The study area is located in Mu Us sandy land (107°4’–109°26’ E, 38°22’–39°15’ N) in northern China ( Figure 1 ). The sandy land has a temperate continental climate, with a mean annual temperature and a mean annual precipitation of 6.2°C and 250mm, respectively (Liu et al., 2021). Artemisia ordosica, Salix psammophyla, and other psammophytes grow on sandy land. The sandy land contains various soil types, such as Kastanozems, Solonchaks, and Histosols (Liu et al., 2021). This area has begun implementing aerial seeding restoration measures since the 1980s, mainly by sowing seeds of Hedysarum laeve Maxim and Hedysarum scoparium.

We selected 32 sites in Mu Us sandy land, including 30 sites in different years for aerial seeding restoration (1983–2015, excluding 1986, 1988, and 2003), one mobile dune site (2017), and an original site (top community and undisturbed >30 years). We established a 10 m×10 m plot at relatively flat area in each site and set up three 1 m×1 m quadrats along the diagonal at each plot.

We conducted the vegetation and soil survey in August 2017 (plant growing season). We investigated, recorded, and collected the plant species in each quadrat and harvested the aboveground biomass of each quadrat by drying it to constant weight in an oven at 65°C. The soil cores were taken at 60 cm depth in each quadrat because precipitation can affect soil moisture up to 60 cm deep (Yu et al, 2018). We thoroughly mixed the samples from the same site, and stored them at 4°C to determine soil nutrient content.

We identified healthy and pest-free individual plants of each species found in sites and determined seven functional traits associated with plant adaptation to drought conditions. Functional traits included plant height (H), leaf area (LA), leaf dry mass (LDM), specific leaf area (SLA), leaf carbon content (LCC), leaf nitrogen content (LNC), and leaf phosphorus content (LPC). H was estimated by measuring the natural vertical height of fifteen individuals of each species. We sampled five whole leaves from each plant and measured LA using a leaf area meter (LI-3100 Area Meter, LI-COR, Lincoln, United States). LDM was obtained by drying and weighing leaves at 65°C. SLA was calculated as the ratio of LA to LDM. We used an elemental analyzer (Euro Vector EA3000; Milan) to determine LCC and LNC and the ammonium molybdate spectrophotometric method to determine LPC (Yan et al., 2019).

Based on the method proposed by Jin and Qian (2022), we obtained the phylogenetic information of species and constructed phylogenetic trees. This method is based on the super phylogenetic tree of 74,531 species of vascular plants worldwide, which can be quickly obtained fine phylogenetic trees in the study. First, we consulted all plant names recorded in the survey using the World Plants (WP) database (https://www.worldplants.de) to obtain recognized species names. Then, we used the ‘phylo.maker’ function in the ‘V.phylomaker2’ package to construct phylogenetic relationships for the species based on the collated list in this study (Jin and Qian, 2022).

We selected two climate factors and six soil nutrient indicators as environmental factors. For climate factors, we selected mean annual temperature (MAT, °C) and mean annual precipitation (MAP, mm), extracted for each site from 30 arc-seconds spatial resolution raster from the WorldClim (v2.1) database (http://www.worldclimate.org/) using ArcGIS software (v10.7) (Fick and Hijmans, 2017). The climate data from the WorldClim (v2.1) database ranged from 1970 to 2000. Moreover, we selected total nitrogen (TN), total organic carbon (TOC), available phosphorus (AP), total phosphorus (TP), nitrate-nitrogen (NO₃ ^--N), and ammonium nitrogen (NH₄ ⁺-N) as six soil nutrient indicators. We measured TN using the Semimicro-Kjeldahl method; TOC using the potassium dichromate heating oxidation method. The AP was determined by the sodium bicarbonate (NaHCO₃) leaching-Mo-Sb colorimetric method; TP by the alkali fusion-Mo-Sb colorimetric method; NO₃ ^--N and NH₄ ⁺-N by an AA3 continuous flow analytical system (Zhang et al., 2017; Su et al., 2022).

We characterized plant diversity from taxonomic, functional, and phylogenetic diversity. Taxonomic diversity was represented by species richness. Functional diversity was represented by Euclidean distance of function traits among species in the community. Phylogenetic diversity was represented by phylogenetic distance among species in the community.

To investigate the response of plant taxonomic, functional, and phylogenetic diversity to aerial seeding restoration, we used a general linear model to construct a response curve of diversity of three dimensions to aerial seeding restoration time respectively. Previous studies have found that the plant diversity change with the restoration time at the late restoration period showed a threshold effect (Liu et al., 2021). In this study, we used a log-linear model to fit the corresponding curves and calculated the response threshold of plant diversity in three dimensions to aerial seeding restoration time. To explore the response of soil nutrients to aerial seeding restoration, we evaluated the relationship between soil nutrients indicators (including TN, TOC, AP, TP, NO₃ ^--N, and NH₄ ⁺-N) and restoration time using a general linear model.

