SUMMARY

Fire as a global ‘herbivore’: the ecology and evolution of flammable ecosystems William J. Bond

1 and Jon E. Keeley

2,3

1 Department of Botany, University of Cape Town, Rondebosch, South Africa

2 U.S.GeologicalSurvey, Western Ecological ResearchCenter, Sequoia-Kings Canyon National Parks,Three Rivers, CA93271-9651, USA

3 Department of Ecology and Evolutionary Biology, University of California, Los Angeles, CA 90095, USA

It is difficult to find references to fire in general

textbooks on ecology, conservation biology or biogeo-

graphy, in spite of the fact that large parts of the world

burn on a regular basis, and that there is a considerable

literature on the ecology of fire and its use for managing

ecosystems. Fire has been burning ecosystems for

hundreds of millions of years, helping to shape global

biome distribution and to maintain the structure and

function of fire-prone communities. Fire is also a

significant evolutionary force, and is one of the first

tools that humans used to re-shape their world. Here,

we review the recent literature, drawing parallels

between fire and herbivores as alternative consumers

of vegetation. We point to the common questions, and

some surprisingly different answers, that emerge from

viewing fire as a globally significant consumer that is

analogous to herbivory.

Parallels between fire and herbivory

Ecologists and biogeographers generally assume that plant distribution, abundance and, therefore, community composition, structure and biomass, are determined largely by climate and soils. This is implicit in current attempts to model species range shifts in response to climate change [1]. However, nearly 50 years ago, Hair- ston et al. [2] suggested that the properties of ecosystems are instead determined by the regulation of herbivores by predators. In the absence of predators, herbivore popu- lations would proliferate, consuming such large quantities of vegetation that plant communities would be trans- formed to those tolerant of herbivory rather than those best able to compete for resources. Critics claimed that terrestrial plants are largely inedible so that, even without predators, herbivores could seldom consume enough to transform ecosystems [3]. The effects of fire are, in many ways, analogous to those of herbivory, but have been missing from the trophic ecology literature. Although usually treated as a disturbance, fire differs from other disturbances, such as cyclones or floods, in that it feeds on complex organic molecules (as do herbivores) and converts them to organic and mineral products. Fire

Corresponding author: Bond, W.J. ([email protected]). Available online 3 May 2005

www.sciencedirect.com 0169-5347/$ - see front matter Q 2005 Elsevier Ltd. All rights reserved

differs from herbivory in that it regularly consumes dead and living material and, with no protein needed for its growth, has broad dietary preferences. Plants that are inedible for herbivores commonly fuel fires.

How does fire, unconstrained by low food quality, fit the predictions of Hairston et al. [2] as an ecosystem consumer that is unconstrained by predators? Here, we discuss the ecology of flammable ecosystems, using the term ‘consumer control’ for ecosystemsin which fire orherbivores significantly alter biomass, the mix of plant growth forms, and species composition in ecosystems. We contend that consumer control is important ecologically, biogeo- graphically and evolutionarily when the consumer is fire.

Fire and consumer control of ecosystems

Polis [3], in a review of the ‘green world’ hypothesis, argued that terrestrial vegetation is determined largely by climate, locally modified by low-nutrient soils, with consumer control by herbivores sometimes occurring but being localized in space and time. How can the global importance of consumers (herbivores and fires) versus resources (climate and soils) in shaping vegetation be evaluated? A useful alternative to meta-analyses of experimental studies (often limited in space, time, taxonomic bias and reportage counts) is to compare potential versus actual ecosystem properties for a given locality. If an ecosystem differs greatly from its resource- limited potential properties, then it is a candidate for ‘consumer control’, be it either by herbivory or fire (Figure 1). Frequent fires reduce the height of the dominant plants (Figure 2) and, therefore, the position, but not necessarily the amount, of leaves and canopy photosynthesis. Woody plant biomass, rather than primary productivity, is therefore the more revealing measure of consumer control by fire. The problem is how to measure potential biomass, the ‘carrying capacity’ of trees at a site, against which actual ecosystems can be measured.

Dynamic global vegetation models (DGVMs) can be used to provide an approximation of climate-limited potential biomass [4]. DGVMs are complex models, analogous to global climate models, which ‘grow’ plants according to physiological principles using climate and soil physical properties as input [5,6]. The models predict vegetation responses to global change and can simulate

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005

. doi:10.1016/j.tree.2005.04.025

TRENDS in Ecology & Evolution

0

200 Plant available moisture

T re

e b

io m

a ss

High

High

Low Low

Actual

Consumer control

Climate potential

Figure 1. Assessing consumer control of tree biomass. The extent of consumer

control of an ecosystem can be measured as the difference between tree biomass at

‘climate potential’ and the actual tree biomass. Large differences between potential

and actual woody biomass suggest significant consumer control of the ecosystem.

‘Climate potential’ can be viewed as the carrying capacity of a site for trees.

