019 - Modelling human behaviour in wildfire evacuation with Erica Kuligowski

Hello, everybody. Welcome to session 19 of the Fire Science Show. As you may notice, we've talked quite a bit about evacuation and human behavior on the podcast. I had really nice interview with Dr. Mike Kinsey on the cognitive biases and how they impact the decision-making process in evacuation. I had a really interesting chat with a professor Enrico Ronchiand professor Ruggiero Lovreglio on the future of acquisition modeling. And in these talks I was really trying to emphasize this line that distinguishes modeling movement from modeling behavior. it's something really important to me. Show you this difference that what happens when you just simulate pre movement times and simulate the movement of your agents in an evacuation modeling and what happens when you start modeling human behavior, their choices, and how that impacts the output of the simulation. And I think is really important because it's one of those. Paradigm shifting things in the industry. If we were able to easily model human behavior, we would end up with a completely different fire safety engineering as a whole, because understanding human behavior is fundamental to fire safety engineering. And if we can make a huge leap in this field, well, the whole fire safety engineering improves a lot. And today another interesting discussion on evacuation modelling, I have another world-class expert and she actually has quite a lot of achievements in terms of modeling human behavior. A lot of successful modeling actually. She did a lot of studies on previous disasters, like World Trade Center, for example, and now have having moved to Australia, studying a bit of bush fires and, uh, a lot of interesting things that you're going to end at the end ofthe interview. So let's not wait anymore. Please welcome with me, Dr Erica Kuligowski from RMIT in Melbourne and yeah, let's spin the intro and jump into the interview. Hello everybody. Today, I'm here with Dr Erica Kuligowski from RMIT university in Melbourne. Hi, Erica. Great to have you here.

Hi, thank you for having me.

So let's start with your interesting twist in a career. You've spent like two decades in NIST and now moved into Australia in the middle of pandemic. So how was the experience?

What was I thinking? Oh, yes. So yes, I moved during a pandemic. It was interesting. Um, there was a bit of a delay in waiting for my visa, but I'm so excited to be here. You are absolutely correct. It's been an interesting career. I started at NIST in 2002, and then just last year made a decision to move to Australia and work at RMIT University. I'm a Vice Chancellor, Senior Research Fellow here in the Civil and Infrastructure Engineering discipline, and focusing on, um, wildfires or bushfires and evacuations. So very excited about this new change.

That's really cool. When I've followed your career from the early days of my own engineering, I always saw connected you with this field of understanding what people do in fires or what people think in fires. And I've connected this with building fires. You obviously investigated hurricanes and other catastrophic evens in US as well. And now this switch to wildfire science, was there a particular reason to, to switch into that? Or how did you jump into this new, exciting field?

Yeah, that's a great question. I have been actually interested in looking at wildfires for quite a number of years because my background isn't fire protection engineering, as well as sociology. And you know, as an FPE, , we didn't really delve into wildfires. I still always had an interest in understanding people's responses to fires of all kinds. And I think when we started seeing more and more people lose their lives in the U S and the fires becoming more intense. And covering larger areas and the California and other places out west. And then certainly here in Australia, I started thinking that it would be really interesting as well as I think I could have impact in translating all of the information and research I've done and building fires and large-scale hazards to the wildfire area. And you know, everyone asks what gets you out of bed in the morning? For me, it's trying to make a difference and hope hopefully reducing lives lost and disasters. And I think there's a lot of work to be done in the wildfire evacuation area.

Especially that on one hand you see huge fires in which the community could escape their location. And of course there is damage. There is destruction because it's a cataclysm, but not that many people lose their lives. And then you see some images of some groups of people that are just jumped by the fire or that attack them unprepared. And you see this mass loss of lives or this traumatic events where people like drive miles through fires multiple times. And you get to wonder why this one community have succeeded, and save their lives and why the other did not. And I'm not sure why maybe after an hour of talking with the I'll understand it a bit more, but there is something, that in some cases we succeed, in some cases we miserably fail. Is this something you also observe in this field?

I definitely, wouldn't say it's a failure. There's so many different aspects that go into evacuation. There's obviously lot of factors about the community itself, how they live day to day, as well as their experiences with fire, their awareness about their risks and the threat. And then also the environment around them. So what's going on with the fire? What cues are they getting? What types of warnings are they receiving? Who are they receiving the warnings from? But I think the another really big thing. Is how fast the fire is spreading and how if officials are able to kind of understand and try to predict where it's going, and it's all really about timing, it's not all about that, but that's an important part is are our people warned with enough time to leave and, fires can be with the change of winds, for example, go in a different direction or move or spread faster. So it's really tough for officials to be monitoring, making predictions and then warning communities with enough time. And then as I mentioned, the communities are bringing a lot to the situation. And how did they think about fires? How do they perceive it? How do they perceive their own risk? And then that has plays a role in how they make their own decisions. So there's a lot at play.

