An experimental seperated v-shaped fire conducted in the CSIRO Pyrotron. Photo: Andrew Sullivan, CSIRO.
New research is developing the first two-dimensional fire simulation model that can operate in faster than real time, while incorporating intrinsic, fire line dynamics.
A pioneering Bushfire and Natural Hazards CRC project has highlighted the role pyroconvective interactions and spot fire dynamics play in the spread of fires across a landscape.
By A/Prof Jason Sharples from the University of New South Wales, Dr James Hilton and Dr Andrew Sullivan from CSIRO. This article was first published in Issue Four 2019 of Fire Australia.
Fires that burn in close proximity can influence each other due to pyroconvective interactions between individual fires. The same processes can apply to different parts of a single fire line. A typical example is when intense spotting causes many fires to form and coalesce. Interactions between individual spot fires and other parts of the main fire perimeter can increase local rates of spread in unexpected directions, potentially producing broad, flaming zones that can entrap firefighters and increase the likelihood of extreme bushfires.
By combining advanced mathematical modelling with laboratory and numerical experimentation, this research is providing insights into the physical drivers of these interactions.
The knowledge gap
Current operational fire simulation methodologies cannot account for dynamic modes of fire propagation (see definitions, page 31) that are driven by complex interactions between the fire and the local atmosphere. Nor can they explain basic fire spread patterns, such as the observed parabolic rounding at the head of a wind-driven fire line. This project developed modelling techniques that address these shortcomings.
Fire behaviour in dry eucalypt forests in Australia is characterised by spot fires—new fires ignited by the transport of burning debris, such as bark, ahead of an existing fire. This also applies to many other vegetation types, but to a lesser extent. Under most burning conditions, spot fires have little influence on the overall spread of a fire, except where spot fires can overcome hurdles, such as topography or breaks in vegetation. However, in severe bushfires, spot fires can become the dominant propagation mechanism—the fire spreads as a cascade of spot fires that forms a ‘pseudo’ front, way ahead of the main fire front.
It is well known that multiple individual fires can affect the behaviour and spread of all fires present. Similarly, different parts of a single fire line can influence each other, particularly when the fire line develops certain geometric configurations. In such instances, fire spread is driven by the combination of extrinsic influences, such as wind and terrain, and intrinsic effects that arise when different fires or different parts of a fire interact through pyroconvective (and other) processes (see definitions). In some cases, intrinsic effects can become significant and result in distinctly dynamic modes of fire propagation, such as what occurs in junction fires, for example, where fires meet and merge. These dynamic modes challenge an assumption behind all existing operational fire behaviour models—that fires spread at a quasi-steady rate. These operational models also assume that different fires, or different parts of a single fire line, essentially burn independently; this neglects any potential influence of intrinsic effects. These models cannot account for potential dynamic interactions, which may significantly influence the spread of a fire. The inability to accurately predict fire behaviours can place firefighters at risk and hinder issuing effective warnings to the general public.
Some aspects of dynamic fire behaviour can be modelled using coupled fire atmosphere models, but these models are too expensive for operational use. To meet this operational need, this research is developing computationally efficient, two-dimensional fire-simulation methods, an innovation which, for the first time, can account for key intrinsic dynamics of fire propagation.
How this research was conducted
This project treats a spreading fire as an evolving interface between burnt and unburnt ground. Previous research has also investigated this, but the methods used often encountered difficulties when fire lines merge or when isolated pockets of unburnt vegetation remain, which typically occur when spot fires coalesce (see Figure 1). A methodology that can successfully deal with these complexities is crucial—which is why our research team chose to use the level set methodology.
The level set method forms the basis for the development of efficient propagation models that use physically simplified proxies to account for complicated dynamic effects. To develop the model, the research team is conducting a series of experiments using the CSIRO Pyrotron facility.
These experiments target specific fire line configurations, such as parallel fire lines, v-shaped ‘junction’ fires, ring fires and multiple spot fires.
These laboratory experiments were also complemented by field experiments conducted in Portugal by international collaborators, under the auspices of the Portuguese Foundation for Science and Technology’s ‘Project fire whirl: vorticity effects in forest fires’.