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005388

potential vegetation for any given location (Figure 2). According to these simulations (e.g. Figure 3), vast areas of wooded grasslands in Africa and South America, and smaller areas of grassy ecosystems and shrublands on all vegetated continents, have the climate potential to form forests. Closed forests, which currently cover a quarter of the land surface on Earth, would more than double in extent if world vegetation was as ‘green’ as it could be [4]. These simulations contradict current perceptions that consumer control is of negligible importance in terrestrial ecosystems [3,7]. The biomes most at variance with climate potential are C4 grasslands and savannas, especially in more humid regions, such as Brazilian cerrados and the wetter regions of Africa. These are the most frequently burnt ecosystems in the world, burning several times in a decade and some burning twice a year [8,9]. Thus, fire is the prime candidate for consumer control of large parts of the world. Past or future changes in the extentoftheseecosystems,orspecieswithinthem,cannotbe understood without understanding the ecology of fire.

0

5000

10000

15000

20000

25000

Zimbabwe 1 Z

B io

m a ss

g m

– 2

South Africa

Figure 2. Changes in woody biomass in savanna long-term burning experiments. The un

and the shaded bars indicate biomass where fire has been excluded for 35 years or more.

Woody biomass simulated by the Sheffield DGVM for ‘fire off’ is indicated by the filled

www.sciencedirect.com

The implications of Figure 3 are even more significant when it is recognized that the mismatch with climate potential is based only on biomass and not on changes in species composition. The simulations cannot identify those ecosystems in which fires change species compo- sition without significantly altering the biomass of trees, such as conifer forests [10] or Australian eucalypt communities [11]. The full extent of fire-controlled vegetation, defined as ecosystems that are altered in structure, composition and functioning when fire is released or suppressed, is much greater.

Fire and consumer control of species composition

Trophic cascades, measured as large changes in species composition, are an expected consequence of predator removal in ecosystems where consumers have the poten- tial to proliferate in their absence. Although evidence for trophic cascades in terrestrial ecosystems is disputed [7], cascading changes in species composition are common- place where fire is the consumer. For example, in tropical forests, a single fire can reduce woody plant richness by a third to two-thirds depending on fire severity and can have negative impacts on a diverse array of faunal components [12–15]. Changes in fuel distribution and microclimate after a tropical forest fire increase the probability of more fires and conversion of forest to scrub and grassland [12,15].

By contrast, for ecosystems with a long history of fire, there is concern over the cascading consequences of anthropogenic fire suppression. In tall grass prairies, and comparable grasslands elsewhere, fire suppression has led to the loss of as many as 50% of the plant species [16,17]. Small herbaceous plants with high light require- ments for growth and seedling establishment are the worst affected. Changes in faunal composition have also been reported, for example, in dry dipterocarp woodlands, where fire suppression has resulted in a marked loss of termite species [18]. Even greater species losses occur where fire suppression leads to complete biome switches, such as from savannas to forests [19,20]. There is, as yet, no global synthesis of species turnover in different

imbabwe 2 Venezuela

Site

North America

shaded bars indicate aboveground woody biomass in frequently burnt treatments

Sites are ranked, from left to right, according to increasing plant available moisture.

squares. Modified, with permission from the New Phytologist Trust, from [4].

TRENDS in Ecology & Evolution

Bare C3 C4 Ang CropGym

(a)

(b)

Key:

1

2 3

4 5

6 7 8

9

10

Figure 3. A comparison of global biome distribution at climate potential (a) versus

actual vegetation (b). Biomes are represented by the cover of the dominant plant

functional type: C3 grasses or shrubs; C4 grasses or shrubs; Ang, angiosperm trees;

gym, gymnosperm trees (mainly conifers). The numbers indicate sites where fire

has been excluded for several decades. All the higher rainfall sites showed a

successional tendency to form forest following suppression of fire. The map of

potential World vegetation, limited only by climate, was simulated using a DGVM

(using global climate and soil databases). The map of actual vegetation was sourced

from ISLSCP: (ftp://daac.gsfc.nasa.gov/data/inter_disc/biosphere/land_cover/); repro-

duced, with permission from the New Phytologist Trust, from [4].

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005 389

ecosystems and under different fire regimes following fire release or suppression. We would expect a continuum of responses from near-complete species replacement follow- ing biome switches to negligible changes in ecosystems where fires, although predictable, are infrequent. The Yellowstone fires of 1988, for example, caused no loss or gain of species in this landscape [21]. Thus, it is not yet possible to draw a global map to show the extent of ecosystems whose species composition would change significantly if fires were suppressed.

The variable nature of fire as a consumer control

Flammable ecosystems include boreal forests, eucalypt woodlands, shrublands, grasslands and savannas. Why, if fire is such an influential consumer, is there such a diversity of growth form mixtures in flammable ecosys- tems? Fire ecologists have looked first to the diversity of fire regimes for answers. A fire regime includes the patterns of frequency, season, type, severity and extent of fires in a landscape (Box 1). Vegetation consumed and patterns of fire spread vary across landscapes, and different fire regimes produce different landscape pattern- ing and select for different plant attributes. It follows that changes in fire regimes, within a given landscape, should have major ecosystem consequences.

