Aims: We attempt to review the conceptualisation, science and classification of biomes and propose to limit the definition of a biome to potential natural vegetation as determined by general environmental variables.

Results: Classifying the distribution and abundance of vegetation types on earth has been a central tenet of vegetation science since Humboldt’s classic studies in the early 1800s. While the importance of such classifications only grows in the wake of extreme changes, this review demonstrates that there are many fundamentally different approaches to define biomes, hitherto with limited efforts for unifying concepts among disciplines. Consequently, there is little congruence between the resulting maps, and widely used biome maps fail to delimit areas with consistent climate profiles.

Conclusions: Gaps of knowledge are directly related to research avenues, and suggestions for defining and classifying biomes, as well as modelling their distributions, are provided. These suggestions highlight the primary importance of the climate, argue against using anthropogenic drivers to define biomes and stabilize the concept of biome to escape from the current polysemy. The last two decades have seen an emergence of new approaches, e.g., using satellite imagery to determine growth patterns of vegetation, leading to defining biomes based on the objective, observable qualities of the vegetation based on current reality.

climate, climax vegetation, ecozone, formation, global, potential natural vegetation, terrestrial, typology, vegetation classification, zonal

Mapping and classifying the distribution and abundance of the world’s organisms, and shifts in their distribution and abundance, is the only means to understand species’ response to numerous factors stressing those organisms (climate change, pollution, habitat loss, etc.). Examining species’ distributions has been a central tenet of the organismal sciences for 200 years, with the understanding that distributions follow rules and that if we can model those rules, we can predict responses to changes as well as look back historically (e.g., shifts during the Pleistocene, Davis and Shaw 2001; Loidi et al. 2012). Concepts like biomes (developed throughout this review and defined under Final remarks), ecozones, and formations have described such distributions at the regional and global scale due to the importance of this global scale for conservation biology (Chytrý et al. 2020) and answering basic ecological questions (Mucina 2019). For example, the biome concept has been used to examine diversity-productivity (Madrigal-González et al. 2020) and species-area (e.g., see Dengler et al. 2020) relationships, quantify temporal dynamics (Wang and Fensholt 2017), model historical distributions and shifts following climate change (Rowland et al. 2016), and identify areas most affected by human impact and showing evolutionary convergence in plant form and function (Ringelberg et al. 2020). Indeed, our need for understanding regional distributions is growing quickly as we struggle to understand the effects of a rapidly changing climate, land use changes and other anthropogenic impacts on ecosystems (Shukla et al. 2019). While some of the major drivers of distributions (i.e., ‘the rules’) have been identified (e.g., climate factors and phylogeny), and those drivers have been related to certain biota (e.g., southern hemisphere taxa) and biotic traits (e.g., life form, growth form and life history), an accepted classification and map are lacking (Moncrief et al. 2015; Higgins et al. 2016). That is, we have yet to define these regional distributions with metrics and subsequently map the world’s biomes, despite the obvious and pressing need. One reason is that the knowledge and tools for such work have recently exploded and a concerted effort to incorporate these many new ideas has not been achieved (although see Mucina 2019 and Procheş 2020). Toward that goal, we attempt to review here the history, science and classification of biomes and biome-like concepts and propose a conceptual frame in order to build a global biome classification usable for mapping.

We first show the development of the biome concept, then the historical antagonism between the two main approaches to delineating biomes (floristics and physiognomy), and then focus on the development of the physiognomical approach and from that approach to the concept of a biome.

This conceptual review demonstrates that there are many fundamentally different ways to define biomes, hitherto with limited efforts for unifying concepts among disciplines. Consequently, there is little congruence between the resulting maps (Higgins et al. 2016), and widely used biome maps fail to delimit areas with consistent climate profiles (Moncrieff et al. 2015). The approach of the formation concept by Faber-Langendoen et al. (2012), EcoVeg, integrates growth form and global climates as modified by regional edaphic factors but is not yet in a form that can be modelled. The highly clarifying contribution of Procheş (2020) establishes the concept of biome clearly separate from biogeographic concepts such as phytochoria.

