National and international policy frameworks, such as the European Union's Renewable Energy Directive, increasingly seek to conserve and reference ‘highly biodiverse grasslands’. However, to date there is no systematic global characterization and distribution map for grassland types. To address this gap, we first propose a systematic definition of grassland. We then integrate International Vegetation Classification (IVC) grassland types with the map of Terrestrial Ecoregions of the World (TEOW).
We developed a broad definition of grassland as a distinct biotic and ecological unit, noting its similarity to savanna and distinguishing it from woodland and wetland. A grassland is defined as a non-wetland type with at least 10% vegetation cover, dominated or co-dominated by graminoid and forb growth forms, and where the trees form a single-layer canopy with either less than 10% cover and 5 m height (temperate) or less than 40% cover and 8 m height (tropical). We used the IVC division level to classify grasslands into major regional types. We developed an ecologically meaningful spatial catalogue of IVC grassland types by listing IVC grassland formations and divisions where grassland currently occupies, or historically occupied, at least 10% of an ecoregion in the TEOW framework.
We created a global biogeographical characterization of the Earth's grassland types, describing approximately 75% of IVC grassland divisions with ecoregions. We mapped 49 IVC grassland divisions. Sixteen additional IVC grassland divisions are absent from the map because of the fine-scale distribution of these grassland types.
The framework provided by our geographical mapping effort provides a systematic overview of grasslands and sets the stage for more detailed classification and mapping at finer scales. Each regional grassland type can be characterized in terms of its range of biodiversity, thereby assisting in future policy initiatives.
Grasslands have historically been an area of expansion for human land use (White et al., 2001), and much of the world's highly productive grassland has been converted to crops, mixed farming and artificial pastures (Suttie et al., 2005). In temperate grasslands, this conversion occurred prior to the 1950s (Millennium Ecosystem Assessment, 2005), and the percentage of protection for this biome is lower than for all other biomes (Hoesktra et al., 2005). A current wave of agricultural expansion is occurring in the tropics, with many tropical savannas and grasslands undergoing change (Gibbs et al., 2010). Growth of agricultural sectors in South America (Gavier-Pizarro et al., 2012), southern Africa (Maeda et al., 2010), North America (Landis & Werling, 2010), and Asia (Qiu et al., 2010) heralds new pressures on global grassland ecosystems. Future threats to grasslands also appear high, given a need to feed a rapidly growing human population (Foley et al., 2011).
These threats challenge governments, business and civil society to develop policies that address conversion pressures on global grassland ecosystems and seek to balance development with conservation. However, decision-makers currently lack a framework within which to monitor global grassland biodiversity for both biological uniqueness and total historical distribution. One promising initiative is the International Union for the Conservation of Nature's (IUCN) proposed Red List of Ecosystems, where the likelihood that an ecosystem will persist into the future is assessed (Rodríguez et al., 2010). However, the projected completion date of the global Red List assessment is 2025 (Rodríguez et al., 2012; Keith et al., 2013), and policies are being implemented today. For example, the European Union's Renewable Energy Directive (EU RED) restricts imports of biofuels feedstock harvested from areas containing significant biodiversity and/or carbon stock (European Commission, 2009). A clear intent of this policy is to conserve grassland biodiversity, but the policy cannot be operational on a global basis without a global grassland distribution map as a foundation.
To address this gap, we present a framework for defining world grassland types and a methodology for mapping their geographical distribution. We propose the combination of two systems: the International Vegetation Classification (IVC), to give clarity to the definition of grasslands (Faber-Langendoen et al., 2014), and Terrestrial Ecoregions of the World (TEOW), to provide an initial global geospatial characterization (Olson et al., 2001). By combining these two systems, we generate a systematic, spatially explicit framework that broadly accounts for grassland biodiversity (as vegetation types) and the spatial ecological complexes (as ecoregions) within which grasslands occur. This approach provides a better platform for decision-makers to advance grassland conservation.
Defining grassland: challenges in developing a common framework
A primary obstacle to developing and implementing effective grassland conservation policies is the wide spectrum of grassland definitions. Unlike forests, for which the United Nation's Food and Agriculture Organization (FAO) provides a clear definition (5 m in height, 10% or more canopy cover, > 0.5 ha, and not under agricultural or other non-forest land use; FAO, 2010), grasslands are variously defined (e.g. Gibson, 2009; and see the FAO's compilation of definitions http://www.fao.org/agriculture/crops/thematic-sitemap/theme/spi/gcwg/definitions/en/). This profusion of definitions may be due to the greater difficulty in characterizing the limits of grasslands, a less persistent canopy structure, more frequent disturbance regimes, and their occurrence within a physiognomic continuum between forests and deserts.