Consider the conifer forests of southwestern North America. In these semi-arid landscapes, forests have long

www.sciencedirect.com

been shaped by a fire regime of frequent relatively low intensity (low flame height and temperature) surface fires. These forests share attributes with subtropical grasslands in that fires are ignited by frequent lightning strikes at the beginning of the monsoon season, when the fuels are at their driest. However, primary productivity in conifer forests is lower than in mesic savannas owing to their greater aridity and this translates into lower fire frequency, lower fire intensity and greater heterogeneity in ‘feeding patterns’ of the fire [22]. As a consequence, opportunities exist for the occasional establishment of trees that persist to form low-density forests. Fires exhibit a sort of ‘selective herbivory’, consuming herbaceous surface biomass but leaving the dominant overstorey trees untouched. Following human settlement during the early 20th century, these conifer landscapes have been managed with a policy of total fire suppression, which is a fortuitous experiment on how fire controls vegetation structure, and has resulted in near-total fire exclusion. Forests that naturally burned at rates of once or twice a decade have now gone unburned for more than a century [23], resulting in major shifts in ecosystem structure and function. Tree density has increased by an order of magnitude or more, with major losses in the herbaceous understorey and species diversity. In addition, the absence of fire has resulted in changes in many ecosystem components. Of profound management importance is the fact that fire suppression has lead to fuel accumulation and this has set the forest on a different trajectory such that, when fires do occur, they now feed as massive forest- consuming ‘monsters’, rather than in the manner of ground-dwelling herbivores.

Most work on fire regimes is constrained to particular landscapes and ecosystems. There is no global synthesis on what determines fire regimes in world ecosystems. We do not yet understand the synergies and relative import- ance of ignition, dry periods, the properties of vegetation as fuel, or topographic barriers to fire spread in determin- ing which fire regimes occur where. This seriously under- mines our ability to predict the consequences of global change for fire-affected ecosystems or to interpret past changes in the distribution of flammable ecosystems. What is clear is that different fire regimes select for different plant attributes and similar fire regimes select for similar attributes. Savanna ecologists worldwide find similar plant traits with similar fire responses [24]. Ecologists working in Mediterranean-type shrublands find convergent fire-related plant traits on different continents [25,26] but these are different from those of savannas. Transgressing from one fire regime to another seems to be as difficult as finding commonalities between insect and mammal herbivory, because the biology of the ‘organisms’ is so different.

Fire and community assembly

Hairston et al. [2] predicted relatively little competition between plants where herbivores proliferate in the absence of predators, because plant growth would be limited more by consumption than by resources. Instead, community assemblages would comprise species that are best able to persist and thrive in the face of repeated

Box 1. Fire regimes

Gill [61] introduced the concept of a fire regime, which we have

modified to include: (i) fuel consumption and fire spread patterns; (ii)

intensity; (iii) severity; (iv) frequency; and (v) seasonality.

Fuel consumption and fire spread Fires consume a range of fuel types, which has profound

impacts on ecosystems. Surface fires spread by fuels that are

close to the ground, such as grass or dead leaf and stem

material, whereas crown fires burn in the canopies of shrub- and

tree-dominated associations. Ground fires burn soils that are rich

in organic matter. They can be ignited by lightning strikes and

can smolder for long periods until changes in the weather favor

surface or crown fires.

Some forests have a heterogeneous mix of surface fires,

crown fires and unburned patches, which is important to

ecosystem processes such as tree recruitment. For example, in

the mixed conifer forests of the Sierra Nevada in California,

patches of high-intensity fires produce light gaps that are

important for tree regeneration [62]. These gaps also accumulate

fuels at a slower rate and thus have a greater probability of

being missed by fires until saplings reach sufficient size to

withstand them [63].

The ecological importance of fire size varies with the

ecosystem and also with different species in the system. For

example, chaparral shrublands commonly experience large

crown fires that can completely denude tens of thousands of

hectares. This poses no threat to the plant species in these

ecosystems because regeneration is entirely dependent upon

endogenous processes (Box 2). However, mixed conifer forests

in the western USA are potentially more sensitive to fire size.

Historically, these forests have burned with a mix of surface

fires, which left dominant trees alive, and crown fires, which

killed all trees within small patches from a few hundred square

meters to a few hundred hectares. Reproduction of the dominant

trees requires gaps generated by crown fires, but they must be

within dispersal distance of parent trees. When crown fires are

very large, regeneration is negatively impacted.

Intensity Fire intensity refers to the energy release or, more loosely, to

other direct measures of fire heating or behavior, such as flame

length and rate of spread. Fireline intensity, which is the energy

per length of fire front, is increasingly used as a standard for fire

intensity.

Severity Although fire intensity is a measure of immense importance to fire

fighters, ecologists are often more interested in fire severity, broadly

defined as a measure of ecosystem impact. In forested ecosystems,

tree mortality is commonly used as a metric for fire severity;

however, other metrics are used in shrublands where all above-

ground plants are consumed.

Frequency Fire frequency is the occurrence of fire for an area and time period of

interest. There are complications with assessing fire frequency that

involve complex fire behavior at different spatial scales with different

limitations. Fire rotation interval is the time required to burn the

equivalent of a specified area, whereas fire return interval is the time

interval between fires at any one site [10].

Season Fire season is dictated by the coincidence of ignitions and low fuel

moisture. This is usually the driest time of the year, which varies with

regional climate. In many ecosystems, humans have greatly altered

fire season by providing ignitions outside the natural lightning storm

period.

