263 - Developments in modelling WUI fires with Arnaud Trouvé

Wildfire Season And Guest Setup

Wojciech Wegrzynski 0:00

Hello everybody, welcome to the Fire Science Show. Somehow in the summers I usually drift the podcast towards the wildfire and the wildland urban interface fire problems. I guess that's the flavor of the season. Unfortunately, we've seen smoke covering a football stadia in the US a few weeks ago. We've seen terrible fires in France, fires in southern Europa. It's kind of horrible how the world drifts towards bigger and bigger outdoor fires. Well, on this note, while the problem is on the rise, we also have a great group of researchers devoted to this type of problems in the fire science. And uh today I'm highlighting contributions from Professor Arnaud Trouvé from Maryland, who was also kind of distinguished by the Combustion Institute and invited to deliver a plenary talk at the recent Combustion Institute Symposium in Kyoto. I was actually last week and uh Arnaud delivered uh his keynote there. It's a big thing actually, the Combustion Institute Symposia are huge conferences. I think that's a conference which gets like two and a half thousand papers submitted to it is madness, it's giant. But uh having a fire researcher being highlighted with a plenary lecture, that's a great distinction for Arnaud, but also a great distinction for the fire science and the importance of it. Arnaud's research at UMD is touching the world and urban interface itself, so modeling fires which transition from wildlife, from forests into into the cities, and the specific challenges related to modeling that transition and that spread of fire through communities, an increasingly important thing for building our ability to understand the mechanisms of the fire of those fires and building resilience uh for the communities which are threatened by those fires. So a lot of important stuff in this episode. And outside of just modeling, I think Arnaud in this interview really captured well the distinction between having a pure wildfire, you know, like a forest of bush fire, and a fire which enters the community. I think in this in the way how he models them and the in the way how he looks at the problem, the distinction is as clear as ever. So that's something to look for in this conversation. But anyway, uh I think that doesn't need much more introduction. It's very relevant to all in the field, and unfortunately, it's the current reality if you look at the news right now. On that note, let's spin the intro and jump into the episode. Welcome to the Fire Science Show. My name is Voyage Vingchinski, and I will be your host. The Fire Science Show podcast is brought to you in collaboration with OFR Consultants, a multi-award-winning independent consultancy dedicated to addressing fire safety challenges. OFR is the UK's leading fire risk consultancy that this year celebrates its tenth anniversary. As experts in fire engineering, they are fully committed to delivering preeminent expertise to protect people, property, and the environment. With over 30 chartered engineers and a team of fire researchers at their core, they continually explore the challenges that fire creates for their clients and society so that the best research, experience, and diligence can be applied for effective tailored solutions. In 2026, OFR will grow its team once again and is keen to hear from industry professionals who want to collaborate on fire safety features this year. Get in touch at OFRConsultants.com. And now back to the episode.

Why Combustion Still Matters

Wojciech Wegrzynski 4:14

Hello everybody, I am joined today by Professor Arnaud Trouvé from University of Maryland. Hey Virgin. Good to see you. I know you're getting ready for a very hot uh Kyoto in uh just a week or two, uh, where you are the plenary speaker at uh combustion symposium. Congratulations, Arnaud! That's that's that's a big thing. Thank you very much.

Arnaud Trouvé 4:37

I'm I'm looking forward to this, and I hope I can do a good job at representing fire research.

Wojciech Wegrzynski 4:41

Yeah, let me help you a little bit practice uh what you want to tell there by by explaining this to my uh my fantastic audience. I mean, I just have to recognize the fact there's uh a fire plenary speaker at the combustion symposium, is it's really reassuring. We're also a significant part of that world. Uh a few weeks ago, I had enormous pleasure to interview Professor Bart Mercy uh regarding uh his combustion model developments after his award uh acceptance uh talk at the IFSS outstanding talk, really. Uh I think we can agree on that. Yes. Wow. Uh but anyway, some people are were asking me after the interview and like what does it mean to us? What does it mean to engineers? Like, does uh what's changing? What's the large-scale image of this progress? We spent a lot of time, and Bart could talk for 10 hours about the the the details of the progress in the models that he's developing, the class of models he's developing. But from the bird's eye view, how does the progress in modeling combustion impact fire engineering?

Arnaud Trouvé 5:54

Yes, this is an excellent question. And I I should uh um first mention that I came to uh fire protection engineering when joining the department at the University of Maryland in 2001, and I came from combustion science. So I thought combustion is everything. Okay. And of course, fire is first and foremost of a combustion problem. However, uh you have to become very humble after saying this because you realize that uh in modeling, most of the time, researchers and practicing engineers for sure are using design fire curve where the combustion problem is bypassed. Okay. And for typical design calculations, this is sufficient for most of the time. And so this uh brings the question of what is the value of combustion research with some of the topics covered by Dr. Mercy. Well, I think there is value, of course. And uh the the value first, even when you are using a design fire curve, practitioners very often do their safety calculation based on quantities like smoke visibility. Well, that means we have a soot problem. Well, a soot problem is a combustion problem. It's actually a combustion problem that is unsolved in the combustion science community, so we don't have a good solution for that. But one of the main quantities of interest to fire engineers, which is visibility through smoke, is actually a byproduct of combustion. So something that we haven't solved, at least in my uh career lifetime, I would say is the suit problem. But that's still an open question. It is a combustion problem. Now, even if you put the suit aside, there are, of course, a number of projects, and I'm thinking about forensic uh type of problems, uh, fire investigation problems, where you are not going to be satisfied with design fire curve. You're going to try to reconstruct the growth of the fire, the credit of the fire from room to room if you are looking at a building problem. And so there are applications where you want to go uh you know deeper than what you're typically doing for design calculations, uh, because you have a kind of unique project that requires it. But you know, I I think I would agree that combustion is uh often bypassed in uh fire models, but you you see it coming up uh in different ways, and I think the smoke is the best example where we would like to understand how much smoke is produced by a fire, and that's that's a combustion problem.

