The study area was located at the Linze Inland River Basin Integrated Research Station (LIRBIRS, 39° 21’ N, 100° 07’ E, 1,367 m above sea level) of the Ecosystem Research Network of the Chinese Academy of Sciences, in Pingchuan Town, Linze County, in the central part of the Hexi Corridor, Gansu Province, China ( Figure 1A ). The study area has a temperate continental desert climate, characterized by sparse precipitation throughout the year, which is concentrated in summer, as well as a dry climate, long day lengths, and intense solar radiation; the annual precipitation is 124.3 mm and the average annual temperature is 7.6°C. The annual average wind speed is 3.2 m·s^-1, the maximum wind speed is 21 m·s^-1, and the prevailing wind direction is northwesterly (Zhao et al., 2003). The precipitation and maximum wind speed dynamics during the fencing period ( Figure 1C ), which were determined using meteorological data obtained from the desert ecosystem meteorological long-term experiment sample sites at LIRBIRS, indicated fluctuations but no significant annual increases or decreases. The study area is dominated by semi-fixed dunes with a simple plant community structure and few species. Xerophytes and semi-shrubby vegetation dominate. The xerophytes and ultra-xerophytes do not typically include short-lived or annual plants. Representative plants include Reaumuria songorica, Nitraria tangutorum, Calligonum chinense, Haloxylon ammodendron, and Agriophyllum squarrosum, which exhibit a mixed spatial distribution, constituting a unique patch vegetation pattern. The dominant herb species are Artemisia scoparia, Suaeda glauca, Halogeton arachnoideus, and Eragrostis minor, with sporadic Zygophyllum kansuense and Allium mongolicum.

The importance value, Margalef richness index (Ma), Shannon–Wiener diversity index (H), and Pielou evenness index (E) were selected to express the Reaumuria songorica–Nitraria tangutorum community diversity through a statistical analysis of the community quantitative characteristics using the following formulas:

where the relative density is the density of a species/sum of the densities of all species; the relative height is the plant height of a species/sum of plant height of all species; the frequency of species i in the sample was first calculated as the ratio of the number of species i in the sample to the total number of species in the sample. then, the frequencies of all species in the sample square were calculated.

Margalef richness index:

where S is the number of species in the sample square; and N is the number of all plants in the sample square.

Shannon–Wiener diversity index:

where P_i is the ratio of the number of individuals, n_i, of species i in the sample to the total number of individuals, n, of the species in the sample, P_i = n_i/n and i=1, 2, 3,…

Pielou index:

The data were prepared and statistically analyzed using Microsoft Excel and SPSS19.0 software. The significance of the difference between each soil nutrient parameter in different years of fencing and between different soil layers was tested by one-way ANOVA. Redundancy analysis (RDA) was conducted using Canoco 4.5 software to study the relationship between plant community diversity and environmental factors such as soil nutrients and water content. The correlations between biomass characteristics of the shrub and herb layers, and between species diversity and density, were analyzed using the performance analytics and GGally package in R R Core Development Team 2021 and plotted using Origin 2018.

The number of species, height, and vegetation density in the plant community tended to increase with the number of years of fencing, and the community gradually evolved from simple to complex ( Table 2 ). The plant community was divided into two different levels according to the growth form: a shrub layer and a herbaceous layer. The community had a total of six families, nine genera, and nine species, with two families, two genera, and two species of shrubs (22.2% of the total species richness), and five families, seven genera, and seven species of herbaceous plants (77.8% of the total species richness). Zygophyllaceae play an important role in the vegetation species, accounting for 33.3% of all species.