Grasslands might well be expected to be dominated by grasses, but the term often has a broader meaning when set in the context of defining a comprehensive set of ecological vegetation types (such as grassland versus forest, desert, tundra or wetland). In that context, the concept still emphasizes dominance by grasses or grass-like plants (graminoids) and the lack of trees, but the full suite of growth forms may include grasses, other narrow-leaved grass-like herbs (i.e. non-woody graminoids) and even forbs (broad-leaf herbs). Perhaps the more technically appropriate term is ‘herbland’ [similar to UNESCO's (1973) ‘Herbaceous Vegetation’], but ‘grassland’ is the most popular, given that grasses are by far the most typical component and because forbs are often mixed within or patchily distributed among grasses (Davies et al., 2004). In his comprehensive review of major grasslands regions of the world, Coupland (1979, p. 22) defined ‘grassland’ as referring to ‘ecosystems in which the dominant vegetative component is comprised of herbaceous species’. Sometimes the term grassland is used even more inclusively to encompass herbs and shrubs (White et al., 2001); grasses and shrubs can form intricate mixes, and dominance may alternate between the two within the span of years or decades. In some cases, grasses may overtop shrubs (Faber-Langendoen et al., 2012).
Here, we consider the various concepts of grasslands and provide a synthesized definition based on previous work. First, we clarify the term ‘grass’, which we define broadly as an herbaceous monocot with narrow leaves, sometimes referred to as a graminoid. Raunkiær (1934, in Mueller-Dombois & Ellenberg, 1974, pp. 458–459) defines ‘grass’ as ‘a caespitose or reptant hemicryptophyte life form’. Box (1981, p. 162) defines it as graminoids that are, ‘narrow-leaved herbs…growing from generally well-developed underground rootstocks which may be either perennial (e.g. rhizomes) or annual…classified as bunched (cespitose), or spreading (sward-forming), and rooting’. The primary taxonomic members are Poaceae, but they may also include Cyperaceae, Restionaceae and other narrow-leaved monocots. We consider grasslands to be dominated by these members, while often containing, and sometimes dominated or codominated by forbs. A dominant or co-dominant is any species or growth form with at least 10% cover (Faber-Langendoen et al., 2012). Grass dominance is clearly expressed when grasses have greater than 25% grass cover (Kucera, 1981) but may be as low as 10% cover if they exceed that of all other growth forms. Shrub cover in grasslands is typically < 25%.
Second, we distinguish largely native or natural grasslands from cultural grasslands. Natural grassland ecosystems are thought to have had a global distribution for at least 15 million years (Jacobs et al., 1999). The widespread expansion of C4 grasses, which developed with seasonal climatic aridification and/or atmospheric change and which grow exclusively in open terrestrial areas, is seen in the macrofossil and pollen record as far back as the Miocene. Additionally, herbivore dental morphology has been shown to have co-evolved with the newly available C4 grasses, substantiating the existence of widespread climax grassland ecosystems prior to the Anthropocene (Coupland, 1992; Jacobs et al., 1999; Edwards et al., 2010). Grasslands today range from strongly cultural, human-created systems, such as exotic grass pastures, to those largely shaped by more natural ecological processes of climate, fire and native grazers (FAO, 2005). For example, Mongolian grasslands have been managed as pasturelands since before the days of Genghis Khan (Li et al., 2006). In Australia, native grasslands are recognized by their component species, distinct from recently introduced exotic pasture grasslands (Lonsdale, 1994; Ash et al., 1997). But, the distinction between natural and cultural grasslands is not always black and white: the western North American grasslands are often referred to as rangelands (which include both shrublands and grasslands) and are often managed as such, but currently they form a continuum of natural (native), semi-natural (naturalized exotic), and cultural (intensive pasture) grasslands. For our purposes, we define native or natural (including semi-natural) grasslands, as those where non-human ecological processes primarily determine species and site characteristics. In other words, the vegetation is composed of a largely spontaneously growing composition of plant species shaped by both geophysical (site) and biotic processes (Küchler, 1969; Westhoff & van der Maarel, 1973). Natural vegetation forms recognizable groupings that can be related to ecological site features. Human activities influence these interactions to varying degrees (e.g. logging, livestock grazing, fire, introduced pathogens), but do not eliminate or dominate the spontaneous processes (Westhoff & van der Maarel, 1973). As with forests in the FAO definition, we exclude cultural grasslands, which are primarily planted and maintained for agricultural reasons (pasture, hay, intensive livestock production). Although these distinctions can sometimes be problematic, they are also consistent with the approach of the Ecosystems of the World project, which provided separate descriptions of natural (Coupland, 1992) and managed grasslands (Breymeyer, 1987).
