A biogeographic model of fire regimes in Australia: current and future implications

ABSTRACT

Aim Patterns of fire regimes across Australia exhibit biogeographic variation in response to four processes. Variations in area burned and fire frequency result from differences in the rates of ‘switching’ of biomass growth, availability to burn, fire weather and ignition. Therefore differing processes limit fire (i.e. the lowest rate of switching) in differing ecosystems. Current and future trends in fire frequency were explored on this basis.

Location Case studies of forests (cool temperate to tropical) and woodlands (temperate to arid) were examined. These represent a broad range of Australian biomes and current fire regimes.

Methods Information on the four processes was applied to each case study and the potential minimum length of interfire interval was predicted and compared to current trends. The potential effects of global change on the processes were then assessed and future trends in fire regimes were predicted.

Results Variations in fire regimes are primarily related to fluctuations in available moisture and dominance by either woody or herbaceous plant cover. Fire in woodland communities (dry climates) is limited by growth of herbaceous fuels (biomass), whereas in forests (wet climates) limitation is by fuel moisture (availability to burn) and fire weather. Increasing dryness in woodland communities will decrease potential fire frequency, while the opposite applies in forests. In the tropics, both forms of limitation are weak due to the annual wet/dry climate. Future change may therefore be constrained.

Main conclusions Increasing dryness may diminish fire activity over much of Australia (dominance of dry woodlands), though increases may occur in temperate forests. Elevated CO₂ effects may confound or reinforce these trends. The prognosis for the future fire regime in Australia is therefore uncertain.

INTRODUCTION

Australia is the most fire prone of all continents, but fire regimes vary widely, ranging from high frequencies and relatively low intensities (e.g. 1–5-year recurrence intervals and < 10,000 kW m⁻¹; Gill et al., 2000; Williams et al., 2002) in tropical savannas of the north, to low frequencies and high intensities (e.g. > 100-year recurrence intervals, > 10,000 kW m⁻¹; Gill & Catling, 2002) in tall open forests of the cool, temperate south. These examples represent relatively moist ecosystems at continental extremes of latitude. By contrast, the bulk of the continent between these extremes is dry (arid or semi-arid) but still remarkably fire prone (Allan & Southgate, 2002; Russell-Smith et al., 2007).

Russell-Smith et al. (2007) divided the continent into six zones that reflect a gradient of rainfall seasonality, with recent, annual fire activity a positive function of rainfall seasonality. The dynamics of biomass growth (fuel) and availability to burn (dryness) govern these trends, resulting in an immense annual area of fire in the northern wet/dry tropics. Russell-Smith et al. (2007) also detected a strong influence of human activity on fire, with both positive (i.e. increased ignitions) and negative effects (i.e. intensive pastoral and agricultural land use) evident. Broadly similar conclusions emerge from a recent analysis of southern African regimes (Archibald et al., 2009).

There is general debate about the relative importance of these drivers (climate, fuel, ignitions), and the related influences of land management, in shaping fire regimes (Westerling et al., 2006; Marlon et al., 2008). Linkages between these drivers and underlying biophysical processes should explain variability in fire regimes within and between ecosystems (Archibald et al., 2009). In particular, such an understanding should indicate which processes are likely to have the dominant influence on fire in particular ecosystems.

The drivers of fire regimes are reasonably well known for some regions within Australia. For example, variations in litter fuel dynamics in temperate eucalypt forests affect the season, frequency and intensity of forest fires (Walker, 1981; Raison et al., 1983). Drought is associated with major fires in southern, forested regions (Ellis et al., 2004; Verdon et al., 2004), whereas above average rainfall leads to burning of very large areas in arid, central Australia (Allan & Southgate, 2002). Seasonal timing of ignition can cause changes to area burned, intensity and spatial heterogeneity of fire regimes in tropical savannas (Russell-Smith et al., 1997; Gill et al., 2000). Cross-regional integration of these disparate effects is, however, lacking. Insights via exploration of large-scale fire patterns in relation to key processes are also restricted by the limited temporal scope of data (e.g. Russell-Smith et al., 2007).

Australia is predicted to be prone to substantial shifts in climate, including increases in temperature and evaporation, and highly variable and uncertain changes in rainfall (CSIRO, 2007). Much of the focus on future fire under climatic change in Australia has been on forecasting changes to fire danger indices (e.g. Williams et al., 2001; Hennessy et al., 2005; Lucas et al., 2007; Pitman et al., 2007). Cary (2002) used landscape-scale simulation to show that future increases in fire danger may translate into significant increases in area burned in forested environments of south-eastern Australia. However, in addition to fire weather, climate change has the potential to alter ignition rates and the availability of fuel by, respectively, changing the incidence of lightning and plant productivity and decomposition of dead plant material. Other influences such as the effects of elevated CO₂ on plant growth and the effects of humans on ignitions add further complexity. Simulation using dynamic vegetation models (e.g. Thonicke et al., 2001; Scholze et al., 2006; Lenihan et al., 2008) offers potential for an exploration of future Australian fire regimes via an integration of these influences, but such models are currently hampered by coarse spatial resolution and use of inappropriate plant functional types for Australia.

This paper presents a conceptual model of the key drivers of fire regimes, and uses it to characterise the biogeography of fire across Australia as a basis for predicting the effects of global change on fire regimes. Given that the fundamentals of the model are universal, insights from its application to Australia may have wider applicability (e.g. exploration of global patterns of fire, Bowman et al., 2009). The aim is to:

• 1

  Illustrate that variations in fire regimes across ecosystems reflect a systematic variation in rates of a common set of key processes, rather than idiosyncratic local effects.

• 2

  Predict whether climate change will have equal effects on fire regimes throughout Australia, and if not, where the highest levels of change may be anticipated.

