Ancient grasslands at risk

See allHide authors and affiliations

Science  08 Jan 2016:
Vol. 351, Issue 6269, pp. 120-122
DOI: 10.1126/science.aad5132

Embedded Image
Ancient grasslands.

Mosaics of forests and grasslands—as shown here for the southern Patanal, Brazil—are common in the tropics and subtropics. The grasslands have long been interpreted as secondary vegetation produced by deforestation, and many are targets for “reforestation” (1). However, they have a rich endemic ancient biota adapted to frequent fires.


Concerns over deforestation have led to attempts to identify suitable areas for reforestation around the world (1). The most ambitious effort to date is the World Resources Institute (WRI) Atlas of Forest and Landscape Restoration Opportunities (1). This map is linked to a global plan to reforest degraded lands to offset anthropogenic CO2 emissions. The immediate target is the reforestation of 1.5 million km2 by 2020 (2, 3). Vast areas of open grassy vegetation have been identified as suitable for reforestation. But are all these grasslands secondary products of deforestations? Recent research shows that grasslands are often ancient and highly biodiverse, but it remains difficult to distinguish between primary and secondary grasslands on a large scale. Reforestation efforts thus risk converting ancient tropical grasslands to plantations.

Research on tropical grassy biomes has been remarkably neglected relative to forests (4, 5). This neglect has been attributed to the widespread perception that they are human artifacts created and maintained by human-lit fires. Yet, there is evidence for a much more ancient origin of fire-maintained grassy vegetation in regions where the climate and soils can support closed forests (see the figure). Tropical grassy biomes (grasslands and savannas) include ecosystems with a rich plant and animal life entirely restricted to open habitats (4, 6). They provide important ecosystem goods and services to the ~500 million people who live in them (5). Furthermore, the full consequences for the Earth-atmosphere system of converting grassy ecosystems to forests have not been evaluated (7). For example, changes in albedo can offset any gains in stored carbon as the reflective surface of grasslands, especially in the dry season, are converted to the dark absorptive surfaces of forests.

A key clue to the antiquity of grasslands is the presence of species unique (endemic) to open grassy habitats. The cerrado (savannas) of Brazil are ranked 12th on a list of global biodiversity hot spots with about 6000 plant species. Once considered poor in vertebrate species, cerrado is now proving to have a rich, endemic fauna of reptiles (8). The grassland biomes of southern Brazil and South Africa are also rich in plant species, with an estimated 3000 to 4000 in an area of 137,000 km2 in southern Brazil and 3800 in 112,000 km2 in South Africa (4, 9). The Atlantic coastal plain of North America is also species-rich, with some 6000 plant species (nearly a third of the North American flora), of which 29% are endemic (10, 11). Most of the endemic species occur in pine savannas and communities embedded within them. All these systems depend on frequent fires to maintain them and are replaced by closed forests where fires are suppressed.

Underground trees.

Many tropical grassland species have large underground storage organs that enable them to survive frequent grass-fueled fires. Some woody species with tall tree relatives form “underground trees,” with massive belowground branches supporting short aboveground stems that resprout rapidly after a fire.


The Atlantic coastal plain grasslands of North America have only recently been recognized as a significant biodiversity hot spot (11). Noss attributes this oversight to misconceptions and myths, such as the myth that fires can only be ignited by humans. Similar misconceptions are widespread throughout the tropics. In Madagascar, for example, unique for its remarkable endemic forest fauna and flora, the grasslands that cover most of the island were long thought to be produced by deforestation in the aftershock of human settlement a few thousand years ago (12). This narrative of deforestation as the source of grasslands was powerful enough to put off researchers from exploring the diversity of grasslands. Recent grass collections have revealed a remarkably rich grass flora, with some 40% of the 580 grass species endemic to the island. These levels of endemism are among the highest in the world for a grass flora (12). Despite clear evidence for the antiquity of Madagascar's grassland biota, it remains highly uncertain how much of the island's grasslands is a result of deforestation.

Tropical grassy ecosystems are highly flammable. They are most extensive in seasonal wet/dry climates, where high rainfall supports high productivity in the rainy season, whereas the long dry season produces highly flammable grass fuels. Humid tropical grasslands burn several times in a decade; some regularly burn twice in a year. The grassy biomes of Africa alone accounted for more than 70% of the world's burnt area from 1997 to 2009.

Plants living in these highly flammable environments survive by retreating underground, developing large underground storage organs (USOs) that resprout rapidly after a burn (6). Nongrassy herbs (forbs) with USOs commonly flower in the first growing season after a burn before they are smothered by grass growth. Among woody plants, thick bark, insulated buds, clonal spread, and juvenile forms with USOs are adaptations to frequent grass-fueled fires (13).

“Underground trees” are a peculiar growth form characteristic of frequently burnt savannas in South America and Africa (14). These plants have massive underground branches protected from surface fires (see the figure). The leaves are borne on short stems that resprout from the buried branches after fires. Underground trees have evolved multiple times from different lineages in South America and Africa in open grassy habitats with frequent fires and, often, poor soils (13, 15).

Underground trees (geoxylic suffrutices) have close relatives that are true trees. Using a dated phylogeny, Maurin et al. (15) have shown that most underground trees in fire-maintained savannas in Africa originated between 6 and 2.5 million years ago, with peak divergence in the past 2 million years (15). These times of origin are remarkably similar to those obtained from phylogenetic studies of fire-adapted woody legumes in South American savannas (13). They are also consistent with fossil evidence for the spread of tropical grassy systems from the late Miocene and with charcoal records from marine cores that show a sharp increase in fire activity over the same period (16).

Thus, fire-dependent tropical grassy systems predate human deforestation by millions of years. Nevertheless, this is an intriguingly late origin from an Earth history perspective. The causes of this late spread of savannas are currently an area of intense research interest.

A distinctive feature of the flora of tropical grasslands is a marked intolerance of shade. Fire-suppression experiments have led to the elimination of long-lived forbs as a result of shading by an accumulation of dense undecomposed grass litter. Old-growth grassland plants, especially long-lived forbs with USOs, are eliminated by forest expansion or afforestation with pines and eucalypts. Secondary grasslands, formed after deforestation or removal of plantations, have completely different species compositions from primary grasslands (4, 6). Forbs with USOs and underground trees are particularly vulnerable, with no recovery to old-growth conditions after decades of natural succession.

These studies are revealing striking differences in old-growth versus secondary grasslands. However, the scale of analysis does not yet permit large-scale identification of old-growth grasslands. It is not yet possible to map old-growth grasslands at the scale of the WRI Atlas of Forest and Landscape Restoration Opportunities (13). The Food and Agriculture Organization (FAO) uses tree cover and height to classify forest and ignores the herbaceous layer (5). Yet there are profound functional differences between open woodlands with a highly flammable grassy understorey and closed forests that lack fuel to burn. Ultimately, old-growth versus secondary grasslands will have to be classified by identifying characteristics of the herbaceous layer, such as grass composition, that are visible from satellites and that are also proxies for primary grasslands.

There are many fundamental unanswered questions about grassy biomes, their function, origins, and antiquity. The recent spate of papers on these fascinating systems suggests that this neglect is coming to an end. There has been considerable progress in identifying old-growth grasslands and their ecological requirements, and this bodes well for restoring forests and the services they provide while also maintaining ancient tropical grassy ecosystems. It would be a travesty if ancient grasslands are replaced by ill-conceived forest plantation projects because of misconceptions about the origins of tropical grassy systems.


  1. World Resources Institute maps;

Navigate This Article