Third, we clarify the limits of grassland along an ecotone from grassland to woodland. We set a literature-based threshold for grassland with respect to tree cover, beyond which trees become a co-dominant and/or diagnostic part of the plant community concept, exerting disproportionate influence on competition for canopy cover and subsurface resources (White, 1983; Scholes & Hall, 1996; Scholes & Archer, 1997; House et al., 2003; Lock, 2006; Bucini & Hanan, 2007). In the temperate region, tree savannas are more restricted in area and often closely related to or included within the concept of woodlands (Faber-Langendoen et al., 2012). When tree cover exceeds 10% in temperate regions, we exclude it. In the tropics, tree savannas are extensive and overlap with open savannas or grassland. The canopy cover threshold is notoriously variable for tropical wooded grasslands or tree savannas, and varies from low (25%) (UNESCO, 1973; Mueller-Dombois & Ellenberg, 1974), to high (75%) (Mucina & Rutherford, 2006). We used a 40% canopy cover threshold to distinguish between tropical grassland (including wooded grassland) and tropical woodland, with tropical wooded grasslands having a continuous grass layer, trees < 8 m in height, a simple two-layer structure, between 10 and 40% canopy cover, and open grassland having < 10% tree cover. Similarly in need of differentiation are shrublands, defined as where shrubs > 0.5 m tall have > 25% shrub cover (or if < 25% cover, shrubs have at least 10% cover and exceed herbaceous cover), and tree cover is < 10% (Faber-Langendoen et al., 2012) (see Table 1 for a comparison with definitions provided by Lock, 2006).
Table 1. Adapted version of Lock's (2006) table comparing intercontinental (African and South American) variations on the definition of savanna. Our grassland concept includes these three grassland types
Environment and structure
Approx. equivalent South American term(s)
Our review of the cerrado literature suggests that ‘cerrado sensu stricto’ also fits with wooded grassland, but may have canopy cover up to c. 70%. Thus, contra Lock (2006), we would not equate all of the cerrado sensu stricto as ‘woodland’. Similar issues may exist in Africa where e.g. Lock places both Miombo woodland and Miombo savanna in the woodland category.
Single dry season > 4 months. Trees with crown cover < 40%, > 10%. One tree layer. Grasses narrow-leaved, tussock-forming and xeromorphic. Single dry season > 4 months. Fires regular, often annual. Tree-dominated vegetation; crown cover at least 40%. Usually only one main tree layer. Woody climbers and epiphytes absent or very scarce. Grasses narrow-leaved, tussock-forming, often xeromorphic.
Single dry season > 4 months. Bushes (multi-stemmed, short stature) < 40%, > 10%. One shrub layer. Grasses narrow-leaved, tussock-forming and xeromorphic.
Open bushland, bushed grassland, savanna bushland, bush savanna
Campo sujo, sabana arbustiva
Single dry season > 4 months. Woody plants with canopy cover < 10%. Grasses usually tussock-forming and xeromorphic, at least in Africa. Fires regular. Natural grasslands often in sites with seasonal waterlogging, shallow soil or high metallic ion concentrations.
Grass savanna, savanna grassland
Campo limpo (no large woody plants), camp sujo, sabana abierta, sabana lisa
Finally, wetlands are excluded where graminoids and other herbaceous vegetation occur in a matrix with wetland species, including aquatic plants, forbs and mosses. We suggest that although these wetlands may technically meet certain aspects of the grasslands definition, they are typically composed of a range of non-grass vegetation and better treated as part of global wetland definitions, such as that of the Ramsar Convention (Matthews, 1993).
In summation, we propose the following definition of grasslands for global application. A natural or semi-natural grassland is defined by the following characteristics: (1) a non-wetland formation; (2) vascular vegetation has at least 10% cover; (3) graminoids have at least 25% cover (but if < 25% cover, graminoids exceed that of other herbaceous and shrub cover); (4) broad-leaved herbs (forbs) may have variable levels of cover and dominance; (5) shrubs have < 25% canopy cover; (6) and trees: (i) in temperate zones, typically have < 10% canopy cover, are < 5 m tall and single-layered, or (ii) in tropical regions, typically have < 40% canopy cover, are < 8 m tall, and are single layered.
Beyond this basic physiognomic definition of grassland, reference can be made to the floristic composition of a division and lower levels of the IVC hierarchy. For example, decisions about how to classify wooded tropical grasslands with > 40% cover could factor in the degree to which specific grassland species are dominant in the ground layer.
Natural grasslands occur around the world and have been characterized using a number of methods (see Appendix S1 in Supporting Information). For global characterizations, the methods can be grouped into four types: vegetation composition; ecological and economic assessment; ecosystem mapping; and remote sensing classification. The vegetation approach stresses the importance of species and growth forms as a primary expression of a terrestrial ecosystem and uses plant species assemblages to classify stands into plant community types (e.g. ‘associations’, ‘alliances’) and, combined with physiognomy, into broader vegetation types (e.g. classes, divisions, formations) (UNESCO, 1973; Ellenberg, 1988; DiGregario & Janssen, 1998; Faber-Langendoen et al., 2014). The ecological and economic assessment approach characterizes global grassland ecosystem health through an analysis of pressures exerted on the ecosystem, and also reports on the connection to human well-being (Coupland, 1979; White et al., 2001; Suttie et al., 2005). The ecosystem mapping approach emphasizes the geographical or landscape delineation of ecosystem boundaries based on patterns present in biophysical factors, such as climate, landform and, sometimes, floral and faunal evidence (Holdridge, 1967; Uvardy, 1975; Walter & Box, 1976; Schultz, 1995; Bailey, 1996; Olson et al., 2001). The remote sensing method uses the vegetation approach in combination with satellite imagery to create global land cover datasets describing generalized spatial patterns in vegetation, abiotic and anthropogenic features on the Earth's surface (Defries et al., 1995; Loveland & Belward, 1997; Bontemps et al., 2011).