Prior to fencing, the shrub layer was dominated by N. tangutorum, with R. songorica becoming the dominant species from the second year of fencing and the herbaceous layer becoming dominated by S. glauca in the early stage of fencing (densities of 0.11 m^-2 and 0.13 m^-2, respectively). The dominant species in the middle stage of fencing (4–6 years) was A. scoparia (density = 0.24 m^-2 in year six), whereas the dominant species in the late stage of fencing (7–9 years) were A. scoparia and H. arachnoideus (density = 0.43 m^-2 in year seven). The density of H. arachnoideus increased to 1.48 m^-2 in the ninth year of fencing; however, the perennial herb Z. mucronatum and the grass E. minor colonized at this stage, with the density of the latter reaching 0.83 m^-2. Moreover, the density of the perennial herb Z. kansuense increased to 0.17 m^-2 at this stage, and the plant community gradually stabilized ( Table 2 ).

The community density increased significantly over the study period, and reached a maximum value of 40.14 times the initial community density in the ninth year of fencing. For example, the density of the shrub layer increased from 0.07 to 0.57, and the density of the herbaceous layer increased from 0.32 to 1.40. Notably, the density of H. arachnoideus increased from 0.21 to 1.48 in the late period ( Table 2 ). Development of the vegetation community was characterized by the successful invasion of annual and perennial herbs and a gradual increase in the number of species.

The species composition within the Reaumuria songorica–Nitraria tangutorum community was simple, with plant diversity from the early–middle–late stages of fencing first increasing, then decreasing, then increasing again ( Figure 2 ). The community richness and diversity showed a significant increasing trend in the early stage ( Figures 2A–C ); the richness and diversity in the third year of fencing were 1.56 and 2.44 times higher than the initial values. The diversity index showed a significant decreasing trend, with the lowest value in the middle stage. The richness and diversity indices showed a significant increasing trend in the later stage, and the diversity of perennial as well as annual herb layer plants also increased significantly, causing the plant community structure to gradually stabilize.

The soil water content of the Reaumuria songorica–Nitraria tangutorum community increased then decreased over time ( Figure 3 ). Soil water content increased from 5.25% to 7.74% in the middle of the fencing period, and reached a maximum in the fifth year of fencing; it then decreased from 7.74% to 5.54% in the later years. The soil water content also increased then decreased with increasing depth of the soil layer. In the 0–40 cm soil layer, the soil water content increased significantly with depth; in the middle layer (40–100 cm), the water content was higher than that in the other layers; and in the deep layer (90–150 cm), the soil water content gradually stabilized. In general, the soil water content of the community increased slowly during the study period; however, an abnormally high value appeared in the 5th and 6th year of fencing because of high precipitation in these years. The soil water content then gradually stabilized in the late period, but remained higher than that in the early fencing period.

SOM and TN contents decreased in the middle period and recovered in the later period ( Figures 4A–C ). The available phosphorus and available nitrogen contents reached maximum values in the middle of the fencing period, and were 1.87 and 4.70 times higher in the fourth year than in the ninth year of fencing, showing a trend of increasing then decreasing, before finally reaching a stable state ( Figures 4D, E ). The pH value first increased then decreased over the study period ( Figure 4F ), falling to a minimum value of 7.61 in the fourth year of fencing before gradually stabilizing in the later years of fencing. Furthermore, the nutrient content in the shallow (0–10 cm) soil was higher than that at 10–20 cm, showing clear “surface aggregation.”

According to Table 3 , the Eigen values of the first two axes of the community were 0.745 and 0.163, explaining 74.5% and 16.3% of the species-environment relationships, respectively, with a cumulative variance that explained 90.8%. Moreover, the Eigen values of 0.823 and 0.115 for the first two axes of the shrub layer explained 82.3% and 11.5% of the species-environment relationships, respectively, with a cumulative variance that explained 93.8%. The Eigen values of 0.952 and 0.046 for the first two axes of the perennial herbaceous layer explained 95.2% and 4.6% of the species-environment relationships, respectively, with a cumulative variance that explained 99.9%. Lastly, the Eigen values of the first two axes of the annual herbaceous layer were 0.811 and 0.153, which explained 81.1% and 15.3% of the species-environment relationships, respectively, with a cumulative variance that explained 96.4%. This reflects the ranking information more completely, indicating that the selected environmental factors can better reflect the distribution pattern of plant communities, and that the ranking results are credible.