When Design Fires Stop Working

Wojciech Wegrzynski 8:30

Well, yeah, I I I agree with that completely. And myself, uh every now and then you reach a project at which your design fire is just insufficient. It's like so obviously wrong, you know, the design curve that you put into the building, a standardized one, let's say, you know, uh, that uh you start questioning it. It's it's just you know a tool with within uh a tool set. And I also see a lot of people try to scale up, you know, they do some sort of pyrolysis modeling in FDS to have some sort of a spreading fire, and that gives opens up new possibilities in in modeling. Uh and suddenly this is extremely challenging because how how can you be sure that the model that you've developed is is valid, that that your solution is close to reality, right?

Arnaud Trouvé 9:20

Right, that that's a challenge. But the other example I wanted to mention is uh the problem that I'm going to talk about in Japan, which is the the problem of wild underburn interface, fires or urban fires. In that case, you are in a quite unusual situation where the fire in a given home is going to allow to burn from ignition to burnout, usually without firefighter intervention. So you have a very long fire event, or even a few hours sometimes. And uh you're going to uh have a problem on the modeling side and also on the engineering side to uh really uh frame your problem to come up with the equivalent of a design fire curve, but now at the level of the entire building and the entire structure. And you're going to have several tens of megawatts of fire for an hour-long uh event. And uh, we don't have a lot of experimental data on this because this is an unusual problem. And uh modeling is also um can uh provide insights on uh the heat release rate produced by a burning structure. But again, you have to now uh use uh fire growth and uh and room-to-room fire spread techniques uh that are going to require something that is more sophisticated than what we're using in the usual design fire curve in fire engineering. So that's again, that's that's a combustion problem, and it's a combustion problem that will include underventilation. And uh, you know, that's uh to echo what Dr. Mercy was talking about. When you have underventilation, then you can have flame extinction. Yes. And then uh, you know, your combustion problem now has become much more complex. And here underventilation is very important uh in structure-to-structure fire spread in uh we in our urban communities because it's going to control the size of the flame that is ejected from windows or from doors or from roofs, and so that controls in part uh spread from one structure to the next. So these are issues, you know, these are combustion issues, these are issues that we are trying to frame and to understand. I mean, it's interestingly, in the case of we and urban fires, we are still at the level of trying to agree as a community of what the problem should look like and what where we should put our efforts. And so that's part of the discussion.

Wojciech Wegrzynski 11:49

Uh two two notes. One uh I was uh when you were mentioning the design fire for a building and how you have to model from compartment to compartment and all the you know tiny details that's gonna happen in in inside of it. For me, an equivalent that draws in my head if I model my compartment fire, that's if I you know model the crib uh stick by stick with the gaps in between. So that's the level of detail I would have to go in my compartment fire to to match this uh you know kind of a problem that you have now. Instead, I have a design fire, not model my crib. But the the the class of models that you're investigating is not yet there to have like uh a household uh you know model. You you have to go stick by stick with in here, meaning compartment to compartment. And the underventilation that you brought up, this is a very fundamental problem. And indeed, when we were talking with with Park, this also came up as one of the main goals of all of this. Because if you have a design fire curve, it doesn't care if it's underventilated, overventilated, you put the curve, it's not responsive to the surroundings. But uh in reality, you will have that, and we know that it's changing a lot of stuff in the combustion. I guess we can discuss this further. I remember a paper, um I think it won the Biggestone Award, 20 large uh scale informal settlements on fire, and they have burnt 20, like in layout of four by five informal settlements. And I remember in the middle of them, they they were kind of underventilated because the oxygen was eaten by the surrounding settlements. So already at the scale, not a whole city, but at the scale of this experiment, they observed some sort of that. Do you think it's an important phenomenon in urban scale uh fires, conflagrations when we reach these houses that that you suddenly eat a lot of oxygen in your surroundings to the level where it starts uh influencing? Or you meant more like uh global equivalence inside the buildings that changes?

Arnaud Trouvé 13:51

Right. So I think it's important to differentiate between indoor fires and outdoor. In indoor fires, when you have on the ventilation, let's assume you have a fire that starts in the room and and the windows and the doors are closed, then uh there may be a little bit of uh, you know, oxygen pathways through leakage paths, but um, but not enough to sustain the fire. So the fire will basically self-extinguish. So that's important. If the door is open, then the fire will actually be fed with supplied with oxygen and will be healthy. So ventilation controls basically the heat release weight in that case. Outdoor fires, when you have a lot of fuel that is generated at a given location, maybe by several structures or a large area of wild biomass that is going to start burning, then yes, you don't have access to oxygen at certain locations, but the flame, the flame geometry will change. So that's what it means. You can still have basically a flame that will burn all the fuel that is released by uh thermally degrading fuel sources, but the the geometry of the flame will be different. Okay. So I think to me, this issue uh sometimes I read in papers people mentioning underventilation for outdoor fires, but I think it's quite different. It's actually you're going to have an impact on the flame geometry, but not an impact on the heat release rate. Uh I'm not aware of evidence that actually outdoor fires can uh you know uh be uh significantly affected by extinction because of not even enough oxygen. So I tend to differentiate between problems where the flame shape is different and problems where the heat release rate is different, it's not the same.