Can you draw a parallel to hurricane management in here because you don't evacuate communities when the hurricane is at the Caribbean sea or something? You have to time it.

That's right. You can definitely draw parallels. Although I would say that there are differences as well. So parallels of course it's heading toward a community. They have to make predictions. There's, what's called the cone of uncertainty. So the closer you are to the hurricane, the more certain you are on the direction, the further away the cone becomes larger. But with fires, the prediction is a bit more difficult. I would say. Or we don't have the tools yet, right. That are getting to that ability to draw this cone of uncertainty. And the timelines are also different. A fire could start and could have a difficult day with winds and all of a sudden it's spreading into a community where people could have up to three days in a hurricane to evacuate. And so we're talking about definitely different timescales. So while we have to make the same types of decisions in these different events. We have different timescales and officials maybe don't have the same tools at their disposal as they would for hurricane.

I've watched, webinars with you about, modeling this behavior in wildfire evacuations. And you've also shown the, some reasons from Chimney Tops fire investigation. That's also published. I'm going to drop that these resources in the show notes sleep stunned by analyses you've done on what drove people to the decision of the evacuation. And it was really a surprising in a way. Can you bring the listeners a bit on the research and then I'll tell you what surprised me.

Absolutely. Yes, absolutely. So what I do in these types of events is I try to understand what influenced people to evacuate or not. We surveyed quite a number of people who evacuated that fire. And we asked them different things about themselves, their experiences with fires and their demographics, you know, who they are, gender, income, education. And then we asked them quite a bit about the event itself. We asked them to tie the types of cues that they witnessed, what warnings they received, if they received any at all. And then how did they perceive these cues around them and make a decision? And I think you're going to hear me talk about perceptions quite a bit, because it's someone's perception of their threat and risk that have quite a bit of influence on the decision they made. And by threat perception. I mean, do people think that a fire is likely in their area and then risk perception, do they think, or how likely do they think that the fire will negatively impact them and the people around them, injury or death even. And so if people have lower perceptions of threat and risk, research has found that they're less likely to evacuate. And I think about perceptions as kind of, I try to think about it visually. I think about it as like a measuring cup, if you will. And some people come into a fire with a lot of liquid in that glass and some people come to that fire just a bit. And if you have, if you reach a certain threshold in that measuring cup, then you're likely to then say, okay, I'm going to evacuate now. Right? I've reached a certain threshold in my perceptions for some communities, maybe especially communities that don't have a lot of experience with fires. They come to the fire with less liquid in their cup, right? So it takes a lot more cues around them. I'm seeing other people evacuate, maybe even seeing the fire itself for them to reach that threshold and then make a decision to evacuate. So in that study, not only did I try to understand I myself, I had a team of people worked with me. We tried to understand the factors that influenced both people's perception of risk and their decision to evacuate because. Following the theory that I use as the foundation for my work, it really is people feeling at risk, understanding the threat and then taking action. So there's quite a bit of factors that can influence our perception of risk that can raise our perception of risk. We've found from research that gender plays a role. So actually women are more likely, and I really disliked gender findings because you know, it's, sometimes it has more to do with role than it is with gender. But women have been found to have a higher perception of risk. People with shorter residence time. So they've lived in an area for a shorter amount of time, people who have prior knowledge and experience and fires. And I want to come back to experience in a moment cause that's a really complicated. People who have seen very intense cues are more likely to have higher perceptions of risks. So if they see others evacuating, if they see flames or embers in their neighborhood, right. That helps them to realize, oh my gosh, this is a real threat to me. And also the types of information that they receive. So if they receive information from a trusted source and the source is not the same for everyone, unfortunately, they're more likely to feel at risk. And then if they receive information about not only that a fire is occurring, but what are the consequences of this thing, right? How can it affect me personally, then they're more likely to feel at risk and take action. There's other things aside from risk perception that can affect our decision to evacuate. Like if people have pets and livestock, they really have to think about what they will do with them. And research has shown that it reduces the likelihood of evacuating. Having children increases the likelihood of evacuating and self-efficacy, that's another really important one. And that's, people's belief in their ability to perform actions that can reduce their risk. Right. They feel like they can have influence.

They can the fire

In a way, right? So if I have taken action in my home then maybe I'm more likely to feel like it's a safe place to be. So self efficacy plays quite a big role in that it can increase people's likelihood to stay. And I think that has to do with, it's also decreasing people's perception of risk, right. That I can do something to protect myself. And lastly, if people think that evacuation isn't an effective way to reduce their risk, then they're more likely to stay in place. So I know I threw a lot of things at you

I was prepared.