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005390

www.sciencedirect.com

defoliation. One of the striking features of the fire ecology literature is that there are many studies on life-history traits that enable the persistence of species in a given fire regime (Box 2), but few on resource acquisition and competition. The consistency with the predictions of Hairston et al. [2], that competition will be of minor importance in consumer-controlled ecosystems, seems to have gone unnoticed.

The plant traits that are important for fire persistence are different in communities that experience different fire regimes. In crown-fire regimes, where all woody biomass is consumed, there are numerous studies of the mode of recovery from burning (vegetative sprouting or non- sprouting), fire-stimulated recruitment, time to first reproduction and the persistence of seedbanks to the next fire [20,26,27]. These plant traits, together with the patterns of fire consumption, especially its frequency, are widely used for predicting compatible species assemblages [26,28]. However, community membership is seldom attributed to competitive interactions with other plant species, except when those species change the disturbance regime [29].

In surface-fire regimes, such as savannas, fires feed selectively, consuming plants in the grass layer but not trees taller than 2–4 m. The coexistence of trees and grasses has been attributed to niche differentiation, with grasses being the more successful competitors for resources in the soil surface, and trees accessing resources in deeper soil layers [30]. An alternative idea, consistent with consumer-controlled ecosystems, is that tree cover is limited by demographic bottlenecks at different life- history stages in tree growth [30,31]. Fire would be a major cause of these bottlenecks in frequently burnt savannas, reducing seedling establishment and prevent- ing saplings from emerging from the ‘fire trap’, the flame zone produced by grass fires. Vertebrate herbivores have analogous effects, suppressing seedlings by heavy brows- ing with rare burst of recruitment when plants are released from herbivory [32]. The niche differentiation hypothesis predicts no changes in tree cover from fire suppression (or herbivore exclusion) because tree cover is limited by resource competition. But many long-term fire exclusion experiments (Figure 2) show that tree cover is limited by fire. In these instances, consumer control, rather than resource competition, determines tree cover [33].

Fire as an evolutionary agent

There are few studies of the evolution of fire-adaptive traits, and many plant traits have been uncritically labeled as ‘fire adaptations’ without any rigorous analysis either as to the functional importance of the trait, or its phylogenetic origin. For example, post-burn sprouting is often seen as a ‘fire adaptation’, but sprouting per se is a widespread trait in angiosperms. Evolutionary interpret- ationsoftheloss or gain ofsprouting in different fireregimes make no sense without phylogenetic analysis [34,35].

Among the most compelling new studies are those exploring the evolution of flammability. In a debate echoing that over whether plants have evolved to promote herbivory (and just as controversial), ecologists have asked whether plants in fire-maintained ecosystems

Box 2. Life histories shaped by fire

Of the many traits that can be interpreted as being of functional

importance in fire-controlled environments, two have captured most

attention: sprouting and fire-triggered seedling recruitment. Sprouting

is the vegetative regeneration that occurs following the destruction of

living tissues. This can be either from roots or stems following the

death of all aboveground tissues, or along stems where branches

have been killed. Sprouting is a widespread trait in woody species

and is not closely tied to fire-prone environments [35]. One

exception is Pinus, a genus in which sprouting is rare and

apparently derived in crown fire ecosystems [40]. However,

sprouting from basal lignotubers that are produced as a normal

development stage is a combination much more commonly found

in fire-prone Mediterranean-climate ecosystems [25]. Sprouting in

the context of other life-history characteristics represents complex

patterns that have recently been reviewed elsewhere [35].

Many species in fire-prone environments with stand-replacing fires

have seedling recruitment restricted to the first postfire year [20,27]. In

flammable southern hemisphere shrublands, many species produce

serotinous fruits that open following fire and disperse seeds that

readily germinate following the wet season rains [64]. In comparable

shrublands of the northern hemisphere, serotiny is relatively rare. In

both hemispheres, many species produce seeds that are dormant and

accumulate in the soil. Germination is triggered by either heat or

smoke (or charred wood) [65]. Heat-stimulated germination is typically

in hard-seeded species that have a physical seed coat barrier to water

uptake. Germination is triggered by heat shock from fire, or by high soil

temperatures on open sites. There is a marked phylogenetic pattern in

that certain plant families are associated with either one mode or

another; for example, heat-stimulated germination is widespread in

Fabaceae, Cistaceae, Convolvulaceae and Sterculiaceae, and smoke-

stimulated germination is lacking [65]. Heat-stimulated germination is

globally widespread in numerous fire-prone ecosystems. Chemical

stimulated germination is triggered by smoke and/or charred wood. It

has, so far, been found to be important in only three Mediterranean-

climate shrublands, California chaparral [65], South African fynbos [66]

and Australian heathlands [67]. In California, the plant families in which

this germination mode is found are generally not the same as in the

southern hemisphere shrublands, indicating that this trait might have

convergently evolved.

Fire-stimulated flowering is another mechanism for post-burn

seedling recruitment [20]. Flowering occurs in the first postfire year

on resprouts from bulbs or rhizomes, followed by abundant seedling

recruitment in the second postfire year. Most species continue to

flower sporadically in later years, thus there is no obligate dependence

on fire for flowering. One exception is the South African fynbos

geophyte Cyrtanthus ventricosus, which germinates within days of a

fire, regardless of the season, and remains dormant until flowering is

again stimulated by smoke from another fire [68].