Other recent challenges to the biome concept include the finding that vegetation structure and function of the same biome on different continents can differ substantially; for example, savannas (Lehmann et al. 2014; Moncrieff et al. 2016; Muldavin et al. 2021). Moreover, under the same climatic conditions, different plant formations such as savannas and tropical deciduous forests are possible. These alternative states are modulated by complex interactions between climate, soil conditions, herbivores or disturbances such as fire (Breckle 2002; Bond et al. 2005; Moncrieff et al. 2016; Langan et al. 2017). In the case of alternate states under the same climate, one also needs to decide whether these two formations represent one or several biomes. An additional dilemma is that there is no agreement regarding which kinds of edaphic, hydrological, orographic or disturbance-driven deviations should be considered important enough to count them as separate biomes. For example, Olson et al. (2001) recognize mangroves as a biome but not riparian forests. Likewise, mountain areas with their compressed sequence of vegetation belts are not treated consistently and are often simply overlooked (Woodward 2004) or included into a general category (Mucina 2019) most likely due to spatial representation. Lastly, as humans have shifted the biosphere into a new geological epoch, one can question whether it still is appropriate to consider only the potential natural vegetation in a biome map (Ellis et al. 2010; Ellis 2011). The idea of Ellis (2011) to add degree and type of human alterations as an additional dimension to a revised biome concept is appealing.

It has become apparent that a close one-to-one relationship between climate type and physiognomic types has some weaknesses, as different floras show disparity from predicted convergence. For example, the asymmetry between Northern and Southern Hemispheres was initially pointed out by Willdenow (1811), who focused primarily on the floristic differences between hemispheres, rather than on vegetation sensu stricto. This dissimilarity was extensively addressed by Troll (1948) who published his famous average continent, and later by Box (2002), who argued that under similar climatic envelopes, particularly under non-tropical conditions, the physiognomy of vegetation is often different between the latitudes. This dissimilarity is more pronounced in extratropical areas and is in contradiction with the principle of ‘similar-climate, similar-physiognomy’. Thus, climatic differences between the halves of the planet do not sufficiently explain vegetational divergences, and it is likely that a long period of separation and evolutionary history plays a major role in floristic-phylogenetic dissimilarities (Hopper 2009).

The necessity to consider physiology and plant functional types became apparent. Box (1981) attempted to define more precisely the physiological limits of major plant forms and developed an understanding of plant functional types based on physiological limits imposed (largely) by climate. He defined one hundred plant functional types based on climatic filters and used to create biome types based on a culmination of the types available within zones. The advancements in physiological classifications using plant functional types and climate were still, however, formalised within a deterministic framework largely informed by expert opinion (Ni 2001; Bunce et al. 2002; Reu et al. 2011; Lehmann et al. 2011; Moncrieff et al. 2015). Additionally, in spite of these developments in our understanding of physiological tolerances and functional types, a resurgence of purely floristic and compositional approaches also occurred around this period (1980’s).

In recent studies, the importance of phylogeny and floristic divergence in producing different physiognomic profiles within similar climatic envelopes has been highlighted (Pennington et al. 2004; Moncrieff et al 2014, 2015). Plant traits are not only determined by current environmental conditions, but are also the result of inherited ancestral adaptations to past conditions, i.e. “niche conservatism” (Donoghue 2008; Crisp et al. 2009; Loidi 2018). The divergent physiognomy observed in the altitudinal belts of mountains between different geographical areas provides an example. Zonation in tropical mountains does not match the well-known model for temperate and Mediterranean mountains, as initially pointed out by Humboldt. While climatic differences due to latitude could explain part of these differences, we cannot ignore the floristic-phylogenetic divergences (Troll 1961: Stocker 1963) and that additional abiotic and biotic interactions are likely important in driving the evolutionary pattern.

The underlying principles of strict relationships between climate and vegetation used to develop the majority of earlier schema have been re-evaluated by a number of researchers. Bond et al. (2005) showed that fire was a controlling factor in the distribution of physiognomic types across large areas of the world and this helped to explain some of the anomalies encountered under purely deterministic approaches. Crisp et al. (2009) showed that phylogenetic history and continental movements constrained the expression of plant functional types limiting the direct relationships between climate and functional types. Phylogenetic lineages were rarely found to colonise new biomes (Crisp et al. 2009). Montcrieff et al. (2016) highlighted that not only did evolutionary and biogeographic history limit functional convergence, but that ecosystem engineering could occur due to feedbacks between vegetation and climate along with disturbance, thus altering the conditions for plant function influencing the expression of traits. Top-down pressure from large herbivores was shown to alter the combination of traits expressed beyond the effects of climate or local edaphic conditions (Lehmann et al. 2011).

