At the community level, the effect of the shrub on herbaceous biomass (RII mean ± SE: 0.02 ± 0.03) and richness (RIImean ± SE: −0.06 ± 0.02) was almost neutral among all treatments, exceptRII on richness at W40N0 treatment which was slightly negative (see result of sample t-test in Fig. 2). Watering and N fertilization did not significantly affect RII on biomass and richness (Table S1).
The RDA results showed that height and LDMC contributed to explain how RII varied at the species level in response to watering and N fertilization (R2 = 0.22; Fig. 3a). Four groups of species were identified with the cluster analysis conducted on RDA species scores (Fig. 3b and Table S2). Axis 1 opposed mostly group C, including the tallest species (all rhizomatous perennial grasses, e.g.,L. secalinus and P. flaccidum ), to group A including the shortest ones (mostly annuals, e.g.,Euphorbia humifusa and C. puberulum ), while group B (including mostly annuals, e.g., Incarvillea sinensis and S. viridis ) and group D (including all bunchgrasses, e.g., Cleistogenes squarrosa, Agropyron mongolicum, and Stipa glareosa ) contained species of intermediate heights. Axis 2 opposed mostly group D, including species with the highest LDMC values, to group B with the lowest ones, while groups A and C contained species with intermediate LDMC values.
Watering and N fertilization had contrasting effects on the biomass of the four cluster groups in the quadrats away from the shrubs (Fig. 4). The biomass of Group A increased with watering (F = 21.85, P< 0.001), in particular from W20 to W40 treatments (see Tukey test in Fig. 4). The biomass of groups B and C increased with N fertilization (F = 6.77, P = 0.03 and F = 15.73, P < 0.001, respectively), in particular from N10 to N60 treatments for both groups. The biomass of Group D only decreased with important N fertilization in the W40 treatments (significant water × N interaction; F = 11.28, P = 0.02). Contrasting responses to watering and N fertilization among cluster groups were also observed for RII . RII of group A decreased with N fertilization (from overall neutral RII values in N0 (0.15 ± 0.13) and N10 (−0.10 ± 0.13) treatments to negative RII values in N60 (−0.41 ± 0.12) treatments; F = 10.92, P = 0.004), whereas that of group C increased with watering (from neutral RII values in W0 (0.02 ± 0.12) and W20 (0.11 ± 0.14) treatment to positive RII values in W40 (0.41 ± 0.09) treatment; F = 6.98, P = 0.03; see the results of Tukey test in Fig. 4). RII of Group D strongly increased with N fertilization in the W40 treatment, but slightly decreased with N fertilization in the two other water treatments (significant water × N interaction; F = 21.24, P < 0.001; see the results of Tukey test in Fig. 4). Watering and N fertilization had no significant effects on RII of group B, which values were overall slightly negative (−0.17 ± 0.06). The responses of RII s for each cluster group were consistent with the positions of these groups and the direction of treatment arrows in the RDA (Fig. 3 and Fig. 4). This was in particular the case for group C, located on the left of axis 1, in the same direction than all W40 treatment arrows, and for group A, located on the right of axis 1, in the opposite direction than all N60 treatment arrows. Finally, group D was located on the upper part of axis 2 in the same direction than the W40N60 treatment arrow and the opposite direction of the W40N0 one.