Microclimate and soil nutrients jointly shaped contrasting elevational patterns of understory biodiversity and community assembly in subtropical karst forests

Karst mountains are global biodiversity hotspots shaped by complex edaphic and microclimatic heterogeneity. However, how different life forms in the understory respond to elevational gradients in these ecosystems remains unclear. By investigating species and phylogenetic diversity and structure of shrub and herbaceous communities across an elevational gradient in a subtropical karst mountain, we aimed to explore patterns of diversity and community assembly mechanisms. Shrubs and herbs showed contrasting elevational responses, with shrub diversity peaking at mid-elevations and herbaceous diversity declining along elevation and stabilizing at higher altitudes. Phylogenetic structure differed between shrubs and herbs, indicating divergent assembly processes shaped by both stochastic and deterministic processes, with stochasticity playing the dominant role. Microclimatic variables, particularly soil temperature and moisture, had contrasting effects on the two life forms by promoting herbaceous diversity while suppressing shrub diversity. In addition, soil total phosphorus showed a negative association with both species and phylogenetic diversity, while soil pH was positively associated with diversity in both layers. These findings underscore divergent strategies of understory plants in response to environmental gradients and offer a framework for predicting biodiversity dynamics and guiding conservation strategies in fragile montane ecosystems. • Shrub diversity peaked at mid-elevation, whereas herbaceous diversity declined along the gradient. • Microclimatic factors promoted herb diversity, but suppressed shrub diversity, in karst understories. • Soil total phosphorus had a negative effect on diversity in both plant layers, while soil pH had a positive effect. • Both stochastic and deterministic processes shaped community assembly, with stochasticity prevailing.