The study was conducted in the southeastern Gurbantunggut Desert near the Fukang Desert Ecosystem Observation and Experiment Station, Chinese Academy of Sciences. The region has a typical temperate continental climate, with dry and hot summers and quite cold and snowy winters. The temperatures vary considerably during the year, ranging from a minimum of -42.2°C to a maximum of 44.2°C. The annual average wind speed is 3.2 m/s, with prevailing northwesterly winds (Zhang et al., 2007). The average annual evaporation is approximately 1000 mm, while the average annual precipitation is only 70-150 mm, part of which occurs in the form of snowfall, accounting for approximately 18.7% of the total annual precipitation (Zhou et al., 2012). Rapid snow melting in the early spring (March to April) initiates a brief bloom of ephemeral germination, but most of the time, only sparse shrubs and small trees grow on the surface (such as H. ammodendron and H. persicum), with less than 10% vegetation cover (Zhou et al., 2012). Sandy soils are the main soil type in the study area, with small particles, which flow easily in the wind and have good permeability and are thus susceptible to nutrient loss and become infertile (Zhang and Chen, 2002). Water from oasis recharges groundwater from its outer desert, so that the groundwater is buried at a depth of only about 3-4 m on the southern edge of the Gurbantunggut Desert, but more than tens of meters in its hinterland (Chen, 2017). Moreover, seasonal dynamics in groundwater existed. According to the observation of the wells, groundwater rises in the period of snow-melting. After that, the groundwater level decreased and remain stable, without significant change. H. ammodendron and H. persicum are constructive species in the Gurbantunggut Desert (Wu et al., 2019a), both belonging to the Chenopodiaceae. The two species, with similar canopy and assimilated branches morphology, have developed adaptive features in structure and physiology, which are mainly manifested in the degeneration of the assimilated branches into green assimilated branches to carry out photosynthesis and have higher efficiency (Li and Li, 1981; Guo et al., 2004) (Figure 1). However, the dominant topography of their distribution is significantly different. H. persicum is distributed at the top of sandy dunes with lower soil nutrient and salt contents due to its high nutrient conservation rate (Qian et al., 2008; Yue et al., 2020). H. ammodendron occupies interdune depressions, where soil nutrient and salt contents are higher because of high nutrient uptake efficiency (Yue et al., 2020; Li et al., 2023).

From the southern edge of the Gurbantunggut Desert to the hinterland, the depth of groundwater gradually increases, resulting in a natural groundwater gradient. In October 2015 and April 2017, 25 groundwater observation wells were drilled in the southern Gurbantunggut Desert, with groundwater depth variation range of 4-16m, and four areas with groundwater depth of 4m, 7m, 13m, and 16m were selected for experimentation (Figure 2). The dunes have an average height of 11 meters (Dai et al., 2014), so for H. persicum, the depth of groundwater is 15m, 18m, 24m, and 27m. Based on information on the groundwater gradient and the changes in plant communities, four typical study sample sites were finally identified. The coordinate locations and altitude are shown in Table 1 .

The experiments were conducted in June-July 2017, when the plants were in full growth. Three 100 m*10 m plots were set at the depth of each groundwater depth. H. ammodendron and H. persicum in the plots were counted, and the actual number of individuals was converted to vegetation density, in No/ha ( Table 1 ). At each sample point, five H. ammodendron plants and five H. persicum plants with similar morphology and health status were selected. Their healthy, pest-free assimilated branches were collected from different directions, mixed and stored in kraft envelopes at low temperature, for a total of 40 plant assimilated branches samples.

Along with the plant sampling, soils were sampled from different layers under the canopy of the two species. After stripping off the litter, soil samples were collected with a 5 cm diameter soil auger at depths of 20, 40, 60, 80, 100, 140, 180, 220, 260 and 300 cm. Four replicate groups were taken for each layer. In total, 320 soil samples were collected. Each soil sample was divided into two parts, one in a Ziplock bag for the determination of soil physical and chemical properties and the other in an aluminum box for the determination of soil water content.

The measured assimilated branches indicators included assimilated branches organic carbon (LOC), total nitrogen (LTN) and total phosphorus (LTP) and their stoichiometric ratios. Before the measurement, assimilated branches samples were dried to constant weight in an oven at 70°C, ground with a grinder (NM200, Retsch, Haan, Germany), passed through a 0.15 mm sample sieve and stored in sealed bags. Assimilated branches organic carbon was determined by the potassium dichromate-sulfuric acid oxidation method. Assimilated branches total nitrogen was determined using a fully automatic Kjeldahl analyzer (The Swedish FOSS, Kjeltec™8400). The total phosphorus of the assimilated branches was determined using a molybdenum antimony anti-colorimetric method, and their C/N, C/P and N/P were calculated using elemental values.