Not all species in fire-prone environments have life histories that

have been shaped by fire. In Californian and Mediterranean Basin

shrublands, many species have seedling recruitment that is restricted

to fire-free conditions [69]. They have a suite of reproductive traits,

including seed dispersal and seed germination behavior, which are

quite distinct from species with fire-stimulated seedling recruitment. In

both ecosystems [69,70], these non-fire types are from older lineages

and are derived from taxa that had origins under a different climate. It

has been suggested that these traits are no longer adaptive and

represent historical effects and species sorting processes [70]. An

alternative view is that these life-history syndromes are adapted to

habitats that still exist in fire-prone landscapes, and the coexistence of

fire-type and non-fire types is promoted by natural variability in fire

frequency [69].

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005 391

have evolved flammability. Are there benefits for flam- mable plants that outweigh the costs to survival of burning more fiercely? Theory predicts that flammability could, indeed, evolve if fire spread from a flammable plant to kill its neighbors, and if the progeny of more flammable mutants were more likely to recruit into the gaps created [36,37]. In these models, flammability acts as a ‘niche constructing’ trait [38,39], modifying the local environ- ment to the benefit of the flammable genotype. This hypothesis makes the testable prediction that flammable morphology and fire-stimulated recruitment should be correlated traits, and there is some support for this prediction in pines [40]. In Pinus, serotiny (the retention of seed in cones which open after a fire), a fire-recruitment trait, is correlated with dead branch retention, a flamm- ability trait. Plants that retain dead branches are more likely to carry a fire into the canopy than are plants that self prune. Schwilk and Ackerly [41] tested whether these traits showed correlated evolution in pine phylogeny. Using a set of ‘supertree’ phylogenies, the authors found strong support for the predicted association between serotiny and dead branch retention, and also between these and other ‘fire-embracing’ morphological traits, such as thin bark, early maturation age and more flammable foliage, which would be expected in these stand- replacing fire regimes [40]. It would be intriguing to explore the evolution of flammability in other taxa and other ecosystems. Has ‘niche construction’, via the evolution of flammability of common species, played a part in the spread of the flammable formations in which they are contained?

www.sciencedirect.com

Studies of trait evolution, and the origins of the woody flora of savannas, are hampered by our lack of under- standing of the key traits needed to survive in grass- fuelled fire regimes. Traits that are common in crown-fire regimes are rare or absent in savannas [40]. In productive grassy ecosystems, fires are too frequent to provide safe sites for seedlings and fire-stimulated seedling recruit- ment, including serotiny, seems to be an exception. Fires are too frequent for the evolution of woody non-sprouters and sprouting is the norm [31,42,43].

Trees that survive anthropogenic fires in tropical forests tend to be those that have thicker, insulating bark [12]. Although trees in savannas are often thick barked, regeneration of new plants is perhaps the main obstacle for maintaining populations. Seedlings and saplings face frequent and severe fire damage in mesic savannas. Between fires, seeds have to germinate and seedlings have to acquire bud and root reserves to resprout to survive the next fire. Given that fires occur several times in a decade, seedlings would need to acquire the ability to resprout rapidly. A recent study in Brazil confirms this conjecture [44]. The only consistent differences between seedlings of sister species from forest and savanna habitats was the greater allocation to coarse roots, associated with sprouting, in the savanna species and more plastic light responses [44].

Frequent fires also select for a peculiar sapling growth form, with pole-like stems and swollen underground roots (Figure 4, [45,46]). The pole-like stem facilitates rapid bolting towards a height that is out of reach of surface fires and the roots provide the resources to resprout if the stem fails to reach a fire-proof size before the next burn.

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005392

Saplings can be trapped in the flame zone for decades. In longleaf pine, Pinus palustris, from the southeastern USA, saplings bolt only once and fire-induced sapling mortality is high [47] presumably because they lack comparably large root reserves. In the polymorphic African tree, Acacia karroo, pole-forming saplings occur in frequently burnt savannas whereas plants form cage- like architectures where fires do not occur [46] (Figure 4). The contrasting sapling architectures suggest alternative evolutionary trajectories when plants are exposed to selection by different types of consumer (i.e. fire versus browsers).

Fire and the origin of biomes

The large area occupied by flammable biomes, especially in the tropics and sub-tropics, has often been attributed to anthropogenic burning. Although anthropogenic fires

Figure 4. Fire, herbivory and tree architecture in savannas. (a) Grass-fuelled fires are very

and saplings. However, flame heights are too low to cause significant damage to em

architecture, such as this Acacia karroo from a South African savanna. (c) The cage-lik