• 3

  Predict the consequences of other global change effects.

Will vectors of change (climate, CO₂ and humans) have congruent or antagonistic effects on fire regimes? If congruent effects are likely, only the magnitude of change will be uncertain. If antagonistic effects are likely, both the direction and magnitude of change will be uncertain.

THE BIOPHYSICAL BASIS FOR PREDICTION OF VARIATION IN FIRE REGIMES

Extremes of moisture availability govern the growth/accumulation of biomass/fuel (wetness) on the one hand and its ability to burn (dryness) on the other (Huston, 2003; Bond & Keeley, 2005; Pausas & Bradstock, 2007; Russell-Smith et al., 2007; Archibald et al., 2009). Fire is therefore related to moisture in a nonlinear way (Fig. 1). The effects of dryness are only important following periods of high moisture and resultant plant growth. Temporal fluctuations in moisture availability will therefore influence fire frequency. Characteristic patterns of mean and variance in available moisture are therefore likely to have a fundamental influence on variation of the fire regime (Gill et al., 2002; Russell-Smith et al., 2007). In particular, moisture may influence fire regimes through determination of fuel types, via selection of key plant species that are highly influential (‘fuel species’; Gill et al., 2002).

Figure 1

Potential relationships between moisture, productivity and fire frequency (after Bond & Keeley, 2005; Pausas & Bradstock, 2007). Corresponding trends in Australian tree cover and domains of strong limitation by woody (vertical hatching) and herbaceous fuels (horizontal hatching) or weak limitation (unhatched) are indicated.

A small number of large fires account for the bulk of area burned over time in many ecosystems (Boer et al., 2008; Cui & Perera, 2008). Significant area burned, via large fires, can only occur when available fuel is spread across large areas. Connectivity of available fuel is important due to the contagious nature of fire spread (Peters et al., 2004; Pueyo, 2007). Landscape- and regional-scale patterns of moisture fluctuation will influence the probability of large fires and thus area burned via effects on the production and drying of contiguous fuel. Temperature, wind speed and humidity directly influence the rate of spread of fires and the degree to which fires can bridge fuel discontinuities (Catchpole, 2002; Peters et al., 2004; Boer et al., 2008). As a consequence, area burned will also be influenced by these ‘fire weather’ variables, as recognized in models of fire behaviour and fire danger rating systems (Catchpole, 2002). Climate will also determine the frequency of lightning and the ensuing probability of ignition (e.g. Price & Rind, 1994).

Biomass production (B), its availability to burn (A), fire weather (S) and ignition (I) represent a hierarchy (Fig. 2) of conditional processes governing fire (Archibald et al., 2009). They can be regarded as four hypothetical ‘switches’ that must be simultaneously activated for fire to occur. Area burned will be a direct function of the spatial scale (i.e. the contiguous area) over which they jointly prevail. Should any switch be ‘off’, fire will not occur. Fire can therefore be constrained in differing ways, through the effect of turning different switches ‘off’. Some switches are ‘off’ more often than others in different ecosystems, resulting in fundamentally different fire regimes. Identification of the ‘limiting’ switch (i.e. the switch activated least often) and its relationship to differing ecosystem characteristics is the key to understanding these variations.

Figure 2

Influences of biogeographic factors (climate, soils, habitats, plant functional types) on fire regimes via four ‘switches’ (biomass growth, B; availability for burning, A; ambient fire weather, S; ignitions, I). Potential effects of changing climate, human activity and atmospheric CO₂ are indicated by dashed lines.

The rate of switching in each case will be a function of interactions between climate (mean and variability) and soils (fertility, texture, depth). This will shape habitats and the selection of available plant functional types (Fig. 2). Therefore, both the limiting switch and its inherent switching rate will vary among Australian ecosystems as a result of patterns of biogeographic variation in these factors (Hutchinson et al., 2005).

Biomass (B)

Surface or near-surface fuels (primarily dead leaf material) are critical for the spread of fire in most Australian vegetation types (Catchpole, 2002). Fuel quantity is therefore affected by foliar cover (Walker, 1981) and leaf attributes. A primary distinction can be made between herbaceous and woody plants because of differences in the characteristics, rate of supply and arrangement of material derived from these life-forms.

Tree cover will be a primary determinant of the type of surface fuel. Tree cover is positively correlated with patterns of available moisture that arise from climate and interactions with soil depth/texture (Beadle, 1981; Williams et al., 1996; Specht & Specht, 1999; Fensham et al., 2005). Fire regimes may also reduce tree cover below edaphic/climatic potential (e.g. Liedloff & Cook, 2007).

Eucalypt open forests (i.e. canopy projective foliage cover > 30%; Gill, 1994; Specht & Specht, 1999) occupy areas with relatively high moisture availability (Fig. 1). Litter from trees forms the principal surface fuel (Walker, 1981; Raison et al., 1983), augmented by variable contributions from shrubs and herbs. Fuel quantity and extent is a therefore a function of the turnover of perennial foliage with major variations in cover and load of surface fuel driven primarily by the degree of tree and shrub cover (a function of moisture availability) and time since the last fire (Walker, 1981; Specht & Specht, 1999; Berry & Roderick, 2002).

Woodlands (10–30% tree cover; Specht & Specht, 1999) are common in drier environments (Fig. 1), and cover vast areas of the interior (Beadle, 1981; Groves, 1994). There is interplay between woody litter (trees) and herbaceous fuels in woodlands. Litter fuel from trees is typically discontinuous because of relatively low cover, so that overall fuel availability is strongly influenced by fluctuations in herbaceous biomass. Connectivity of surface fuels is provided by herbaceous material (e.g. grasses and herbs; e.g. Hodgkinson, 2002; Prober et al., 2007, 2008).