Wojciech Wegrzynski 15:39

Also, I think it's important to account that they would mostly happen in a strong wind.

Arnaud Trouvé 15:44

Right.

Wojciech Wegrzynski 15:44

That that's when I hear about catastrophical urban fires. They have they usually happen uh with some strong wind present, or if they're strong enough, they create their own wind. And and if you have a few meters per second, that indeed is that's a lot of cubic meters of air coming through. I but I I agree. Like I've never seen uh maybe I'm limited in in my literature studies, but I've never seen someone actually quantify like you know equivalence rays or or uh oxygen concentrations in proximity of those large experiments, especially that it's very hard to carry those experiments.

Arnaud Trouvé 16:20

Yeah. I think the the the best experimental analogy that I have seen on this problem is uh the wood creep fire done by uh uh Dr. Sarah McAllister in Missula, because she has a regime where, for example, the wood crib burning at an elevated position with air coming from the bottom, then you would have individual flames around the individual wood sticks. Okay, so you have well ventilated. And you put the same crib on the floor, and there's less oxygen coming in, and then you have now a flame that sits at the outside of the crib as an envelope around the crib, and that's so from a compartment fire perspective. If you define the compartment as the wood crib, you are either overventilated or underventilated. And the flame shape changes, right? And uh and the heat feedback to the wood sticks also changes, so there is a kind of an impact on the heat release rate that way, but the flame is still alive, okay? So you don't have uh the same thing as in a compartment fire that doesn't have oxygen coming in.

Wind Firebrands And Structure Loss

Wojciech Wegrzynski 17:24

Let's look further into uh urban and wild and urban interface um fires. In modeling those, how different is is the problem from the perspective of combustion and and broader speaking, from the perspective of modeling the fire itself compared to let's say compartment fires? I think a compartment fires in this interview could be a good reference point because I assume uh most of the uh people in the audience would be engineers who have done some FDS calculations or at least you know are exposed to that. So I think compartment is it's a safe space to return to for us all. So how how much uh how much it differentiates when you go outdoor?

Arnaud Trouvé 18:05

Well, an obvious difference is the presence of of the wind, right? So so the wind plays a dominant role as you pointed out. Usually fire disasters are related to extreme uh wind events, and so the wind plays a major role, and uh that's something that you have to factor in into uh your your thinking, your analysis, and your simulations. But uh the other obvious difference is the role of fire brands. Uh so these fire brands basically they don't have much of a role in uh building fires, uh, but they play a dominant role in wide-on-urban interface or urban fires. They are often thought to be the dominant factor of uh structure loss or fire penetration into target structures. And the examples of this is that when you see these images showing a structure that is completely destroyed by the fire, but it is surrounded by a garden that is intact. Okay, and uh that's a good evidence that uh this structure was not thermally exposed by uh radiation or by uh fire, flame, what is called flame contact, which is convectivity transfer, so but was basically attacked by firebrands that have uh you know basically ignited something in the close vicinity of the structure or have penetrated directly into the structure through open vents. And uh so the role of firebrands, which is probably something that is um not very well understood today, is very important and is a special feature of these outdoor fires.

Wojciech Wegrzynski 19:41

Uh is it very different when it's purely wildlife fire versus when it uh touches the city versus it's just purely a city conflagration or or those uh problems kind of intertwine uh each other?

Arnaud Trouvé 19:55

It is different because you do have firebrands in wild and fires, but often uh these firebrands uh are going to be only able to survive in flaming or smoldering mode where they are dangerous uh at short distances. And so you have a lot of visualization that shows these firebrands flowing and uh in front of the flame and landing on the ground, possibly ignited vegetation, but you see that they are just a few meters ahead of the main fire front. And so their role is minimal because uh basically the firefront will propagate to that location before these firebrands have any significant dynamical impact. And that's different in the you know, in the case of structures, these firebrands are going to uh be able to ignite uh fires in the vicinity of the envelope of a building and uh are often responsible for fire penetration and fire loss.

Wojciech Wegrzynski 20:54

In modeling these types of fires, what what scale are are we talking about? What's the scale of the problem of interest? Is it more like what you just described, the ignition problem and just modeling how they trigger uh house fires? Is it like uh a few houses interacting with each other? Or is it like you would like to mount in the whole model the whole mountain range with with cities in it?

Arnaud Trouvé 21:19

Like yeah, well, thank you for this question because this is uh uh one of the main messages I want to convey uh in in Kyoto is uh we have a multi-scale problem. A lot of these scales play a role, and we have in fact different modeling tools to uh capture these scales.

