Fire regimes: moving from a fuzzy concept to geographic entity


Pyrogeography Working Group, Meeting 1, Australian Centre for Ecological Analysis and Synthesis, Brisbane, Australia, June/July 2011

Understanding ecological processes at a global scale relies on the use of interrelated biogeographic concepts such as the ‘biome’, ‘vegetation type’ and ‘climate zone’. The familiarity of these concepts amongst ecologists and biogeographers probably masks their inherent complexity and the difficulty in arriving at workable and agreed definitions. These classical concepts have become refined and operationalized over the last 200 yr, and they can now be broadly reproduced by process-based dynamic global vegetation models (DGVMs) based on a small array of plant functional types.

‘…the complexity of fire regimes, and the difficulty of reducing a fire regime to a single variable…, befuddle attempts to use simple statistical models to predict climate change effects’

An important omission in these foundational biogeographic concepts concerns fire disturbance, which is increasingly recognized as an integral part of the Earth system (Bowman et al., 2009). Fire is recognized as being an emergent property of climate and vegetation type (Bradstock, 2010; Krawchuk & Moritz, 2011), with distinct patterns in time and space. The ‘fire regime’ is now a central concept in fire ecology, capturing variation in fire frequency, season of burning, intensity (energy released), severity (the biological impact) and spatial scale and pattern. While the distributions of biomes are now well described at global and continental scales, detailed descriptions of fire regimes at similar scales remain unavailable.

The need to understand the geographical distribution of fire regimes motivated the formation of a Pyrogeography Working Group (‘Pyrogeography: integrating and evaluating existing models of Australian fire regimes to predict climate change impacts’;, of the Australian Centre for Ecological Analysis and Synthesis. The recent first meeting of the working group focused on characterizing the diversity of fire regimes across an entire continent, using Australia as a case study. While such an exercise initially seems rudimentary, it represents an essential step in the quest to move the inherently ‘fuzzy’ concept of fire regime up to the same footing as vegetation type and biome.

Do emerging conceptual models overlook the complexity of fire regimes?

Only recently has there been impetus to form unifying theories of the biogeography of fire at continental and global scales, primarily with the work of Meyn et al. (2007), Parisien & Moritz (2009), Bradstock (2010) and Krawchuk & Moritz (2011). Conceptual models developed by these researchers focus on the ‘switches’ (sensuBradstock, 2010) that must be activated for fires to occur, including: sufficient biomass, availability to burn, ambient fire weather suitable for fire spread, and ignitions (Fig. 1). The relative importance of ‘biomass’ and ‘availability to burn’ is demonstrated by the nonlinear relationship between fire activity and both productivity and aridity (Fig. 2); in dry, unproductive environments, fire is typically limited by a lack of biomass, while in wet, productive environments, fire is limited by high fuel moisture. The recent work by Krawchuk & Moritz (2011) is the only attempt to validate such conceptual models with extensive spatial data sets of fire activity.

Figure 1.

The range of timescales over which climate and weather variables operate to affect the occurrence of fire, assuming the ‘four-switch’ model of Bradstock (2010).

Figure 2.

Conceptual diagram of the hypothesized, nonlinear relationship between aridity and primary productivity (x) and fire frequency (y). In arid, low-productivity environments, fire activity is limited by sufficient biomass to burn, while in mesic, high-productivity environments, fire activity is limited by conditions conducive to fire (e.g. low fuel moisture content, high temperatures and high wind speed). The result is that fire activity tends to be highest at intermediate levels of aridity and productivity. Modified from Bond & Keeley (2005), Bradstock (2010) and Krawchuk & Moritz (2011).

A key issue emerging early in the meeting was that existing conceptual models generally only relate to fire frequency, representing just one component of a fire regime. The importance of other fire regime components was demonstrated by contrasting the fire regimes of tall eucalypt forests and temperate rainforests that are often juxtaposed in the wetter parts of south-eastern Australia. David Bowman (University of Tasmania, Australia) pointed out that the classic conceptual diagram showing variation in fire frequency across a productivity and aridity gradient (Fig. 2; Bond & Keeley, 2005; Bradstock, 2010; Krawchuk & Moritz, 2011) fails to account for the startlingly different fire regimes between these two vegetation types. The difference lies not in fire frequency, as they have similarly infrequent fires (typical intervals in the order of centuries), but in the intensity of fire when it does eventually occur.