(d) Shows a young sapling of Terminalia sericea, a common African savanna tree specie

sapling growth above the flame zone while also ensuring the ability of the plant to resp

www.sciencedirect.com

have extended areas of flammable vegetation [48,49], there is now evidence that natural fires occurred long before humans and that flammable ecosystems pre-date anthropogenic burning by millions of years [4,50]. Stable isotope evidence shows that C4 grassy ecosystems, the most extensive flammable formation worldwide, first appeared between 6 million years ago (Ma) and 8 Ma [51]. Cerling et al. [51] hypothesized that this was due to decreasing atmospheric CO2, but recent studies of paleoatmospheres do not support their assumption of low concentrations of CO2 during the late Miocene [52,53]. The spread of grasses has often been attributed to coevolution with mammal grazers. However it is hard to see how, by consuming grass, grazers would promote the spread of grasslands at the expense of forests. By contrast, grass-fuelled fires are known to promote the spread of grassy ecosystems by carving holes in forests [29]. Strong

frequent in mesic savannas, posing a problem for the recruitment of tree seedlings

ergent trees. (b) Many trees in frequently burnt savannas have pole-like sapling

e architecture of A. karroo occurs where fires are rare but browsers are common.

s The combination of pole-like stems and reserves stored in swollen roots facilitate

rout after repeated fires.

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005 393

candidates for feeding the fires are the highly productive and highly flammable C4 grasses of the humid tropics, which began replacing forests from the late Tertiary onwards [54–56].

Recent studies in Australia link fire to the origin and spread of the biomes of that continent. Bowman [11] set out to explain the distribution of the small fragments of ‘rainforest’ distributed as an archipelago in the vast seas of eucalypt formations. The rainforests are closed for- mations of fire-intolerant trees with shade-tolerant under- storey species. By contrast, the eucalypt formations support shade-intolerant understoreys of flammable grasses or shrubs. Similar anomalous forest patches in a matrix of flammable grasslands are widespread in Africa and South America [4,57]. Generations of ecologists have debated the determinants of these alternative biome types, citing aridity, fire, nutrient-poor soils and anthro- pogenic burning. Bowman critically evaluated the evi- dence for all of these for Australia. Both biomes occur across a wide rainfall gradient with ‘rainforest’ (a misnomer) petering out where rainfall drops below 600 mm yK1. Contrary to popular perceptions, the flam- mable formations are not ‘arid’. Indeed, eucalypts are less tolerant of drought than are many rainforest trees sharing the same climate [11]. Bowman concluded that the only consistent difference between trees of the two formations was in their fire response: eucalypts have a remarkable ability to survive and thrive under frequent fires, whereas rainforest trees are killed by repeated burning [11,58]. The implication is that if you could ‘switch off’ fires for a long enough period, large areas of Australia would support an entirely different ‘rainforest’ flora. This process is underway in wet eucalypt formations, where fires have been reduced and rainforests are invading [59]. However, continent-wide replacement is unlikely because the rate of rainforest colonization is slow in drier climates and on nutrient-poor soils compared with the frequency of fires [59].

Recent molecular systematic studies have found that the phylogenetic structure of legumes was caused as much by ecological setting as by tectonic history and the location of land masses [60]. The implication is that speciation and subsequent dispersal of some taxa are confined ecologically, within biomes, as much as by dispersal problems across oceans. The appearance of grass-fuelled fires might be an example of an intra- continental ‘vicariance’ event, segregating taxa into those that could tolerate fires and those that could not. New molecular tools, coupled with focused ecological studies, promise new insights into the evolutionary history of those parts of the world where fire uncouples biomes from their climate-limited potential.

Conclusions

We have shown major similarities between fire and herbivory and argued for a more-inclusive view of top- down, or, in this instance, consumer control of biomes. We believe that the global extent of fire as a consumer, its many parallels with herbivory, its role in selecting for particular plant traits and in the evolution of biomes, is worthy of much wider attention from ecologists.

www.sciencedirect.com

There is an added incentive for greater understanding of fire as a globally important consumer. Climate change, habitat fragmentation, the unprecedented transport of highly flammable plants to novel settings, and the ubiquitous overlay of human impacts on fire regimes demand a new level of synthetic understanding for our peaceful coexistence with this charismatic beast.

Acknowledgements We thank Jeremy Midgley, C.J. Fotheringham, Ian Woodward, Guy Midgley, Ross Bradstock, David Keith, Dave Bowman, Malcolm Gill, Alan Andersen and Herve Fritz for useful discussions. Jeremy Midgley and anonymous reviewers provided useful comments on the article.

References

1 Thomas, C.D. et al. (2004) Extinction risk from climate change. Nature 427, 145–148

2 Hairston, N. et al. (1960) Community structure, population control and competition. Am. Nat. 94, 421–425

3 Polis, G.A. (1999) Why are parts of the world green? Multiple factors control productivity and the distribution of biomes. Oikos 86, 3–15

4 Bond, W.J. et al. (2005) The global distribution of ecosystems in a world without fire. New Phytol. 165, 525–538

5 Cramer, W. et al. (2001) Global response of terrestrial ecosystem structure and function to CO2 and climate change: results from six dynamic global vegetation models. Glob. Change Biol. 7, 357–373

6 Woodward, F.I. and Lomas, M.R. (2004) Vegetation dynamics – simulating responses to climatic change. Biol. Rev 79, 643–670

7 Polis, G.A. et al. (2000) When is a trophic cascade a trophic cascade? Trends Ecol. Evol. 15, 473–475

8 Barbosa, P.M. et al. (1999) An assessment of fire in Africa (1981–1991): burnt areas, burnt biomass and atmospheric emissions. Global Biogeochem. Cycles 13, 933–950