The measured soil characteristics included major chemical and physical properties, including soil organic carbon (SOC), total nitrogen (STN), total phosphorus (STP) content and their stoichiometric ratios, pH, water content, and electronic conductivity (EC).

Soil samples that had been air-dried for two weeks were subsequently ground to a fine powder and passed through a 0.25 mm sieve before being stored in sealed bags. Soil organic carbon was determined by the potassium dichromate-sulfuric acid oxidation method. Total nitrogen was determined using a fully automatic Kjeldahl analyzer (The Swedish FOSS, Kjeltec™8400). Total phosphorus was determined using the sulfuric acid-perchlorate-boiled-molybdenum antimony colorimetric method. Their C/N, C/P and N/P were calculated using elemental values. The soil water content (SWC) was determined using the drying method. Soil pH and electrical conductivity (EC) were measured using the electrode potential method in the supernatant of 1:5 soil−water mixtures.

The Shapiro-Wilk test is used for normality testing. Levene’s test was used to test for the homogeneity of variance between groundwater. For the differences in soils and plants ( Tables 2 , 3 ), if the variance is normally distributed and the variance is homogeneous, the t-test is used, otherwise the Mann-Whitney U test is used. Two-way ANOVA was used to determine the effects of groundwater depth, soil depth, and their interaction on soil properties. When the variance was equal, comparison of means using Duncan’s multiple comparisons was performed. When the variance is unequal, comparison of means is performed using T2 Tamhane’s test (α = 0.05). Differences between species were tested for assimilated branches and soil ecological chemometrics using t tests (α = 0.05). The linear regression method was used to determine the relationship between assimilated branches or soil characteristics and the depth of groundwater (soil samples from the same site at different depths were averaged and plotted as the sample points). The relationship between assimilated branches indicators and soil characteristics for H. ammodendron and H. persicum was determined using redundancy analysis (RDA). Data were collated and processed using Microsoft Excel 2019. Analysis and graphing were performed using Origin 2022, Canoco 5, and R 4.2.3.

According to Table 2 , significant differences were observed between the assimilated branches ecological stoichiometry and soil characteristics of the two plants (P < 0.05), with the exception of LTP and soil N:P. Notably, H. persicum exhibited significantly higher LOC, assimilated branches C:N and C:P values than H. ammodendron, while LTN and N:P were significantly lower. Furthermore, the mean assimilated branches N:P for both were less than 11, emphasizing their potential nitrogen limitation.

In terms of soil characteristics, H. ammodendron exhibited significantly higher values of SOC, STN, STP, SWC, pH and EC than H. persicum (P < 0.05). The stoichiometric ratios (C:N, C:P, N:P) were smaller in H. ammodendron, with significant differences observed only in C:N and C:P (P < 0.05). Soil pH was alkaline for both species, with values exceeding 9 and showing minimal variation (CV < 4%).

Table 2 shows that the variation in assimilated branches stoichiometry was comparatively small, with coefficients of variation ranging between 9% and 20% and mostly exhibiting minor differences between the two plant species. In contrast, other soil properties displayed moderate to strong levels of variability, as presented in Table 3 . Specifically, the variability of SOC, STN, and STP in soils under the canopy of H. ammodendron exceeded that of H. persicum. Meanwhile, SWC in soils under the H. ammodendron canopy exhibited the highest variation, with a coefficient of variation exceeding 115%.

The results presented in Figure 3 demonstrate significant variations in the assimilated branches ecological stoichiometric characteristics of both plant species across different groundwater depths (P < 0.05), with the exception of H. persicum N/P ratios. In particular, the assimilated branches LOC of H. ammodendron increased significantly at deeper groundwater depths. The highest LTN and LTP values were observed at 13 m depth of groundwater, concomitant with a significant decrease in assimilated branches C:N and C:P values. Conversely, at 7 m depth of groundwater, H. persicum exhibited the lowest LOC and the highest LTN and LTP values, displaying significantly lower assimilated branches C:N and C:P ratios compared to other depths. It is speculated that groundwater depths of 7 m and 13 m may represent pivotal points for understanding the ecological stoichiometry of H. ammodendron and H. persicum, respectively.

