Biomes are constructs for organising knowledge on the structure and functioning of the world's ecosystems, and serve as useful units for monitoring how the biosphere responds to anthropogenic drivers, including climate change. The current practice of delimiting biomes relies on expert knowledge. Recent studies have questioned the value of such biome maps for comparative ecology and global-change research, partly due to their subjective origin. Here we propose a flexible method for developing biome maps objectively. The method uses range modelling of several thousands of plant species to reveal spatial attractors for different growth-form assemblages that define biomes. The workflow is illustrated using distribution data from 23,500 African plant species. In an example application, we create a biome map for Africa and use the fitted species models to project biome shifts. In a second example, we map gradients of growth-form suitability that can be used to identify sites for comparative ecology. This method provides a flexible framework that allows a range of biome types to be defined according to user needs. The method also enables projections of biome changes that emerge purely from the individualistic responses of plant species to environmental changes.

The dataset contains growth-form preference matrices (.csv files) for Africa based on the TTR-SDM and MaxEnt, respectively, and biome maps (.tif files) that were produced from these matrices. An R script that illustrates how to produce a biome map from a preference matrix is also provided.