9 Dwyer, E. et al. (2000) Characterization of the spatio-temporal patterns of global fire activity using satellite imagery for the period April 1992 to March 1993. J. Biogeogr. 27, 57–69

10 Agee, J.K. (1993) Fire Ecology of Pacific Northwest Forests, Island Press

11 Bowman, D.M.J.S. (2000) Australian Rainforests: Islands of Green in a Land of Fire, Cambridge University Press

12 Cochrane, M.A. (2003) Fire science for rainforests. Nature 421, 913–919

13 Cleary, D.F.R. and Gennert, M.J. (2004) Changes in rain forest butterfly diversity following major ENSO-induced fires in Borneo. Glob. Ecol. Biogeogr. 13, 129–140

14 Barlow, J. and Peres, C.A. (2004) Ecological responses to El Niño- induced surface fires in central Brazilian Amazonia: management implications for flammable tropical forests. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, 367–380

15 Laurance, W.F. (2003) Slow burn: the insidious effects of surface fires on tropical forests. Trends Ecol. Evol. 18, 209–212

16 Leach, M.K. and Givnish, T.J. (1996) Ecological determinants of species loss in remnant prairies. Science 273, 1555–1558

17 Uys, R. et al. (2004) The effects of different fire regimes on plant diversity in southern African grasslands. Biol. Conserv. 118, 489–499

18 Davies, R.G. (1997) Termite species richness in fire-prone and fire- protected dry deciduous dipterocarp forest in Doi Suthep-Pui National Park, northern Thailand. J. Trop. Ecol. 13, 153–160

19 Peterson, D.W. and Reich, P.B. (2001) Prescribed fire in oak savanna: fire frequency effects on stand structure and dynamics. Ecol. Appl. 11, 914–927

20 Bond, W.J. and van Wilgen, B.W. (1996) Fire and Plants, Chapman & Hall

21 Turner, M.G. et al. (2003) Surprises and lessons from the 1988 Yellowstone fires. Front. Ecol. Env. 1, 351–358

22 Covington, W.W. and Moore, M.M. (1994) Southwestern ponderosa forest structure. Changes since Euro-American settlement. J. For. 92, 39–47

23 Rollins, M. et al. (2002) Landscape-scale controls over 20(th) century fire occurrence in two large Rocky Mountain (U. S. A.) wilderness areas. Landscape Ecol. 6, 539–557

Review TRENDS in Ecology and Evolution Vol.20 No.7 July 2005394

24 Scholes, R.J. and Archer, S.R. (1997) Tree-grass interactions in savannas. Annu. Rev. Ecol. Syst. 28, 517–544

25 Keeley, J.E. (1986) Resilience of Mediterranean shrub communities to fire. In Resilience in Mediterranean-type Ecosystems (Dell, B. et al., eds), pp. 95–112, Dr. W. Junk

26 Pausas, J.G. et al. (2004) Plant functional traits in relation to fire in crown-fire ecosystems. Ecology 85, 1085–1100

27 Whelan, R.J. (1995) The Ecology of Fire, Cambridge University Press 28 Bradstock, R.A. and Kenny, B.J. (2003) Application of plant functional

types to fire management. J. Veg. Sci. 14, 345–354 29 Brooks, M.L. et al. (2004) Effects of invasive alien plants on fire

regimes. Bioscience 54, 677–688 30 Sankaran, M. et al. (2004) Tree-grass coexistence in savannas

revisited: insights from an examination of assumptions and mechan- isms invoked in existing models. Ecol. Lett. 7, 480–490

31 Higgins, S.I. et al. (2000) Fire, resprouting and variability: a recipe for tree-grass coexistence in savanna. J. Ecol. 88, 213–229

32 Prins, H.H.T. and van der Jeugd, H.P. (1993) Herbivore population crashes and woodland structure in East Africa. J. Ecol. 81, 305–314

33 Tilman, D. et al. (2000) Fire suppression and ecosystem carbon storage. Ecology 81, 2680–2685

34 Verdu, M. (2000) Ecological and evolutionary differences between Mediterranean seeders and resprouters. J. Veg. Sci. 11, 265–268

35 Bond, W.J. and Midgley, J.J. (2003) The evolutionary ecology of sprouting in woody plants. Int. J. Plant Sci. 164 (Suppl. 3), S103–S114

36 Bond, W.J. and Midgley, J.J. (1995) Kill thy neighbour: an individual- istic argument for the evolution of flammability. Oikos 73, 79–85

37 Kerr, B. et al. (1999) Re-kindling an old flame: a haploid model for the evolution and impact of flammability in resprouting plants. Evol. Ecol. Res. 1, 807–833

38 Odling-Smee, F.J. et al. (1996) Niche construction. Am. Nat. 147, 641–648

39 Schwilk, D.W. (2003) Flammability is a niche construction trait: canopy architecture affects fire intensity. Am. Nat. 162, 725–733

40 Keeley, J.E. and Zedler, P.H. (1998) Evolution of life histories in Pinus. In Ecology and Biogeography of Pinus (Richardson, D.M., ed.), pp. 219–249, Cambridge University Press

41 Schwilk, D.W. and Ackerly, D.D. (2001) Flammability and serotiny as strategies: correlated evolution in pines. Oikos 94, 326–336