That's what we look at when we do these studies, we try to understand. What's going on around them, but also what they're bringing to the fire and how it influences their decisions to evacuate. And the reason we care about that is because not only do we want to model it, and I'm assuming you'll have some questions for me about that, but I think Lauren Moore and they've done quite a bit of work in Australia where they've developed archetypes and these are kind of types of people, if you will. Right. And we know if we have this type of person in this neighborhood, maybe we'll need to educate more or work with them more, or maybe they'll need more prompting to evacuate, or maybe they'll need more assistance. So we're trying to understand the types of people that exist in various communities. Can we understand what they're likely to do? And then can we make sure that we're providing and by we, I mean, emergency officials. Providing them with additional assistance or additional prompting, or can we target the messaging in a certain way to help people understand their risk and take action in the safest way possible.

It's brilliant. You're doing what Facebook has been doing with advertising for like the last decade, but

right, but it's for life safety.

this work and it works. So that's great. That's great. thing kept ringing in my head I was wondering about this type of let's say herd behavior of people. And I see that often in buildings. It's often it matters when the first person starts to evacuate and the other, see them, that person and start running away. Now I wondered if the same thing can be observed in communities, to what extent this is influenced by them. And. I also wondered how would it work across community? Because if you have a raging wildfire, it can devour a town after town, I assume. So to what extent, like destroying one city would promote the evacuation of another city. Once it moves to what extent this is a factor it's really fascinating to me.

I love that. You're so fascinated by it because it's, I am too, I really enjoy talking to you about this. So yes, social influence absolutely plays a role in buildings as you mentioned. And that's why, just distributing a pre evacuation time is. Just curious, right. Because it's kind of all right. You're a bunch of people in a room and we're just going to give them all a different pre evacuation time in our simulation model and see what happens. But we know that people in the same room are definitely in the same space are influenced by what people are doing around them. So it's very artificial to be thinking about pre evacuation delay time in that way. If we see someone, and they're packing up their belongings and they're taking whatever's going on seriously. And especially if we feel like they're a credible source to us. So I know them, they're my colleague. I talked to them every day and they think this is serious. Then people are more likely to start preparing themselves and to evacuate. In my dissertation, in the World Trade Center, it was really these early actors that played quite a role. And not only did they start. Preparing and getting themselves ready for safety in the World Trade Center, they also told other people to evacuate and they started assisting other people. And that played quite a big role. And these early actors were people who had fire experience, who were fire wardens, people in management positions. So these people were stepping up and they were used to stepping up even people with military training. So that played quite a big role. And I would say the same is true for large-scale disasters. If people see their neighbors take you know, preparation, they might go over there and ask, Hey, what are you doing? Like what's going on? And that's kind of that milling that goes on, right? That plays quite a role in our decision-making and people need confirmation of the threat also. And so when they receive a message, but then they see their neighbor preparing that's confirming, right. Someone else is taking this seriously. Maybe I should take it seriously as well.

And can you go opposite way? Like you've mentioned that people who have livestock do not want to evacuate people who have bigger self efficacy, then they can try to let's say survive the fire at their place. So this would be like the negative actors in here. If we consider the evacuation as the general good.

exactly right. That's not always the case. Sometimes it's not the safe thing to do, but yes, when an evacuation is called for, if you're noting that there are some people who would stay and if people see that type of behavior, it can also influence their decision to stay as well. So it definitely goes both ways. But you know, there's a lot that needs to be taken into account. Maybe that's not negative behavior for them, maybe that is the right decision for them. If they have a number of livestock and they need to be there and make sure that they're taken care of. But you also mentioned if a fire is occurring in one location and you know, what about this other location? And it's something that I'm really interested in. I know that there are people, especially in the bush here in Australia who are very aware, right. That fires are occurring. They look on their map, their apps, the maps, to understand where the fires are related to them. And if a fire is occurring nearby to them, they're monitoring that's likely to happen in some of these communities that are very familiar with fire. Some people are not. It just depends. And there's all that, all those factors that I mentioned before to you. but I think also media plays a role too, and that if we're showing a large evacuation in one community, it could effect what people are doing in this other community. It's interesting. I've never done research on that to understand, you know, the role of, okay, you're seeing this video of this community over evacuating. What does that mean for you? The other issue is whether or not they're seeing people freely evacuating, or if they're seeing people in the next community in a lot of congestion that certainly can play a role as well. Does that mean I need to leave now and get out of here? So the same doesn't happen to me or does this mean I'm not going to go at all.

Assume from this relations, you can actually tell if people are evacuating because they've planned it or they are escaping for their lives. You know, that there's a, there's quite a bit difference. And if you see a video of someone driving through a forest or a bush, which is on fire and there's fire on the left fire on the right wall of smoke in front, I would not ever would like to see myself in such a situation, especially with my family.

agree. It's very scary. It was a very scary. situation and it might be that maybe they wait, they wanted to wait and see, and that's a very common behavior. And that goes back to what you and I were talking about confirming the risk. If people didn't, if they weren't able to that to confirm the risk, if their risk perception didn't reach that threshold, then maybe they waited to see if something was going on and left at the very last moment. Or maybe they just didn't receive the warning in time.