Three Scales And Three Toolsets

Arnaud Trouvé 21:37

So we don't have a single tool that can go from the community scale to the uh indoor structure scale. So you have basically three types of scales, in my view. Uh, you have the indoor fire scales that again are important to explain how much heat is going to be released by uh a structure that is burning. You have the power. Soil level scales, that means that's an outdoor fire scale where you look at one structure or maybe a few structures at the most, and you look at the immediate environment around these structures to understand vegetation to structure spread or structure to structure spread. And the most important thing is fire penetration. How does the fire transition from being an addoor outdoor fire to being an indoor fire? And that's something that is not enough appreciated, I think, by our research community, where if you look at structure ignition investigations, typically people look at the start of a flame outside of a structure. And when you have a flame outside of the structure, you know you are in bad shape. Okay, because there is supposedly usually nobody to extinguish it. But you still have to continue the analysis to understand how this small flame grows and how it's going to penetrate inside the structure, maybe by breaking a window, maybe by you know igniting some combustible siding if you have combustible siding or combustible roof. But you need to understand this fire penetration that transitions the problem from an outdoor fire to an indoor fire. So back to your question, you have indoor fire dynamics, that's one scale level. You have outdoor or parcel level scales that explain how fire penetrates from outdoor to indoor. And then you can zoom out, you have the landscape scale, the community level scales, where you look at the interaction of the fire with the local wind, with the topography, things at larger scales, which are usually the scales at which fire risk tools used by practitioners who operate.

Wojciech Wegrzynski 23:44

I love especially that uh indoor scale and uh the example that you've given. I I have kind of similar observations in relationship to vehicle fires. Like uh a lot of people quantify that ignition of subsequent vehicle in a car park fire happens when I don't know, a bumper trim ignites or a gasket ignites. Well, that doesn't mean yet the car is on fire. Yes, there is a flame that's somewhere living on that vehicle, but it's not yet, you know, contributing eight megawatts to the overall fire. And then in the same in here, if you have a little fire on your porch, in consequence, yes, it may mean that the house is gonna be lost, but in that moment, how much it contributes to the entire wildfire or urban interface fire is insignificant at that point in time.

Arnaud Trouvé 24:36

Yes, I agree. And um this actually um, I'm sorry I'm going to preach for my choir, but this actually is this way of looking at the problem is helped by uh what people call uh sometimes computational thinking. Because when you are putting a model, a computational model together, you really have to put an entire puzzle together. So all the pieces have to be there. Hopefully they kind of fit together, but at least all the pieces have to be there. And when you think about a computational model to describe the entire dynamics that you find in this conflagration problem, then you realize that uh the ignition of a small fire in the vicinity of a target structure is not actually the time at which you can use a design fire curve for the full structure. The design fire curve for the full structure describes already an indoor fire as being nighting. And so you suddenly realize that you have a problem of linking the small flame outside of your target structure to uh an ignition of something inside. And so that leads you to realize that you have to actually understand fire penetration.

Wojciech Wegrzynski 25:46

Now let's talk about the models. Like for all three scales, uh what kind of tools are you using to model the fire? Is it all CFD models or is it different models per scale? Uh and how do you link them?

Arnaud Trouvé 26:01

Yeah, so uh we use CFD for indoor fire dynamics, of course, that's a classical approach. We use CFD also for parcel level analysis. Uh I would say that the CFD approach here uh is still missing a few models to actually describe completely the problem, especially at the level of fire penetration. So a lot of CFD studies at the parcel level to date have focused on calculating the formal exposure on a structure due to a passing fire, or but it doesn't calculate the response of the target structure and the fire penetration yet. And so that's we have challenges there, but it's still a CFD analysis. But when you go to the community scale level, then we use non-CFD models, the fire risk models that we use. So in the US, we are using uh NFIR, Agnar, and SWIFT. There are three groups in the US who are developing models that are uh have uh as applications, uh uh Wii and urban fires. So these models, all of them are non-CFD because we have coarse resolution. Typically, uh, you know, we have fuel vegetation maps at a level of 30-meter resolution. So, 30 meter, you're not going to capture most of the important phenomena relevant to fire dynamics. So you have to use reduce order models. And so that's one challenge. How can we fit the physics of fires into a code that has 30-meter resolution that can capture now the all the details about the large-scale aspect of the problem, the topography, the vegetation, the structure maps, the their footprints at least, and the roads and everything? But you you get everything at 30-meter resolution.

Wojciech Wegrzynski 27:50

So is the knowledge coming from the indoor studies and then through the partial level studies? Is this used to create new models, reduced order models for those uh coarser uh studies? Or what what is there a link? Is there a connection and benefit to to study in greater detail and then create reduced order models later on?

Arnaud Trouvé 28:13

Yes, there is. Uh, but there are also different flavors. There are models that claim to be physics-based. I mean, that's the kind of model I work on. There are models that are going to be more, I would say, data science-oriented, uh, using uh machine learning algorithms or different uh techniques of that kind. Uh uh cellular automata is very, very popular. Uh but these models also at that level of cellular automata, they use some aspects of the physics too. So I wouldn't put a strong line between models that claim to be physics-based and models that do not. But back to your question, yes, there are quantities, for example, at the level of firebrand transport, where you we are going to use probability density functions or travel distance of firebrands that are going to be parametrized in terms of wind velocity and heat release rate of the fire. These PDFs, these probability density functions, it's a statistical representation of the problem. They come out from CFD. And so they are differently informed by CFD. Another example is spot fire ignition by firebrands, use a probability of ignition model that is also a product of experimental investigation and CFD. So there is a transfer of information from high-resolution simulations done in CFD to low resolution simulations done with uh simplified models.

What 30 Meter Grids Miss

Arnaud Trouvé 29:36

Yes?

Wojciech Wegrzynski 29:36

How does it look like a cell 30 by 30 meters? That's a nice plot to build your house, actually, 30 by 30 meters. How does it look from the perspective of the model? Because it's it's just one cell at that point. So so what what can you tell about the the fire phenomena happening there? What are you trying to generate? Average heat release rates, etc.