Anthropogenic influences were also considered important to the extreme they severely influence a majority of terrestrial ecosystems of the world. The term “anthromes” has been coined to designate human influenced systems (Ellis and Ramankutty 2008). Clark et al. (2010) suggested that natural selection factors or competition could exceed the influence imposed by climate on composition and expression of functional traits. Further, Jiang et al. (2017) proposed the importance of climatic fluctuations and predictability, and their complexity, as being highly relevant to the structure and functioning of vegetation types, which could also explain divergences based on average climate records. Plants require different strategies and tactics to cope with differing levels of predictability within their environment (Jiang et al. 2017). The non-deterministic nature of plant functional responses provided evidence that alternative stable states of biomes may occur within the same location and under the same climatic conditions, dependent on historical and current influences (Bond et al. 2005; Lehmann et al. 2011; Buitenwerf and Higgins 2016; Moncrieff et al. 2016).

With improved access to geographical information systems and higher computing power over the last two decades, an emergence of more top-down approaches to defining the boundaries of major vegetation types and biomes became a possibility. Such approaches were able to use satellite imagery to determine growth patterns of vegetation, for example NDVI, leading to defining biomes based on the objective, observable qualities of the vegetation (Pennington et al. 2004). Woodward et al. (2004) assessed physiognomic and phenological types based on climate envelopes using remote sensing. Such approaches were viewed as a progress as they allowed interpretation without circularity, but also placed the focus on existing vegetation compared to potential. With global databases, satellite imagery and GIS abilities, more intensive and complex interrogation of global patterns and species functional responses have become achievable. Dynamic Global Vegetation Models (DGVM) were first used by Prentice et al. (1992), who argued that models based on deterministic considerations would be more robust than those based on correlations. Reu et al. (2011) extended the DGVM concept by investigating the survival of plant functional trade-offs across a range of global conditions and regional climatic constraints. Scheiter et al. (2013) proposed that, in order to move away from the fixed plant functional type paradigm, DGVMs should be based in community assembly, competition and coexistence theory.

In this section we offer a summarized conceptual proposal of the term Biome. The proposal combines historic evolution with more recent contributions to the concept, trying to safeguard a necessary stability in the use of the term in order to prevent a “babelization” which we consider entirely inconvenient. In science, concepts can evolve, while avoiding change to the original conceptual underpinnings (semantic area). Similar to the term “species”, which has been used for centuries while the information carried in it has increased enormously (from morphology to current genetics), but we apply it to the same objects as the ancient botanists. If there is a horizontal displacement, i.e., a change in the group of objects included within the concept, excluding some objects and including new objects, that is a change in the meaning (semantic area) and confusion is likely. Science has to stick to the highest terminological accuracy so that the well-known concepts can be enriched but not changed. If there are new concepts, new terms have to be coined to name them. In the case of biome, the most recent version of this term appearing in the literature is that of the Global Ecosystem Typology, issued by the IUCN (Keith et al. 2020). In it, the term biome (functional biome) is used for level 2 of the proposed classification and is based on an imprecise definition with conditions such as “main ecological drivers” and “main ecological functions”. Of the seven biomes recognized for the terrestrial domain (1-tropical-subtropical forests, 2-temperate-boreal forests and woodlands, 3-shrublands and shrubby woodlands, 4-savannas and grasslands, 5-deserts and semi-deserts, 6-polar-alpine, 7-intensive land-use systems), only four, 1, 2, 5 and 6, are determined by climate in a very loose way. Two others are largely miscellaneous units composed by vegetation types representing seral stages associated with disturbance regimes, often fires, and the last one is the unstable and heterogeneous “biome” of intensive land use systems. This conception of biomes is far from the ones based on a climatic or on stable environmental factors determinism. Additionally, different criteria are adopted for the biome definitions (Keith et al. 2020): climate, disturbances, human influence, and that could be considered a source of inconsistency. If a new conceptual entity is proposed, perhaps a new name should be proposed to avoid confusion.

The inclusion of human influences in the conceptual framework of biome has the following objections:

Human influence is relatively new, with notable influences on terrestrial ecosystems beginning approximately 11,000 years ago when the Neolithicum age started and agriculture and cattle raising arose (Lubbock 1913). Before that point, the impact of humans was that of a medium-sized mammal. After that time, these activities expanded throughout the world at very different paces and intensities, transforming the territories in numerous ways, but global change has only occurred in the last few hundred years. In any case, human influence in terrestrial ecosystems has been and is enormous, and manifests in a complexity of ways, depending on geographical conditions, technology and cultural variability.