You have also mentioned the previous experience, and I was wondering how to touch on that, because this previous experience, actually in buildings, you have this evacuation drills, you know, it's easier to train and you will not make a, an evacuation drill of a town or would you?

You would, you would.

Yeah. But, how to build this previous experience without exposing people to like real threat, is there a way

It's a great, great question. And it's actually something that I'm looking into now on a current project, but I have so much to say about previous experience. It's such an important question. I'm glad you've asked it. So let me first say that you can do a full-scale evacuation of a town and some communities have done it. I was in Colorado before the pandemic and a community did a full scale evacuation drill. The warnings were sent out at a certain time. Everybody knew ahead of time. And they got in their cars and they traveled the one route out and then they also practice the other routes. And then they came back to a community center and discussed kind of a debrief on how it went, but not every community can do that. Not every community does that. So I agree. It's definitely not as frequent as would occur in a building evacuation being able to practice it. It takes an entire community out of commission for quite, quite a bit of time. It takes a lot of time, right. But it is possible. But previous experience is difficult because it can mean so many different things to different people. So the problem is, and I'm putting my own. Research under scrutiny here, but a lot of times in surveys, we ask very basic questions about experience. We just say, have you ever been in a fire before? Or we ask, have you ever evacuated, how many times have you evacuated? But that doesn't get to the how important previous experience can be. People could have evacuated and nothing happened to their house. So they could have said, well, that's a false alarm, right? They could have perceived it as that. I didn't need to go. People could have not evacuated and nothing happened to their house, People could have evacuated with plenty of time and it was not a traumatic experience for them. People may have evacuated in very dangerous conditions. Could have been, as you mentioned, the fire, or it could have been heavy congestion that is also very traumatic for people. Sitting in traffic, the fires over there, they're stopped. So that, that could be very traumatic. People may have evacuated and came back to a damaged or destroyed home. Maybe they didn't evacuate, but they encountered extremely dangerous conditions when they were sheltering in place and they could have lost a loved one in a fire had personal injury. So it's like previous experience spans quite a range of different consequences if you will. And we're not in a lot of different studies we're not digging into the nuances there. And so in recent study of the Kincade fire, which was the 2019 fire in California. We did ask that question about previous experience. Unfortunately, there wasn't enough variance in the response, right? So we couldn't include it in our regression, but we analyzed the end of the survey comments, right? That open box at the end of our survey. And we just ask people to provide additional information. And so many people talked about the fire that had occurred two years prior to 2017 Tubbs fire. And we found out from analysis of those comments that Tubbs fire influenced people's awareness of the fire that actually occurred in 2019. It some people prepared early and were ready to go. That they're go kit ready there. You know, no nonsense. As soon as they were even warned, they didn't even wait for the orders. Then they went, it influenced people's evacuation decision. They said, we're not waiting around. 2017 was terrible. And then it also influenced people's departure time. So some people said to us that I sat in heavy traffic and it was very traumatic. So I left early this time. And I guess in conclusion, right? To your question, I think that there's quite a bit of range of experience. We need to delve into that and understand what it is about that experience to then understand its role in future behavior. Because I think we're, we've not peeled the onion, if you will, we've not really kind of understood the ins and outs of that variable. There have been some studies that have looked at this and delved in a bit more. But not enough, I think. And so it's something that I'm working on at the moment. And then you also asked about, okay, well then what does it mean for the future? How do we make sure that people understand their threats and risk? And I don't really have a great answer to that. Although I'm thinking of a few, I've got some ideas in my head but it's really helping people to understand. And I'm sorry, I keep coming back to this, but the threat and the risk. So how can we help people to understand, okay, this is how fast a fire can spread. This is what a fire, is going to look like. If you're in the middle of it, here's what will happen if you wait until. Here's how, and obviously I don't expect to immerse someone, but here's how hot these things can get right in the heat. But just to help people understand if they stay behind, this is the scenario that you could possibly be faced with and it's dangerous, right? And as climate is changing, as the fires are becoming more intense as they're, the winds are picking up and maybe fires are spreading faster or burning more land area. This is what you need to know, not just about the fires that have occurred in the past, but this is what you need to know about the fires that are likely in the future and how they could possibly affect you. And what keeps coming up for me is consequences. And this is what they have known in risk communication for years and years. If people do not understand the consequences of these events to themselves and their families, it's hard for them to get to a place where they're going to leave early. Not wait and see,

One thing that's very relatable to me. I wonder if he ever investigated this. In Poland, we had a lot of Polish tourists captured by the fires in like Greece or Turkey this year, some years ago in Italy. Where. You basically, you could be prepared for a fire at home, but you're then exposed to completely different environment, different challenges vegetation that you are unfamiliar with the five dynamics you may have never seen, even if you witnessed a crown fire in Poland, that's not the same as bushfire in Turkey. Enrico in our interview also brought that we need to understand where are the hotels, who are in these places, and to what extent that changes, because if you are talking about the inhabitants of the wild land, urban interface, it's people who can train, but tourists, it's not possible to train them, but yet you have to account for them in your preparedness. And it was this something that you've studied or maybe again, this is something that can come from hurricane.

