**3. Results**

For the whole plot (20 ha), evaluation of the genetic distribution of neighboring trees with *C. chinensis* demonstrated a scale-invariant relationship between phylogenetic distance of neighboring trees and *C. chinensis* (Figure S1; Figure S2). Based on data from all subplots, the relationship of phylogenetic distance of *C. chinensis* and the percent of neighbors within different annuli suggested that neighbors of *C. chinensis* were either more likely to be closely related or more distantly related. Specifically, *C*. *chinensis* had more individuals of Group 1 (more closely-related) and Group 6 (more distantly-related) than expected from a null distribution. Groups 3 and 4 represented intermediate relatedness to *C. chinensis* and had the lowest percent of individuals adjacent to *C*. *chinensis* across

all observed scales (Figure S2). However, in the successional forests, the results indicated that the percentage of neighbor trees of *C. chinensis* was higher than the values expected from a null model, except for the most closely related species (i.e., Group 1), at all observed scales (Figure 4a; Figure S3). Meanwhile, in the mature forest, the frequency of neighbors around *C. chinesis* for the phylogenetically closely related species (Group 1,2,3,5) and most phylogenetically distantly related species (Group 6) were random. But for the groups 4, the frequency of those species was higher than null model expected (Figure 4b). Group 1 was significantly more abundant depending on the scale across all observed scales (Figure S4).

**Figure 4.** Distribution of neighbor trees within a 20 m radius to *Castanopsis chinensis* based on phylogenetic distance (six ordinal groups) in succession forest (**a**), and mature forest (**b**). Horizontal axis is group number representing increasing phylogenetic distance of neighbors to *C. chinensis*, while the vertical axis represents percent of neighbors within that group. Dashed lines represent the upper (97.5%) and lower (2.5%) confidence envelope for a null (neutral) hypothesis, while the solid blue line represents observed distributions for each of the six phylogenetic groups, the purple line represents the standard error.

The results of Mantel test showed that in the whole plot, there was no significant positive or negative correlations existed between geographic distance between *C. chinensis* and other trees, and phylogenetic distance between *C. chinensis* and other trees (*p*-value = 0.644>0.05, Observation value = 0.002). However, in mature forests, negative relationship was found between geographic distance between *C. chinensis* and other trees, and phylogenetic distance between *C. chinensis* and other trees (*p*-value = 0.027 < 0.05, Observation value = −0.017). In the successional forests, the results indicated that correlation between geographic distance between *C. chinensis* and other trees, and phylogenetic distance between *C. chinensis* and other trees was positive (*p*-value = 0.039<0.05, Observation value = 0.018).

Seven traits of 194 species used in this study were also tested by *K* values, and the *K* values of all 7 species were between 0 and 1. We used *p*-value to judge the significance of the functional character phylogenetic signal by comparing the *K* value with the random distribution. The results of testing phylogenetic signal showed that five of seven traits exhibit significant phylogenetic signals (*p* < 0.05), leaf thickness was marginally significant (*p* = 0.06), and leaf area was not significant (*p* = 0.956; Table S1).