Availability to burn (A)

In forested systems, drought alters the moisture status of compacted litter beds derived from woody plants and thus availability to burn in horizontal (e.g. differing aspects) and vertical (e.g. within litter bed) planes. This changes the propensity for large fires to develop by altering the connectivity of fuels across landscapes. Droughts may also temporarily increase litter fall from trees and shrubs in both temperate and tropical environments (e.g. Pook et al., 1997; Cook, 2003). By contrast, in grass and herb fuels, dry periods initiate rapid leaf death (curing) of perennial and ephemeral species (e.g. Catchpole, 2002; Spessa et al., 2005).

Ecosystems in the northern, monsoonal tropics experience prolonged annual wet and dry seasons, whereas those in southern, temperate regions may experience severe drought on a multidecadal cycle, influenced by El Niño oscillations (e.g. Cullen & Grierson, 2009). Pronounced water deficits are a regular seasonal feature of arid and semi-arid environments (Hutchinson et al., 2005).

Fire spread (S)

Severe ambient weather conditions (high temperatures, wind speeds and low humidity) result in the rapid spread of fires, irrespective of fuel type (Catchpole, 2002). Ease of ignition and flame transfer are increased by high temperatures and low humidity. Wind directly affects flame length and depth, and propagation of fire via embers. Area burned will therefore be directly influenced by these variables (Gill et al., 2002; Boer et al., 2008). Fire weather is a function of latitude and rainfall, with the annual incidence of severe fire danger conditions declining with increasing rainfall and latitude (Gill & Moore, 1990; Williams et al., 2001; McCaw & Hanstrum, 2003; Lucas et al., 2007).

Ignition (I)

Definitive data on ignitions from lightning versus anthropogenic sources are elusive, due to the remoteness of many regions and difficulty of attribution to sources (e.g. McCaw & Hanstrum, 2003). Many official records indicate large proportions of ignitions of unknown origin. Lightning ignition rates (I) generally vary from being common annually in the tropics to rare in the cool temperate zone (Kuleshov et al., 2006). Anthropogenic ignitions are a function of population density and land use, with rates high near urban development (Gill & Williams, 1996; Bradstock & Gill, 2001) and low in dry pastoral/agricultural landscapes (Noble & Grice, 2002; Russell-Smith et al., 2007).

CURRENT FIRE REGIMES IN AUSTRALIA: PATTERNS OF LIMITING ‘SWITCHES’

Potential influences of fuel, fire weather and ignitions on fire regimes are coupled (Fig. 2). Grass/herbaceous fuels will be important where moisture deficits (low to moderate tree cover) and high levels of fire danger are severe and regular (Fig. 1). Woody/litter fuels (moderate to high tree cover) occur where such effects are less severe and regular (Fig. 1) or where trees can access other sources of water (e.g. tropical savannas; Bowman & Prior, 2005). Growth and curing of grasses and herbs (perennial and ephemeral) respond rapidly to fluctuations in available moisture. C₄ grasses may be highly influential in this regard, due to their productivity under pronounced conditions of seasonal variability (Murphy & Bowman, 2007; Osborne, 2008). As a result, fire regimes in woodlands dominated by grass/herbaceous fuels will be governed by fluctuations in biomass growth. Moisture variation will therefore strongly influence fire regimes.

By contrast, in forested systems where woody litter fuels are prominent, regular patterns of fuel dynamics occur that are governed by rapid accumulation after fire (Walker, 1981; Raison et al., 1983). The amount of fuel is therefore less tightly coupled to short-term moisture variations. Sufficient litter is usually present for propagation of fire at most times except for very soon after fire (i.e. fuel is non-limiting). Fire will therefore be limited primarily by fluctuations in availability to burn and propagation potential, governed, respectively, by drought and ambient weather at the time of ignition.

Ecosystems can be arranged (Figs 1 & 3) along axes representing either biomass/fuel (switch B) limitation, or a combination of fuel availability (switch A) plus fire spread potential (a function of ambient weather, switch S). This latter combination is consistent with fire danger rating systems [e.g. the McArthur Forest Fire Danger Index (FFDI); Catchpole, 2002]. Thus high values of fire danger indices, conducive to a rapid rate of spread and large area burned, are improbable without both severe drought and ambient weather (e.g. Bradstock et al., 2009).

Figure 3

Potential minimum length of the interfire interval (IFI) as determined by: (a) biomass growth (B); (b) fire weather (i.e. availability of dry biomass plus influence of wind, temperature and humidity on rate of spread of fire, A + S) in relation to trends in available moisture, as indicated by direction of arrows (see Figure 1). Different forest and woodland types are indicated (tropical forest, TF; temperate dry sclerophyll forest, DSF; cool temperate wet sclerophyll forest, WSF; temperate grassy woodland, TGW; arid woodland, AW). Strong limitation by B and A + S is indicated by vertical and horizontal hatching, respectively, while the unhatched area indicates weak limitation (see Fig. 1).

This approach contrasts the sensitivity of fire regimes to relative effects of biomass growth (B) on the one hand and fire danger (A + S) on the other (Fig. 3). These represent, respectively, the indirect (‘bottom-up’) or direct (‘top-down’) influence of climate (Fig. 2). Predictions of potential fire regimes (i.e. the limits of average interfire interval) can be made, with realization of potential being dependent on the rate of ignition (i.e. the fourth ‘switch’: Fig. 2).