Arnaud Trouvé 30:00

Yes, so I would say I have to distinguish between the wylon and the uh urban areas. Okay. In the wylon, having a 30-meter size uh cell is relatively okay because that captures, I would say, changes in the in the land cover, changes in the topography. These scales are usually not uh you know small. So at 30-meter resolution you capture that. So you can assume uniform vegetation in a 30 by 30 meter cell and calculate through the rate of spread model the not only the the propagation of the fire, but its heat release rate as well. And you have something that you have an average approach, but that looks okay. In the case of wild uh urban interface uh areas or urban areas, then we are in trouble because you have scales like the size of a structure, the size of uh the separation distance between structures, the size of a road uh that is a fuel break that are not going to be correctly captured at 30 meter resolution. The road is interesting. The the road is either ignored by the model, so it has zero meter size, or it has 30 meter width, which is not actually realistic either, right? So you realize that you have smaller scales in the Wii and urban conflagration problem that require higher resolution. So we need to go better than 30 meters, we need to go to at least 10, maybe 5 meter resolution to be able to capture these scales. If not, we have to put models that do that, but it's it's much harder. So I would recommend going to higher resolution. It is not simple, but remember that unlike CFD, we are solving a problem in 2D when we are spreading the fire on a projected horizontal map. And so it is not as costly as CFD in that sense. But this issue of scale is a problem. I think we would need to have cells that have a size that describes correctly the structure size. So you are quasi-uniform inside that cell, the separation distance between structures, um, even you know, we recognize the importance of what is called in the US zone zero fuel. So the fuel that is within 1.5 meters of your structure, because that fuel, if ignited, is going to break the windows and then provide the direct pathways for fire penetration. And so that means that we have important features of our problem that are 1.5 meter scale. Uh, I don't think uh we are there yet to claim that we should go to that resolution, but that gives you an idea of the size. So I would say the size of the road, the size of the structure, the size of the separation distance between structures, and ultimately at some point the size of the zone zero region are important sizes that one way or another we need to put into our models.

Wojciech Wegrzynski 33:00

How do you consider the the problem of the ignition in that case? Like when is the ignition of your next 30 by 30 patch uh triggered? And does the model consider like the fire moving through that, or just considers the the whole cell being like completely unified, one control volume and it's all burning?

Arnaud Trouvé 33:23

Yeah, the model considers that things are uniform at the level of the cell. So right now, this is a problem. Again, at 30-meter resolution, this is not good. You have a structure that is even 900 meters square or is ignored. Okay, it's well. And so we want to be somewhere in between. I mean, typical structures would have a footprint on the order of 100 meters square. So that's what we we need to have as a resolution. So after that, the there is a model for the the burning and the flame shape of that structure. So you're going to have a model for the radiation. Uh, typically we use simple radiation models that come from uh you know zone modeling uh using a point source model or uh uh or at best uh uh an inclined uh you know cylinder uh solid cylinder called flame model. This type of level of models that have been in the literature for many decades now, but are pretty useful. Uh we use also uh an estimate of the flame shape to calculate the heat transfer by flame contact, and then we use firebrand models where we estimate how many firebrands are released by a burning structure or burning vegetation. We transport them through a probability density function model, and then we land them on the ground and we have a quality of ignition model to decide after a certain accumulation typically whether we ignite to spot fire.

Wojciech Wegrzynski 34:48

How is the gas phase considered in here? Is it like a continuous fluid and you solve it like kind of in a CFD, or you rather use like individual Gaussian plumes for every, I don't know, cell that's burning and just have a summary of those?

Arnaud Trouvé 35:06

Uh no, I would say there is no gas phase uh solution in this one. So it's just 2D. Yeah, it's the gas, there's a wind calculation that is an input to the problem. So you have uh an estimated of wind velocity uh in every cell, but that wind is basically prescribed. It can be time dependent given by a weather forecast or by your uh you know atmospheric science calculation. But it's a one-way coupling. So the wind uh in uh these models at this point doesn't is not affected by the presence of the fire, which is probably one of the main weaknesses. Uh and something again that is not done today in uh in uh one-on-one fires. There are more there is two-way couplings in many models, but something that for in the case of Wii and urban conflagration simulators, we haven't done yet.

Wojciech Wegrzynski 35:59

Which is again interesting because those would happen largely in in wind conditions and also like the air density is gonna change uh downwind. Uh so interesting problem, uh problem for sure. How about the the terrain and and the you know slopes, etc.? How are those accounted? Because that is again an important feature of of uh urban.

Arnaud Trouvé 36:22

Right. So uh I would say the topography is represented by these models the same way that it's represented by one on fire. So we have the topography uh given by uh the uh the the fuel maps. However, the models are often ignored. So, for example, if I look at the probability density function models I have referred to for fire brands, they should be affected by topography, but typically they've been developed on the assume flat terrain, and we use the same uh models for more complex terrain. So this is a big open area, open question where we are not accounting correctly for the topography in terms of its effect on fiodynamics.

Why Rate Of Spread Fails

Wojciech Wegrzynski 37:05

Uh maybe we can move back to another class of models uh which are rate of spread models. Uh how do those differ? Because I know they're also used a lot in practice. So the models that uh that you you talk about and rate of spread models, how much they have in common and and how much they differ?

