Traits explain canopy tree occurrence along regional environmental gradients; a subset combine to be useful.

Trait-Species Distribution Models (trait-SDM) help to understand the importance of plant strategies to niches, assess their generality across species and provide a path to predicting species distributions from a shortlist of traits. Yet published trait-environment associations show considerable inconsistency. Region-scale models may leverage more species, traits, trait ranges and climatic gradients, than at local scales, while retaining biogeographic coherence. Here we fit trait-SDMs using multilevel models with six traits and over 90 eucalypt tree taxa for plots along environmental gradients across 120,000 km2, 7—15 oC MAT and 700—1590 mm MAP (5th, 95th percentiles). We found stem sapwood density, bark thickness, seed mass and maximum height were useful predictors in a multi-trait model of environmental responses to temperature, water deficit, soil depth and pH. They explained 9-19% of variance between species in environmental responses. We found less support for specific leaf area and leaf size. Species occurred unimodally along environmental gradients. Trait-environment terms indicated species with dense stems were more likely in drier climates, thicker bark eucalypts in warmer climates and that both thinner bark and larger seeds increased occurrence in shallow soils. Taller species were more common and more likely to occur towards sites that were warmer and wetter than average for our region. Useful trait-based prediction of species occurrence at regional scales is attainable and trait-SDMs help to test trait-based theory about realised niches. Single trait models reflect the maximum potential explanation of niche differentiation by a trait, while multi-trait models represent integrated phenotypes.