To explore the relationship between plant taxonomic, functional, and phylogenetic diversity and environmental factors, we used Pearson correlation analysis to calculate the correlation between the plant diversity of three dimensions and soil nutrients and climate factors. Further, to clarify the relative contributions of time, soil nutrients, and climate factors to the diversity of three dimensions during the aerial seeding restoration process, we conducted the variation partitioning analyses to divide the separate and coupled contributions of the three factors affecting the diversity of three dimensions.

All calculations were conducted in R 4.0.5 (R Core Team, 2021). We calculated plant taxonomic, functional, and phylogenetic diversity. Taxonomic diversity was calculated using the ‘vegan’ package (Oksanen et al., 2020), functional diversity based on Euclidean distance of functional traits among species, and phylogenetic diversity based on phylogenetic distance among species was calculated using the ‘picante’ package (Kembel et al., 2010). The ‘lm’ function in the ‘stats’ package was used to curve fit the diversity of three dimensions and restoration time, and the ‘pettitt.test’ function in the ‘trend’ package was used to calculate the response threshold of the diversity of three dimensions to the aerial seeding restoration time (Pettitt, 1979). Soil nutrients and restoration time were fitted to the general linear model by the ‘lm’ function in the ‘stats’ package (R Core Team, 2021). Pearson correlation analysis was carried out based on the ‘cor’ function of the ‘corrplot’ package, and variation partitioning analyses was further conducted based on the ‘varpart’ function of the ‘vegan’ package (Friendly, 2002; Oksanen et al., 2020).

The log-linear model showed that plant taxonomic ( Figure 2A ), functional ( Figure 2B ), and phylogenetic diversity ( Figure 2C ) showed a similar nonlinear response pattern to aerial seeding restoration time (p < 0.01). Plant taxonomic, functional, and phylogenetic diversity increased rapidly in the early restoration period. After 14 years of restoration, the plant diversity of three dimensions increased slowly and gradually reached saturation.

The general linear model showed that different soil nutrient indicators had different responses to restoration time ( Figure 3 ). TN ( Figure 3A ), TOC ( Figure 3B ), and NO₃ ^--N ( Figure 3C ) showed a highly significant increase with restoration time (p < 0.01). TP ( Figure 3D ), AP ( Figure 3E ), and NH₄ ⁺-N ( Figure 3F ) had no significant changes (p > 0.05).

Pearson correlation analyses showed that plant taxonomic and functional diversity were significantly positively correlated with MAP (p < 0.05), highly significantly positively correlated with TN、TOC、NO₃ ^--N (p < 0.01), and significantly negatively correlated with AP (p < 0.05), but had no significant correlation with TP、NH₄ ⁺-N and MAT ( Figure 4 ). Plant phylogenetic diversity was highly significantly positively correlated with MAP、TN、TOC、NO₃ ^--N (p < 0.01), but had no significant correlation with TP、AP、NH₄ ⁺-N and MAT ( Figure 4 ).

Variation partitioning analyses showed that plant taxonomic, functional, and phylogenetic diversity was mainly affected by the interaction of restoration time and soil nutrients, which together explained 38% of taxonomic diversity variation ( Figure 5A ), 21% of functional diversity variation ( Figure 5B ), and 24% of phylogenetic diversity variation ( Figure 5C ). In addition, both plant functional and phylogenetic diversity were affected by climate factors (3%), while taxonomic diversity was not significantly affected ( Figure 5 ).

Improving soil nutrients help restore degraded ecosystems. Our study found that the aerial seeding restoration significantly increased soil nutrients in Mu Us sandy land ( Figure 3 ). TN ( Figure 3A ), TOC ( Figure 3B ), and NO₃ ^--N ( Figure 3C ) showed a highly significant increase with restoration (p < 0.01). The increase of soil carbon and nitrogen with ecosystem restoration was also demonstrated in the study of Yang et al. (2023a). Plant litter, roots, and secretions are important sources of soil nutrients, and more leaf and fine root litter and soil organic matter input caused by vegetation restoration promoted carbon and nitrogen accumulation (Yang et al., 2021). However, this study also found that TP ( Figure 3D ) and AP ( Figure 3E ) did not significantly change with restoration. Soil phosphorus is mainly formed by rock weathering and is affected by soil parent material and biogeochemical processes (Lu et al., 2022). Because the parent material and climate of sandy land were similar during the restoration process, the change of soil phosphorus content was not significant in the short term. Our research showed that soil carbon and nitrogen were restored significantly during the restoration process. However, the restoration of phosphorus is not obvious, which should be paid attention in future restoration.

