Ontogenetic divergences in soil nutrient preferences between arbuscular and ectomycorrhizal trees regulate community assembly in subtropical forests

Subtropical and tropical forests are characterized by a species‐rich understorey with diverse arbuscular mycorrhizal (AM) tree species and a hyperdominant canopy with ectomycorrhizal (ECM) trees. Divergent nutrient acquisition strategies and habitat preferences among tree species with different mycorrhizal types are essential for forest biodiversity maintenance and community assembly. However, previous studies investigated tree nutrition competition mainly at the seedling stage and only focused on a few species with manipulated experiments. In this study, we explored distributional preferences and growth responses to soil nitrogen (N) and phosphorus (P) nutrients for AM and ECM tree species across three ontogenetic stages (seedling, sapling and adult). We established 1200 seedling quadrats within six 1‐ha plots and monitored seedling survival and growth annually from 2008 to 2022, and also surveyed sapling and adult performance and distributions within a 50‐ha plot from 2016 to 2021. We found that AM tree species maintained consistent preferences for inorganic N and consistent negative associations with organic N across all ontogenetic stages. In contrast, ECM species exhibited stage‐dependent resource specialization, with seedlings preferring inorganic N and saplings showing inorganic P preference, while the spatial distribution of adult ECM trees was significantly and negatively correlated with soil organic P. This high ontogenetic plasticity in ECM foraging strategies suggests a competitive advantage in response to the spatiotemporal heterogeneity of soil nutrients, thereby maintaining the local abundance of ECM plants in subtropical forests. Synthesis : These results establish ontogenetic trajectories of nutrient niche differentiation, illustrating how stage‐specific mycorrhizal foraging strategies between AM and ECM trees across different ontogenetic stages regulate community structure via edaphic filtering and drive community‐level resource partitioning.