We observed significant correlations between different plant groups and environmental factors ( Figure 5 ), with the correlation coefficients between different plant groups and soil and environmental factors equal to one. The Eigen values of axis 1 ranged from 74.5% to 95.2%, and those of axis 2 ranged from 4.6% to 16.3%. For the plant community ( Figure 5A ), the diversity, richness, and evenness indices were all affected by soil TN, SOM, available potassium, and pH, which exhibited positive correlations, and the community density was positively correlated with SOM. For the shrub layer ( Figure 5B ), the diversity and evenness indices were positively correlated with soil nutrients, such as available nitrogen and available potassium, soil pH, and deep soil water content, and both indices were positively correlated with the plant growing season temperature and precipitation. In terms of the perennial herbaceous layer ( Figure 5C ), the diversity and evenness index were positively correlated by soil pH, TN, and SOM, but negatively correlated with maximum wind speed during the growing season. As for the annual herbaceous layer ( Figure 5D ), the density, diversity, richness, and evenness indices were positively correlated with SOM, TN, and pH. The growth of annual herbs, as recorded by plant height, was greatly influenced by the shallow soil water content, whereas the diversity and uniformity of annual herbs were more influenced by the soil nutrients.

The duration of fencing and the plant community characteristics (height and density) of the shrub and herbaceous layers showed a significant positive correlation ( Figure 6 ). Shrub height was significantly and positively correlated with perennial herb height and annual herb density, and the shrub layer density was positively correlated with perennial herb height and annual herb density. The increase of shrub layer plant density also had a promotional effect on perennial and annual herbs. When shrubs promote plant growth and settlement, or increase the species richness and diversity of herbaceous plants, a nursing effect is observed between species (Cahill, 2008). This nursing effect of shrubs directly promotes growth of the herbaceous layer, and occurs frequently in arid environments, which is an important reason for plant diversity.

We observed a positive correlation between plant density and the number of species ( Figure 7 ). This is because greater vegetation density changes the habitat conditions, resulting in an increase of surface roughness, which reflects the influence of vegetation density on soil wind erosion, and leads to an improved soil environment, the survival and growth of more plant seeds, and an increase of plant diversity (Dong et al., 1996).

Most of the species in the studied community belonged to the family Zygophyllaceae, with relatively few other plant families present. This is typical of arid desert areas, where shrubs and semi-shrubs dominate and herbs mostly exist as companion species. This is attributed to the dry climate and strong wind erosion in arid desert ecosystems, where shrubs and semi-shrubs can better adapt to the stress of abiotic environmental factors (He and Zhao, 2004). Moreover, R. songorica replaced N. tangutorum as the dominant species in the shrub layer after fencing. This is mainly because R. songorica, as a medium forage grass with better palatability, is more susceptible to human grazing or wildlife activities than N. tangutorum, which is easily eaten by sheep and camels in the marginal areas of desert oases where plant species are very scarce. Therefore, R. songorica gradually recovered and eventually became the dominant species in the shrub layer after natural fencing and elimination of grazing and wildlife disturbance. The results of this study showed that the implementation of fencing measures can significantly promote the recovery of desert vegetation diversity. In the later stage of fencing, the species richness increased significantly, and Z. mucronatum, H. arachnoideus and E. minor began to colonize plots, which inevitably affected the growth of existing dominant herbs. As a result, the importance value of the previously dominant species, A. scoparia, decreased. However, from the perspective of colonizing species, annual herbs still dominated, which indicates that the recovery of desert plant communities is a slow and gradual process, and community succession in the enclosed study area was still in the transitional stage at the end of the study period.