Significant interspecific differences were also observed in the assimilated branches ecological stoichiometry of the two plant species. Specifically, compared to H. ammodendron, H. persicum exhibited greater LOC at all groundwater depths, while LTN was significantly lower at the 4 m, 13 m, and 16 m groundwater gradients (P < 0.05). Additionally, the assimilated branches N:P values of H. persicum were less than those of H. ammodendron. Assimilated branches C:N and C:P ratios were greater in H. persicum than in H. ammodendron, except at the 7 m groundwater depth.

By two-way ANOVA, Table 4 indicates significant differences in soil ecological stoichiometry with varying groundwater depths and soil depths, except for soil C:N ratios. The results indicated that SOC in H. persicum soils, as well as soil N:P in H. ammodendron soils, were both significantly influenced by groundwater depth and soil depth (P < 0.05), but with a weak interaction effect between the two (P > 0.05).

The C:N ratios in H. ammodendron soils were found to be influenced by groundwater depth (P < 0.05) but were not significantly affected by soil depth or their interaction (P > 0.05). In contrast, the GWD, SD, and their interaction had no significant impact on the C:N ratios of the soil of H. persicum (P > 0.05).

Among various groundwater and soil depths, the soil ecological stoichiometry of both species was found to be significantly heterogeneous ( Table 4 and Figure 4). The levels of SOC, STN, soil C:P and soil N:P decreased with increasing soil depth, which was more pronounced in shallow soils than in deeper soils. The variation in C:N with groundwater depth and soil depth was not uniform, indicating that the influencing factors may be intricate. In H. persicum, SOC, STN, soil C:P and soil N:P decreased with increasing groundwater depth in shallow soils, while STP increased significantly at 7 m groundwater depth in the deeper layers, which may be related to the higher content of LTP in H. persicum assimilated branches at that depth. On the other hand, SOC variation in H. ammodendron soil was the greatest at 13 m groundwater depth, and STP content slightly increased at the deeper layer, being significantly higher at 4 m depth than at other depths. The differences in soil C:N at each depth were not significant in the shallow layers but diverged with increasing depth of the soil layer. The variation in soil C:N mainly indicated an increase with increasing groundwater depth. The two species also exhibited differences in soil ecological stoichiometry. Significantly, H. ammodendron had higher concentrations of SOC, STN, and STP than H. persicum. Additionally, the stoichiometric ratios of H. persicum were more extensively varied than others. H. ammodendron had significantly higher levels of SOC, STN, and STP in shallow soils and more pronounced changes with increasing soil depth.

Except for H. persicum assimilated branches LTN, the nutrient contents (LOC, LTN, and LTP) of assimilated branches increased with deeper groundwater ( Figure 5 ). Furthermore, assimilated branches stoichiometric ratios varied with groundwater depth. The C:N ratio of assimilated branches increased with groundwater, with a greater increase in H. persicum than in H. ammodendron, and changed more rapidly. The C:P ratio of assimilated branches remained relatively stable, while the N:P ratio showed a decreasing trend in both species. The slopes of the linear relationships for H. ammodendron and H. persicum LTP and C:N showed significant differences, while those of LOC and N:P did not, indicating that they changed at similar rates.

As groundwater depth increased, SOC, STN, and STP in the soils of H. persicum and H. ammodendron decreased significantly Figure 6 . While the slope of the linear fit for H. ammodendron SOC was mostly similar to that of H. persicum, the change rates of STN and STP were greater. Soil C:N did not vary significantly with groundwater depth. The C:P of soil under the H. persicum canopy was significantly lower than that under the H. ammodendron canopy, although there was less variation within the soil under H. ammodendron. The rate of decrease in N:P of the soil under both species was similar with increasing groundwater depth.

As shown in Figure 7 , the RDA ranked the assimilated branches stoichiometry characteristics and soil factors separately for the two plant species. The bubble diagram revealed clear differentiation of assimilated branches characteristics across varying groundwater gradients. H. persicum accounted for 37.29% and 7.15% in Axis I and Axis II, respectively. The cumulative amount of plant assimilated branches characteristics explained in the first two axes was 44.44%. For H. ammodendron, the corresponding values were 54.66%, 10.27%, and 64.93%. The first two RDA axes effectively reflected most of the relationships between plant assimilated branches ecological stoichiometric characteristics and soil factors, primarily determined by Axis I.