Yes. Yes. And there have been some studies in the bushfire area about tourists and actually work in Australia that came out of the, quite intense 2019 - 2020 fires. That study to talk to not just people who lived in the area, but visitors as well. And I'm doing some work on bushfire maps. I'll come back to that in a second. But yes, visitors, tourists are very important, as you mentioned, and as Enrico mentioned to understand where they're located in your community, because that goes back to some of the targeted messaging that you and I talked about in the very beginning. Is to understand who is where, and that there might need some additional prompting or additional information. The Chimney Tops Two fire that you asked me about earlier happened in an area where not only are their resident residential, but they're also cabins and places that people rent out and also a hotel as well. And we were kind of thinking about, well, can you, as soon as you enter this place, whatever it is, you have fire information, maybe on the back of the doors or something of your hotel you're in fire season. This is what you need to know. These are your routes out because that's the issue. It's not just awareness and threat perception, and risk perception, right? Which potentially could be lower if your visitors and you don't know the risk, but also once you decide to evacuate, where do you go? If you're not as familiar with the routes and in some of these locations, it's quite windy, mountainous roads which. Difficult to navigate, especially in smoky or hazy conditions. So it's important for people to understand their threat, their risk, but also what do they do once they're told to evacuate which is really important for visitors. The one thing that I have been working on with a colleague in a geospatial in the School of Science here at RMIT is bushfire maps. Because you go from one state to another and they use different maps. The people may be very familiar with the map in their area. They go to a different place. It's a completely different map. How are they using them? How are they making sense? And so we're, we're thinking, we're looking at that and trying to understand how people, both residents. And she has a survey that's asking tourists how they use bushfire maps and make sense of their own threat and risk and decisions based on that. So I'm very excited to see what comes out of that work. But we've also been talking with different organizations here about standardizing some of these maps, right? So when we're talking about visitors going from one place to another, it'd be good to keep the information that they receive as familiar as possible to them as they go from one place to another. Targeted information, familiar information, information that they are expecting and that they know what to do with when they receive are all really important pieces. And then as you mentioned, the officials knowing kind of where they're located. So if they need additional assistance or prompting, they know where to go.

Okay. Now I would like to lower you into the region of computer models and modeling this because it's something really interesting to me. And I know that a lot of people in the audience are interested in modeling actually, behavior of people in fires. And this is something that crosses from wildfires to building fires all areas of fire engineering. So tell me, how are you modeling this? Is this, modeling of decisions or is this just some stochastic distribution of times or events that come to tell me a bit more on, on how the model did

Absolutely. Yes. And I'm in really enjoying all of these great questions because very, very important questions. So yes. There are a range of different models that are available. So in some cases, we are modeling this at the aggregate level where we're interested in a distribution time for delays of households, everything, you know, is at the household level. And we're interested in the relationship between density and speed on the roads based upon the road capacity to calculate evacuation time.

Sorry, to drive to parallel to a building. That would be what we usually do in the buildings with pre evacuation time distributions and fundamental diagrams

That's right. Exactly. That. And I don't know. Did Enrico, did you guys mention with, with Rino. Do you focus on buildings or did you also talk about wildfire? Everything, everything under the

I'm really surprised with not going to evacuating space colonies in that chat it was fantastic. So we've touched a bit of everything.

that's great. Okay. Well, I'm working with, Enrico and international team on the WUI-nity model. And so that is right now an aggregate model, but what's great about it's a platform that can plug and play, right? So more sophisticated models can be included into WUInity if we wanted to do like agent-based simulation. So WUInity he is one of those unique models in that it models the fire, the pedestrian movement. You know, movement from the house to the vehicle, for example, and that uses the movement speeds that we're very familiar with and building horizontal spaces and then traffic behavior by traffic movement, if you will. I have to be careful with my use of behavior. So this is capacity of the roads and you know, speeds based upon, as you mentioned, the fundamental diagram, and here we're very interested in calculating evacuation time. There are other types of models. There are more sophisticated models. One at RMIT that's been developed by Narendra Singh, for example, that is using agent-based modeling techniques, to try to predict what people are doing based on the environment around them. And so that goes from one range to another and everything in between. The problem with the models, I would say at the moment is they have to make. Number of assumptions or, obviously we're all probabilistic in that. We, we're not going to say, everybody that does, this is automatically going to do this. There's probabilities involved, but there are other assumptions that we have to make because we don't have any data on that. For example, route choice, the models that I know of either you define it as a user, which is predetermining the, the outcome before you run the scenario or it's show shortest or fastest route. But we know from hurricane research, for example, that people are very likely to use a familiar route. So it's understanding some of these behaviors is what I'm trying to do actually at RMIT is to collect data everywhere from the evacuation decision, all the way to the destination. Cause there's a number of decisions that people make in between those two to understand the behavior associated with that and start getting some of that into the models, even the agent based models that make some assumptions. But I want to talk a little bit too about why we care about evacuation simulation and link it to buildings. Because, that we use evacuation models and buildings to compare with fire models, or maybe we use a model that incorporates both, although they're rare to under I'm sorry.

