier months may be the result of osmotic or turgor regulation of the leaf. This suggested that when leaf water potentials gradually became negative during dry periods, more osmotically active carbohydrates such as sugars were partitioned to the leaf to maintain leaf turgor and leaf hydraulic water status in H. ammodendron (Woodruff et al., 2015).