Exploring changes of plant diversity of three dimensions simultaneously with ecological restoration help to comprehensively understand biodiversity and conservation. Similar to the response of soil nutrients, our study found that plant taxonomic, functional, and phylogenetic diversity increased significantly with aerial seeding restoration ( Figure 2 ). This is consistent with previous research, which showed that ecological restoration improved biodiversity in different dimensions (Ribeiro et al., 2018). The reason may be that the soil of sandy land was poor in the early restoration period, plant growth and colonization were difficult, and the taxonomic diversity was low. Moreover, due to strong environmental filtering, the species have near genetic relationships and similar functional traits (Qin et al., 2016). Therefore, plant functional and phylogenetic diversity in the early restoration period was also low. Plant litter and root exudates increased with aerial seeding restoration and improved soil ecological conditions (Wang et al., 2016). The continuous increase of soil nutrients, such as carbon and nitrogen, promoted the colonization and growth of various plants, and the taxonomic diversity increased ( Figure 3 ). Meanwhile, the improvement of sandy land resource availability provided a larger shared ecological niche for plants, and provided opportunities for functional traits and phylogenetic differentiation of species, leading to an increase in plant functional and phylogenetic diversity (Galland et al., 2019).

However, we found that in the middle and late restoration period (after 14 years), the plant diversity of three dimensions gradually reached saturation ( Figure 2 ), while soil nutrients continued to increase ( Figure 3 ). The results showed that the restoration of soil nutrients lagged that of plant diversity during the aerial seeding restoration process. The reason may be that the plant demand for soil nutrients increased with plant diversity in the late restoration period. At this time, plant diversity may be limited by some soil nutrients, such as phosphorus ( Figure 3 ). The restriction of soil nutrients can reduce the niche differences among species in the community, leading to the emergence of redundant species with similar functional traits and near genetic relationships, and make the community tend to be stable (Ricotta et al., 2020). Therefore, plant taxonomic, functional, and phylogenetic diversity of species showed saturation trends.

When considering the relationship between plant taxonomic, functional, and phylogenetic diversity and single factors in soil nutrients and climate during the restoration process, it was found that plant taxonomic, functional, and phylogenetic diversity had a positive relationship with TN, TOC, and NO₃ ^--N, indicating a positive feedback effect between plant and soil ( Figure 4 ). The diverse litter and organic matter contributed by plant can promote the increase of soil C and N content during restoration process (Yang et al., 2021). Functional traits, such as leaf carbon and nitrogen content, may also have significant effect on litter quality and further influence soil decomposition rate and soil nutrients (van Der Putten et al., 2013; Münzbergová and Šurinová, 2015). Moreover, genetic differentiation within species may also impact soil biota and further affect soil communities (Münzbergová and Šurinová, 2015).

On the contrary, soil provides a better environment for nutrient absorption and growth of plants, thus promoting the colonization and growth of different plant species and improving plant diversity from three dimensions through feedback affecting functional trait differentiation and phylogenetic distance (van Der Putten et al., 2013; Münzbergová and Šurinová, 2015). AP was negatively correlated with plant diversity in three dimensions ( Figure 4 ). The reason is that phosphorus is the limiting factor of soil nutrients in this area. The higher the plant diversity, the more phosphorus absorption and utilization, and the lower soil phosphorus availability (Zemunik et al., 2015; Yang et al., 2021). Given the different effects of different soil nutrients on plant diversity, species can be selected for the future ecological restoration of Mu Us sandy land, such as planting nitrogen-fixing plants to promote the restoration of sandy land.

Further, this study integrated restoration time, soil nutrients, and climate factors to explore their relative contributions to plant taxonomic, functional, and phylogenetic diversity and found that the plant diversity of three dimensions was mainly affected by the coupling effect of restoration time and soil nutrients during the restoration process ( Figure 5 ). The meta-analysis on vegetation ecological restoration conducted by Zhao et al. (2023) on the Loess Plateau found that restoration time was the main factor affecting soil organic carbon after vegetation restoration. This suggested that restoration time improved plant diversity in multiple dimensions, mainly by improving soil nutrients.

In addition, in terms of climate factors, MAP was also significantly positively correlated with plant taxonomic, functional, and phylogenetic diversity ( Figure 4 ). A study conducted in the Inner Mongolia grassland also showed that increasing precipitation increased diversity in three dimensions (Hu et al., 2022). Sandy land is located in a semi-arid area, and plants are more sensitive to precipitation change. Increasing precipitation can weaken the environmental filtering caused by water stress and improve the taxonomic diversity by regulating the availability of soil nutrients and water so that the functional traits tend to be diversified (Hu et al., 2022; Yang et al., 2023b). Increased precipitation may also promote the emergence of distant species with different traits through competition, resulting in increased phylogenetic diversity (Yang et al., 2023b).