Groundwater depth was found to be the most significant influencing factor, with SWC, pH and STP being ranked highest for their roles in the variation in assimilated branches stoichiometry in both H. persicum and H. ammodendron. In both species, LOC, LTN, and LTP were positively correlated with groundwater depth. Among the soil factors, H. persicum showed a significant positive correlation between soil pH and LTP, as well as between STP and LTN and LTP, while a significant negative correlation between STP and assimilated branches C:P was observed. In H. ammodendron, soil pH was found to be significantly negatively correlated with LTN and LTP. The relationship observed between STP and assimilated branches traits was opposite in H. persicum compared to H. ammodendron. Further significant correlations were also observed between assimilated branches traits such as LTP and N:P, LTN and N:P, and LTN and C:N.

H. persicum and H. ammodendron LOC, LTN and LTP are lower than the global average for shrubs (461.16 g/kg, 20.1 g/kg, 1.77 g/kg) (He et al., 2019). Assimilated branches C:N and C:P were much higher than average (23.8 and 300.9) (He et al., 2019). Soil SOC, STN, and STP were enormously lower than the global terrestrial average (11.12g/kg, 1.06g/kg, 0.65 g/kg) (Luo et al., 2017). Soil N:P (0.25 ± 0.03, 0.22 ± 0.02) and C:P (2.59 ± 1.90, 1.89 ± 0.11) were much lower than the average in China (2.15, 25.77) (Li et al., 2014). The soil C:N (17.24 ± 2.60) under H. persicum was higher than the Chinese average (12.01), while the soil C:N (10.12 ± 0.94) under H. ammodendron was lower than the Chinese average (Luo et al., 2017).

C, N and P are important elements for plant growth, development and genetics (Huang et al., 2021). Leaf C:N and C:P can be used as a measure of the plant’s ability to sequester carbon as it absorbs N and P nutrients (Tian et al., 2021). Attributed to lower rainfall and slower nutrient absorption, there was less nutrient storage in the study area, resulting in reduced nutrient accumulation in the assimilated branches. In barren habitats, plants have a conservative survival strategy of maintaining high C:N and C:P ratios to increase nutrient use efficiency to maximize productivity (Zhang et al., 2020).

Assimilated branches N:P can be used as an indicator of vegetation function and nutrient limitation (Tian et al., 2021). Compared to the global mean (He et al., 2019), the assimilated branches N:P of the two Haloxylon species was low, which may be the result of low LTN content. According to Gusewell’s 10/20 theory, both H. ammodendron and H. persicum assimilated branches N:P ratios are mostly less than 10 and show N limitation (Gusewell, 2004). Nevertheless, the supply of plant N and P was obviously insufficient. Plants also have the ability to maintain stable relationships between their intrinsic elemental content and ecological stoichiometry when faced with dramatic changes in environmental conditions (Elser et al., 2010). Therefore, the nutrient limitation of plants should be considered in a wider range of environmental factors rather than being measured by N:P alone.

The element content of soils and their stoichiometric ratios are important indicators of soil composition and quality (Tian et al., 2009). Soil carbon and nitrogen are mainly derived from the decomposition of vegetation litter (Luo et al., 2017). The foliage of H. persicum and H. ammodendron is degraded, and litter production is much lower than the national average (An and Li, 2014). Soil microbial diversity and biomass are also significantly reduced in arid habitats, and nutrient turnover is slower (Liu et al., 2022). Plants living in barren habitats always have a higher nutrient acquisition cost, so there is often a resorption of nutrients in senescing withered parts, which implies the adaptability of desert plants (Veneklaas et al., 2012). Soil nitrogen can also be derived from rainwater leaching and deposition of atmospheric nitrogen. However, the low precipitation (less than 200 mm per year) in the study area and the longer distance from cities or factories limit soil nitrogen replenishment through this route (Zhou et al., 2012). A combination of many factors results in the reduction of carbon and nitrogen input to the soil. STP is mainly derived from rock weathering (Tian et al., 2021), and the low rate of rock weathering due to drought and low rainfall in the study area contributes to the low soil STP content.

The nutrient content of soil is mainly affected by litter input, while the physical properties are mainly affected by topography and groundwater supply (Zhou et al., 2012; Chen et al., 2021). As groundwater conditions deteriorate, desert plants adopt stricter defense strategies and produce fewer nutrients in the litter (An and Li, 2014). Increased LOC lead to stimulation of soil organic matter turnover and more carbon input (Li et al., 2018), stimulating microbial activity and soil organic matter decomposition, ultimately resulting in reduced carbon stability; thus, the SOC decreases (Chen et al., 2021). The leaf C:N increasing, so relative N content of litter decreases and causes decomposition to consume soil N (Enriquez et al., 1993). Consequently, the STN content of soil decreases, N limitation increases and leaf N:P decreases significantly. This is also consistent with the results of the RDA: GWD is significantly negatively correlated with soil STN and branches N:P. As structural components of organic matter, C and N have relatively fixed ratios in the process of accumulation and consumption (Huang et al., 2021). As a result, soil C:N did not vary significantly among different groundwater depths and soil layers.

