Vegetation and soil succession determine the water-use strategy of Caragana korshinskii in a chronosequence of revegetated areas in Ningxia, Northwest China

Revegetation is considered as the most effective way to control sandstorms and prevent further desertification in water-limited regions (Li et al. 2004). A number of studies have shown that revegetation can effectively slow the spread of desertification, reduce wind erosion, and promote local habitat restoration (Wang et al. 2013; Huang et al. 2016). For example, the Shapotou sand-fixing revegetation system established in 1956 has remained stable after more than 60 years of succession. The local eco-environment, such as the diversity of plants and animals, and the soil physical and chemical properties have been improved (Li et al. 2004). However, some problems emerged in recent years, including declines in groundwater and the mortality of early revegetated sand-binding shrubs in some areas, which directly influenced the sustainability and benefits of this ecological restoration project (Li et al. 2013). The main reason for these problems was that we cannot determine how much water the revegetated plants use (Li et al. 2013), as water is the crucial factor determining vegetation restoration and eco-environmental reconstruction in revegetated desert ecosystems (Schwinning & Ehleringer 2001; Li et al. 2014). Thus, quantifying the amount of water used by the plants and finding the main influencing factors will have important scientific and practical significance for future vegetation construction and management.

Plants have gradually evolved special water-use strategies to adapt water deficits through the long-term coevolution and natural selection in arid and semi-arid areas (Wu et al. 2014). A number of studies have found that plants with different functional types and life stages had different water-use strategies (Dawson & Pate 1996; Wu et al. 2016; Zhou et al. 2017; Tiemuerbieke et al. 2018). For example, most opportunistic grasses were utilizing water only available in the short term in the upper soil layers, while shrubs could use water from both upper and deeper layers (Rossatto et al. 2012). This plasticity in water uptake patterns has been considered as an important strategy to allow the co-existence of a diversified plant community in water-limited systems (Yang et al. 2011). However, the water-use strategy may vary with seasons for many species (Eggemeyer et al. 2008; Wu et al. 2016). For example, Quercus ilex, Arbutus unedo, and Phillyrea latifolia at the Prades Holm oak forest in southern Catalonia mainly extract soil water during the cold and wet seasons but increased their use of groundwater during the summer drought (Barbeta et al. 2015). Banksia prionotes trees in Mediterranean-type ecosystems of south Western Australia use the majority of the water from deeper sources in the dry season, but during the wet season most of the water was derived from shallower soil layers (0–40 cm) supplied by plant upper lateral roots (Dawson & Pate 1996). Juniperus ashei in the Edwards Plateau of the USA exhibits a similar switch in water sources (McCole & Stern 2007), which indicates that the plant water-use strategy is a dynamic, flexible process that can change with vegetation and soil succession. Furthermore, different plants in the same habitat may utilize water from different sources. The semi-arid pinyon–juniper Juniperinus osteosperma in southern Utah, USA, takes up summer precipitation, while Chrysothamnus nauseosus mainly uses groundwater (Flanagan et al. 1992). Also, the same plant species growing in different habitats may have different water uptake patterns. Zhou et al. (2011) demonstrated that Nitraria tangutorum and Artemisia ordosica depended mainly on shallow soil water at the Hanggin Banner site, while at the Dengkou and Minqin sites both species obtained water from deep soil layers or groundwater. The water source of Populus euphratica near the river channel originated from river water, while on sand dunes, it mainly used groundwater and multi-layer soil water (Si et al. 2014). Although the water use by single plant species or coexisting plant species in a certain time period has been well studied in diverse ecosystems, we still have limited knowledge about long-term plant water-use strategies at many water-limited ecosystems.

Recent studies have showed that the age-related revegetated plant water-use patterns might fluctuate in arid desert areas (Matzner et al. 2003; Kerhoulas et al. 2013; Song et al. 2015, 2016). Old plants may have deeper roots and, therefore, greater access to deeper water sources than young plants (Dawson 1996; Kerhoulas et al. 2013; Song et al. 2016). For example, Kerhoulas et al. (2013) reported that old (large) Pinus ponderosa Dougl. trees primarily used soil water deeper than that used by young (small) trees, which may help these trees alleviate the impacts of climate warming on their health. Song et al. (2016) reported that the younger Mongolian pine trees (10- and 22-year-old) only used the soil water, whereas the older trees (32- and 42-year-old) utilized both soil water and groundwater in the Keerqin Sandy Land. Moreover, Song et al. (2015) reported that the 21-year-old Mongolian pine plantation trees had the highest water-use efficiency, followed by the 41-year-old plantation trees and the 9-year-old plantation trees. This finding implies that the succession of vegetation will have a very important impact on plant water-use strategies. The succession of vegetation and soil will have a certain impact on local ecological and hydrological processes. Such as in the Shapotou sand-fixing revegetation system, biological soil crust and a sub-soil layer were formed during the vegetation succession, which led to an interception of soil water availability in the topsoil and a dry sand layer in the deep soil layer, so deep-rooted shrubs were gradually eliminated from the community (Li et al. 2004, 2013). Thus, it is important to reveal water-use strategies of plants at different ages, which will help us better understand the adaptability of the revegetated plants in arid desert areas and their feedback mechanisms (Kerhoulas et al. 2013; Wu et al. 2014). However, relatively few studies have concentrated on how the plant water-use patterns change during their life cycles to cope with vegetation–soil succession at different time scales.

