Aim Ecologists have generally agreed that β‐diversity is driven at least in part by ecological processes and mechanisms of community assembly and is a key determinant of global patterns of species richness. This idea has been challenged by a recent study based on an individual‐based null model approach, which aims to account for the species pool. The goal of the present study is twofold: (1) to analyse data sets from different parts of the world to determine whether there are significant latitude–β‐diversity gradients after accounting for the species pool, and (2) to evaluate the validity of the null model. Location Global. Methods A total of 257 forest plots, each being 0.1 ha in size and having 10 0.01‐ha subplots, were used. We conducted four sets of analyses. A modified version of Whittaker's β‐diversity index was used to quantify β‐diversity for each forest plot. A randomization procedure was used to determine expected β‐diversity. Results The number of individuals per species, which characterizes species abundance distribution, alone explains 56.8–84.2% of the variation in observed β‐diversity. Species pool (γ‐diversity) explained only an additional 2.6–15.2% of the variation in observed β‐diversity. Latitude explains 18.6% of the variation in raw β deviation in Gentry's global data set, and explains 11.0–11.6% of the variation in standardized β deviation in the global and three regional analyses. Latitude explains 33.2–46.2% of the variation in the number of individuals per species. Main conclusions Species abundance distribution, rather than species pool size, plays a key role in driving latitude–β‐diversity gradients for β‐diversity in local forest communities. The individual‐based null model is not a valid null model for investigating β‐diversity gradients driven by mechanisms of local community assembly because the null model incorporates species abundance distributions, which are driven by mechanisms of local community assembly and in turn generate β‐diversity gradients.

SignificanceIdentifying and explaining regional differences in tropical forest dynamics, structure, diversity, and composition are critical for anticipating region-specific responses to global environmental change. Floristic classifications are of fundamental importance for these efforts. Here we provide a global tropical forest classification that is explicitly based on community evolutionary similarity, resulting in identification of five major tropical forest regions and their relationships: (i) Indo-Pacific, (ii) Subtropical, (iii) African, (iv) American, and (v) Dry forests. African and American forests are grouped, reflecting their former western Gondwanan connection, while Indo-Pacific forests range from eastern Africa and Madagascar to Australia and the Pacific. The connection between northern-hemisphere Asian and American forests is confirmed, while Dry forests are identified as a single tropical biome.