The scattered survey of plant communities often fails to conform to the characteristics of both continuity and discontinuity of community distribution. In recent years, the research platform of forest dynamics plot provides a new way for the study of vegetation classification. The 10.24 hm2 forest dynamics plot in Huangshan was divided into three clusters by using multivariate regression tree, indicator species and redundancy analysis (RDA) with topographic factors and species importance values as variables. Castanopsis eyrei was the species with the highest importance value in the three associations, which indicated that the investigated forest dynamics plot belongs to the alliance of C. eyrei. The α-diversity index was used to measure the species diversity of associations at local scale, and PER-SIMPER was introduced for the first time to infer the assembly mechanism of evergreen broad-leaved forest community in Huangshan.The results showed that the association classification in the community is mainly affected by elevation and convexity, while slope and aspect exhibit little influence on association classification. Niche and dispersal limitation jointly dominate the assembly of plant community in Huangshan. The species diversities in different associations are different. There is a significant difference in α-diversity between groups distributed in low elevation ridges and groups distributed in valleys. The results of this study provide a basis for decision-making on local forest vegetation management and protection.

forest dynamics plot; multivariate regression tree; ordination; association; community assembly