Brett Murphy (South Dakota State University, USA) noted that the complexity of fire regimes, and the difficulty of reducing a fire regime to a single variable (e.g. fire frequency or typical fire intensity), befuddle attempts to use simple statistical models to predict climate change impacts on fire regimes. For example, Krawchuk et al. (2009) modelled the global distribution of fire over a 10-yr period (i.e. fire present or absent) as a function of climate. They predicted reduced frequency (‘retreat’) of fire in the wetter parts of south-eastern Australia. However, the occurrence of extremely intense crown fires, such as the recent Black Saturday (7 February 2009) wildfires, are of greater ecological significance. These infrequent, extremely intense wildfires are linked to rare combinations of fire conditions that cannot be accounted for by simple correlative modelling exercises focussed on mean annual conditions. Grant Williamson (University of Tasmania, Australia), Matthias Boer (University of Western Sydney, Australia) and Mark Cochrane (South Dakota State University, USA) suggested that modelling the frequency of extremely intense fires requires a shift in focus from mean annual fire conditions to the tails of frequency distributions (e.g. the 99th percentile of the Forest Fire Danger Index (FFDI) vs mean annual FFDI).

The need for process-based models of fire regimes

Geoff Cary (Australian National University, Australia) and David Bowman indicated that climatic variables driving fire regimes tend to be so highly correlated that conventional statistical analyses may not partition out the individual climatic effects. For example, a number of conceptual models of fire activity (e.g. Meyn et al., 2007; Parisien & Moritz, 2009; Bradstock, 2010; Krawchuk & Moritz, 2011) emphasize the importance of climate and weather variables operating at a range of temporal scales (Fig. 1). The need to understand the relative importance of short-term weather and long-term climate is highlighted by the uncertainty with which decreasing rainfall, productivity, and fine fuel loads will offset increasing fire danger in some parts of the world (Williams et al., 2009).

Process-based models of fire regimes may effectively circumvent this problem. These include landscape fire succession models (LSFMs) such as EMBYR (Ecological Model for Burning the Yellowstone Region) and FIRESCAPE (Keane et al., 2004), which simulate the linked processes of fire and succession in a spatial domain, and DGVMs, which simulate vegetation processes as a function of environmental variables at large spatial scales. Recent advances in the spatially explicit modelling of fire within DGVMs (e.g. Prentice et al., 2011) provide strong promise that these two modelling paradigms can be integrated and fire regimes can be realistically simulated under a range of global change scenarios. A clear advantage of DGVMs is that they provide the means to account for processes, such as increasing atmospheric CO2 concentration, that are likely to increase productivity and fuel production.

Geoff Cary pointed out that existing LSFMs are not easily transferable between strongly contrasting systems (e.g. desert vs temperate forest), with extensive re-parameterization necessary, precluding continental-scale predictions of fire regimes at this stage. DGVMs are better able to simulate vegetation and fire processes at large spatial scales, but their ability to realistically recreate many characteristics of fire regimes has yet to be established. Hence, a key research challenge is to evaluate the extent to which existing process-based models can adequately predict the diversity of fire regimes seen at continental and global scales.

Defining fire regimes and validating fire regime models

An issue arising repeatedly during the meeting was the difficulty of defining fire regimes at large spatial scales, and generating data sets for validating fire regime models. Fire activity and season of fire are now routinely derived using satellite-based fire detections (e.g. Krawchuk & Moritz, 2011), although a weakness inherent in this approach is the inability to provide data on the return time of fire in infrequently burnt systems, given that consistent satellite records are only available since the 1970s. There is some scope for satellite-based detection of fire intensity (e.g. fire radiative power (FRP); Wooster et al., 2005) and fire severity (e.g. differenced normalized burn ratio; Cocke et al., 2005); however, the reliability of these indices in a wide range of systems, and exactly how they relate to ground-based measurements of intensity and severity, have yet to be established.

Conclusions and challenges emerging from the workshop

Making the fire regime a usable concept remains a research challenge. A key outcome of the meeting was the consensus that the Australian continent could be used as a case study to develop a framework for describing and mapping the diversity of fire regimes at a continental scale, by identifying and describing major fire regime types through synthesizing existing vegetation maps, expert knowledge, literature reviews and satellite-derived fire activity products, and by reconciling phenomenological descriptions of fire regimes with conceptual models. It remains to be demonstrated empirically how closely fire regime (cf. fire activity) is linked to climate, although the more detailed simulation of fire using process-based models such as DGVMs is likely to facilitate our understanding of the complexity of fire regimes and their key drivers. The challenge left for the working group is to develop a framework to evaluate the extent to which fire regimes are driven by climate and other environmental variables, and whether these fire−environment relationships concord with predictions of the group of conceptual models recently developed, and predictions of process-based models. It remains to be seen if fire activity can be described in climate space in the same manner as the Holdridge life zones (Holdridge, 1967).