*2.3. Species Composition Measurement*

In each plot, we placed three 50 cm × 50 cm quadrates as replicates to survey the species by recording their names, coverage (%), and richness (i.e., the number of species) and then classified all the species into the four plant functional groups: graminoids, forbs, legumes, and shrubs (if any). The coverage was estimated visually, but significantly positive correlation of coverage and species richness confirmed the data reliability (R2 = 0.418, *p* < 0.001).

### *2.4. Data Analysis*

Firstly, the effects of slope aspect on leaf [N], [P], and N:P ratio were detected using a linear mixed model by treating "site" as the random factor (LMM) for the mixed samples, with species-level value as a replicate unit and one common species, *Anaphalis lacteal*, that occurred in each of the six plots. Correlations among leaf [N], [P], and N:P ratio were evaluated in each site using linear models. Comparison of the leaf N:P stoichiometry between different PFGs was also conducted using the LMM. Secondly, the effects of slope aspect on species richness and coverage and their correlations with leaf N:P stoichiometry (using plot-level means) were analyzed using LMM.

All the variables were log10-transformed. All analyses were performed with R version 4.0.3 (R Core Team, 2020) in RStudio version 1.3.1093 (RStudio Team, 2020).

**Table 1.** Mean value of leaf nitrogen (N) and phosphorus (P) concentrations of the 40 measured species. PFG represents plant functional group.


### **3. Results**

*3.1. Variations in Leaf [N], [P], and N:P Ratio across the Slope Aspects*

The leaf [N], [P], and N:P ratios ranged from 10.09 to 42.96, 0.89 to 3.10, and 8.10 to 23.63 mg g−<sup>1</sup> in Hezuo and from 4.36 to 33.53, 0.536 to 2.33, and 4.14 to 21.21 mg g−<sup>1</sup> in Maqu, respectively. Overall, the leaf [N], [P], and N:P ratios were all significantly lower in Maqu, with higher elevations than in Hezuo (Figure 2A,C,E).

**Figure 2.** Comparisons of leaf [N] (**A**,**B**), [P] (**C**,**D**), and N:P ratio (**E**,**F**) between sites and among slope aspects. SFS, WFS, and NFS represent south-, west-, and north-facing slope aspects, respectively. Different letters represent significant differences at the level of *p* < 0.05. Variables are in log10 scale.

Although leaf [N] and [P] were not significantly different between slope aspects, leaf N:P ratio was significantly lower on the NFS than on the SFS for the mixed samples (Figure 2). The one common species, *Anaphalis lacteal*, showed the same patterns as those of the mixed samples (Figure S1). Moreover, significant correlations between leaf [N] and [P] and between the leaf N:P ratio and leaf [N] rather than [P] were found in each plot (Figure 3). In addition, legumes showed a significantly higher leaf [N] and [P] than forbs and graminoids and a higher leaf N:P ratio than forbs (Figure S2).

### *3.2. Correlations of Leaf Traits with Species Richness and Coverage*

Significantly higher richness of all species and forbs and lower richness of legumes were found on the NFS than on the SFS and WFS (Figure 4A–D); however, the richness of graminoids and plant coverage for each PFG and for all species did not differ significantly among the slope aspects (Figure 4E–H).

**Figure 3.** Linear regressions between leaf [N] and [P], between leaf [N] and leaf N:P ratio, and between leaf [P] and leaf N:P ratio in Hezuo (**A**–**C**) and Maqu (**D**–**F**). Solid and dashed lines indicate significant and non-significant regressions, respectively. SFS, WFS, and NFS represent south-, west-, and north-facing slope aspects, respectively. Variables are in log10 scale.

**Figure 4.** Comparisons of species richness (**A**–**D**) and coverage (**E**–**H**) for each plant functional group and the whole community among slope aspects. SFS, WFS, and NFS represent south-, west-, and north-facing slope aspects, respectively. Variables are log10-transformed.

