Biome classification influences the projected rate of future biome transitions

Aim. Biome classification schemes are widely used to map biogeographic patterns of vegetation formations on large spatial scales. Future climate change will influence biome patterns, and vegetation models can be used to assess the susceptibility of biomes to experience transitions. However, biome classification is not unique, and various classification schemes and biome maps exist. Here, we aimed to assess how the choice of biome classification schemes influences current and projected future biome patterns. Location. Africa, Australia, Tropical Asia Time period. 2000-2099 Major taxa studied. Tropical vegetation Methods. We used adaptive dynamic global vegetation model version 2 (aDGVM2) to simulate vegetation in the study region. We classified vegetation into biomes using (1) a classification scheme based on the cover of functional types, (2) a cluster analysis based on the cover of functional types, and (3) a cluster analysis based on trait patterns simulated by the aDGVM2. We compared the resulting biome maps to multiple observation-based biome maps and quantified differences in projected biome changes under the RCP8.5 scenario for the different classification schemes. Results. As expected, biome patterns were strongly related to the scheme used for biome classification. The highest data-model agreement was derived for a cluster analysis using 21 simulated traits. Traits related to size were most important for classification. Considering all classification schemes, the area projected to undergo biome transitions under climate change varied between 16.5% and 32.1%. Despite this variability, different schemes consistently showed that grassland and savanna areas are most susceptible to climate change, whereas tropical forests and deserts are stable. Our results demonstrate that traits simulated by aDGVM2 are appropriate to delimit biomes. Main conclusions. Studies projecting biome patterns and transitions under current and future climate should consider applying different biome classification schemes to avoid biases in such projections caused by biome classification schemes