Arnaud Trouvé 37:24

Yeah, so the the rate of spread models that are used to simulate wine on fire spread are uh a well-established technique. They are pretty powerful, but they kind of assume uh some continuity in the fire dynamics. They don't apply very well to the problem of ignition of structures where you have these basically discrete events. So there is um a problem in uh you know merging, in applying, I would say, the rate of sprint models to uh the problem of uh fire propagating through uh Wii or an urban community. That being said, the earliest example of a fire propagating for a city after uh an earthquake is a model that was developed in Japan by a researcher called Hamada, and the Hamada model uh uses a rate of spread type of formulation. It's not usually framed that way, but you can basically reformulate the model in terms of a rate of spread. And in that case, that looks similar to uh uh a vegetation fire. But uh I think it's important to recognize that in um in we in urban areas, uh you're going to have something that is less continuous, less smooth than what you have in the Rhineland. You're going to have structures that are going to ignite and suddenly the heat release is going to change dramatically, so it's more intermittent. And also with structures, they burn for one hour. So that means your fire zone is not going to be a few meters deep, it's going to be actually a few hundreds of meters deep. So you can have structures that are going to burn over a fire zone thickness, quote unquote, over a region that can be uh half a kilometer or even a kilometer wide. And so you need to capture that physics. And that's not described uh by the rate of spread. The rate of spread approach works well for a thin firefront. If you are moving into a more distributed combustion zone to use combustion terminology, uh then uh you need to have a more elaborate uh description of your problem.

Wojciech Wegrzynski 39:32

So kind of valid for a bush fire or or a crown fire when it's rather fast, burns through fuel quite rapidly, and not for a lasting fire that creates products over a prolonged uh time, right?

Arnaud Trouvé 39:46

Yes. I mean if you think about it, the the rhetoric spread describes the leading edge motion of the fire. So if your fire front is thin, then the only thing you need is a leading edge motion. And you can draw a line and that's enough. Okay. Even if that line is a few meters thick, who cares? I mean, that's uh but now if uh the leading edge plays a role, it's important, so the outside fire perimeter plays a role. But if you have structures behind that perimeter that are going to burn for an hour or even longer, then you need to keep track of that in your model. So you're going to have a more distributed zone. It's not a thin flame front, and you are not in a in a situation where the depth of your fire is much smaller than the fire line contour.

Wojciech Wegrzynski 40:33

But the the structures deep within the fire front, there are they still influencing the spreads, the yeah, they are in firebrands. Okay, and they they fly all over that and they add to the total number, which increases the probability of ignition of subsequent elements.

Arnaud Trouvé 40:51

Yeah, yeah, they still play a role, a dynamic role on what's happening in the front.

Wojciech Wegrzynski 40:55

That's really interesting. When when you try to model those those fires, uh, I had guests in the podcast, uh, I think it was uh Professor Simoni who said that that the fire, once it ignites the first structure, it completely changes the the spread because of this increased production of fire brands and and the the changing the overall dynamics of this. Is this also something you can now see in the models that once the fire is is traveling through the wildland, uh it's one fire, but then as it meets the city, it changes completely.

Arnaud Trouvé 41:29

Yeah, it does change completely. The the models uh show that, and what I think Professor Simony was um referring to is the importance of the edge of communities. Uh that's where you want to actually harden your structures because uh you want to prevent these first structures on the edges from start burning. I mean, that's where you need to have a line of defense, yes.

Wojciech Wegrzynski 41:53

How much we are limited by data right now in this? And I know it's it's very important for you. Uh starting with your efforts at MacFP and everything, validation and then you know, having good models, uh they they require good experiments. Where do we get the knowledge to build those models from? Like how how much we know about those fires? It's it i i it's hard for me to imagine someone starting a controlled experiment where they are allowed to burn down a mountain in a village, you know, just to see that behavior, and they would be still limited to a single wind they got on that day. So where do you get your knowledge from?

Arnaud Trouvé 42:34

So we have to get more data from fire disasters. Yeah, that's the only way, right? I mean, uh when we have uh, I mean, I'm using an analogy here. When we when we have an airplane crash, okay, uh, people don't study airplane crashes by flying airplanes and crashing them on the ground. They they study them by analyzing disasters because there is a lot of useful information there, and you want to make the most of it because of the impact of the loss. And so, you know, going back to your question, we have to analyze fire disasters, but unfortunately, we don't have a lot of data from fire disasters. So typically uh we have uh observation data that are satellite-based, and they are coming to us with a spatial resolution of a few hundred meters and a temporal resolution that is a few hours, typically six or twelve hours. And so that's not enough for us to understand the fire dynamics. And so we need more data, not only to model, to have validation data to check our models, but we need more data just to understand the problems. One of the problems in we and urban fires is that, and that has been reviewed by laboratory and full scale tests. So, tests performed, for example, by UL, by in the US, by uh IBHS. Um Is that uh you realize that there is a lot of vulnerabilities in structures. There are many ways a fire can penetrate a structure, especially in the US, where we have a lot of our buildings that use uh combustible materials. And so the first thing we need to know is reduce the parameter space to see okay, what are the more probable fire scenarios that explain fire penetration and fire loss. And until we do that, we're going to have a limited number of people, and I'm one of them, trying to do their best effort at studying this problem, but studying in the end different scenarios. And so the first thing we need to do is understand the problem better with having more observation data and also validation data for the models. And I'm a combustion person. When I talk about data, I want to have them at the scale of the flame. So that means meter scale, and that means also I would settle down with 10 meters or something of that sort, but something comparable to the kind of resolution we have in the models.

Data Poor: Validation And Windows

Arnaud Trouvé 45:07

Um and also having to be able to understand fire propagation, to have something that gives me uh some idea of how things change, uh, I would say per a few tens of seconds at the most per minute, something like that, but not you know a data point every uh six hours.