What does this research mean?
The research has highlighted the critical role that pyroconvective interactions play in many aspects of fire propagation. For example, it has discounted radiative heat transfer as the main driver of the local increase in the rate of spread associated with junction fires—the observed effects were able to be accurately modelled using pyroconvective interactions alone. It has also highlighted the complex role that pyrogenic vorticity has in dynamic fire propagation.
Pyroconvective interactions can significantly affect fire behaviour—even in very basic patterns of fire spread, such as a line fire being driven by a uniform wind. The familiar ‘parabolic head’ shape that develops at the fire front is due to differences in the pyroconvective indraft along the fire line. Fire spread simulators now used operationally in Australia don’t account for pyroconvective interactions, and cannot accurately model the development of a basic, straight-line fire.
To address these types of shortcomings, this research initially attempted to model pyroconvective interactions using fire line curvature. Fire line curvature served as a good predictor of dynamic fire spread in some, but not all, cases. The research then considered a very simple idea—to treat each point on the fire line as an independent ‘sink point’ for horizontal airflow.
This means that each point along a fire line creates its own radially symmetric indraft wind, the strength of which is determined by the intensity of the fire at that point. Considering the fire line as a whole, the indraft effects created by each of the individual points combine to produce an overall indraft wind, which is referred to as the ‘pyrogenic flow’—that is, the flow of air created by the fire itself. This pyrogenic flow can then be added to the ambient wind flow to more accurately model the spread of the fire. This model is called the pyrogenic potential model, due to its similarity to models for determining the electrostatic potential of an array of electric point charges.
The pyrogenic potential model is a highly simplified, coupled fire atmosphere model. The influence of the pyrogenic flow depends critically on the geometry of the fire line (see Figure 2). For certain configurations, the pyrogenic flow locally increases the rate of spread of certain parts of the fire line. For example, in a junction fire it produces the rapid advance of the point of intersection, as has been observed in experiments and numerical simulations (Figures 2a and 2b). The pyrogenic flow causes two straight parallel fire lines to ‘draw in’ towards each other (Figure 2c) and accounts for the higher rates of spread encountered as a closed arc of fire collapses upon itself (Figure 2d).
Ongoing research will continue to develop the pyrogenic potential model. For example, the model was recently extended to account for other near-field effects, such as localised sources of vertical vorticity (see definitions). This has provided a way to model complex modes of dynamic fire propagation, such as vorticity-driven lateral spread (see definitions), which is highly efficient at triggering zones of deep and widespread flaming, consisting of many coalescing spot fires. This can increase the likelihood of pyrocumulonimbus (fire thunderstorm) development.
These state-of-the-art, near-field models are easily applied in fire simulation models such as Spark.
Research enhances predictive capability
The recognition that fires can mutually influence each other’s spread has been used in various contexts, such as prescribed burning and backburning. This research provides a theoretical basis for such practices, enabling a more quantitative understanding of their effects.
By accounting for pyroconvective interactions between different fires or different parts of a fire line, this research improves the estimation of the overall power of a fire. When combined with research into the atmosphere’s role, it can alert forecasters to the likelihood of a fire transitioning into a more extreme event, such as a pyrocumulonimbus storm. This could better inform burning operations and help to avoid unexpected fire blow-ups.
Ultimately, the research enables the modelling of key aspects of fire behaviour that was previously only possible using computationally expensive, coupled models. The nearfield techniques that form the pyrogenic potential model allow complex modes of fire propagation to be simulated within operational timeframes.
The research team is discussing the potential uses of this research with end users, including use in firefighter training materials and equipping fire behaviour analysts with technological tools to assess the likely progression of bushfires and the potential for escalation.
A word from the emergency services
“This project has demonstrated the feasibility of using computationally efficient approaches as a proxy for more costly approaches to modelling dynamic fire behaviours. Investigation toward incorporating this work into trial operational models and training is both progressing and very promising. This exciting work is directly relevant to strongly enhancing our understanding and predictive capacity in terms of dynamic fire behaviours.” Brad Davies, Fire Behaviour Analyst, NSW Rural Fire Service.