We chose to develop our map of global grassland distribution using a combination of the vegetation approach, represented by the IVC, and a spatially explicit landscape-based approach, as manifested in the TEOW framework. Both systems offer a robust, hierarchical approach to describing global grassland biodiversity.
Materials and methods
We evaluated each ecoregion within the TEOW framework for grassland characteristics and integrated IVC grassland divisions to develop a global distribution map of world grassland types, and reported IVC grassland types if they currently occupy, or historically occupied, at least 10% of an ecoregion (Fig. 1). The IVC is a non-spatial vegetation-based classification system that describes a hierarchy of terrestrial ecosystems using the EcoVeg approach, as described in Faber-Langendoen et al. (2014). This technique uses a combination of physiognomic, floristic, ecological and biogeographical patterns to organize vegetation patterns into an eight-level hierarchy, and has been used as the basis for a vegetation classification standard in several countries and continents (Baldwin & Meades, 2008; Federal Geographic Data Committee, 2008; Faber-Langendoen et al., 2009; Navarro, 2011; Sayre et al., 2013). For the purposes of this study, we focused on two of the higher levels of the IVC hierarchy: the formation and the division (Table 2). In the IVC, a formation represents combinations of dominant and diagnostic growth forms reflecting macroclimatic factors incorporating elevation, seasonality, substrates and hydrological conditions (Faber-Langendoen et al., 2014). Nested within each formation is a set of divisions, which describes broadly uniform growth forms and a broad set of diagnostic plant species at large biogeographical scales, reflecting continental gradients in climate, geology, substrates, hydrology and disturbance regimes (Faber-Langendoen et al., 2014). At both the formation and division levels, the ecological and vegetation types include a range of tree savanna, shrubland and grassland types. For example the IVC division Patagonian Grassland & Shrubland includes reference to both grassland and shrubland. Grassland and shrubland are grouped together at the IVC division level because they strongly overlap in floristic composition, growth forms and biogeography. However, within the hierarchy, a lower level distinction is eventually made between the grassland component and other components. Faber-Langendoen & Josse (2010) drafted a set of formation and division levels that contain grasslands, but in preliminary form and without geographical distribution information. Here we build on that study, providing an updated version of the global divisions that contain grasslands.
Table 2. Comparison of our two classification methods for ecosystems – one based on vegetation and ecological pattern without spatial constraints, the other based on biodiversity and ecological pattern with spatial constraints. See also Fig. 1
International Vegetation Classification: vegetationally constrained hierarchy
Terrestrial Ecoregions of the World: spatially constrained hierarchy
Formation Combinations of dominant and diagnostic growth forms that reflect global macroclimatic conditions as modified by altitude, seasonality of precipitation, substrates, and hydrologic conditions (Federal Geographic Data Committee, 2008; cf. ‘formation-type’ and ‘biome-type’ of Whittaker, 1975; Lincoln et al., 1998), e.g.
Tropical Grassland, Savanna & Shrubland
Temperate & Boreal Grassland & Shrubland
Major Habitat Type/BiomeVegetation structure, ecological dynamics and environmental conditions (Wikramanayake et al., 2002), e.g.
Division Combinations of dominant and diagnostic growth forms and a broad set of diagnostic plant species that reflect biogeographical differences in composition and continental differences in mesoclimate, geology, substrates, hydrology, and disturbance regimes (Federal Geographic Data Committee, 2008).
EcoregionRelatively large units of land containing a distinct assemblage of natural communities and species, with boundaries that approximate the original extent of natural communities prior to major land use change (Olson et al., 2001).
Macrogroup A vegetation unit defined by ‘moderate sets of diagnostic plant species and diagnostic growth forms that reflect biogeographical differences in composition and sub-continental to regional differences in mesoclimate, geology, substrates, hydrology, and disturbance regimes’ (Federal Geographic Data Committee, 2008) cf. Pignatti et al. (1994), Brown (1982).
TEOW is a spatial system of 867 ecoregions nested within a set of 14 global biomes. An ecoregion is a large complex of ecosystems with roughly equivalent biophysical characteristics and species compositions. Importantly, ecoregions are bounded at a regional scale, with boundaries synthesized from previous ecosystem delineation efforts and from the use of biophysical and remotely sensed data (Olson et al., 2001). The TEOW framework separates broadly distinct sets of ecosystems, and includes detailed ecological characterizations describing grassland composition. We found TEOW to be a comprehensive dataset that can describe grassland distribution globally.
