Data analysis

Slope aspect was divided into four slope directions: (1) shaded slope (0°–45° and 315°–360°); (2) half-shaded slope (45°–90° and 270°–315°); (3) half-sunny slope (90°–135° and 225°–270°); and (4) sunny slope (135°–225°). Defined values were assigned to the qualitative variables, aspect and position. The sunny slope was assigned 0.3, the half-sunny slope was assigned 0.5, the half-shaded slope was assigned 0.8, and the shady slope was assigned 1.0. The top of the slope was assigned 0.4, the mid-slope was assigned 1.0, and the downslope was assigned 0.8.

The number of individual understory plants per plot and the Shannon–Wiener diversity (α-diversity) and its components (i.e., species richness (S), Shannon–Wiener index (H′), and evenness) were used to describe vegetation biodiversity. These indices were calculated with PC-ORD version 5.0³⁸, using the following equations:

where N_i is the number of individual understory plants per plot, S is the species richness, and p_i is the relative abundance of species i, calculated as the number of plants of species i as a proportion of the total number of plants of all species in the jth plot.

Duncan's multiple comparison tests were used to test for differences in overstory and understory characteristics with soil and topographic factors among the three vegetation types. Pearson correlation analysis was used to detect relationships between environmental factors and overstory and understory flora. Differences in understory flora species composition were tested with the means of the pair-wise multi-response permutation procedure (MRPP), using a presence/absence species data matrix with rank transformed Sorensen (Bray–Curtis) distance measures. MRPP provides a T statistic that describes the separation between groups (the more negative T is, the stronger the separation is) and its associated significance⁴⁵. An A value is provided to estimate the within-group similarity, with values ranging from 0 to 1, where 1 indicates identical items within the group and 0 is equal to chance expectations. The P value is used to indicate the significance level of the corresponding T statistic. Indicator species analysis (ISA) was performed to test the indicator species in each vegetation type. The indicator value of each species was assessed for significance in relation to an α-value (α = 0.05) with a Monte Carlo test based on 9,999 permutations. MRPP and ISA were performed using PC-ORD version 5.0.

Canonical correspondence analysis (CCA) was performed to examine the influence of environmental factors on understory plant distribution. We used CCA with forward-selected explanatory variables to test the effects and significance of each variable. Furthermore, variation partitioning was conducted to partition the pure effects of soil factors and topographical variables. CCA was performed with the vegan package of R 3.6.2.