Abstract
Biomes are globally distributed, structurally and functionally similar vegetation units defined without reference to plant species composition. The boundaries between biomes are presumed to correspond with species turnover and changes in biogeochemical cycling. Determining the controls of biome distributions is thus critical for anticipating global change impacts. Historically, climate and soils have been understood to adequately explain the global distribution of biomes. Convergent evolution and environmental filtering are assumed to be pervasive, ultimately resulting in deterministic patterns of vegetation structure and function in relation to prevailing environmental conditions. Recent studies have highlighted significant problems with this view of biomes. Firstly, systematic structural and functional divergence within biomes has been identified when comparing environmentally similar, yet floristically distinct regions. Secondly, climatic determinism is being further undermined by evidence suggesting multiple stable biome states are possible for some combinations of climatic drivers. We argue that biomes remain useful and necessary constructs for organizing our knowledge of how ecosystems function and for predicting how they might respond to change. However, biome concepts should acknowledge the limits to predictability from environmental conditions alone and the influence of historical processes on modern vegetation patterns. We discuss how direct mapping of plant structure and function, the incorporation of insights into biome evolution and a new generation of vegetation models will lead to improvements in the concept of what biomes are, where they occur, and efforts to predict their distribution.