To provide an ecologically based geographical distribution of the IVC grassland types provided by Faber-Langendoen & Josse (2010), we listed all ecoregions where the grassland component of an IVC division occupied at least 10% of the given ecoregion. The linkage of TEOW to IVC was completed through an iterative process of comparing and contrasting ecosystem characteristics. A review of TEOW characterizations (http://worldwildlife.org/science/wildfinder/), literature review, outreach to regional grassland experts, and consultation of geographical datasets assembled from remotely sensed data, ecosystem delineations and ecological characterizations was completed to resolve the grassland characteristics of each ecoregion and determine to which IVC division it belonged (see Appendix S2).
Here, we provide an example of the synthesis between TEOW and IVC for the IVC division Eastern Eurasian Cool Semi-Desert Scrub & Grassland. The division is described as extending ‘from Kazakhstan to China, including Mongolia and central China. It is dominated by perennial bunchgrass, ranging from forest steppe to semi-desert steppe and into the montane regions of Tibet’ (Faber-Langendoen & Josse, 2010, Appendix B ‘Division Description and Richness Sum’). The characterization of each ecoregion in the TEOW framework was reviewed for comments on the types of ecosystems present as well as dominant and diagnostic plant species. If there was a lack of clarity on either herbaceous cover or grass species presence, external datasets were consulted. The map of Ecosystems of Mongolia (Guinin, 2005) was used to determine the grassland/desert ecotone of northern China and southern Mongolia, as well as clarify the presence of grassland in the Altai montane forest and forest steppe. Data from the Econet project from WWF-Russia (Pereladova, 2002) were used to evaluate dominant/diagnostic grass species occurrence throughout the central Asian region. Finally, data from Globcover 2009 (Bontemps et al., 2011), Sun (1989) and Zhao & Herzschuh (2009) were used to evaluate the presence of grasslands in the Qiadam Basin semi-desert ecoregion just north of the Tibetan Plateau. The remotely sensed data and literature review indicated that a majority of the ecoregion contains arid desert with little herbaceous cover; however, the eastern section contains enough herbaceous cover (10% grass cover) with grass species present to qualify under the grasslands definition.
Although we typically excluded wetlands from our study, we did include several divisions in the tropical freshwater marsh formation in South America (found in the Pantanal, Humid Chaco, and Guayaquil flooded grasslands ecoregion) because these divisions contain a complex of upland grassland types or grasslands flooded during a short period of time, in enough proportion (at least 10% of an ecoregion contains grassland) so as to qualify as a grassland ecoregion.
Our approach identifies all ecoregions where IVC divisions contain grassland types that form a dominant component of the ecoregion. In a few cases, the original ecoregion boundary was altered because the ecoregion encompassed an overly broad geographical range of vegetation types (grassland and other vegetation). For example, the northern half of the Canadian Aspen forests and parkland ecoregion was removed to isolate the grassland-dominated southern half. The Montana Valley and Foothills ecoregion was split in half to recognize the cool semi-desert climate in the Montana valley as distinct from the temperate mixed grass vegetation and climate of the Montana foothills.
We matched TEOW with IVC divisions and identified 49 taxonomically and spatially distinct grassland types, creating a new global biogeographical representation of Earth's grassland types. Our review led to several additions and refinements to grassland IVC divisions. Faber-Langendoen & Josse (2010) originally described 56 IVC divisions. In this new analysis, 20 IVC divisions were either refined or added to reflect new ecosystem information obtained throughout the analysis. The new or updated IVC divisions reflected climatic differentiation or a more accurate geographical nomenclature (i.e. Pampean Grassland & Shrubland, or semi-arid Pampa). Of the resulting 65 IVC divisions (nested within the nine IVC formations), grassland components were extensive enough in approximately 75% (49/65) (Table 3, Fig. 2) to be displayed within the TEOW distribution map. These IVC divisions were aggregated from 145 ecoregions (Fig. 3). The remaining approximately 25% (16/65) of IVC divisions with major grassland components were not mapped (Table 4) onto TEOW, either because of the patchy and diffuse nature of grasslands in the IVC division, or because the scale of grasslands in the IVC division is smaller than the regional scale at which the ecoregions were drawn. These were frequently Pacific island grasslands, high elevation grasslands, or continental grasslands occurring in a much larger matrix of non-grassland.