In addition, the way in which humans have influenced ecosystems has also been heavily influenced by the natural conditions inherent to them. This has to do with profitability of the environment in question; with humans particularly concentrating modifications within highly fertile environments and leaving highly infertile landscapes much less disturbed (YODFELS as opposed OCBILS of Hopper 2009; Hopper et al. 2021). For example, compare and contrast human occupation and use within deserts with the seasonal tropical forest, or on the tundra with the Mediterranean evergreen sclerophyllous forest areas. Modern technology is pushing towards a homogenization of the land uses and species composition and thus of the ways that humans transform natural ecosystems. We can nearly grow tomatoes in the Arabian desert by means of intensive irrigation and we can grow oranges in the tundra if we provide a formidable greenhouse and fertilizers. Thus, human influence is very diverse and is changing with technology, population growth and time and space. We therefore propose that human influence should not be considered as a defining element for biomes and that the creation of Anthromes is counterproductive (Ellis and Ramankutty 2008; Ellis 2013). Concepts that include the influence of humans within a definition of a biome such as anthromes are likely to have a blurring effect that will only be temporary in nature as technology and human uses advance and change over the decades. The biome should be a concept restricted to nature in the first instance. If we do so, we can use the biome concept to assess the degree and type of human alteration on a given site just by comparing with the corresponding biome. This has been also argued in favour of the Potential Natural Vegetation concept (Loidi and Fernández-González 2012).

Another point is disturbances as a main factor in defining biomes. This is also ill-defined because many of these disturbances are human induced (e.g. grazing, browsing, etc.) or the disturbance regime was altered. We need to take particular care if considering the use of disturbances as separating nature versus human disturbances can be highly complex.

The numerous challenges for developing a global biome classification are synonymous with understanding the diversity of life on earth, which essentially point to knowledge gaps, but those challenges also point to opportunities. In simply trying to understand the diversity of life on earth, Hortal et al. (2015) lists seven shortfalls: (i) Linnean shortfall (not all species have been discovered), (ii) Wallacean shortfall (lack of knowledge of species geographical distributions, especially less common species), (iii) Prestonian shortfall (lack of knowledge about species dynamics in space and time), (iv) Darwinian shortfall (lack of knowledge about evolutionary lineage of species and traits), (v) Raunkiaeran Shortfall (lack of knowledge of ecologically relevant species traits), (vi) Hutchinsonian Shortfall (lack of knowledge about species’ tolerances), and (vii) Eltonian Shortfall (lack of knowledge about species’ interactions). These gaps of knowledge are directly related to research avenues for defining and classifying biomes, as well as modelling their distributions. Models provide one way to move the concept of biomes forward in testing the concept – testing the importance of variables like climate, disturbance, and phylogeny. Therefore, refining models of biomes toward prediction will allow the concept of a biome to be tested and ultimately define a biome.

In basic agreement with Mucina (2019), a biome is a large-scale container concept that includes a series of elements that belong to these three categories:

• A biome encapsules all the biological diversity that can be found within its limits: plants, animals, fungi, etc.
• A biome encompasses all the forms of assemblages of these species: populations, communities (coenoses), landscapes.
• A biome encompasses the processes taking place in the frame of the two aforementioned components: ecosystem functioning, dynamic processes, evolutionary processes, disturbances, etc.

The limits of a given biome in comparison with neighbouring biomes are given by:

• Physiognomy, dominant life-forms. Deciduous forests vs. evergreen forests, steppe vs. desert, etc.
• Regional climate or climatic zone. Ever rainy tropical vs. seasonally rainy tropical, boreal vs. temperate, summer rainy vs. winter rainy, etc.
• Ecological factors. Soil fertility, natural disturbance regime, etc.

As an integrative concept, the biome should in first principles be defined by natural features: natural biota (flora, fauna, etc.), natural ecosystems, natural landscapes. Natural is considered when human influence is less apparent at the level of noticeable ecosystem modification.

JL and JTH wrote much of the initial draft but all authors JL, JTH, SF and SL contributed greatly to the writing and construction of the final document.

JL has been supported by the funds of the project IT936-16 for research groups of the Basque Government.

• John Hunter (Corresponding author, jhunter8@bigpond.com), ORCID: https://orcid.org/000-0001-5112-0465
• Scott Franklin (Scott.Franklin@unco.edu), ORCID: https://orcid.org/000-0003-3922-8857
• Sarah Luxton (Sarah.Luxton@nt.gov.au), ORCID: https://orcid.org/0000-0002-4714-9799
• Javier Loidi (javier.loidi@ehu.eus), ORCID: https://orcid.org/0000-0003-3163-2409