Predicting the unpredictable: how computer modelling now helps emergency services plan for disaster evacuation

“I was up at seven and we had white ash coming from the sky.”

That ash was from a nearby bushfire burning towards Melinda Quinlan’s property in the country town of Cudlee Creek, just hours into the New Year, 2015.

The blaze started in the town of Sampson Flat – about 15 kilometres away – and quickly blew towards several other populated Adelaide Hills communities.

Melinda has lived in the area for almost two decades, and grew up in South Australia’s south-east, which is also challenged by annual forest and grassfires. Her experience of hazardous events was enough to know it was time to leave.

“Bushfires are not alien to me,” Quinlan says, “And when Sampson Flat started, I was concerned only because of where it was and the way the weather was travelling, which meant it was coming towards us.”

Melinda Quinlan has prepared her property to be able to stay and defend in the event of a nearby fire / Credit: Supplied

“On the Friday night before we were impacted, I walked to the top of the hill [on our property] and all I could see was a glow along the hill opposite us, on the other side of the gorge.

“I came back down the hill and said to the kids, ‘If it’s ‘snowing’ [ash] tomorrow, we’re leaving.’ I watched it on and off during the course of the night and was up at seven the following morning… and I said the best thing to do now is to leave and we’ll deal with whatever the consequences are.”

While Melinda decided what action to take, emergency response agencies would have been guessing what people like her might do. To assist in creating a smooth evacuation, these groups try to predict human behaviour, such as how quickly departures might occur, and road conditions that will be encountered when doing so. They also seek to understand the emotional decisions that influence a decision to leave: is it one person, or a whole family involved in the evacuation? How old are they? Are there people with disabilities? What about pets, livestock and other animals? Are they going to assist neighbours?

Should I stay or should I go?

Whether to evacuate one’s home and leave a community is an incredibly difficult decision to make. Some delay the decision for as long as possible, and for those without practice or experience, evacuation often takes longer than anticipated.

In Australia, there’s no requirement to leave a property in the face of a hazardous event (in the USA, authorities take that decision out of individuals’ hands with compulsory evacuation orders).

While flooding hazards – like those experienced by the eastern states during recent La Niña seasons – kill many more people than bushfires, blazes are a regular fixture in the world’s driest inhabited continent.

The Black Saturday bushfires claimed 173 lives and razed two thousand homes in February 2009. This led to a royal commission and nationwide changes including a major overhaul of Australia’s bushfire danger rating system.

In Australia, there’s no requirement to leave a property in the face of a hazardous event.

A review found human readiness for those fires ranged from people having sophisticated plans, to no preparation at all. Over a third of those who initially stayed to defend their homes eventually evacuated because of higher perceived levels of danger, equipment failure or because their home caught fire.

A decade later, the Black Summer fires across Australia resulted in thirty-three deaths, the destruction of thousands of homes, and the death and displacement of almost three billion animals.

Of 80 recommendations made by the Royal Commission into National Natural Disaster Arrangements following the Black Summer fires, seven were related to evacuations. These included periodic review and updates of evacuation plans in relation to evacuation routes, educating communities including ‘seasonal’ populations, ensuring essential services and supplies, and the use of nationally consistent terminology.

A family standing in front of a large tree outside
Ebony Jenkins and her family. They’ll evacuate in line with advice from emergency services. / Credit: Supplied

That education component, says lifelong Adelaide Hills resident Ebony Jenkins, might help improve understanding of the risks of bushfire.

In the past 30 years, Jenkins estimates she and her family would have three ‘decision days’ to leave their property for a safer location each bushfire season.

For her, it’s an easy decision to make: pack a bag of valuable items, bundle the kids, dog and husband into the car, and drive either an hour towards the Fleurieu Peninsula or to Adelaide’s metropolitan area, whichever is away from the fire.

“I don’t think evacuation is a straightforward call.”

Ebony Jenkins

Often, these fires have been tens of kilometres away, with no guarantee they would blow towards her home. In recent years, the situation has been more serious, as her property was close to the 2021 Cherry Gardens fire that burnt through 2,700 hectares and razed two nearby homes.

She describes the Cherry Gardens fire – “just over the hill” from her home – as the scariest she’s had to consider.

“But I don’t think evacuation is a straightforward call,” Jenkins says.

“I think it’s because through our area there are so many horse properties and a lot of people. With the [Cherry Gardens] fire some of our friends stayed because they’re on big properties, and they’ve got horses they couldn’t get out in time.

“For one of our friends, the fire came within 700 metres of his property, and he stayed, and he was on a dead-end road, so it could have been really bad.