Table 3. International Vegetation Classification (IVC) formations and divisions with substantial grasslands showing distribution by Terrestrial Ecoregions of the World (TEOW). TEOW where grassland occupied < 10% of the ecoregion are not shown
Alpine Scrub, Forb Meadow & Grassland
Australian Alpine Scrub, Forb Meadow & Grassland
Australian Alps montane grasslands
Central Asian Alpine Scrub, Forb Meadow & Grassland
Altai alpine meadow and tundra
Altai montane forest and forest steppe
Central Tibetan Plateau alpine steppe
Eastern Himalayan alpine shrub and meadows
Karakoram-West Tibetan Plateau alpine steppe
North Tibetan Plateau-Kunlun Mountains alpine desert
Northwestern Himalayan alpine shrub and meadows
Qilian Mountains subalpine meadows
Sayan Alpine meadows and tundra
Southeast Tibet shrublands and meadows
Tian Shan montane steppe and meadows
Tibetan Plateau alpine shrublands and meadows
Western Himalayan alpine shrub and Meadows
Yarlung Tsangpo arid steppe
European Alpine Vegetation
Alps conifer and mixed forests
New Zealand Alpine Scrub, Forb Meadow & Grassland
South Island montane grasslands
Boreal Grassland, Meadow & Shrubland
Eurasian Boreal Grassland, Meadow & Shrubland
Faroe Islands boreal grasslands
Scandinavian Montane Birch forest and grasslands
Tropical Montane Shrubland, Grassland & Savanna
African (Madagascan) Montane Grassland and Shrubland
Madagascar ericoid thickets
African Montane Grassland and Shrubland
Angolan montane forest-grassland mosaic
East African montane moorlands
Eastern Zimbabwe montane forest-grassland mosaic
Ethiopian montane grasslands and woodlands
Ethiopian montane moorlands
Jos Plateau forest-grassland mosaic
Rwenzori-Virunga montane moorlands
Southern Rift montane forest-grassland mosaic
Brazilian-Parana Montane Shrubland and Grassland
Campos Rupestres montane savanna
Guianan Montane Shrubland and Grassland
Indomalayan Montane Meadow
Kinabalu montane alpine meadows
New Guinea Montane Meadow
Central Range sub-alpine grasslands
Southern African Montane Grassland
Drakensberg alti-montane grasslands and woodlands
Drakensberg montane grasslands, woodlands and forests
Western Eurasian Cool Semi-Desert Scrub & Grassland
Alai-Western Tian Shan steppe
Eastern Anatolian montane steppe
Kopet Dag woodlands and forest steppe
Western North American Cool Semi-Desert Scrub & Grassland
Great Basin shrub steppe
Montana Valley grasslands
Snake-Columbia shrub steppe
Wyoming Basin shrub steppe
Warm Semi-Desert Scrub & Grassland
Australia Warm Semi-Desert Scrub & Grassland
Great Sandy-Tanami desert
Eastern Africa Xeric Scrub and Grassland
Masai xeric grasslands and shrublands
Northern Acacia-Commiphora bushlands and thickets
Southern Acacia-Commiphora bushlands and thickets
Somali Acacia-Commiphora bushlands and thickets
North American Warm Desert Scrub & Grassland
Table 4. International Vegetation Classification (IVC) formations and divisions containing grassland types without a corresponding Terrestrial Ecoregion of the World (TEOW) due to their fine-scale, patchy distribution. The IVC formations are italicized, while the IVC divisions are regular type
Alpine Scrub, Forb Meadow & Grassland
Southern African Alpine Vegetation
Eastern North American Alpine Scrub, Forb Meadow & Grassland
Western North American Alpine Scrub, Forb Meadow & Grassland
Tropical Montane Shrubland, Grassland & Savanna
Caribbean and Central American Montane Shrubland and Grassland
Vancouverian and Rocky Mountain Grassland & Shrubland
Eastern North American Grassland, Meadow & Shrubland
Western North America Interior Sclerophyllous Chaparral Shrubland
Southeastern North American Grassland & Shrubland
European Grassland & Heath
Boreal Grassland, Meadow & Shrubland
North American Boreal Grassland, Meadow & Shrubland
The distribution and variety of grasslands that are dominant in at least one ecoregion is highly variable across the world. South America and Africa have the highest number of mapped dominant grassland divisions, with 16 and 12, respectively. North America and Oceania have the lowest numbers, each with four mapped divisions, and each had six and five divisions, respectively, that were not mapped because, to the best of our knowledge, they are not dominant in any ecoregion (Table 5). Eurasia contains nearly double the number of total grassland ecoregions of the other continents, with those ecoregions aggregated into nine divisions, and with two more divisions that were never dominant within an ecoregion, and therefore were not mapped. Many of the unmapped grassland divisions may have historically only occurred as small patches within the ecoregion, and may only be dominant at very local landscape scales, such as dry hillslopes, glades and rocky grasslands in an otherwise forested landscape.
Table 5. The distribution of International Vegetation Classification (IVC) grassland types per continent. We describe 49 global IVC grassland divisions, with the Mediterranean Basin Dry Grassland IVC division occurring in both Africa and Eurasia
IVC divisions with dominant grassland types per continent
Some IVC divisions are more diverse ecologically than others, as nearly half of ecoregions identified as grasslands occur in just seven IVC divisions. These seven IVC divisions contained 72 ecoregions. The remaining set (42) of the IVC divisions corresponded to the other half of the total (73) ecoregions, indicating that most IVC divisions roughly correspond to a similar scale ecoregions are delineated (Fig. 4).