Currently, the stable isotopic technique is used to determine plant water-use strategies (Ehleringer & Dawson 1992). The hydrogen and oxygen stable isotope compositions of water (δD and δ¹⁸O) extracted from plant stems accurately reflect the isotope ratio of soil water used by plants, and no isotopic fractionation occurs during soil water uptake by roots (Barbour 2007). By comparing and analyzing the isotopic characteristics of xylem waters and soil water at different soil layers, linear mixing models and the software IsoSource can be used to calculate the contribution of each potential water source by plants (Phillips & Gregg 2003; Phillips et al. 2005). This technique has been widely applied in the fields of forestry, ecology, and hydrology (Gazis & Feng 2004). Extensive studies have also been conducted to examine the plant water-use source in many desert areas, including Southwest America (Schwinning et al. 2002), West Asia (Gat et al. 2007), and Northwest China (Cheng et al. 2006; Huang & Zhang 2015). Yet, little is known about the water-use strategies of different-aged C. korshinskii in a chronosequence of revegetated areas.

C. korshinskii is an important shrub species for vegetation rehabilitation in North China for its high ecological and economic values (Zheng et al. 2004). They were planted in dune areas after straw checkerboard sand barriers were erected on shifting sand surfaces, and now over 40,000 hm² of C. korshinskii were revegetated in Northwest China (Li et al. 2004). However, investigations on C. korshinskii have mainly focused on its seed germination (Zheng et al. 2004), above-ground properties and root biomass (Li et al. 2004), evapotranspiration (Huang & Zhang 2015), and its effects on long-term soil water dynamics. Although our previous studies addressed that C. korshinskii mainly tapped water from the 100 cm soil layer in a mixed plantation of C. korshinskii and A. ordosica (Huang & Zhang 2015), questions still exist about the water uptake in monocultures of C. korshinskii at different ages during the growing season. Moreover, quantitative studies of soil and plant succession are also lacking. Thus, in this study, C. korshinskii growing at four different-aged (established on 1976, 1987, 1996, and 2005) revegetation sites in a desertified steppe ecosystem of Northwest China was selected, and the isotopic composition of xylem and soil water was analyzed. The aim of this study was to examine how plant water utilization strategies vary during the long-term soil–plant succession. Our goal was to provide a scientific basis to understand plant water-use mechanisms and to achieve sustainable management of water and soil resources in arid revegetated desert areas.

We conducted our study in northern Yanchi County (37 °04′–38 °10′N and 106 °30′–107 °41′E), which is located at the southwest fringe of Mu Us Sandy Land in the Ningxia Hui Autonomous Region, China (Figure 1). The climate at the site is characterized by temperate continental semi-arid monsoonal climate. The mean annual precipitation is 292 mm, with over 70% of the total precipitation occurring from June to September. The elevation is 1,295–1,951 m above sea level, the mean annual temperature is 7.5 °C, and the lowest and highest monthly mean temperatures are −8.7 °C in January and 22.4 °C in July. The mean potential pan-evaporation during the growing season (April–October) is 2,710 mm. The windy season lasts from September to April, with an average wind velocity of 2.8 m s⁻¹ (Liu et al. 2015).

Over the past decades, the study region has undergone a sharp increase in the extent of desertification since 1950s. This process was created mainly by extensive fuel wood gathering and overgrazing (Liu et al. 2015, 2016). In order to restore the desertified lands, the local government implemented a conservation program in the 1970s in which many native plant species, including the Caragana shrub, were selected for long-term afforestation to stabilize the mobile sand land. This ecological shelter was extended in 1976, 1987, 1996, and 2005. Other government measures, such as enclosures to exclude animals and the banning of grazing, an artificial shrub-land ecotone was established in Ningxia, Northwest China. Now, the Caragana plantations account for 33.37% of the entire sandy grassland, the environment has improved, and the stabilized sand surface has created conditions that support the colonization of a number of herbaceous species. Until know, the original desertified lands have evolved into a complex, man-made, and natural desert vegetation landscape through the large propagation of C. korshinskii (Liu et al. 2016).