42 Hoffman, W.A. (1998) Post-burn reproduction of woody plants in a neotropical savanna: the relative importance of sexual and vegetative reproduction. J. Appl. Ecol. 35, 422–433

43 Williams, R.J. et al. (1999) Fire regime, fire intensity and tree survival in a tropical savanna in northern Australia. Aust. J. Ecol. 24, 50–59

44 Hoffmann, W.A. and Franco, A.C. (2003) Comparative growth analysis of tropical savanna and forest trees using phylogenetically-indepen- dent contrasts. J. Ecol. 91, 475–484

45 Gignoux, J. et al. (1997) Alternative fire resistance strategies in savanna trees. Oecologia 110, 576–583

46 Archibald, S. and Bond, W.J. (2003) Growing tall vs. growing wide: tree architecture and allometry of Acacia karroo in forest, savanna, and arid environments. Oikos 102, 3–14

47 Glitzenstein, J.S. et al. (1995) Effects of fire regime and habitat on tree dynamics in north Florida longleaf pine savannas. Ecol. Monogr. 65, 441–476

48 Pyne, S.J. (2001) Fire: A Brief History, University of Washington Press 49 Bowman, D.M.J.S. (1998) The impact of Aboriginal landscape burning

on the Australian biota. New Phytol. 140, 385–410

www.sciencedirect.com

50 Scott, A.C. (2000) The Pre-Quaternary history of fire. Palaeogeogr. Palaeoclimatol. Palaeoecol. 164, 297–345

51 Cerling, T.E. et al. (1997) Global vegetation change through the Miocene/Pliocene boundary. Nature 389, 153–158

52 Pagani, M. et al. (1999) Late Miocene atmospheric CO2 concentrations and the expansion of C4 grasses. Science 285, 876–879

53 Pearson, P.N. and Palmer, M.R. (2000) Atmospheric carbon dioxide concentrations over the past 60 million years. Nature 406, 695–699

54 Keeley, J.E. and Rundel, P.H. (2003) Evolution of CAM and C4 carbon concentrating mechanisms in plants. Int. J. Plant Sci. 164 (Suppl. 3), S55–S77

55 Bond, W.J. et al. (2003) The importance of low atmospheric CO2 and fire in promoting the spread of grasslands and savannas. Glob. Change Biol. 9, 973–982

56 Sage, R.F. (2001) Environmental and evolutionary preconditions for the origin and diversification of the C4 photosynthetic syndrome. Plant Biol. 3, 202–213

57 Filho, A.T. and Ratter, J.A. (2002) Vegetation physiognomies and woody flora of the cerrado biome. In The Cerrados of Brazil (Oliveira, P.S. and Marquis, R.J., eds), pp. 91–120, Columbia University Press

58 Fensham, R.J. et al. (2003) Effects of fire and drought in a tropical eucalypt savanna colonized by rain forest. J. Biogeogr. 30, 1405–1414

59 Russell-Smith, J. et al. (2004) Rain forest invasion of eucalypt- dominated woodland savanna, Iron Range, north-eastern Australia: I. Successional processes. J. Biogeogr. 31, 1293–1303

60 Lavin, M. et al. (2004) Metacommunity process rather than con- tinental tectonic history better explains geographically structured phylogenies in legumes. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359, 1509–1522

61 Gill, A.M. (1975) Fire and the Australian flora: a review. Aust. For. 38, 4–25

62 Stephenson, N.L. et al. (1991) Restoring natural fire to the sequoia- mixed conifer forest: should intense fire play a role? Proc. Tall Timbers Fire Ecol. Conf 17, 321–337

63 Keeley, J.E. and Stephenson, N.L. (2000) Restoring natural fire regimes in the Sierra Nevada in an era of global change. In Wilderness Science in a Time of Change (Cole, D.N. et al., eds), pp. 255–265, USDA Forest Service

64 Lamont, B.B. et al. (1991) Canopy seed storage in woody plants. Bot. Rev. 57, 277–317

65 Keeley, J.E. and Fotheringham, C.J. (2000) Role of fire in regeneration from seed. In Seeds: The Ecology of Regeneration in Plant Commu- nities (Fenner, M., ed.), pp. 311–330, CABI

66 Brown, N.A.C. et al. (2003) Patterns in the seed germination response to smoke in plants from the Cape Floristic Region, South Africa. Sth. Afr. J. Bot 69, 514–525

67 Roche, S. et al. (1998) For everything a season: smoke-induced seed germination and seedling recruitment in a Western Australian Banksia woodland. Aust. J. Ecol. 23, 111–120

68 Keeley, J.E. (1993) Smoke-induced flowering in the fire-lily Cyrtanthus ventricosus. Sth. Afr. J. Bot. 59, 638

69 Keeley, J.E. (1998) Coupling demography, physiology and evolution in chaparral shrubs. In Landscape Disturbance and Biodiversity in Mediterranean-type Ecosystems (Rundel, P.W. et al., eds), pp. 257–264, Springer

70 Herrera, C.M. (1992) Historical effects and sorting processes as explanations for contemporary ecological patterns: character syn- dromes in Mediterranean woody plants. Am. Nat. 140, 421–446