Wojciech Wegrzynski 45:25

Uh what's the time scale in the simulation? So 30 by 30 meters is the spatial resolution. What's the temporal one? Like a minute, a second, an hour?

Arnaud Trouvé 45:34

That's actually a question uh that uh is interesting because uh some so there's a difference there between models that are quote unquote physics-based and models that are more databased. Models that are database sometimes they have a temporal resolution that is uh five minutes or ten minutes. Okay. Whereas in the physics-based models that I'm working on with uh people like Dr. Golner at Berkeley and and Chris Lautenberger, also at uh Cloudfire Incorporation, that's the name of his company. Uh I want to acknowledge them. I don't want to necessarily, but I want to acknowledge their contribution in my thinking. Then we use a current Friedrich Louis condition that is similar to the condition that we use in in CFD. And then we have basically a time resolution that is a few seconds. That's to be able to keep up with the travel of firebrands that go at the speed of uh the wind and uh or the spread or the fire that goes at the rate of spread.

Wojciech Wegrzynski 46:31

So that's pretty much the time that it takes for your firebrand to fly above your 30 by 30 meter cell.

Arnaud Trouvé 46:39

Right. Yes. If you have 10 meter wind, it will take a firebrand three seconds to the curly meter. So that's that's the kind of time scale you need to resolve to correctly capture the motion of his firebrands. So we have high temporal resolution. Well, we don't have high spatial resolution, but we do have high temporal resolution in our models.

Wojciech Wegrzynski 46:59

Yeah. I mean, your example of having a few hundred square meters image or data point every six hours is like if I studied compartment fire and uh I had uh maybe uh two thermocouples in the entire building and and recorded them every like the 15 minutes. I mean, I would get a data point or two, but is it peak? Is it like middle of the fire? Is it dying fire? Is it ignition? I would have no idea.

Arnaud Trouvé 47:28

Yeah. If you if you combine maybe that few measurements with a deep understanding of your problem, then maybe you're in okay shape. But in our case, we still struggle to identify the main scenarios for fire spread. Okay, and so I mean, I have hypotheses, for example, that are supported by some of my colleagues and some of the experimental uh evidence. I think uh the windows is one of the main pathways really for fire penetration. And and sometimes our colleagues in Europe are saying, oh, you know, in the US you have a problem that is different than ours because we don't have combustibles uh in the materials to build our houses. Uh so we don't have combustible walls, we don't have combustible roofs, and that's true. But I I always reply, but you do have you do have windows and you do have ornamental vegetation, trash cans, a car in the vicinity of your window, and when this thing burns, it's going to break the window. You know, so so to me, the windows is a universal weakness of our structures, uh, especially if they are single pane and with plain glass. I mean, there are different types of windows. Uh, they are particularly bad if you have a flammable uh combustible frame, as we do often in the US. Uh, but you know, that's at this point I'm speculating, and I'm sure some people in your audience may disagree or may uh actually uh qualify what I'm saying. And so what we need is more data and a community-level analysis of this data to agree on what are the more likely fire scenarios.

Wojciech Wegrzynski 49:03

I had a beautiful episode with the Sarai people on their experiments from uh building to building flame spread, and windows were a significant part of that investigation. So I think that that's a good data point made for this purpose, actually. But uh yeah, one of the many we need. Uh do you get more information from fires in urban settings? Because you could technically have access to like CCTV, they like we've seen an amazing amount of videos from Palisades, for example, because there's so many cameras there. But it must be hard to make stuff out of this random scat of information.

Arnaud Trouvé 49:42

Yes, I I should say that uh I'm not the one doing this work, I'm learning from my colleagues. So I think there are people at least uh uh Alex Maringidez has done fire investigations, uh like the campfire for NIST. Uh we have people at uh UL FSI, Dan Goram, for example, uh, that have been doing investigations for the LA fires. So I'm reading their reports and trying, you know, because I'm I'm really hungry, uh, have a big appetite for data. And these people are doing their best of using any data available and explaining in their reports what what we know. Uh and so, but this is not systematic enough. So, you know, uh first, there are satellites observations that unfortunately are not in the public domain that have the data that we would need. Okay. Uh military type of satellites, for example. So that would be great if we could access that. But there is also an effort in the US to deploy uh low-altitude uh satellites networks to be able to monitor wild on fires, urban fires everywhere in the US, uh, and uh with enough resolution now so that we can actually see what's going on. So I think in the future this problem is going to get better and we're going to file get the data that we need.

Watching Fires With Satellites And Drones

Arnaud Trouvé 51:00

But at this point, we are we are struggling.

Wojciech Wegrzynski 51:02

Because right now you would either have uh geostationary satellites, which would be extremely far away. I don't even know if you do remote observations from geostat because that's a really, really far orbit. So it's it's just far away, or ones that are close, but they orbit the Earth. So they have to be above the point that is interesting for you, which is not necessarily gonna happen every 40 minutes when they circle. Uh it's this six by twelve hour window that you mentioned.

Arnaud Trouvé 51:27

Right. Yeah. You either have uh good temporal resolution but poor spatial resolution, or relatively good spatial resolution but poor temporal resolution. You don't have both. Unless you're a CIA denied.

Wojciech Wegrzynski 51:40

Yeah, well, that would that would be nice if the if the data exists, if we could if we could get anything we can do as a community to do better, I guess developing our models better, so the less data points are are more valuable for you.