Three 10 m×10 m quadrats were randomly selected in each sand-binding revegetation plots established in previous years (1976, 1987, 1996, and 2005). The crown size, height, basal diameter, coverage, and biomass of the C. korshinskii community were measured. In each sampling point mentioned above, three 1 m×1 m quadrats were designed randomly to investigate the species richness, density, and coverage of the herbaceous plants. Soil water content (SWC) was measured with the oven drying method, and they were taken at 11 different depths: 0, 20, 40, 60, 80, 100, 120, 150, 180, 210, and 250 cm with three replicates at each depth and then dried at 105 °C for 24 h. At the each above 10 m×10 m vegetation quadrats of different revegetation establishments, soil samples were obtained from the surface 0–20 cm by a 5-point sampling method; thus, we have three soil sample replicates in each revegetation establishments. Each sample was air-dried, crushed, and passed through a 2 mm sieve, and the physical and chemical properties of each sample were analyzed in the laboratory. Particle size was analyzed using a MS-S light scattering apparatus (Malvern Instruments, Malvern, UK). Soil bulk density was measured with the ring-cutting method (Nanjing Institute of Soil Research, Chinese Academy of Sciences 1980). Soil organic carbon (SOC) was determined according to the Walkley–Black dichromate oxidation method (Nelson & Sommers 1982). Total nitrogen (TN) was measured with a Kjeltec System 1026 Distilling Unit (Tecator AB, Sweden). Total phosphorus and potassium were measured with the standard analysis methods developed by the CERN (Liu 1996). Available nitrogen was determined using the alkaline diffusion method, available phosphorus was determined using the Bray method, and available potassium was determined by means of flame spectrometry after extraction with 1 mol/L NH₄OAc (ISSCAS 1978). All fieldworks were carried out in August, 2017.

Three mature C. korshinskii plants with an average height and breast height diameter in each revegetated site were chosen for isotopic analysis. Twig samples were collected every month during the growth season from April to October in 2017. They were cut from live branches of the three selected individuals and then immediately stored in the small vials sealed with Teflon-lined screw caps and parafilm. Plant samples were collected at a fixed time (10 AM on the 20th of every month) in each collection day and then taken to the laboratory immediately. Concurrent with plant tissue sampling, soil samples from each revegetated site were collected with a bucket auger at 10 depths (20, 40, 60, 80, 100, 120, 150, 180, 210, and 250 cm) around the three selected plants. They were also sealed in vials with Teflon-lined screw caps and parafilm. Plant and soil samples were stored in the freezer (−20 °C) until water extraction. Precipitation was measured with a standard rain gauge, and the rainwater was collected with a polyethylene bottle. In each precipitation, rainwater samples were immediately enclosed in air-tight glass vials, wrapped with parafilm, and stored in the freezer (4 °C) until stable isotope analysis.

Plant and soil properties at different revegetated sites were compared by single factor analysis of variance (one-way ANOVA), and Tukey's test was used for post hoc multiple comparisons. SPSS 13 (SPSS 13.0 Inc., Chicago, IL, USA) and Origin 7.0 software (OriginLab Corporation, Northampton, MA, USA) were used for data analysis and graphing. The SWC was divided into three layers (0–60, 60–120, and 120–250 cm) according to the coefficient of variation (CV): (1) 0–60 cm shallow soil layer: CV was larger than 20%. (2) 60–120 cm middle soil layer: CV ranged from 10 to 20%. (3) 120–250 cm deep soil layer: CV ranged from 0 to 10%. Thus, for the convenience of subsequent comparison and analysis, the above three potential soil water sources were identified and the IsoSource mixing model was used to calculate the proportions of plant water sources from different soil layers (Phillips & Gregg 2003; Phillips et al. 2005). The fraction increment was set at 1%, and the mass balance tolerance was set at 0.1%. The boosted regression tree (BRT) model was used to model the relationship between environmental variables and plant water sources in different revegetated sites. The BRT model is a recently developed technique, combining the advances of traditional regression models and machine-learning methods (Tonkin et al. 2015). Unlike traditional statistical approaches where a single parsimonious model is fitted, BRTs are an ensemble method whereby many simple models are combined to improve model performance (i.e., Boosting), using recursive binary splits to relate responses to predictor variables (i.e., Regression trees) (Elith et al. 2008). BRTs are robust to variable collinearity, variable outliers, and missing data and can include both categorical and continuous variables. BRTs have performed well in determining the important independent variables and solving the problems of classification and forecasting (Aertsen et al. 2011), and has become increasingly popular in hydrology (Naghibi et al. 2016), soil science (Martin et al. 2009), ecotoxicology (Lampa et al. 2014), and ecology (Elith et al. 2008; Golivets et al. 2019).