“But it’s a difficult decision when you need to move animals, and big animals, fast.”

Model behaviour: Predicting how people will respond to hazards

While educating, and potentially motivating, people to safeguard their lives by leaving a hazard area is important, the task of reviewing and updating evacuation plans – and predicting human behaviour that might transpire during a hazardous event – is far more complex.

Whether a person chooses to stay or leave, or how many community members may evacuate, is an important consideration for government planners and emergency services.

Will an access road be blocked because it’s full of cars driving the other way? Where are the places homeowners will remain to defend their property? How will behaviour change when there’s a fire burning 50 kilometres away, instead of five?

Whether a person chooses to stay or leave, or how many community members may evacuate, is an important consideration for government planners and emergency services.

For midweek evacuations, they need to consider how parents will get their kids from schools, and how busy town centres will become when people start congregating.

So how do authorities predict the unpredictable?

Fortunately, there’s a model for that.

Dr Erica Kuligowski at RMIT’s School of Engineering specialises in evacuation and emergency communications in fires and other hazardous events.

At this year’s Australasian Fire and Emergency Service Authorities Council conference, she presented a program, named WUI-NITY, to delegates from emergency agencies across the nation.

WUI-NITY is an amalgamation of WUI (for wildland-urban interface) and Unity3D, a game engine popularised by smartphone apps like Pokémon Go and Call of Duty Mobile. In essence, the software simulates the behaviours of communities confronting a bushfire. Within the program, fire behaviour, pedestrian response and traffic movement are modelled to aid real-world response planning.

A woman sitting a computer terminal
Dr. Erica Kuligowski is a Vice-Chancellor’s Senior Research Fellow in the School of Engineering at RMIT, studying evacuation and emergency communications during bushfire, building fires, and other hazards. / Credit: RMIT

Through ongoing verification and validation protocols, developers refine the software to accurately emulate real-world actions by humans and hazards.

Modelling programs are intended to remove the guesswork for response planners.

“There are many factors that influence what people do, we actually have a good sense of that,” Kuligowski tells Cosmos.

“It’s just that in different types of fires, people are getting different types of information.

“There are certain factors that we know influence evacuation behaviour and it depends on the fire scenario, the community, their experience and their training.”

Choosing to stay

At Cudlee Creek, Melinda Quinlan is an example of a person whose experience leaves them better placed to stay and defend against a hazard. Although she evacuated during the 2015 Sampson Flat fire, she stayed to defend her property during Black Summer in 2020.

To an outsider, that might seem unusual: History shows the Sampson Flat fire was contained within a few days and burnt through 20,000 hectares. On the other hand, Black Summer notoriously razed over thirteen times that amount of land across South Australia, including around townships just kilometres from her home. But circumstances during Black Summer were different, and while she stayed to defend her home, the fire ground did not spill into the main Cudlee Creek township, unlike those five years earlier.

“I have no hesitation saying for the next fire, I will still be here, I will still be fighting that fight.”

Melinda Quinlan

However, many homes in surrounding towns were destroyed by falling embers. In Lobethal, just eight kilometres south, a drive down the main street would appear quite normal in the aftermath, save for the confronting sight of singular properties that had burnt to the ground.

“Being here, because you’re on the ground, you know what you have available to fight, you know what you can use to fight the fire that’s coming your way,” Quinlan says.

“I have no hesitation saying for the next fire, I will still be here, I will still be fighting that fight.

“The better set up that you have to allow the CFS access to your property, the more likely they are to come and help defend you. That’s the golden thing. We’re a good community here in Cudlee Creek and we have a really good CFS team. Most of us know several people on that team and they’re familiar with the properties they’re defending.”

From macro to micro

While Kuligowski’s WUI-NITY platform looks at interactions at a community scale, her colleague Dr Dhirendra Singh’s work focusses on individual actions during a hazard.

Singh is also at RMIT, and is a visiting scientist for natural systems modelling at the CSIRO’s Data61 project.

Using his background in software engineering, his current assignment with the CSIRO is the development of evacuation modelling software that can be used by government authorities to project individual actions during a hazard.

Dhirendra Singh speaking at AFAC 22
Associate Professor Dhirendra Singh is the technical lead on the ongoing CSIRO/RMIT Evacuation Modelling project (SEEKER) that is building a national evacuation modelling capability for understanding and assessing risk to communities during natural disasters in Australia. Credit: AFAC

“The idea is you basically build representations of complex systems bottom up,” explains Singh.

“You try and represent those behaviours of individuals and the interactions of those individuals, between themselves and with the environment. And then when you simulate these behaviours, you can then observe sort of complex system level phenomenon.”

