Tree architecture is crucial to maximizing light capture, determined by carbon allocation of individual trees, and consequently characterizes species-specific growth strategies. Its variation and associated life-history strategies have been examined in tropical and temperate forests, but not in subtropical forests. Moreover, a similar architectural pattern was found using a hierarchical Bayesian model in a tropical forest, which differed from most of previous studies. Here, we employed a hierarchical Bayesian model to examine tree architecture differences and associations with adult stature and light requirement among 59 subtropical co-occurring species. Architectural variations among tree species with different seed dispersal and leaf phenology types were analyzed. Most species showed similar architecture in the height of the lowest foliage-tree height relationships (F-H) and the long side of crown- tree height relationships (W1-H), but some species showed interspecific variations in tree height-stem diameter relationships (H-D) among the 59 co-occurring species in the subtropical montane forest. Trees developed deeper and larger crowns at mid-elevation compared to the tropical and temperate forests. Parameters of H-D relationship differed in leaf phenology and dispersal types, and intercepts of F-H relationship and W1-H relationship differed in leaf phenology. Large-statured species had more slender stems, and shallower and narrower crowns at small sizes, but similar crowns at large sizes. Light-demanding species
showed weak correlations between architectural variables and light requirement but exhibited wide crowns at the intermediate sizes. In general, size-dependent architectural differentiation was driven mainly by adult stature and light requirement in subtropical forest. Coexistence species showed different life-history strategies in light capture, which may help provide options in forest thinning and harvesting in subtropical forest. Speciesspecific tree architectural models of 59 co-occurring species represent three-dimensional (3D) structure of this subtropical forest accurately, but also support for future terrestrial laser scanning (TLS) data analysis.