BRT models were operated in R (version 3.6.2) using the ‘gbm’ package (Elith et al. 2008). In the BRT analysis, we chose Gaussian as the error structure for the loss function because of the attribution of our response variable. By trial-and-error, we found that the best performing parameters for our data set were 0.005 for the learning rate, 0.5 for the bag fraction, and 5 for the tree complexity. We selected 12 environmental variables that could be associated with vegetation and soil succession according to a detailed literature review and dengue expert knowledge. These variables represent the vegetation growth, soil moisture, and soil physical and chemical properties. The growth characteristics of vegetation include the crown width, basal diameter, coverage, and the number and coverage of herb species. The soil physical properties included soil texture (content of clay + silt), bulk density; soil moisture properties included soil moisture in shallow, middle, and relative stable layers; and soil chemical properties included SOC and TN.

As presented in Table 1, the height and basal diameter of C. korshinskii increased with age. After 12 years of revegetation, mean plant height and basal diameter were 0.89 m and 1.03 cm, and they increased to 1.68 m and 1.76 cm after 41 years of revegetation. However, plant crown size, coverage, and biomass increased first and then decreased. In the 2005 revegetated site, crown size, coverage, and biomass were 1.78 m², 12.88%, and 58 g/m², respectively, and reached a maximum of 3.58 m², 23.71%, and 92.10 g/m² in the 1996 revegetated site. Then they gradually decreased to the previous state in the 2005 revegetated site. However, the difference between different age-related revegetation sites was not significant. The richness, abundance, and coverage of herb species increased linearly with time (Table 1), and five species found after 12 years of revegetation were Agriophyllum squarrosum, Bassia dasyphylla, Eragrostis poaeoides, Scorzonera divaricate, and Corispermum declinatum. However, the increase rate was lower in the older revegetated areas, which means that the revegetated ecosystem gradually becomes stable with vegetation succession, and the species composition reached its maximum capacity. The abundance and coverage of the herbaceous plants reached a maximum of 265.03 /m² and 21.91% after 41 years of vegetation succession. The variations in soil texture and soil nutrients observed in different age-related revegetation sites were significant (Table 2). Such as in the 2005 and 1976 revegetation sites, the contents of clay and silt (<0.05 mm) in the topsoil (0–20 cm depth) had increased from 12.86 to 25.66%. However, the contents of fine and coarse textured soils had decreased from 72.77 to 68.51% and 14.37 to 5.83%, respectively. Soil bulk density has been decreasing since vegetation reclamation, but no significant differences were found among the different sites (P > 0.05). SOC, total N, P, K, and available N, P, K all increased significantly in the period following revegetation. However, the incremental rates were higher in the younger revegetated sites than that in the older revegetated sites.

During the experimental period, the annual precipitation was 193.8 mm, and the precipitation during the rainy season (from April to October) was 166.28 mm, which accounted for 85.8% of the total annual precipitation (Figure 2(a)). The δD and δ¹⁸O values of the local precipitation ranged from −10.21 to 8.87‰ and −86.13 to 59.16‰, respectively, below the global meteoric water line (GMWL) (Figure 2(b)). The distribution of δD and δ¹⁸O had an obvious seasonality from spring to winter, the lowest values were observed in the winter seasons, such as in October 24, 2017, the δD and δ¹⁸O reached its minima values as low as −86.13 and −10.21‰, and the highest values were observed on April 19, 2017, the δD and δ¹⁸O reached its maxima values as high as 59.16 and 8.87‰, respectively. The mean δD and δ¹⁸O values of the local precipitation were −39.2 ± 4.3 and −5.7 ± 0.6‰, respectively.