In this study, the species diversity index was approximately consistent with the species richness, that is, the higher the number of species, the higher the diversity index, which is consistent with other studies in the Hunshandak sandy area (Qi et al., 2021) and western Inner Mongolia (Yang et al., 2017). Species diversity is characterized by both richness and evenness, and commonly indicated by the Margalef richness index and Shannon–Wiener diversity index (Jiang et al., 2022). The Pielou evenness index of plant communities was generally high in the study area, which is conducive to maintaining the stability of the community structure.

The results of this study indicated that fencing promoted an increase in shrub density, and that shrubs had a conservation effect on herbs. This is attributed to the fact that very few pioneer herbs can survive and reproduce in desert ecosystems, which are subject to wind damage, sand damage, and erosion, However, because of the ability of scrub to reduce the wind speed, intercept wind and sand, and improve the local microclimate (Zhao et al., 2006; Zhao et al., 2007), these plants promote the survival and development of herbaceous plants beneath, allowing some non-pioneer herbaceous plants to grow and reproduce normally (Francisco and Robert, 2000). Moreover, the species richness of herbaceous layer communities in desert ecosystems is generally higher than that of shrub layer communities, and competition among species of similar taxa for the same resources is intense, which leads to more obvious heterogeneity in the phylogenetic structure and successional development of herbaceous layer species (Chai et al., 2019).

Soil moisture and nutrients play an extremely important role in the degree of vegetation restoration, which is an important indicator of soil quality and soil health (Cai et al., 2020; Liu et al., 2022). In this study, we found that the moisture content of soil in the surface layer (0–40 cm) was low, that in the middle layer (40–120 cm) varied significantly but was significantly higher than that in the surface layer, and that in the deep layer (120–150 cm) was minimal. This is because the surface layer of the desert ecosystem is dominated by sand grains and has a weak water-holding capacity, whereas the middle layer is deposited with powder and clay grains, and exhibits greater soil water-holding capacity. In contrast, the transpiration rate decreases with increasing soil depth, so the soil water content increases to become the main source of soil water for plants (Lu et al., 2021).

In this study, soil nutrients continued to aggregate in the surface layer over the study period; however, the aggregation effect decreased significantly with increasing soil depth. This surface aggregation occurred because of restoration of the shrub- and herbaceous-layer vegetation after fencing, which protected the sub-surface from wind erosion (Wang et al., 2022). Furthermore, the deposition of fine particulate matter under the vegetation canopy, a process that proceeds continuously, leads to an increase in clay-powder particles in the soil surface layer and the accumulation of soil nutrients, especially SOM and TN. However, the available soil nutrient contents showed different trends. This is because the chemical form of fast-acting nutrients is water-soluble and easily mobile (Du et al., 2007). As the vegetation recovered, the plant density and cover increased each year, which would have greatly increased the consumption of available nutrients, resulting in lower levels.

Species diversity is a key indicator of the number, distribution, and stability of species in a community or ecosystem, and includes the species richness, evenness, dominance, etc. (Carvalho et al., 2020). In this study, we found that both the diversity and density of perennial and annual herbaceous plants were influenced by soil SOM, TN, and pH. The soil nutrient status improved as the duration of fencing increased, which led to an increase of SOM and TN contents, providing conditions more conducive to plant survival. Thus, it is beneficial to promote the survival of surface herbaceous plants as they are important for the recovery of species diversity in arid zone desert ecosystems (Abramsky and Rosenzweig, 1984). We also observed a significant positive correlation between shrub layer diversity and soil properties. As shrubs have longer root systems and deeper roots than herbaceous plants, the deep soil water content played an important role in the development of the shrub layer, and soil water content was mostly regulated by precipitation as a way to promote community development, whereas the available phosphorus and potassium promote the formation and growth of shrub root systems (Li et al., 2009). In summary, the process of plant community growth and succession in arid desert ecosystems is a process of plant adaptation to soil nutrient changes in their habitats, as well as the interaction of different soil environmental factors ( Figure 8 ).