The task of obtaining substantive data on who people are and what they do within their communities has been made easier by massive research exercises that arose from the royal commission into the Black Saturday fires.

Consider traffic – one of the variables also viewed at a community scale by the WU-NITY platform. Singh’s work instead looks at how individual actions play out on the road network.

Such software can try to make sense of what a person might do if presented with an evacuation-triggered traffic jam.

The task of obtaining substantive data on who people are and what they do within their communities has been made easier by massive research exercises that arose from the royal commission into the Black Saturday fires.

This information has been used to inform the models that Singh and his colleagues have developed.

Making a model

So how does this modelling application work?

In simple terms, a map is rendered with a ticking clock in one corner. Small dots representing individual people going about their day move around the map. Operators then introduce a hazard – such as a fire – at a point on the map, which proceeds to expand based on predetermined conditions.

Those dots then begin moving in different ways, indicative of the behaviours that might be witnessed during a real-life hazard.

As Singh points out, rarely when you tell a community of people to evacuate does every person leave.

“In our models, when you tell people to leave, there’s a whole lot of commotion that starts: some people will leave, some people will start going to the fire because there’s some dependents that live in that direction and they want to attend to them first, and so on.

“What that results in is a whole bunch of these intermediate trips, this extra traffic on the roads, and this is on top of any other background traffic that’s already in play and that can give you a little bit of a more nuanced picture of what evacuation for that community might look like.”

“Rarely when you tell a community of people to evacuate does every person leave.”

Dhirendra Singh

When the Black Summer bushfires threatened to turn towards her property, Ebony Jenkins’ difficult decision was not whether to evacuate, but which way to go.

In Singh’s model, she’s a black dot that he’s trying to accurately simulate.

“There were concerns about using the freeway,” Jenkins explains.

“So we went to Adelaide when it started moving towards Hahndorf. That changed our decision-making around leaving to go to there, rather than towards the Fleurieu Peninsula.”

Singh, Kuligowski and their teams have recently engaged communities in regional Victoria to simulate what happens on a typical fire risk day.

A group of people look at birds eye view computer modelling.
Kuligowski and Singh (pictured centre) with colleagues undertaking scenario modelling. Credit: Supplied.

Getting qualitative data from schools, fire services and councils helps build greater realism into their modelling systems. The result should be a clearer understanding of how people might respond to a hazard, and what pressures on local infrastructure might emerge in an evacuation scenario.

The intent is that feeding modelled scenarios back to local governments can support better infrastructure and evacuation planning. Singh says this process is about building dialogue and information sharing with communities.

“Everyone has different ideas about what could or couldn’t happen, and when you put it together [in a model], maybe people can learn something new about what can happen that they didn’t know about previously,” he says.

“The worst thing you can have in a fire is fear. You need to be practical, you need to think in a way that you’re looking beyond what you’re dealing with now.”

Melinda Quinlan

This willingness for community institutions to collaborate with local emergency services and experts like Kuligowski and Singh, is indicative of substantial research into how people treat hazardous events.

Melinda Quinlan expects she’ll still be there when the next fire rolls around. She also knows every hazardous event is unique, and panic is the enemy of survival.

“The worst thing you can have in a fire is fear. You need to be practical, you need to think in a way that you’re looking beyond what you’re dealing with now,” she says.


Read more: What fuelled Australia’s “Black Summer” fires?


“My house was built circa 1890, so it’s been through a lot of bushfires, and it hasn’t burnt yet. It’s got big, thick stone walls, we’ve got high ceilings, we’ve got a huge roof cavity.

“In our situation, we know bushfires are what we live with, [so] how are we going to improve what we’re living in to protect us better when that time comes.

“I’m quite comfortable in this and it will take a lot for me to decide to leave, it would have to be a pretty fierce, intense fire, and I would make sure I got my facts before I left. Because with me leaving, there is no one here to defend this place.”

“It’s very rare that people are going to kind of act only in their own self interest at the expense of others. People are very altruistic when they’re evacuating.”

Erica Kuligowski

Importantly, as Kuligowski points out, one thing that is known about fire behaviour is that people don’t panic.

“So that’s really important,” she says.

“Because we see that [idea] thrown around in the media and we see that also sometimes in an academic context.

“We’ve seen the studies that we’ve done in the sociology of disaster field, it’s very rare that people are going to kind of act only in their own self interest at the expense of others. People are very altruistic when they’re evacuating.”

Also in Cosmos Weekly Issue 78: Future-proofing the PBS and our health (Part 2 of 2).

Please login to favourite this article.