The SWC changed spatially and temporarily since revegetation (Figure 3(a)). The mean SWC in the shallow soil layer (0–60 cm) was 2.2–4.9% at different revegetated sites, but the difference was not significant. However, the SWC increased with depth in the deeper soil layers (60–250 cm), and the mean SWC in the older revegetated areas (e.g., the 1976 site) was 2.5–3.4%, which was significantly lower than that in the younger revegetated areas (e.g., the 2005 site, 5.2–6.9%). The SWC could be divided into three layers (0–60, 60–120, and 120–250 cm) according to the CV variations (Figure 3(a)). The variation of SWC in the shallow layer was large, ranging from 2.2 to 4.9%, and the CV was more than 20%. However, the variation was relatively small at the 60–120 cm soil depth (middle soil layer) because of the less infiltration and higher root density, with the CV ranging from 10 to 20%. The variation of SWC was relatively stable at 120–250 cm soil depth (the deep soil layer), with CV ranging from 0 to 10%. Mean soil water δ¹⁸O values ranged from −11.95 to −2.77‰, but the variation decreased with the increase of vegetation establishment age. Similar to the variation of SWC, soil water δ¹⁸O values increased with soil depth in the shallow 0–60 cm and then decreased with soil depth in the 60–120 cm layer but were stable at the 120–250 cm soil layer, where the δ¹⁸O values were −10.2 to −9.1‰ (Figure 3(b)).

During the growing season in 2017, the SWC fluctuated during the entire measurement period and was mainly affected by precipitation (Figure 4(a)–4(d)). The SWC was higher in August and September because of the increasing precipitation events in this period, but the soil moisture dynamics in different revegetated sites are also different, such as the SWC at the 2005 revegetated site was significantly higher than that at other sites, and the SWC was highest in the 120–250 cm soil profile, which increased from 4.28% in April to 8.07% in August. In the middle 60–120 cm and upper 0–60 cm part of the soil profile, the SWC also exhibited obvious seasonal changes, with mean values of 4.78 and 3.19%, respectively. However, the sensitivity of each layer of soil moisture to precipitation was reduced with increasing site age. For example, in the 1976 revegetated site, the variation of SWC in each layer remained between 2.15 and 3.83%, and the SWC exhibited no obvious seasonality.

The δ¹⁸O values of soil water at the four study sites varied between different months and exhibited a monotonic decline with increasing soil depth (Figure 4(e)–4(h)). In the upper 0–60 cm soil profile, and the soil water δ¹⁸O values were higher and ranged between −5.51 and −3.68‰, −5.75 and −3.76‰, −5.54 and −3.03‰, and −5.21 and −2.11‰ in the 1976, 1987, 1996, and 2005 revegetated sites, respectively. The soil water δ¹⁸O values in the middle 60–120 cm had a similar tendency but were smaller than those at the surface soil layer. However, the variation of δ¹⁸O values in the 120–250 cm soil layer had the largest fluctuations, which respectively ranged between −2.94 and −10.61‰, −11.40 and −8.27‰, −10.31 and −5.69‰, and −8.35 and −3.66‰ with decreasing site age. Furthermore, the variation of soil water δ¹⁸O values in each soil layer in the older revegetated areas was lower than that in the younger revegetated areas. From April to October, the average δ¹⁸O values for C. korshinskii were −10.19 ± 0.42, −7.79 ± 0.37, −6.49 ± 0.51, and −4.43 ± 0.13‰ in the 1976, 1987, 1996, and 2005 revegetated sites (Figure 4(i)–4(l)), respectively. There were significant differences in plant δ¹⁸O values in different revegetated sites, which indicated the diversity and complexity of the water-use strategies of C. korshinskii in different growth stages.

The contribution of each water source (shallow, middle, and deep soil layers) consumed by C. korshinskii at different ages is shown in Figure 5 for each sampling month. The mixing IsoSource model revealed that plants were able to obtain water from the three water sources simultaneously, but in varying amounts. In the 1976 revegetated site, the contribution of shallow soil water to the total water use was 13.3–16.3%, while that of middle and deep soil water was 29.1–37.2 and 49.1–54.8%, respectively. They had slight monthly fluctuations. However, in the 1987 revegetated sites, the contribution of shallow soil water increased to 17.2–21.4% and the proportion of relatively stable soil water decreased to 34.8–50.1%; in the 1996 revegetated sites, the contribution of shallow soil water increased to 18.3–33.1% and the proportion of relatively stable soil water decreased to 32.1–45.2%; in the 2005 revegetated sites, the contribution of shallow soil water increased to 24.8–38.6% and the proportion of relatively stable soil water decreased to 13.1–32.2%, respectively. They all fluctuated seasonally and exhibited a clear shift in water use during the growing season from April to October, especially in the younger revegetated sites. For example, in the 2005 revegetated site, C. korshinskii mainly used shallow and middle soil water from April to June, but the utilization of middle soil water increased rapidly in July and August, and from September to October, the utilization of middle soil water increased slightly (up to 33.6–59.3%).

