Abstract: [Objective]The study was to determine the species composition and diversity of different forest layers and to determine how vertical distribution of trees affect the community structure and its underlying mechanisms with the view of arbor layer's stratification.[Method]Five 50 m×50 m sample plots were investigated in Wanmulin Provincial Nature Reserve, Fujian Province. With maximum light receiving plane (MLRP) method, arbor layer of each plots were divided into three sub-layers from top to bottom, the sub-layer I, Ⅱ and Ⅲ. The sub-layer I and Ⅱ belonged to light receiving layer (LRL), the sub-layer Ⅲ was non-light receiving layer (NLRL). Then the species composition and diversity of each sub-layer were studied using importance value, diversity index and similarity index.[Result]The species richness and diversity increased when the arbor layer decreased from sub-layer I to sub-layer Ⅲ of the five plots. The composition and number of species varied in the same sub-layers of different plots. The maximum tree height of 35 tree species can reach about 30 m even 40 m and they have the potential to distribute at top layer of the community, i.e. sub-layer I, but there were only 10 tree species could successive distribute at all the three sub-layers among all species of all plots. The maximum height of 16 tree species and 4 shrub species was about 25 m and sub-layer Ⅱ was the limit of what they could reach. The maximum height of 17 tree species and 13 shrub species was 16-17 m and sub-layer Ⅲ was the limit of what they could reach. Tree species in LRL were dominant species and important associated species of arbor layer, including evergreen trees Schima superba, Cinnamomum chekiangense, C. camphora, Machilus pauhoi, Tsoongiodendron odorum, Michelia fujianensis, Parakmeria lotungensis and some trees species of Castanopsis, and Liquidambar formosana, Alniphyllum fortunei, Nyssa sinensis and Fraxinus insularis. Tree species only distributed in NLRL included evergreen trees M. skinneriana, Sycopsis sinensis, Photinia davidsoniae, Daphniphyllum oldhami, Ternstroemia gymnanthera, and deciduous trees Albizia julibrissin, Diospyros kaki var. silvestris and Dalbergia balansae. Most shrubs species could only distribute in NLRL except Symplocos sumuntia, Adinandra millettii, Rhododendron henryi and S. dunnii could reach to sub-layer. Priority should be given to dominant species of sub-layer I when determining dominant species of the community, because their basal areas were obviously bigger than trees of other sub-layers. The similarity of sub-layer I and Ⅱ was bigger than that of sub-layer I and Ⅲ or sub-layer I and arbor layer. The differentiation was prominent at the critical plane LRL and NLRL when the similarity was the degree of differentiation and the arbor layer was the reference, suggesting spatial niche separation being prominent at MLRP.[Conclusion]In the five sample plots, dominant species and important associated tree species of arbor layer were mainly distributed in the LRL. Niche differentiation was prominent at MLRP. There were obvious difference between LRL and NLRL in species richness and diversity, which were identified according to MLRP. The differentiation of the LRL and the NLRL has potential practical significance for the selection of target trees in close-to-nature management so that the relationship between the two should be further studied in the future. Very few tree species could distribute in 3 successive sub-layers of each plots among all tree species that could grow to sub-layer I, suggesting that there were almost no particular species could be permanent dominant species in the LRL and species composition and richness in the 5 plots appeared to be caused also by random species replacement, which needs to be further studied in future.

Key words: natural evergreen broad-leaved forest, mid-subtropical zone, arbor layer, stratification, similarity