The IVC division with the largest area, North Sahel Semi-Desert Scrub and Grassland (3,040,000 km2), is composed of only one ecoregion (Sahelian Acacia Savanna) (Table 6), although not all of the area is, or was, grassland. The second largest grassland IVC division is the Great Plains Grassland & Shrubland, which is an aggregation of 15 ecoregions that cover a total land area of approximately 2,980,000 km2. The smallest grassland IVC division is the African (Madagascan) Montane Grassland and Shrubland, which is found in only one ecoregion, with a total land area of 1273 km2.
Table 6. Land area of each International Vegetation Classification (IVC) division with dominant grassland types
African (Madagascan) Montane Grassland and Shrubland
We mapped global grassland ecosystems by linking IVC divisions with their distribution in one or more ecoregions. The TEOW were originally drawn to represent natural communities prior to major land use change (Olson et al., 2001). We suggest that an advantage to using this historical approach of geographical distribution is that it clearly outlines total potential geographical distribution of the individual grassland divisions. Further, the historical approach provides the boundaries needed to complete spatial analyses quantifying total grassland loss, degradation and protection of each division. This spatial information can also be used to evaluate current and historical status of grasslands and their ecosystem processes, including but not limited to hydrological flow, energy cycling, disturbance regimes and ecosystem services.
Our approach capitalizes on the existence of both the IVC and TEOW to develop a platform for global grassland conservation policies. Furthermore, describing grasslands and their spatial distribution through a hierarchy of ecological and vegetation types provides decision-makers with robust information on global grassland biodiversity patterns.
We have demonstrated that an important threshold in spatial scale occurs at the division level, given that species differentiation emerges here. The formation level does not include species, and is only based on climate and distinctive combinations of growth forms. For example, within the temperate grassland formation, there is little to no species overlap between the Eastern Eurasian Grassland & Shrubland division and the Great Plains Grassland & Shrubland division. It is thus at the division level that biodiversity indicators such as species richness, abundance and endemism may begin to be used, facilitating management and policy decisions (Faber-Langendoen & Josse, 2010). However, development of the level below division, that of macrogroup (Faber-Langendoen et al., 2014), would greatly enhance the specificity of the grassland types, because it is more comparable to types used by other widely used classifications. For example, the macrogroup is comparable to the Braun-Blanquet ‘class’ widely used in Europe and elsewhere (Rodwell et al., 2002).
Our criterion of excluding ecoregions with < 10% grassland cover meant that ecoregions where grasses may be common but within woodland systems are omitted. Such ecoregions include the Canadian Aspen forests and parklands at the northern reach of the North American Great Plains division, the Mediterranean dry woodlands and steppe, and the South American dry Chaco. The criterion also led to the exclusion of some semi-deserts, where aridity and edaphic conditions may limit the potential for vegetation growth, but allowed for inclusion of open semi-desert grassland in the North American Chihuahuan desert and the South American Patagonian steppe, and the high montane grasslands of the South American Puna, since at least 10% of their total area corresponds with our grassland definition.
The criterion for trees to have no more than 10% (temperate) or 40% (tropical) cover, and be typically greater than 5 m (temperate) to 8 m (tropical) was critical to evaluating African and Latin American savannas, especially those surrounding the forests of the Congo basin. It resulted in the exclusion of some woodlands that contain a substantial grassland layer, for example African miombo, southern Congolian woodlands, and South American Caatinga, each containing tropical seasonally dry forest or scrub as the dominant vegetation. We included the Cerrado, which historically contained many areas with greater than 40% tree canopy cover, but at least 10% of the ecoregion probably had < 40% tree canopy over 8 m tall with a dominant grass layer, thus qualifying it for inclusion as a grassland ecoregion.
Meanwhile, in the temperate region, many forest–savanna transitional areas, such as the Midwest forest–savanna transition of North America were excluded as grasslands because of the threshold of 10% tree cover with 5 m height. Similarly, the dehesas and montados of the Iberian peninsula (Marañón, 1988; Joffre & Rambal, 2006) were excluded because they contain semi-natural and cultural grassland components in an open woodland matrix, and probably had higher tree canopy cover historically. Nevertheless, a lack of information on the extent and composition of these ecosystems prior to major land use change makes current assessments a challenge.
The IVC places many of the wetland complex/river delta ecosystems in the wetlands formations, resulting in the exclusion of the Nile delta flooded savanna and the Everglades, where grasses may occasionally be dominant. Conversely, there are also grassland types that, while they occur largely within wetland complexes, contain sufficient historical upland grassland communities to be considered grassland by our definition, for example the Guayaquil flooded grasslands in north-western South America.