As presented in Figure 6, the usage of shallow soil water in the 1976 revegetated site was mainly affected by SOC (45.3%) and the content of clay + silt (30%), while the usage of middle soil water was mainly affected by the content of clay + silt (45.3%), SOC (11.2%), deep SWC (10.2%), and TN (9.9%). The proportion of soil water utilization in the deep stable layer was mainly affected by the content of clay + silt (26.8%) and SOC (25.5%). In the 1987 revegetated site, the use of shallow soil water was mainly affected by the content of clay + silt (41.4%) and SOC (26.6%), while the usage of middle and deep soil water was mainly affected by plant basal diameter (43.5 and 46.2%). In the 1996 revegetated site, the three most influential variables were plant basal diameter (34.8%), deep SWC (31.2%), and shallow SWC (19.3%) in the shallow soil water utilization layer, while the usage of middle and deep soil water was mainly affected by plant basal diameter (56.7 and 42.7%) and middle soil layer water content (23.8 and 37.7%). However, in the 2005 revegetated site, the usage of shallow, middle, and deep soil water was mainly affected by shallow (85.9%), middle (95.2%), and deep SWC (39.8%), respectively. Our results indicated that the SWC may determine the plant water-use strategy in the initial stage of plant revegetation, but in the middle stage of vegetation establishment, plant biological characteristics may play a dominant role. However, in the later stage of vegetation establishment, soil physical and chemical properties may be crucial in plant water-use strategies.

Here, we used the space-for-time substitution approach, which has been widely used to infer long-term temporal trends in vegetation and soil succession across ecosystems (Li et al. 2004; Huang et al. 2016; Hu et al. 2019), and found that the reconstruction of the C. korshinskii shelterbelt in degraded sandy lands caused substantial beneficial changes in soil physical and chemical properties. The dominant soil particle class changed from coarse sand to a combination of fine sand and silt + clay. This change showed that the intensity of wind erosion decreased with vegetation reconstruction as the smaller particles are highly vulnerable to wind erosion, and so the silt + clay fractions were gradually being deposited (Chen & Duan 2009). Accompanying this change in soil texture, the content of soil nutrients also increased (Table 2). The changes in soil nutrient contents and their availability were strongly controlled by the changes in the soil particle size distribution (Chen & Duan 2009). The increase in the vegetation cover (e.g., increased litter fall and increased shading that reduced the decay rate of SOC) and biological soil crust also provided important carbon and nitrogen sources to soil systems (Huang et al. 2014). During this restoration process, the habitat changes would further affect the succession of vegetation (Li et al. 2010; 2013). Studies have shown that when the soil moisture content of the shrub root distribution layer is less than 1.5%, the growth of each shrub species begins to be inhibited so that the ground cover of the shrub can only be maintained to a lower value (<10%), and some deep root shrubs will begin to withdraw due to water stress (Li et al. 2004). Therefore, we found that the C. korshinskii began to decline after 21 years of revegetation, and the diversity and coverage of herbaceous plants gradually increased (Table 2). In addition, resource competition between shrubs and grasses was considered as another important factor affecting plant and soil properties. C. korshinskii has showed obvious low-density promotion and high-density inhibition effects on grassland restoration (Liu et al. 2015, 2016). The process of vegetation change at different revegetated sites was the density effect caused by the growth of C. korshinskii community. This effect was consistent with the vegetation and soil succession in the Tengger Desert and the revegetation system in the Loess Plateau of China (Shao et al. 2018). However, the plant diversity in this study was smaller than that in the Tengger Desert. The main reason the diversity was smaller in this area was mainly because the area was planted with monocultures of C. korshinskii, while in Tengger Desert, more than 10 shrub species, including C. korshinskii, A. ordosica, Hedysarum scoparium, Caragana intermedia, Calligonum arborescens, and Atraphaxis bracteata were selected in a mixed plantation. So, in future revegetation projects, a greater diversity of sand-fixing plants should be selected for revegetation rather than monoculture plantations.