for compliance.

rightfor ASET / RSET. Exactly. But Enrico and his team, before I joined their group it's a NIST funded project that WUInity is looked at ASET and RSET for wildfires to come up with a trigger model, which is fantastic. And also this is work that Tom Koba has been doing for years with trigger. And so the point is to run the evacuation simulations, to get an idea of the range, what's the best case scenario and what's the evacuation time. And what's the worst case scenario and this range of evacuation times, and then running a range of fire scenarios and understanding under various scenarios. How long will it take. This community to evacuate so that they know, right. We can put kind of a trigger boundary around a community whereby once the fire crosses that point, , some particular type of fire scenario, we know that the, that this community needs to be warned because we know how long it will take the fire to get to the beginning of the community. We know how long it will take this community to evacuate. And we want to make sure that they're evacuated before the fire gets to that point. So the purpose of these models is to help develop this trigger boundary around communities. It could be one, it could be a range of boundaries understanding that for different fire scenarios, there is some kind of buffer whereby once the fire crosses, the community needs to be warned such that they can evacuate in time.

I think in buildings, it's easier because the mere existence of the fire is already the trigger. And as soon as you have fire in a building, you would like to evacuate it,

exactly right. That's exactly right.

but it, actually becomes more interesting when you go to supertall buildings where, you know, you cannot evacuate the whole building at the same moment. And in this case, such an approach like you've mentioned to assess the threats. Particular groups within the building, you could treat like each floor of a building as a separate community there

That's right. And they do that for phased, for phase evacuation, or, that's evacuate the fire, the area of fire origin, two floors above and two floors below or something like that.

that's great. And, when developing this, types of models, it seems quite, let's say complex, you have to figure out why and when people make a decision to evacuate, will they leave immediately or they need time to prepare something and it's not a simple as and building to take your bag or laptop and run away. It's your home, right?

That's right.

Make a choice. If they run away on a horse bicycle or by car or public transport, right. Where do they go? How do they get there? What did they do when they reached the destination? So that's much more complex than leave to the nearest exit and just wait outside. Right?

100%. And I'm glad you asked that because it's those, all of those decision points in between, right? So the fire starts, or maybe people are warned and then not only do they have to make a decision, they have to decide when they're going to leave their home. They have to decide what mode they will use to evacuate. It might be that a household uses multiple vehicles. It might be that the household uses one car and then they tow their boat or they have a trailer or they're using multiple vehicles. It might be that household needs to use public transport. And then you choose the route, right? If you have a personal vehicle, you have to choose where you're going to go. And it could be that, maybe a route is blocked or it's congested or they use the one that's most familiar. And then finally, as you mentioned, they have to decide where they go for safety. What's even more complicated is an aside from modeling. All of that. It's important to model something that's called intermediate trips. So people may, leave their home for trips that are not related to the evacuation. So what that means if they're still in the fire affected area, but maybe they're picking up supplies from the hardware store, or maybe they're going to pick up their children from school, or maybe they're going to drop off their livestock somewhere or something like that. And it's important to account for those types of trips, but there's been almost, no, I think I've found a few studies on that related to wildfires. What are, what prompts people to take intermediate trips? How many will they take? Those types of things are all very important. And then also above that is, modeling background traffic. So the people who are already traveling through the area in fire effected area before, you know, an evacuation is called as well as shadow evacuation, which is something that I'm grappling with a team also funded by nest led by the University of Florida, where we're looking at GPS data to understand evacuation trajectories from that same fire in 2019 called the Kincade Fire. And we're trying to define all of these different individuals. Are they evacuees? Or within the fire effected area or are they shadow evacuees. That's a term used in hurricane evacuation quite frequently, and that these are people outside of the evacuation zones, but they evacuate anyway. And what that means is there's more potentially, people on the roads, more people in the traffic network that you should be taking into account when you're trying to model all of these different, factors and scenarios. So it's complicated, but not impossible.

so in our building parallel, that would be the guys from the other floors where There is no fire. Yeah. They just chill out and block the staircase. Not very helpful, but, uh, kind of understandable that they do that.

Yes. I mean, yes they might feel that they're at risk, even though they're not on an evacuation floors, so they may decide to evacuate anyway, is that exactly?

