A spark of hope in predicting bushfire behaviour
A lightning strike, an arsonist, a planned burn that gets out of control, a few sparks from farm machinery. We know how a bushfire can start—but there’s no telling where and when it will start.
Even when we’ve detected a bushfire, on the ground or by satellite, there’s no guarantee what path it will take.
Conventional bushfire modelling tells us it will slow as it travels down a slope, and that it will burn at a particular rate in eucalypt forest or grassland.
But what if the fuel type or the terrain changes, or the wind speeds up or shifts direction? What if the fire leaps into the treetops and starts ‘crowning, or a burning tree falls across a road or a creek and establishes a new fire? What about spot fires ignited by embers spat out kilometres ahead of the fire front?
Predicting the unpredictable
Dealing with the unpredictable has always been a challenge for Australia’s state and territory bushfire agencies.
And thanks to climate change, their job just got harder. In southern Australia, the fire season is now longer, extending from October to March, and much of the country has become drier and hotter in recent decades.
More than ever, the pressure is on to better predict where, and how fast, bushfires will start and spread.
To deal with the unpredictability and scale of bushfire in Australia, fire analysts at the NSW Rural Fire Service and its counterparts in other states have been investing more time, effort and resources in the relatively new science of bushfire-behaviour analysis.
Where once they predicted fires using manual calculations based on mathematical fire-behaviour models, fire analysts are now exploring the use of computerised simulation systems, made possible by huge leaps in computing power and some very large datasets, especially the real-time, detailed weather data provided by the Bureau of Meteorology.
With established fire-behaviour models at their heart, computer-based simulators enable experienced analysts to run multiple ‘what-if’ scenarios on their desktop to map out a fire’s many possible paths and rates of spread.
It’s the sort of real-time information needed by state fire services to decide where local bushfire risk is highest, where fire-fighting resources should be directed once a fire has taken hold, and along which routes people can be safely evacuated.
Spark – a game-changer?
For Stuart Matthews, Senior Fire Behaviour Analyst with the New South Wales Rural Fire Service (RFS), running hundreds or even thousands of what-if simulations in parallel—a feature known as ‘ensemble analysis’—and being able to see the results within seconds, is the main attraction of CSIRO’s new fire-simulation package, known as Spark.
When conditions are uncertain, Stuart explains, one simulation is not enough. But being able to instantly see maps and data showing all possible combinations of wind, temperature, fuel, terrain, fire-breaks and humidity is potentially a game-changer.
Spark is not the first simulator being considered by the NSW RFS—they are already using Phoenix RapidFire, a system developed in Victoria. However, they see value in considering more than one simulator, a practice that is common in weather forecasting.
Plugging in to Spark
Stuart Matthews is awaiting the outcome of the Bureau of Meteorology’s evaluation of the four main fire simulators in use across Australia:
- Aurora-Australis (WA)
- Prometheus (developed in Canada, now also used in Tasmania and NZ)
And, while the NSW RFS is not planning to use Spark this fire season, says Stuart Matthews, he can see its potential advantages.
One advantage is that Spark is designed as a framework of plug-in modules that can be easily configured for different users, landscape types and applications such as planning controlled burns, real-time fire response operations, and evacuation planning. Any bushfire spread model can be added to Spark which means that, as new fire models come along, the older parts of the system can simply be replaced.
“Fire researchers targeting a particular aspect of fire behaviour, say, a new spotting model, will be able to put it in to see if it works,” says Matthews.
“And, because it’s been built to be easily reconfigured, you can do a lot of things, including prescribed burning as well as normal operational wildfire prediction.
“You can also do risk-planning studies – you could do a simulation with a proposed fuel treatment, or changes to how a development is laid out. What effects do those changes have on a fire, and on the risk to those assets?
“It’s not just that you can configure the model—it’s also quite omnivorous. You can feed all types of different inputs into it, and you get outputs in a variety of formats, which is really useful for us as we move towards doing ensemble simulations rather than just doing single incident runs.”
Assessing risk is paying dividends
Spark project manager Dr Mahesh Prakash is looking forward to running a second Spark workshop in February 2017 aimed at risk-assessment consultants and other potential commercial users.
Eco Logical Australia is already using Spark to assess risk for the Australian Government Department of Defence and for parks agencies. Spark is helping them work out when and where the risk of bushfire might be too high for bushwalkers or military personnel, and the best routes for safely evacuating a fire-threatened area.
“Every year in Australia, bushfires and other natural hazards cause loss of life along with billions of dollars worth of property, livestock and environmental damage,” says Dr Prakash.
“If risk assessment consultants can use Spark to pre-emptively save lives or reduce property losses, the product will be paying dividends.”