Identifying the environmental conditions that drive biogeographic structure remains a major challenge of biogeography, evolutionary ecology and increasingly, conservation biology. Here, we use multivariate classification trees to assess the biogeographic structure of northeast Pacific (˜ 26–58°N) rocky intertidal species (406 species of algae and invertebrates) from 102 field sites. Random forest analyses are used to assess the importance of 29 environmental variables, encompassing a broad range of potential drivers, to predict biogeographic structure. Analyses are repeated for species with different larval dispersal capabilities and by broad taxonomic categories (invertebrates and algae). Results show that overall biogeographic structure is in general agreement with classic classification schemes, but patterns are variable among species with different larval dispersal capabilities. Random forest models show a very high fit (pseudo r2 > 0.94) and indicate that biogeographic structure can be predicted by a relatively modest subset of variables. Upwelling related variables are the best overall predictors of biogeographic structure (nutrient concentrations, sea-surface temperature, upwelling/downwelling seasonal switch index), but the relative importance of predictors is geographically variable and top predictors are dependent on the type of larval dispersal. Upwelling related variables are more important to predict biogeographic structure for invertebrates with lower-medium dispersal capabilities and algae, whereas species with high larval dispersal (planktotrophic) are better predicted by a different subset of variables (i.e. salinity, precipitation seasonality). Our results lend support to the influence of coastal upwelling in structuring biogeographic patterns and highlight the potential for climate change-induced alterations of upwelling regimes to profoundly affect biodiversity at biogeographic scales.