Abstract: [Objective] Evergreen-deciduous broadleaved mixed forest is one of the main vegetation types in northern parts or high elevations of subtropical regions of China. It is characterized as transitional features from evergreen to deciduous broad-leaved forests. Due to long term and repeated anthropogenic disturbances, these forests were subjected to different extents of degradation. Until now, very few studies have been carried out on mechanisms of restoration of the degraded forests. In this study, by comparing the roles of environmental and spatial factors in restoration and rehabilitation of communities, patterns of variation in community structure and species diversity during the natural restoration in subtropical regions can therefore be analyzed, laying a foundation for deepening our understanding of the mechanisms of natural restoration of this type of forest.[Method] 10 sample plots (20 m×20 m in size) were selected at random and configured as fixed plots for monitoring forest dynamics, respectively in old-growth forest, 35-year-old secondary forest (harvested in 1980) and 20-year-old secondary forest (harvested in 1995) in the Mulinzi National Nature Reserve in Enshi of Hubei Province. For forest communities at different stages of restoration, abundance of woody plants, basal area at breast height, species richness, and richness of rare species (100 individual trees as baseline) were compared. One-way ANOVA and Turkey-HSD were conducted to assess the differences in stem density, basal area, species richness, and rare species richness among the three succession stages. The redundancy analysis (RDA) and non-metric multidimensional scaling (NMS) ordinations were used to differentiate the plots. The environmental and spatial factors (PCNM variables) were treated as explanatory variables to partition the variations at each succession stage. The relative effects of environmental factors and restoration period after disturbance on the four community characteristics were analyzed by performing multivariable regression analysis.[Result] During the process of community restoration, the abundance of woody plants, species richness and rare species richness had a significant decrease, while basal area had a significant increase. Species composition were significantly different among different restoration forests (Adonis, P<0.05). However, all the soil factors did not show any significant differences. Environmental factors showed a greater influence than the spatial configurations. Environmental factors explained 31.4%, 36.8%, 54.1% of species composition variations of 20-year-old, 35-year-old secondary forest and old-growth forest, respectively. Spatial configuration explained 16.8% and 23.2% composition variations for the 35-year-old secondary and old growth forests respectively, while no significant influence on the 20-year-old secondary forest were found. The spatial auto-regression showed that soil pH, litter thickness, elevation, slope gradient, slope orientation and the restoration time played an important role in the restoration of community structure and species diversity.[Conclusion] The community structure, species diversity and environmental features of the subtropical evergreen-deciduous broadleaved mixed forest varied significantly during the secondary succession process after anthropogenic disturbances. Environment, space, and time were the key factors determining community composition and dynamics. The stochastic process (such as dispersal process or prior effect) exerted the greatest influence on community assembly at the early stages of forest restoration, meanwhile, the deterministic process represented by environmental filtering played a dominant role in community assembly of old-growth forest. The differences in species composition between secondary and old-growth forests declined gradually with the progress of secondary succession.

Key words: the subtropical evergreen-deciduous broadleaved mixed forest, natural recovery, secondary succession, environmental filtering, deterministic versus stochastic process, Mulinzi national nature reserve