Fire regime potential was explored for five case studies (Table 1) covering a gradient of tree cover spanning high to low moisture (Figs 1 & 3) and thus degree of limitation by either woody litter or herbaceous fuels. The case studies are: tropical open forest (TF); arid woodlands (AW); temperate grassy woodlands (TGW); dry sclerophyll shrubby forests (DSF); wet sclerophyll forests (WSF). These occur across gradients of latitude, climate (monsoonal tropics to cool temperate) and available moisture. Potential fire regimes in forested communities (tree cover > 30%) tend to be limited primarily by occurrence of severe fire weather. The degree of limitation increases with increasing moisture (Fig. 1 & 3). By contrast, potential fire regimes in woodland communities tend to be limited primarily by biomass growth. Limitation increases with decreasing moisture (Figs 1 & 3). Communities with ‘intermediate’ tree cover (i.e. c. 30%) may be particularly susceptible to factors that alter the balance between woody overstorey and grass/herb understorey (e.g. a change from limitation by B to A or S or vice versa).

Table 1. Attributes of case studies that affect fire regimes in differing Australian ecosystems.

	Tropical open forest (TS)	Temperate grassy woodlands (TGW)	Arid woodlands (AW)	Temperate dry sclerophyll forests (DSF)	Cool temperate wet sclerophyll forests (WSF)
Indicative bioregion and agro-climatic zone (Hutchinson et al., 2005)	Arnhem Coast; Central Arnhem. Hot, seasonally wet/dry (I1)	New South Wales–south-western slopes and Brigalow Belt south. Warm, seasonally wet/dry. Long hot summers and mild winters with significant moisture limits on growth including mediterranean climates (summer dry season, e.g. Adelaide E1) and mid-latitude eastern continental climates with wetter summers and drier winters (e.g. central New South Wales E3)	MacDonnell Ranges Desert – high water limitation (G)	Sydney basin – warm, wet (F3). Long hot summers and mild winters. Jarrah Forest, mediterranean climate (summer dry season, e.g. Perth E1)	Tasmanian west, cold (B2). Very cold winters with short warm summers
Present land uses	Pastoralism, conservation reserves, indigenous	Winter cereals and summer crops, grazing	Pastoralism, conservation, indigenous	Urban, conservation reserves, water catchments, forestry (south-west), mining (south-west)	Forestry, hydroelectricity, conservation reserves
Mean annual rainfall (mm) (Weather station)	1539 (Darwin Post Office)	585 (Dubbo Darling St), 529 (Adelaide West Terrace)	279 (Alice Springs Post Office)	1214 (Sydney Observatory Hill), 868 (Perth Regional)	1212 (Maydena)
Rainfall seasonality. Months of highest and lowest mean rainfall (mm)	Predominantly summer (393 January, 1 July).	Even seasonal distribution (63 January, 43 September, Dubbo). Predominantly winter (72 June, 20 January, Adelaide)	Summer peak (43 January, 9 September)	Late summer–early winter peak (131 June, 69 September, Sydney). Predominantly winter (182 June, 9 January, Perth)	Winter–early summer peak (128 August, 60 February)
Canopy tree cover/composition	> 30%, Eucalyptus, Terminalia	10–30% Eucalyptus, Casuarina, Callitris	< 10–30% Acacia	30–70% Eucalyptus, Angophora, Allocasuarina	Eucalyptus, Nothofagus, Atherosperma
Understorey structure/composition	Annual (e.g. Saga) and perennial grasses	Perennial (e.g. Themeda) and annual grasses, herbs and shrubs	Hummock grasses (e.g. Triodia), tussock grasses (Astrebla), herbs	Diverse shrubs (e.g. Proteaceae, Fabaceae, Myrtaceae), sedges (e.g. Cyperaceae), graminoids	Shrubs (sclerophyll and mesic)
Average number of Very High to Extreme fire weather days per annum (weather station)	> 30 (Jabiru) (Williams et al., 2002)	23 (Dubbo), 18.3 (Adelaide) (Lucas et al., 2007)	> 60 (Alice Springs) (Williams et al., 2001)	7.6 (Sydney) (Lucas et al., 2007) > 10 (Perth) (Gill & Moore, 1990)	2 (Hobart) (Lucas et al., 2007)
Fire season	Winter–spring	Summer–autumn	Spring–summer	Spring–early summer (south-east), summer–autumn (south-west)	Late summer–autumn
Present ignition sources	Predominantly human	Predominantly lightning	Human and lightning	Predominantly human, variable lightning	Predominantly lightning
Present range of interfire intervals (years)	2–5 (Russell-Smith et al., 1997, 2007;Williams et al., 2002)	> 10, often multidecadal in length (Hobbs, 2002;Lunt et al. 2006)	10–80 (Allan & Southgate, 2002)	5–15 near urban areas, 10–25 in remote areas (south-east) (Gill & Moore, 1997;Bradstock & Gill, 2001)	> 20 (Gill & Catling, 2002)

The case studies also represent the influence of particular plant species/functional types and land-use context (e.g. pastoral to urban hinterlands). Examples are discussed using relevant data from a typical region in each instance (Table 1).

Potential fire regimes

TF represents grassy, open forests (i.e. tree foliar cover > 30%; Table 1) that occur in high-rainfall, coastal regions of northern Australia. These are the mesic variant of tropical savannah forests and woodlands (Williams et al., 2002) where moisture availability is sufficient to support relatively high tree cover, consisting of a mix of evergreen, deciduous and semi-deciduous species. TF represents an extreme where the influences of both available biomass (B) and fire danger (A + S) converge so that neither influence is strongly limiting (Fig. 3). Frequency of occurrence of conditions conducive to major fires is close to annual, due to the effects of both regular annual prolonged wet and dry seasons on grass/herb fuels under a monsoonal climate. The wet season provides a concentrated moist period for rapid growth of grasses, while the dry season provides prolonged annual drought (A) and lengthy sequences of days with fire weather (S) suitable for the rapid spread of large, relatively intense fires (Table 1; Williams et al., 2002). Annual grasses such as Saga spp. (= Sorghum; Williams et al., 2002) are suited to these conditions and therefore have the potential to control fire regimes and effectively usurp the influence of litter accumulation from trees.