Thirteen IVC divisions were too limited in distribution to be effectively mapped with our approach, owing to differences in scale between the IVC and TEOW. In some cases, there was no correspondence of TEOW in IVC divisions in montane and alpine elevations, as well as island ecosystems, as small glade or grassy openings in otherwise forested landscapes, or as locally specialized substrate-based ecosystems (e.g. serpentine grassland). Yet because the vegetation types in these places are taxonomically distinct, they warrant a unique IVC division. These divisions highlight the need for a spatial refinement of TEOW, which is a spatial representation of ecosystem expression in a two-tiered hierarchy (biome, ecoregion) nested within biogeographical realms (Olson et al., 2001). The IVC, on the other hand, is a non-spatial vegetation-ecological approach that allows for eight levels of ecological vegetation types. When distinctive combinations of physiognomy, composition and ecology are found to diverge, the IVC is capable of creating a lower level in the hierarchy. As we have demonstrated here, there are significant correspondences between IVC and TEOW with 75% formation/biome and division/ecoregion levels matching when grasslands are dominant components of the landscape. The correspondence decreases where grassland ecosystems occur as small patches or on sites atypical of the TEOW ecosystem regionalization process.
This first approximation of our mapping effort will benefit from further resolving grassland types at finer biogeographical scales. This will take considerable more effort but can build on efforts such as The vegetation of Africa (White, 1983), The vegetation of South Africa, Lesotho and Swaziland (Mucina & Rutherford, 2006), A new map of standardized terrestrial ecosystems of Africa (Sayre et al., 2013), the Ecological systems of Latin America and the Caribbean (Josse et al., 2003), and the Ecological systems of North America (NatureServe, 2009). As noted above, these efforts help establish more detailed concepts at the macrogroup mid-level to attain levels of grassland type specific enough to guide global assessments such as the IUCN Red List of Ecosystems (Keith et al., 2013). Our approach can help set the stage for this work in terms of distribution and definition, as has been done for other major formations, such as tropical dry forest (Miles et al., 2006). A refinement of our analysis of global grassland types could adjust the current spatial boundaries incorporating remote sensing, and spatial analysis to capture the finer scale IVC units at the desired hierarchical level.
Validation of these maps is an important next step. Increasingly detailed land cover and ecosystem maps are becoming available across the globe [e.g. Sayre et al., 2013 for Africa; and landfire for the USA (LANDFIRE, 2013)] and could be used to document the current extent of grasslands. In addition, teams of division-based grassland experts could be formed to create models of grassland extent prior to major land use change.
We believe that our approach of linking vegetation with ecoregions provides a timely framework for policy use. Our global grassland map could support the environmental management goals advocated in the Millennium Ecosystem Assessment (2005), and it may serve as a tool to help convince policy makers and land managers of the conservation value of grassland ecosystems, especially those with high rates of conversion. The definition of grassland that we offer provides a set of criteria that policy makers can use to guide conservation decisions, such as setting sustainability criteria for maintaining native grasslands. The definition also sets the stage for describing the diversity of grasslands around the world through ecosystem mapping, illustrated by the broad-scale distribution map that we provide here.
An analysis of grassland conversion within the mapped IVC divisions is an important next step. Remotely sensed Earth observation data can be used to document explicit rates of conversion due to agricultural activities, such as cultivation of commodity crops and grazing (Ramankutty et al., 2008; Miles et al., 2006).
We hope that the integration of ecoregions and grassland classification will assist ecosystem management efforts through the current period of global economic expansion and population growth. This new map can contribute to improving other quantitative metrics of biodiversity, such as species richness, endemism, abundance, and ecological integrity in relation to the spatial extent where they occur, allowing decision-makers the ability to evaluate species and ecosystems as part of a larger biodiversity matrix of landscape function and process (Noss, 1990, 1999). These metrics may be custom-tailored to the diverse set of grassland types present around the world. Within each grassland division, we can work towards specifying minimum areas of ecological integrity to ensure the persistence of these broad-scale ecosystems in the face of anthropogenic stressors. Finally, combining our dataset and the World Database on Protected Areas (IUCN & UNEP, 2010) can potentially allow us to assess current levels of protection, which would be of value in ensuring that grasslands can continue to provide a variety of high-value ecosystem services.
We would like to thank several experts who generously gave their time to review this manuscript. Robin Abell provided skillful editing of the paper for its content. Roger Sayre and Pat Comer reviewed the paper for its content and our placement of North American grasslands. Gopal Rawat provided important details on Himalayan grasslands. Olga Perelodova provided insights on central Asian grasslands and guided us towards her map of central Asian ecosystems. Chimed-ochir Bazarsad generously provided maps on Mongolian grassland. Eric Dinerstein provided insights on grassland ecology. Mario Barroso offered valuable feedback on Brazilian grasslands. John Benson reviewed our treatment of the Australian grasslands. We also thank three anonymous referees of this journal for their input. All of these contributions have been invaluable to the completion of this paper. We also thank the German Federal Ministry for the Environment, Nature Conservation, Building and Nuclear Safety, who through the International Climate Initiative provided financial support.
Adam Dixon is a Conservation Geographer, interested in observing biodiversity at the landscape to global scale, based at World Wildlife Fund – United States. His work includes applications in land use planning and ecosystem service analysis.
Author contributions: J.M. conceived the idea and provided input into the analysis; A.P.D. led the analysis and writing of the paper. D.F.-L., C.J. and C.J.L. lent ecological expertise and made major contributions to the analysis, writing and editing of the paper.