And to develop these models. What kind of data you need to develop that? Because you've mentioned already, GPS data, you've mentioned the, questionnaires that you give to people and the open-end questions that are probably the most valuable of them. So what else is in your toolbox to calibrate these models?

Yes. So definitely the surveys. I also think that interviews are also important to do with surveys, something I haven't had a chance to do as much, to then provide a little bit more richness in the why's the, how the why. Right. We ask people to answer multiple choice, questions, and surveys, and don't provide much in the way of open ended questions. And as you mentioned, surveys and interviews are individual or household based data, which is good because a lot of these models were trying to simulate decision-making at that level. The GPS data that I mentioned provides more at an aggregate level, just trends, right in evacuation rates evacuation departure times. So that distribution curve or the S curve that people talk about in hurricanes, which is well that's actually for the evacuation numbers too. But for departure time, we want to be able to understand over a certain period of time, when do people enter the traffic network? And then route choice and destination choice is something else we can get from GPS data. But again, that's an aggregate level. We don't understand. Okay. Maybe this unit household, we know everything about them because we don't we know maybe where they're initially located and where they go. Other interesting data could be from social media. And so I'm, I know of some, different studies that have looked at Tweets or Facebook to try to not only understand geo location and as people are tweeting or using Facebook to understand those trajectories, that GPS data can give us, but even extracting information from the tweet, see if we can get information about what they were thinking and what they're doing related to evacuation. That's quite difficult. And speaking with the team at university of Maryland, about some interesting work they're doing with Twitter data. And then lastly, well, not lastly, one of the others is VR and AR, which I'm really interested in. We've done that a lot for buildings as I'm sure you know, and a lot of work at Lund university and University of Canterbury and the work at Massey that Rino's doing and to understand route choice or evacuation time or exit choice. Max Kindateder also NRC, Canada, but we haven't done this kind of work for wildfires. So they use VR and AR for training. Even wildland fire firefighters, which I've seen some of this, the simulations, they're amazing. but we're not using it yet as much, or if at all, to collect data on what people will do. So I'm very interested in digging into that and working with Rino and others too, because the interesting piece about that is you can collect individual decision-making data, right? Immersing people in a particular scenario, and then talking to them afterwards and asking why did you do this? And why did you do that? A lot of the work that's coming out of VR. It's also testing how real people feel the scenario is to get a sense of, okay. This is a simulation we need to maybe treat this. There's a caveat associated with this data, but also, this simulations are becoming more and more realistic. And I think are very useful to collect data, for use in simulation tools.

And what would be the outcome of your modeling? Because when I do my evacuation modeling, my outcome is very simple. That's my available safe evacuation time. And that's pretty much the one thing that I'm looking for in case of modeling these complex behaviors, you are more looking into like outlier scenarios or you're seeking like probabilities, or maybe probably density functions.

That's it. I'm looking for probabilities. I'm looking to understand under which types of scenarios, , people are more likely to do X, Y, and Z, so that we can actually assign an appropriate evacuation time to a community or assign at least a more realistic range of times. So, yes we're trying to be able to predict the likelihood that someone will evacuate and even the time at which they will evacuate. Rino has a model that he's developed from our Chimney Tops Two data that tries to predict the likelihood that people are in a particular state normal state investigating state, vigilant state, those and evacuating state. But yes, it's the probabilities associated with all of this. And the goal is that we can. We don't expect the answer from these things we want to range. And then we can start to identify, as I mentioned the kind of ASET - RSET, those trigger buffers. We want to be able to identify and use these models, maybe in a best case scenario to have to identify some of the best routes that people should choose for evacuation planning. We want to be able to make decisions on when people should evacuate, both during an event and thinking about that for planning purposes, we want to be able to test the plan and put it under some worst case scenarios. But again, these are all, our estimates from our modeling we're going to be able to predict the likelihood, that someone's going to be able to evacuate.

So in the end, if the model works and, the outcome of modeling would be essentially the most effective strategy to get this population out of this place in the, let's say most optimal way,

That's right. So,

circumstances

so that's kind of two things. So we can use the models to predict some of these optimal scenarios and use that for planning. So if I was to tell you, what is your most optimal route here it is. But when we're talking about actually making decisions during the event, we don't want to be simulating optimal behavior. We want to be simulating actual behavior. And that's when we try to put in an understanding of the fire scenario or a range, try to get realistic as possible evacuation timing so that we know when to warn. And so I think the optimum. It's very good for planning, but some of these worst case scenarios are all also very important to stress the evacuation plan to come up with traffic management solutions. Ask what if scenarios? What if people all go to the familiar route? What are we going to do? Right. So not only planning for optimal, but planning for some of the worst case so that we're ready to know. Okay. Here's some of the solutions, here's what we're going to put in place. If some of these worst case scenarios occur. And then of course you need to make decisions during an evacuation about one to warn and that is based on actual or predicted as close as we can get from these models.