In contrast to TF, AW and TGW represent differing degrees of temporal limitation in herbaceous biomass growth (Table 1, Fig. 3a). As in TF, severe fire weather in AW and TGW is non-limiting or weakly limiting (Table 1, Fig. 3b). Variants of AW cover a vast expanse of the continent. The ground layer consists of perennial hummock grasses (e.g. Triodia and Plectrachne spp.) or tussock grasses (e.g. Aristida and Eragrostis spp.) and ephemeral herbs fuels are common (Allan & Southgate, 2002; Hodgkinson, 2002). The tree layer, often dominated by Acacia spp. (Hodgkinson, 2002), is sparse or clumped. Herbaceous cover (perennial and ephemeral) can be sparse in dry times, but also grows vigorously in response to rain (Griffin et al., 1983; Allan & Southgate, 2002; Southgate & Carthew, 2007). Time since fire, in interaction with rainfall, also affects herbage cover, though these effects can be negated after high rainfall (Allan & Southgate, 2002; Southgate & Carthew, 2007). High-rainfall events therefore irregularly provide fuel connectivity by stimulating herbage which is otherwise absent or slow to develop (a ‘fuel irruption’; Southgate & Carthew, 2007). Such events occur at decadal to multidecadal recurrence intervals with consequent effects on fire regime potential (Allan & Southgate, 2002; Hodgkinson, 2002).

TGW is intermediate between TF and AW, reflecting moisture status in deep, well-drained soils on the margins of inland, semi-arid environments (Beadle, 1981; Table 1). Perennial and ephemeral grasses/herbs are prominent, and Eucalyptus, Callitris and Allocasuarina spp. are common as trees. Herbage growth may fluctuate with level of stocking of native and domestic grazing animals and occasional prolonged drought (Noble & Grice, 2002; Prober et al., 2007). Recovery after fire is relatively rapid (e.g. 1–5 years), depending on fluctuations in rainfall (Prober et al., 2007, 2008). The potential for fire is therefore high (e.g. 1–5-year frequency; Fig. 3), because temporal biomass limitation is weak (i.e. biomass fluctuations are rapid and regular). The balance between C₃ (e.g. Poa) and C₄ (e.g. Themeda) grasses in these systems may be important (Prober et al., 2007), with biomass and spatial cover possibly declining if C₄ abundance declines.

In comparison to TF (biomass non-limiting), B in the temperate forest examples (DSF, WSF) is weakly limiting due to the rate of post-fire accumulation of litter. There is sufficient surface litter for propagation of fire at 1–3 years in severe weather (Raison et al., 1983; Morrison et al., 1996; Huston, 2003). Quasi-equilibrium surface fuel loads are reached after about 10 years post-fire (Walker, 1981; Raison et al., 1983). Fire potential therefore tends to be limited by fire weather (Fig. 3b).

In DSF, days of extreme fire danger (i.e. FFDI > 49) are relatively rare (e.g. an average of 1 day per year) in mainland south-eastern and south-western regions (Gill & Moore, 1990; McCaw & Hanstrum, 2003; Lucas et al., 2007). In south-eastern DSF, characterized by non-seasonal rainfall, regular El Niño–Southern Oscillation (ENSO)-related droughts (e.g. 5-year frequency) may provide severe fire danger conditions suitable for major fires (Verdon et al., 2004; Nicholls & Lucas, 2007; Table 1). South-western DSF occurs in a mediterranean climate, with a long annual summer dry period. Severe fire danger episodes are regularly generated (e.g. 5-year frequency) by strong cyclonic activity to the north (McCaw & Hanstrum, 2003; Table 1). Extreme fire danger is rarer in southern Tasmania (Lucas et al., 2007; Table 1), but a positive association between summer dryness and area burned has been shown in western WSF regions (Nicholls & Lucas, 2007). The potential for major fires is therefore regular in mainland DSF (5–10-year intervals, Fig. 3b)and less regular (> 10 years –Fig. 3b) in western Tasmania (WSF).

Realization of fire regime potential

Ignition rates in TF are currently high, due mainly to anthropogenic sources (Russell-Smith et al., 2007), resulting in high-frequency fire regimes (Table 1). Fire regimes are therefore close to their potential limit (Fig. 3). Variations in summer rain may affect the quantity of grass fuels and the length of the dry season, with high rainfall leading to greater area burned (Harris et al., 2008). In AW, ignition rates (anthropogenic and lightning) also appear to be sufficient to saturate opportunities where herbaceous fuel, after curing, is non-limiting, leading to episodes of major fire over vast areas (Heydon et al., 2000; Allan & Southgate, 2002; Russell-Smith et al., 2007).

By contrast, in TGW fire is often constrained through low anthropogenic ignitions and a higher level of availability of fire suppression due to higher population density (e.g. agricultural holdings and small towns), leading to a long-term absence of burning in natural fragments (Table 1). In larger non-fragmented tracts of temperate woodlands, multidecadal fire regimes predominate under the influence of lightning (Table 1). Current fire recurrence intervals are therefore considerably longer than their potential minimum (cf. Fig. 3). High domestic stocking in many regions has degraded perennial grass populations (e.g. Noble, 1997; Hobbs, 2002). Replacement by ephemeral grasses, herbage and woody plants has resulted in diminished fuel continuity and quantity, enhancing the sensitivity of fire potential to rainfall variations and contributing to a decrease in ignitions and area burned (Noble & Grice, 2002).

In DSF ignitions are highly varied (Table 1). In some areas of eastern New South Wales (NSW), for example, anthropogenic ignition rates (sometimes complemented by lightning) are sufficient to exploit most opportunities for major fires, resulting in 5–10-year fire recurrence intervals at landscape scales (e.g. Bradstock & Kenny, 2003), whereas in more remote regions with low rates of anthropogenic ignition, average interfire intervals may be > 20 years under the predominant influence of lightning (Bradstock, 2008). Similar trends apply in other regions such as south-western Australia (Table 1). Fulfilment of fire regime potential in DSF is therefore highly variable (Table 1, Fig. 3).

