Data analyses
We applied a landscape ecological conception (Pickett & Cadenasso, 1995) to analyze the spatial distributions of current and historical individual cushions within communities. For this, we used Fragstats4.2 software (McGarigal & Ene, 2013) to calculate the following six landscape metrics for cushion patches (McGarigal & Ene, 2013; Lustig et al., 2017; Wang et al., 2014): the total area of cushion patches (CA), the edge density (ED), the mean cushion patch area (Area_ Mn), the percent of core area of landscape (CPLAND), the total patch number (NP), and sixth the percentage of like adjacencies (PLADJ). See Text S13 for the details.
To illustrate the abundance of beneficiary species and successional processes in different successional stages (represented by the PJ1 to PJ3 communities) of cushion-dominated communities, we generated stacked graphs at species level for each community.
Since population ID includes information on the study site, including elevation and potential relevant micro-climates, we applied one-way ANOVA to assess the differences in population density and productivity between populations, with population density and productivity as dependent variables and population ID as an independent variable. Tukey HSD tests were applied to assess the significance of differences between populations.
Linear mixed-effects models were applied to assess the following effects. 1) The effects of simulated climate events on seed germination and subsequent seedling survival, with temperature, light availability, water availability and their potential interactions as fixed effects and pot replicate as a random effect. 2) The germination and viability of seeds buried in the natural field, with elevation, checking time and their interaction as fixed effects and tea bag replicate as a random effect. 3) The survival of seedlings transplanted in the natural field, with elevation, micro-habitat and their interaction as fixed effects and plot replicate as a random factor. 4) The effects of simulated extreme climate events on seedling survival, with climate treatment, seedling age and their interaction as fixed effects and pot replicate as a random effect. 5) The effects of beneficiary species on nutrient contents (and stable isotope ratios) of cushion leaves, with beneficiary cover ratio, leaf source (beneficiary-covered or beneficiary-free) and their interaction as fixed effects and sample replicate as a random effect. 6) The effects of beneficiary species on the physiological traits (SLA and LDMC) of individual cushions, with elevation, beneficiary cover ratio, dominating beneficiary species (Kobresia pygmaea orSaussurea leontodontoides ) and their interactions as fixed effects and sample replicate as a random effect. 7) The effects of allelopathical materials on the seed germination and seedling survival, with allelopthical source (aboveground or belowground), extracting mode (aqueous or ethanol), elevation, concentration and their interactions as fixed effects and sample replicate as a random effect. Nutrient content and physiological trait values were square root-transformed to meet assumptions of parametric statistics, while seed germination, viability and seedling survival data were standardized between 0 and 1 by the formula (X -Xmin )/(Xmax -Xmin ), where X is the relevant value of seed germination or seedling survival. The significance of each contrast (Rosenthal & Rosnow, 2010) was assessed using type-I analysis of variance with Satterthwaite’s method for all linear mixed-effects models.
To assess effects of beneficiary species on the performance (including flowering, fruiting, surface death ratio and physiological status) ofA. polytichoides cushions, we calculated Pearson correlation coefficients.
R v.4.1.1 (R. C. Team, 2021) was used for all the above analyses, the lme4 packages was used for the linear mixed-effects modeling (Bates et al., 2015), the ggplot2 package (Wickham, 2016) was used to plot all reported figures and the layout was designed with Adobe Illustrator 2021.