Yeah, I've asked you that because, I'm wondering if one day we could use this approach in buildings and if we did, how could we use it? And I think it's not really very interesting to use it, to get my ASET time more precise, but it's very interesting to actually identify the critical scenarios in which something very bad can happen and then proactively prevent these scenarios because this is how you could potentially save lives.

That's right. Stress, the plan. Absolutely. And Steve Gwynne and I wrote, for, the fire magazine here in Australia to try to think about some of the worst case scenarios that could happen in wildfires. But what we did was we used the Handbook, Daniel Nillson and Rita's chapter on developing occupants scenarios. And so we were using what people are thinking about as far as developing occupants scenarios for buildings, and then expanding that for fire events and communities. And that's just another example of how our knowledge and fire protection can be translated into the WUI space.

That's brilliant. And that's what I try to do here with this podcast, exposing building engineers to wildfire science and evacuation scientists to principles of combustion. And I think it. creates a better environment overall. Erica, um, I'm so thankful for this interview. We're going to wrap up. I see. We can keep talking for hours about this, but it's really it's really a huge, huge topic. We haven't even touched the PADM models and some more in depth things,

Well, to be fair, we did in that, when you talk about risk perception and threat perception, that's like the big foundation of the PADM, but maybe we'll just have me back in a year and we'll talk about

Now maybe in, maybe in a month,

sounds good.

W we'll see. We'll see. So thank you. Thank you very much for coming here and sharing your experiences. What are the next steps for you? So now you're in Australia, you're probably focusing on bushfires preparedness. Tell me what are your nearest plans?

Yeah, that's thanks for asking. So we're hoping to continue the modeling work that that we're doing with Enrico and Steve Gwynne and Max and others NFPA foundation, Guillermo. And so I want to continue this, the simulation route, but I also want to very much focus on the data collection. And so I have a new PhD opening, a scholarship where. That candidate will be collecting data via surveys, interviews, NGPs data, perhaps the hope and trying to kind of use different methods for the same fire or the same set of fires, to start filling in the gaps of, that evacuation timeline, right? All of those decisions. I'm very interested in getting data obtaining additional data on a lot of those decisions along the timeline. And then we can feed those models into WUInity and others. So primarily data collection, I'm really interested in kind of building a great team here and working on evacuation related to bushfires. I'm also interested in floods as well. So there's a lot of. overlap of people's decision-making in different types of disasters and looking to use interesting data collection techniques that we haven't used in the past, like GPS, like, AR VR, like social media data and yeah. Trying to make sense of what people do in these scenarios. Just not just that evacuation decision, but all of the rest of those decisions

That, that sounds like an exciting plan for many years of research.

I know. Right. There's a lot to do.

there's a lot to back. Well, I wish you all the best in solving this and creating even better models. I really appreciate , that the models like WUInity are open source and available to everyone.

Freeware.

Yeah. Freeware. That's great. And then certainly something to appreciate and thank you for sharing your thoughts in, in the podcast.

Well, thank you for having me. This is really fun. And I enjoy talking to you about a topic I'm very passionate about and thank you for all your great questions and giving this topic some air time.

Yeah, I'm very happy as well. I'll definitely see you here in sooner than later.

Sounds good.

Thank you so much. See around.

Okay. Bye.

And that's it. When I'm doing interviews with people who are as passionate as Erica about their research, it's really, really tough to stop discussing the subject of matter. And I'm really happy. I had her in the podcast. I'm really happy, she shared her enthusiasm with all of you, and I hope you picked up something useful from the discussion and for your own practice. I hope you noticed how she d iscussed these aspects that go into modeling behaviors. And what I'm really happy about is that she explained how this behavior modeling can be used in practice in terms of analyzing the threat levels. Preparing scenarios, creating some boundary values after breaching witch and evacuation must be performed of the community, for example, and I'm sure that plenty of this can be also used in modeling evacuation in buildings. Hopefully one day. As I mentioned in the intro able to do this paradigm shifted into behavior, fueled evacuation modeling rather than just movement-based agents models. That would be a great, great future. And I'm looking forward to that. For the end, some other news from the podcast we're approaching episode 20, and it's going to be a great episode. I have a world-class scientist again with me, and we're going to discuss the future of fire safety engineering. So that's going to be really, really interesting. I have also just launched a new website for the podcast. You might want to check this out if you have not seen it. Uh, www.firescienceshow.com and, you can pop comments on there. You can obviously listen. All the episodes, read the show notes, read the transcripts. Hopefully it's useful for all of you, you can also sign to the newsletter. Um, I'm gonna try to do a little bit better job with maintaining the newsletter. So probably that's something , you should or could do. , I will appreciate that. And yeah, for the as usual, please share the knowledge about the podcast with your friends, with your colleagues in your officers, with other professionals, you know, that may benefit from listening to the interesting discussions we're having here and all of you. Thank you so much for tuning in and see you next Wednesday.