In WSF, lightning ignition rates and anthropogenic ignition rates are relatively low (e.g. King et al., 2006; Table 1). These may be insufficient to saturate opportunities for major fires created by irregular summer drought, resulting in fire regimes of multidecadal to century-scale frequency (Table 1, Fig. 3). Limitation of fire regimes by ignition (I) may be generally more common with increasing moisture in forested communities. The degree of limitation will also depend on rates of human ignition as a function of context (e.g. proximity to towns/cities; Bradstock & Gill, 2001).

Fire regimes resulting from any particular combination of B, A, S and I will affect the relative proportions of plant functional types in landscapes, given inherent life-history and regeneration attributes. In turn, this may further influence the size of fires if fuel characteristics (i.e. flammability) differ among available functional types (fuel species). Bowman (2000) postulated that alternative states of flammability (potential dominance by differing fuel species) exist in some Australian ecosystems. Present patterns of fire regimes could reflect such ‘flammable feedback’, in interaction with the influences outlined above.

For this to occur, alternative fuel species must have the potential to occupy large expanses of habitat in order to influence fuel connectivity and fire regimes. Such potential is poorly understood; hence the degree to which ‘flammable feedback’ is currently driving selection between alternative fuel species is unknown. The availability of species in differing ecosystems not only reflects current conditions (soils and climate) but also phylogenetic legacies and the varying Quaternary influences of people (ignitions and species introductions). Nonetheless, landscape flammability may be altered by the introduction of new functional types in interaction with future changes to habitat (Fig. 2).

GLOBAL CHANGE EFFECTS

Influences of global change on fire regimes are potentially varied (Figs 2, 4 & 5), due to the complexity of underlying processes. The influence of climatic change can be distinguished from other influences such as elevated CO₂ and land use (Figs 2, 4 & 5). Climate change has the potential to affect B, A, S and I (Fig. 2), with changes in available moisture (via rainfall, temperature, evaporation) directly affecting B and A, while changes to temperature and wind will affect S. Elevated CO₂ may mostly affect B and possibly A, through alteration of the growth and competitive performance of plant functional types. Human activity can affect B and A via the introduction of new plant functional types. Activities such as grazing and clearing may directly affect B, while there is ever-present potential for human activity to affect ignition (I).

Figure 4

Effects of global change on the interfire interval (IFI) in woodlands (temperate grassy woodland, TGW; arid woodland, AW) via changes to (a) biomass growth, B, and (b) fire weather, A+S. Arrows indicate the direction of effects on potential and current fire regimes. Symbols indicate the effects of climate change (fire danger, f; grass/herb growth, g), elevated CO₂ (c) and exotic species (e).

Figure 5

Effects of global change on the interfire interval (IFI) in forests (tropical forest, TF; temperate dry sclerophyll forest, DSF; cool temperate wet sclerophyll forest, WSF) via changes to (a) biomass growth, B, and (b) fire weather, A+S. Arrows indicate the direction of effects on potential and current fire regimes. Symbols indicate the effects of climate change (fire danger, f; grass/herb growth, g; litter fuel accumulation, li), elevated CO₂ (c) and exotic species (e).

Responses of differing ecosystems can therefore be expected to vary according to the balance of these influences and the nature of fire regime limitation in each instance. Woodlands should inherently be sensitive to factors which affect B whereas forests should be primarily sensitive to factors influencing A and S.

Climate change effects

In woodlands, changes in available moisture have the potential to affect B. Rainfall at stations relevant to the AW and TGW case studies is predicted to either increase or decrease under a range of 2070 scenarios (CSIRO, 2007). The median predictions for the case-study stations (Table 2) are for a decrease in all cases. The consequent decline in available moisture will be exacerbated by predicted increases in evaporation in all cases (Table 2). Increases in summer rainfall and various levels of decrease in cooler season rainfall are generally predicted. Consequent changes to fuel (B; Table 2, Fig. 4b) may include decreased grass/herbage in AW and TGW. Increases in summer rain could favour the growth of the important C₄ grass component in these cases, particularly in non-tropical environments (Murphy & Bowman, 2007). While fire danger may tend to increase in these regions due to climate change (Table 2), such effects may be less significant due to inherent fuel limitation in woodlands (i.e. fire weather non-limiting). Thus there is potential for area burned to either increase or decrease in woodlands due to effects on B but the balance of these effects suggests a decline (Fig. 4).

Table 2. Global change scenarios in case studies in differing Australian ecosystems.

Global change attribute	Tropical open forest (TF)	Arid woodlands (AW)	Temperate grassy woodlands (TGW)	Temperate dry sclerophyll forests (DSF)	Cool temperate wet sclerophyll forests (WSF)
Rainfall (% change)	−1	−9 to −17	−4 to −7 −7 to −13*	−4 to −8 −11 to −19†	−3 to −6
Temperature (°C change)	+1.7 to +3.2	+1.9 to +3.7	+1.7 to +3.3 +1.5 to +2.8*	+1.6 to +3.0 +1.4 to +2.7†	+1.1 to +2.1
Evaporation (% change)	+5 to +10	+4 to +7%	+ 4 to +9 +3 to +6*	+5 to +9 +4 to +7†	+5 to +10
Fire danger (very high to extreme days year^–1)	Increase‡	Increase‡	+4.4 to +20.8§ +1.6 to +11.5*§	+ 0.4 to 6.6 (Sydney)§ Increase likely‡ (Perth)	0 to +0.2 (Hobart)§
Sensitivity (direction of change in mass) of main fuel types to: (1) climate change and (2) elevated CO2	Annual grasses (1) decrease (2) decrease	Perennial grasses and annual herbs/grasses (1) decrease (2) decrease	Perennial grasses and annual herbs/grasses (1) decrease (2) decrease Woody plant litter (1) decrease (2) increase	Woody plant litter and shrub crowns (1) decrease (2) increase	Woody plant litter (1) no change (2) decrease
New plant functional types	Gamba grass	Buffel grass	Tree plantations	Exotic grasses – mediterranean areas
Trend in ignitions		+Anthropogenic	−Anthropogenic	+Anthropogenic	+Anthropogenic