It may be the topography that causes the soil properties to differ between the two plants. Topography is important in controlling hydrological and soil processes because it affects the moisture as well as the accumulation and export of nutrients (Sebastia, 2004). Different topographic leads to the redistribution of soil resources, such as the leaching brings nutrients from the top of the slope to the bottom (Li et al., 2010). Regional groundwater was characterized by high salinity of Cl and SO₄Cl, leading to the accumulation of salts in the soil due to the effect of evaporative concentration (Pan et al., 2020). So, the SOC, STN and EC under H. ammodendron were significantly higher than those under H. persicum.

Moreover, other biological processes also affect soil chemical element content. H. ammodendron mainly absorbs NO^3- from the soil and releases OH^-, leading to an enhanced alkalinity and greater pH of the soil under the canopy. (James et al., 2005). The salt-rich litter is imported and accumulates in the soil under the canopy, resulting in a salt island effect. In addition, H. ammodendron absorbs excess salt ions from the soil to maintain osmotic pressure for survival (Xi et al., 2004). Changes in microbial abundance and metabolic pathways in soil lead to lower turnover rates, then increase accumulation of carbon and nitrogen (Liu et al., 2022). The low C:N under the H. ammodendron canopy allows for faster release of N and facilitates uptake (An and Li, 2014), which may explain why both STN and LTN of H. ammodendron are significantly greater than those of H. persicum.

The results revealed a correlation between assimilated branches ecological stoichiometry. A positive correlation was found between LTN and LTP, which is a general pattern in terrestrial plants reflecting the trade-off among leaf attributes (Rong et al., 2015). LTN and LTP were found to be negatively correlated with leaf C:N and C:P, respectively. These reflect the availability of nitrogen and phosphorus nutrients in the assimilated branches of plants, which determine their survival strategy and growth rate (Poorter and Bongers, 2006). Plants employ a trade-off strategy in their assimilated branches ecological stoichiometry by allocating resources and utilizing them efficiently, which ultimately enhances their adaptation to their respective habitats (Harris, 2003).

The density of plants decreased significantly when the depth increased, which proved that their living conditions deteriorated sharply. We speculate that the different sources of water H. ammodendron and H. persicum at different groundwater depths lead to different degrees of stress, thus showing differences in ecological stoichiometry. The H. ammodendron is a groundwater-dependent plant (Dai et al., 2014). However, for H. persicum, the survival cost of obtaining groundwater is too high, mainly using deep soil water (Wu et al., 2019a). When the groundwater is shallow depth, the roots can directly absorb groundwater, and a large amount of deep soil water is used in the early part of the growing season (Wu et al., 2019a). Various water sources weaken its stress, so C:N is lower in shallow groundwater (Zhang et al., 2020). Due to the increase of depth, the utilization of groundwater by H. ammodendron gradually decreased (Wu et al., 2019a). Precipitation decreased significantly from the edge to the center of the desert, too (Chen, 2017). Soil moisture was reduced by dual effect, drought stress became more serious. Plants adapt to deteriorating water conditions by increasing assimilated branches C:N and nutrient reabsorption (Zhang et al., 2020). Compared with the H. persicum, the C:N of H. ammodendron, has a more pronounced increase. The H. persicum adjust its root distribution to obtain more shallow soil water for more restricted resources, and adapt to the increase in groundwater depth (Xu et al., 2017). In this region, not only is the soil N supply limited, but the P content is also sparse. Under low P supply conditions, plants preferentially allocate P to photosynthetic cells to ensure photosynthetic efficiency for survival (Ding et al., 2021). Under intense drought, plants tend to accumulate more nutrients in branches. With the accumulation of P more than N, the N:P is lower in deep groundwater (Cordell et al., 2001). This storage process leads to a decrease in the assimilated branches N:P ratio. The increase of LOC, C:N and C:P demonstrates that plants have adopted stronger defense strategies to survive in an increasingly drought-stressed environment (Poorter and Bongers, 2006; Rong et al., 2015).

Therefore, we believe that the two Haloxylon adapt to changes in groundwater depths through the allocation of water use and branches ecological stoichiometry.