**Figure 5.** Correlations of species richness of each plant functional group and the whole community with leaf N concentration (**A**–**D**), with leaf P concentration (**E**–**H**), and with leaf N:P ratio (**I**–**L**). Variables are log10-transformed.

The correlations were found to be significantly negative between leaf [N] and richness of all species (R<sup>2</sup> = 0.615, *p* = 0.047) and forb species (R<sup>2</sup> = 0.570, *p* = 0.061), between leaf [P] and graminoid richness (R<sup>2</sup> = 0.362, *p* = 0.093), and between leaf N:P ratios and forb richness (R2 = 0.675, *p* = 0.032) (Figure 5). On the other hand, legume coverage was positively correlated with leaf [N] (R<sup>2</sup> = 0.809, *p* = 0.010), and the coverage of graminoids and all species was negatively and positively correlated with leaf [P] and with marginal significance and significance, respectively (R<sup>2</sup> = 0.051 and 0.704, *p* = 0.103 and 0.029, respectively) (Figure 6).

**Figure 6.** Correlations of species coverage of each plant functional group and the whole community with leaf N concentration (**A**–**D**), with leaf P concentration (**E**–**H**), and with leaf N:P ratio (**I**–**L**). Variables are log10-transformed.

### **4. Discussion**

### *4.1. Variations in Leaf [N], [P], and N:P Ratio across Different Slope Aspects*

Although leaf [N] and [P] did not differ significantly between slope aspects, we found a significantly lower N:P ratio on the NFS, with a mean of 11, and a higher N:P ratio on the SFS, with a mean of 16, which was consistent with a previous study reporting that the lowest N:P ratio occurred on the NFS out of the four slope aspects [39]. Tessier and Raynal [8] pointed out that plants were subject to N-, P-, and co-limitation when the plant N:P ratio was less than 14, higher than 16, and between 14 and 16, respectively. Therefore, our results indicate that plants on the NFS suffered N limitation. The tight correlation of leaf [N] with N:P ratio suggests that leaf N:P is determined by leaf [N] along the slope aspect gradient. A lower soil N availability reported in our previous studies [36]

was probably the reason for the lower leaf N:P ratio on the NFS, which supports the biogeochemical hypothesis suggesting that plant N and P are influenced by the availability of soil nutrients [9]. The lower soil N on the NFS was perhaps due to lower temperatures reducing the decomposition and mineralization of organic material.

In addition, the leaf [N], [P], and N:P ratio were lower in Maqu, which has a higher elevation, which is consistent with previous studies of elevational gradients [14,40], indicating an increase in N-limitation along altitudinal gradients.

### *4.2. Correlations of Leaf N:P Stoichiometry with Species Richness and Coverage across the Slope Aspects*

Significantly higher species richness in each quadrate was found on the NFS than on the SFS, which was mainly caused by the variation in forb richness. These results are consistent with previous studies in which species richness was significantly lower on the SFS [31,41], with the possible mechanism of lower soil water content as the limiting factor.

Community species richness was negatively correlated with leaf [N], which is consistent with the negative effects of N-fertilization on species diversity [42–45]. Many Nfertilization experiments showed considerable species loss, with the underlying improved growth of dominating clonal graminoids excluding forbs through competition [45,46]. The possible mechanisms may be that N-fertilization led to a tendency of P limitation, but graminids are often considered stress-tolerant species due to their low requirement for P and K [47]. Therefore, we found negative correlations of leaf [P] with graminoid richness and coverage because higher leaf [P] indicated the rapid growth of P-sensitive species. The significant negative correlations between leaf N:P ratio and forb richness provide direct evidence that on the natural environmental gradient, a higher N:P ratio is associated with fewer forbs and more graminoids, as hypothesized by Güsewell [6].

In addition, the higher legume richness on the SFS and the positive correlation of legume coverage with leaf [N] were consistent with the higher N:P ratio on the SFS due to their N-fixing functions.