• * Climatic predictions are 2070 90p scenarios from CSIRO (2007) for Darwin (TF), Alice Springs (AW), Dubbo (TGW), Adelaide (TGW mediterranean), Sydney (DSF),
• † Perth† (DSF mediterranean) and Hobart (WSF).
• ‡ Fire danger scenarios are based on Williams et al. (2001) and
• § Lucas et al. (2007) § for 2050.

By contrast, forests are expected to be sensitive to changes in fire danger and frequency of severe drought under climate change (A + S; Fig. 5b). For example, the recent succession of major fire seasons in southern Australia (Ellis et al., 2004; Bradstock, 2008) is coincident with a prolonged period of severe drought and elevated fire danger (Lucas et al., 2007). General increases in future drought and fire danger indices are predicted in most forested regions (Table 2) but the magnitude of predicted change is variable. Hennessy et al. (2005) and Lucas et al. (2007) predicted substantial increases in drought frequency and days of very high to extreme fire danger across broad regions of the south-eastern mainland but very little change for southern Tasmania (Hobart; Table 2). These studies predict an increase in severity of drought and fire danger, rather than a change in drought frequency, as they utilize the current pattern of daily weather as their basis. Changes to ENSO severity are predicted (e.g. CSIRO, 2007) but details of potential changes in ENSO frequency are lacking. These could be crucial to changes in fire activity in many forested regions. Current estimates of change to fire danger may therefore underestimate the potential for climate change to alter fire regimes.

Declining moisture may result in reduced rates of litter fuel accumulation in DSF and WSF (Fig. 5a, Table 2) as demonstrated by comparative trends in litter fuel accumulation under differing present-day moisture scenarios (Walker, 1981; Raison et al., 1983; Huston, 2003). Such effects may tend to reduce area burned. Notwithstanding this possibility, the prognosis for forests is for increasing area burned due to effects of elevated fire danger. The TF case study may be an exception, due to the weak limiting effects of both fuel and fire weather. The potential for climate change to affect either of these influences is more constrained; however, a reduction in grass growth due to drying (Table 2, Fig. 5a) has the potential to decrease area burned (Harris et al., 2008).

There are few detailed predictions of changes to ignition rates from lightning under climate change (e.g. Hennessy et al., 2005). Earlier work suggests that there may be a trend for increasing lightning ignitions (Price & Rind, 1994; Goldammer & Price, 1998) due greater atmospheric instability under global warming.

Other global change effects

Changes to fire regimes stemming from new plant functional types, altered plant growth, land use and ignition rates range from those that are well known (e.g. the effects of exotic grasses) to speculative and highly uncertain effects (e.g. the consequences of elevated CO₂).

In TF and AW there is evidence that new plant species (exotic grasses) may affect fire regimes. In TF, the introduction of gamba grass (Andropogon gayanus), an African exotic, has resulted in a dramatic elevation of fuel loads and consequent fire intensity (Rossiter et al., 2003). The rate of expansion of gamba grass is rapid in places (e.g. the Darwin hinterland), though the bulk of TF remains unaffected. The potential for further spread and consequent changes to fire regimes across northern Australia is high. This situation is paralleled in arid areas of northern and central Australia (AW) by buffel grass (Pennisetum ciliare; Clarke et al., 2005). This species has high drought tolerance, thus providing levels of biomass and spatial connectivity that may exceed that contributed by native grasses and herbage (Clarke et al., 2005).

Reafforestation of agricultural land in the south (TGW and some DSF regions) for timber, paper or fuel, carbon credits, salinity control and habitat restoration (e.g. exotic conifers –Pinus spp.; native and exotic Eucalyptus spp.) may enhance landscape surface fuel connectivity, particularly in heavily fragmented landscapes, where cleared land may otherwise be seasonally fuel free due to cropping and/or grazing. Further fragmentation of natural vegetation through urbanization and agriculture may counteract such trends.

Elevated atmospheric concentrations of carbon dioxide may differentially affect plant growth, via enhanced effects on woody species relative to herbs – particularly C₄ grasses (Wang, 2007; Osborne, 2008). Such effects may be dependent on water and nutrient availability, which may be particularly important in Australia, given the predominant low-fertility soils and aridity (Hughes, 2003; Steffen & Canadell, 2005).

Elevated CO₂ could affect woodlands and TF by altering the balance between grasses, herbs, shrubs and trees (e.g. Banfai & Bowman, 2005; Berry & Roderick, 2006). Major changes in grass/tree ratios in southern African savannas, with resultant effects on fire regimes (i.e. area burned positively related to grass cover), are postulated to be driven by fluxes in atmospheric CO₂ concentration (Bond et al., 2003). Increases in tree and shrub density at the expense of grass/herb cover are implicated in historical declines in area burned in TGW (e.g. Noble, 1997; Noble & Grice, 2002). Elevated CO₂ may increase accumulation of forest litter fuels via an increase in growth and accession of litter. Increases in the C : N ratio of leaves (reduced palatability for vertebrate and invertebrate consumers) could slow decomposition (e.g. Gleadow et al., 1998; Wang, 2007).