Precipitation is considered as the main water source in revegetated desert ecosystems (Schwinning & Ehleringer 2001; Li et al. 2013), and its isotope composition showed seasonal fluctuations, with enriched values occurring in the hot summer months and depleted values in spring and autumn. The reason was that precipitation isotopic composition differences were mainly influenced by precipitation amount, temperature, and local recycling of water vapor (Wu et al. 2016). The δD and δ¹⁸O values of local precipitation fell along or below the GMWL, which suggested the occurrence of substantial soil evaporation enrichments relative to rainwater (Dawson & Pate 1996). These findings were consistent with those of previous studies in other arid and semi-arid communities (Huang & Zhang 2015; Wu et al. 2016; Cui et al. 2017; Zhou et al. 2017). Due to the scarcity and variability of precipitation, the soil water availability in desert ecosystems is often highly variable in space and time (Schwinning & Ehleringer 2001). Vertical soil moisture had obvious stratification characteristics, which can be divided into three layers. We found large fluctuations in the surface soil (0–60 cm) profile as the shallow soil layers are impacted by evaporation and precipitation recharge more frequently than the deeper soil layers. In the middle soil layer, the change of soil moisture was mainly affected by root water absorption, as the root system of C. korshinskii was mainly distributed in the 100–150 cm layer (Zhang et al. 2009). Soil moisture was relatively stable at the deep soil layers as there was no soil water recharge in this layer. Consistent with soil water variations, the isotopic variations of soil water δ¹⁸O and δD values gradually stabilized with increasing soil depth, especially in older revegetated sites, mainly because of the lower soil water evaporation rates, and soil hydrological processes (e.g., infiltration) gradually became weaker with increasing soil depth (Gazis & Feng 2004).

The seasonal fluctuations in SWC and soil water δ¹⁸O were larger in the shallow soil layers, which may be attributed to the precipitation infiltration, soil water redistribution, and the evaporation fractionation in the study area (Huang & Zhang 2015). The isotopic composition of plant xylem water in our study exhibited typical seasonal variability, which suggested that water sources absorbed by plants from various soil depths differed throughout the growing season (Figure 4). During the early spring in April and May, soil water was complemented by snowmelt, which could lead to higher SWC and more depleted soil water isotopic compositions in the shallow soil layers (Huang & Zhang 2015; Cui et al. 2017). Thus, plant roots extract resources from shallow soil layers (when available) to minimize plant energy expenditure (Warren et al. 2015). Another reason may be that the presence of shallow soil water may provide an opportunity for nutrient acquisition during the cool spring months (Williams & Ehleringer 2000). However, during the drier seasons from June to August, less precipitation coupled with higher temperatures and strong evaporative demand was probably the main factor that caused the δ¹⁸O enrichment of soil water in the shallow soil layers (Dai et al. 2015). Therefore, plants switched to obtain water from middle and deep soil layers when suffering prolonged periods of drought. In September and October, the study region had received a large amount of precipitation, so more water was stored in the shallow soil layer (Figure 4). This finding was similar to studies conducted in the savanna ecosystems, where woody plants undergo spatial partitioning and temporal shifts in water uptake patterns to avoid interspecific competition with more shallow-rooted species at the same site (Dawson 1996).

The plant water-use strategy was considered to be an integration of the complex interactions between species of different functional types and the prevailing environmental conditions (Xu & Li 2006). It should be a dynamic, flexible process that can change with vegetation and soil succession. In our study, the seasonal fluctuations of δ¹⁸O in the soil water and plant xylem were not significant in the older revegetated sites (Figure 4), which meant that seasonal variations in δ¹⁸O were not only controlled by water input and evaporation but were also influenced by plant and soil conditions (Li et al. 2013).

In the early stage of revegetation (sites 2005 and 1996), C. korshinskii mainly used soil water from the middle and shallow soil layers, and the contributions of these two water sources can reach up to 70.5–84.5 and 58.3–77.5%, respectively. However, this ratio declined gradually to 50.5–64.3 and 45.3–51.2% in the 1987 and 1976 revegetated sites, while the usage of deep stable soil water increased from 25.8 to 50.1% with vegetation succession. The difference in soil water use of C. korshinskii at different revegetated sites might be related to the regional soil water environment, plant growth characteristics, different root distributions, and heterogeneous soil conditions. For instance, at the 2005 revegetated site, the C. korshinskii usage of shallow, middle, and stable soil water was mainly determined by the SWC at each soil layer (Figure 6). The reason may be that the soil water availability and plant survival were the primary strategy at this stage in arid and semi-arid areas (Li et al. 2014). Some other researches also confirm this conclusion. For example, the main water source for 10-year-old Haloxylon ammodendron was determined by the soil water conditions during the season (Zhou et al. 2017). Hippophae rhamnoides showed plasticity in switching between water from shallow and deep soil layers depending on the availability of soil water (Wu et al. 2016). However, in the middle stage of revegetation, such as in the 1996 and 1987 revegetated sites, the plant water-use strategy was mainly determined by plant base stems and then the SWC (Figure 6), which means that plant characteristics might play a major role in the water utilization in this stage with their different root distributions for better survival and propagation (Si et al. 2014; Wu et al. 2014; 2016; Zhou et al. 2017). Previous results have showed that C. korshinskii have dimorphic root systems at 100–150 cm (Zhang et al. 2009), which enable them to access different water sources during the long-term drought environment. The 5-year-old H. ammodendron mainly used shallow soil water, but when roots reached sufficient depth, they can exploit a deeper water source (Zhou et al. 2017). This means that plants under drought stress may alter their root systems to maintain function and growth (Asbjornsen et al. 2008). It has been reported that Alhagi sparsifolia root could penetrate 12–30 m deep in extremely arid regions, and Calligonum mongolicum have strong root architecture with taproots that extend 3 m in depth and horizontal roots that extend up to 30 m to obtain most of their water from unsaturated soil (Cui et al. 2017).

