Despite the recognition of positive interactions as an important driver of species coexistence and community structure, the underlying mechanism of how facilitation affects assembly processes along stress gradients is poorly explored. Understanding the responses of functional diversity to benefactor species at the extreme end of the stress gradient could provide valuable insight about facilitation-involved assembly mechanisms and contribute to the predictions of species coexistence under climate change.

In the drought-stressed community in the Badain-Jaran Desert, the responses of the local community to the nurse shrub species Calligonum mongolicum Turcz. were evaluated using hierarchical Bayesian models. For the 3-year experiment, summer rainfall in each year formed a natural gradient of drought stress. To evaluate the shrub’s effects on the assembly process along that gradient, individual samples were collected in pairwise under-shrub and open habitats, and four traits related to stress tolerance and resource acquisition were measured simultaneously.

Under moderate drought stress, we observed shifting community-weighted means, broadening ranges and reducing overlaps of functional traits under shrubs. These effects were partly driven by a distinct microenvironment created by shrub plants, in particular the improvement and heterogeneity of soil moisture conditions. However, this influence on trait distributions was strongly dependent on the environmental context, and generally disappeared as drought stress shifted toward its driest end, almost in line with the decreased positive interaction assessed by plant density and species diversity.

This study focused on water-limited community that lies at the driest end of drought gradient and confirmed that facilitation can drive the assembly process through both environmental filtering and niche differentiation. More importantly, these assembly mechanisms are proven to become less efficient under extreme drought stress, which may suggest the occasionally disappearing role of benefactor plants on community assembly and an increasing risk of biodiversity loss in the context of climate change.

This investigation was conducted in a typical desert grassland at the southern edge of the Badain Jaran Desert from 2017 to 2019. During the investigating period, summer rainfall declined from 71.4 mm in 2019 to 48.6 mm in 2017 and 25.6 mm in 2018. Through this investigation, we studied the effect of foundation plant (Calligonum mongolicum, shrub) on local-scale assembly process along an interannual drought gradient in water-limited communities. Species relative abundance was investigated in both under-shrub and open habitats, which can be used to measure the effect size of shrubs on local communities under different stresses. Trait distributions (i.e., community-weighted mean, trait range, trait spacing) among habitats were measured to evaluate the effect of shrubs on community assembly under different drought stresses. Additionally, the effect of shrubs on soil moisture/water content (SWC) and soil organic matter (SOM) under different drought stresses were also investigated, which can be used to determine the underlying mechanism that driving the assembly process.

To measure species relative abundance, a single 2 m × 2 m quadrat was established beneath the canopy of stand-alone Calligonum mongolicum, while another quadrat of the same size was placed in an open area. In these quadrats, species was identified, and the species-specific abundance was recorded.

To measure functional traits at different years and habitats, we randomly selected 2-5 adult individuals for every observed understory species according to their abundance in each selected quadrat at both habitats. Five species were omitted from the analyses because they were too rare to accurately estimate trait values. To obtain robust estimation, more than 30 individuals for each species were sampled in different years and habitats. When the suitable samples of a species did not reach this amount, we randomly collected additional individuals from the same habitats in the fenced plots.

Lateral spread (cm) was measured as the maximum crown width of individuals in the understory community. Height (cm) was measured as the distance between the extended upper boundary of the crown and the ground level. SLA (cm∙g^-1) was determined as the ratio of fresh leaf area to leaf dry mass. LDMC (mg∙g^-1) was determined as the ratio of leaf dry mass to fully rehydrated fresh mass.

To measure soil moisture/water content (SWC), we collected four soil cores (5 cm diameter, 10–20 cm in depth) along the diagonal of each quadrat, and determined them gravimetrically after oven-drying for 48 h at 110°C.

To measure soil organic matter (SOM), we randomly chose 10 pairwise quadrats each year, collected four soil cores within each quadrat, and applied the dichromate oxidation method.