Rates of fire incidence are positively related to population density and proximity to the urban interface (Bradstock & Gill, 2001; Keeley & Fotheringham, 2001). Continuing expansion of existing major urban centres and population drift from cities to rural areas may alter ignition patterns in south-eastern and south-western forested regions in particular (DSF, WSF).

Future trends in fire regimes

Factors that affect the performance of herbaceous/grass functional types (B) will have major effects on Australian fire regimes, because systems dominated by herbaceous fuels (e.g. woodlands) cover the bulk of the continent. Given that a range of effects on herbaceous growth and spatial continuity (B) are plausible, the potential for major changes to Australian fire regimes is high. Nonetheless, synergistic effects of all differing global change factors on fire regimes appear unlikely in many ecosystems. Instead, global change factors are more likely to have antagonistic effects (Figs 4 & 5). Both the direction and magnitude of changes to Australian fire regimes are therefore uncertain, given the difficulty in quantifying the magnitude of the vectors.

The outcome for future fire regimes in woodlands (TGW and AW) will depend largely on the interplay between moisture and elevated CO₂ effects on B. If a decline in moisture occurs as predicted under the median 2070 scenarios (Table 2) then a decline in grass/herb biomass will result. This may complement any negative effect of elevated CO₂ on area burned and fire frequency (Fig. 4). The spread of exotic functional types such as buffel grass could play a contrary role over significant areas of AW due to their effect on landscape connectivity (Fig. 4a).

Enhancement of woody plant growth in woodlands may also be context specific. In arid and semi-arid regions, woody cover is dominated by Acacia, Casuarina and Callitris species, which form litter beds of low flammability (Bradstock & Gill, 1993; Bradstock & Cohn, 2002). In higher-rainfall TGW areas, eucalypts with litter beds of high flammability are more prominent. Negative effects of elevated CO₂ on fire, through stimulation of woody growth, are therefore likely to be more pronounced in semi-arid and arid woodlands. Resolution of the outcome of opposing effects of increasing summer rainfall (favourable for C₄ grasses) and elevated CO₂ (favourable for woody plants) may be particularly significant in TGW. In moist TF regions (grassy woodland/open forest), encroachment of rainforest from fire-protected enclaves, due to CO₂ effects, may also diminish the area burned.

In contrast to woodlands, the balance between fire weather/danger effects (A + S) and fuel effects (B) will determine the nature of future fire regimes in forests (Fig. 5). Because of the inherent limitation of A + S in forests, predicted increases in fire danger make an increase in average area burned plausible. Forecast changes in fire danger are conservative and do not account for crucial changes in major drought frequency that may determine the magnitude of any forcing of fire regimes by climate change. Potential increases in ignitions (I) from urban expansion may be synergistic with fire weather effects in many forested regions.

CONCLUSIONS

Climate change and elevated CO₂ may lead to a decrease in area burned and fire frequency in Australian systems dominated by herbaceous fuels, due to effects on the limiting ‘switch’ (B). Factors such as spread of exotic grasses have the potential to cause contrary effects in these ecosystems. In forests dominated by woody, litter fuels, alterations to B through CO₂ or other effects may be less critical than changes to fire danger (A + S). Climate change therefore has a strong potential to increase area burned in temperate forested regions through an increase in the severity of fire danger, including the contribution of drought.

These examples do not represent fire regime trends in all Australian ecosystems, or all the possible effects of global change. Further extension and refinement of this approach may be appropriate, though the adequacy of understanding of fundamental processes (e.g. fuel dynamics, fire behaviour, ignition patterns and sources) may impose constraints. For example, shrub-dominated communities (e.g. various shrublands, heaths and shrubby woodlands) are of great significance in terms of biodiversity and some (e.g. shrubby woodlands in the south-west) cover an extensive area. Such communities occur across a range of environments comparable to TGW through to WSF.

Fire regimes in some cases may be governed not only by in situ processes but also by characteristics of adjacent communities (e.g. heath patches within tropical savannah or temperate DSF). Differing approaches may be needed to predict future fire regimes at this scale. Ultimately, given the complexity of responses and drivers of fire and potential feedback processes, appropriate numerical simulation models may be required to fully explore the problem.

This overview has also focused on factors governing area burnt by large fires and resultant effects on fire frequency, rather than effects on fire intensity and season and the detail of spatial heterogeneity of regimes. Terrain diversity will influence fire size and spatial patterns of intensity, for example, because of effects of moisture on fuel availability (A) and rate of spread (S). The approach could be extended to explore these components of fire regimes, particularly as information on spatial variation in fire severity accumulates. While these deficiencies are considerable, there is sufficient evidence to indicate that variation in the fire regime among Australian ecosystems can be explained by variations in the rates of four basic drivers. This provides a basis to guide the development of quantitative fire regime models in the future, as well as systematic data collection to document the nature of the processes that shape these drivers.

The concept of spatial connectivity of the drivers of fire underpins the ‘four switch’ concept presented here. Further exploration of the way that landscapes develop ‘flammable connections’ through interactions between fuel, weather and terrain (Falk et al., 2007) will provide an incisive basis for predicting and managing future fire regimes.

ACKNOWLEDGEMENTS

This work was completed with assistance from the Commonwealth Department of Climate Change as part of a national assessment of the consequences of climate change for fire and biodiversity. Input from colleagues engaged on this project is gratefully acknowledged, with particular thanks to Dick Williams. Penny Watson, Alan Andersen and two referees provided invaluable comments on the manuscript.

BIOSKETCH

Ross Bradstock is Director of the Centre for Environmental Risk Management of Bushfires at the University of Wollongong. He has research interests in fire ecology, conservation biology, fire regimes and consequences of global change. He is an adviser on fire management policies and practices to land and fire managers.

Editor: Brad Murray