However, in the later stage of revegetation, such as at the 1976 revegetated site in our study, root distribution may not be a reliable indicator of actual water uptake dynamics (Meinzer et al. 2001). The water use of plants was mainly affected by soil physical properties. Soil texture and SOC were the main factors that influence the water use of C. korshinskii (Figure 6), as soil physical properties were different among different soil layers, which may affect the precipitation infiltration and evaporation. The shallow soil layers had larger particle sizes than deeper soil layers, which allowed rapid infiltration into deep soil layers, which may affect the vertical distribution of SWC, isotopic composition, and the depth of the groundwater table, which indirectly affected plant water sources (Bahejiayinaer et al. 2018). This result agreed with a previous study by Dai et al. (2015), which showed that the spatial variability of soil particle size and porosity results in differences in the plant water-use strategy. It was reported that the water-use pattern of Kalidium foliatum was in disagreement with its root distribution pattern (Gao et al. 2010), and H. ammodendron growing on heavy-textured soil mainly used upper soil water and it might be able to use deeper water sources when growing on sandy soils (Xu & Li 2006). Another comparative study in Dun-huang included three different habitats with different soil types and found that soil heterogeneity determined the water sources and water-use efficiency of desert plants (Cui et al. 2017). Similar results have been reported in other places (Rong et al. 2011; Goldsmith et al. 2012; Penna et al. 2013), which showed that the tree size or age does not affect water uptake. For example, Goldsmith et al. (2012) reported that plant species at different sizes (ages) used water from the same soil layer in a seasonally dry tropical montane cloud forest. Song et al. (2018) reported that the natural Mongolian pine trees consistently use shallow water throughout the growing season in Hulunbuir Sandy Land, regardless of the soil water condition and age. They found that the Mongolian pine trees have a strong lateral root system with a depth of 30–40 cm in the topsoil layer, with the majority of roots only growing in the top 20 cm soil layer (Zhu et al. 2005). Also, our 1987 revegetated site may reach its maximum carrying capacity for vegetation restoration, and C. korshinskii began to degenerate as seen in Table 1. Therefore, the water use of plants was stable and was mainly affected by soil texture and SOC.

Here, we first studied the changes in habitat characteristics of C. korshinskii in a chronosequence of revegetated areas in Ningxia, Northwest China, and then employed stable isotope techniques to determine water sources used by the C. korshinskii trees of different ages. The results showed that the revegetation of C. korshinskii has a positive effect on local habitat restoration. The SWC in each revegetated site could be divided into shallow, middle, and deep layer, and the δ¹⁸O values of soil water at the four study sites varied between different months and exhibited a monotonic decline from the soil surface to the deep soil layers. The δ¹⁸O of stem water of C. korshinskii decreased with the increased vegetation age. C. korshinskii mainly tapped water from the middle and deep soil layers in the older revegetated site. However, at the younger revegetated site, C. korshinskii exhibited a clear shift in water use during the growing season between April and October. In the initial stage of vegetation establishment, the SWC plays a major role in the plant water-use strategy, but in the middle and later stages of revegetation, plant biological characteristics and soil physical properties were the main factors impacting the plant water-use strategy, which indicated that vegetation and soil succession determines the water-use strategy of C. korshinskii in revegetated desert areas.

This work was supported by the Chinese National Natural Scientific Foundation (Grant Nos 41977420), Key Research and Development Program of Ningxia Hui Autonomous Region (No. 2021BEG02009) and the Open Fund of Breeding Base for the State Key Laboratory of Land Degradation and Ecological Restoration in Northwest China and the Key Laboratory for Restoration and Reconstruction of Degraded Ecosystem in Northwest China of Ministry of Education. We thank Russell Doughty at the California Institute of Technology for English editing of the manuscript.

Data cannot be made publicly available; readers should contact the corresponding author for details.