Arnaud Trouvé 51:55

Yeah, well, I think we want to raise the awareness of where are the pain points, and the pain point is lack of data. We are data poor. Uh, and so I think people are changing that. I mean, uh, as you may remember, there's been uh an XPRISE wildfire international competition that actually just finished last month. And uh so and the focus uh uh on the competition was really to bring uh you know new technology to the problem of wine on fires. They were not looking at uh we or urban fires, but the kind of technology they are developing can be can be reused for other fire problems, and uh they are trying to put monitoring systems that bring the kind of resolution that we need for this one, yeah, for sure.

Wojciech Wegrzynski 52:39

I think there is some hope in drone technology, like you know, the types of drones we we we have now uh flying just a little east of Poland. So, you know, that there are drones that are in the air for hours and hours uh doing reconnaissance work. So I think that's also kind of a technology that would probably give you the ability to deploy it near to the wildfire, keep it in air for multiple hours so you have a meaningful data set on propagation across those hours and really good resolution. Maybe that's something that that that could make a difference.

Arnaud Trouvé 53:13

Yes, we have a project at the University of Milan with my colleague, Dr. Rafan Montoya, on on drones application for these type of fires. One difficulty of for the drones is that we are talking about events that take place during strong winds, given storms. So we need to have the drones able to fly under those conditions, and that's that's challenging. I mean, if you look at the LA fires, the firefighting uh airplanes were grounded, you know, on day two because of the winds being too large.

Wojciech Wegrzynski 53:45

On on this note, are firefighters using some sort of that technology to manage their operations? Because they also, I assume their operations must be informed and and and driven by by some knowledge. Do they have access to some observational facilities, or they are also kind of dependent on these satellite images which are uh rare or or or unprecise?

Arnaud Trouvé 54:08

No, they uh they have also their own means. They typically are flying airplanes to uh map the fire perimeter, and they have also ground observations. So they would have you know command center analysts who are going to put this thing together to draw a fire perimeter. But again, in the LA fires, I remember that the fire perimeters were obtained every six hours on day one, and then on day two there was almost none. I mean, uh so it's because they they didn't have access to their monitoring capabilities because of the winds being too strong.

Wojciech Wegrzynski 54:44

And also when the fire hit the city, six hours, like uh that's a whole fire inside of that, like enough to to go through the entire thing. So it's not very helpful when your fire perimeter is five kilometers away. Oh, and now it's beyond. That's that's not not a great resolution.

Arnaud Trouvé 55:00

Yeah, I use as an example the Palisades fire, and most of the spread occurred. The fire burned for more than three weeks, but most of the fire spread occurred during the first two days.

Wojciech Wegrzynski 55:09

Okay, Arnaud, thank thank you. Thank you so much for bringing the the state of art and progress on on wildfire urban interface modeling to the listener of the fire science show. Uh one last message that you would like people to leave uh your lecture in Kyoto with, not not just this interview.

Arnaud Trouvé 55:27

Well, so uh maybe uh a last message is uh about that in addition to data, we need an infrastructure. So, as you know, I mean one of the things um I've been working on is MacFP for compartment files modeling. And so there are this is called MacFP is two things. One of them is um to bring data, reconnect, target experiments for validation of models, but also it's basically a networking effort where the experts come together and actually collaborate and exchange information about where are the difficult points. So I think for uh we in urban fires, we need to have a similar effort where we come together to analyze the data that we have, to see what we know today, to agree on what we don't know and where the efforts should go. Basically coordinate efforts because we are a small community in front of a big problem. And so, one last comment is one first item on this route is um we're going to host at the University of Maryland next month a WUI fire engineering workshop for today is organized by the SFPE Foundation in collaboration with the UL Fire Safety Research Institute and hosted by UMD. And I like this because this is actually the first technical meeting that I'm aware of focused on WUI fires. And uh this problem of Wii and urban conflagration typically has been lumped into uh meetings uh looking at the wine on fire. But I think the problem of Wii and urban conflagration is different from the problem of one on fire, and that's a problem that should have its own community of researchers, its own meetings, and its own infrastructure. And I think we are moving in that direction, but again, uh there's a long road ahead of

Building Infrastructure For WUI Research

Arnaud Trouvé 57:18

us.

Wojciech Wegrzynski 57:18

Yeah, and uh you know, I always hope that uh big urban conflagrations are you know the things for history lessons, you know, the thing that uh we we don't have anymore, but something tells me it's not necessarily the case, especially as the building materials change, uh the way how we build buildings, not just you know, communities with single households, which are probably traditionally a lot of timber, etc. Now we build a lot, awfully lot of timber, you know, in large cities, in in large buildings, in blocks. The cities are super green, filled with vegetation, like uh green walls everywhere, exposed to the longest dry weather periods and peak temperatures ever recorded, you know. All this together tells me that um at some point uh we we may have a really, really nasty urban fire problem, like we've dealt with in in history lessons, and that that kind of scares me.

Arnaud Trouvé 58:18

Yes, I mean we have to learn from past disasters, and we have also to look at what's driving a lot of these disasters are you know wind events, maybe climate change. And uh, you know, when I we look at the heat waves episodes that we are seeing in Europe and in the US uh this summer, we see that things are changing. And so they are there is uh you know traditional zones that felt that they didn't have a fire risk, uh, that are now maybe at risk of uh one on fire or even a conflagration. Uh and so we have to also evolve with our thinking. So we have to learn from the past, and maybe what's more difficult is also try to anticipate the future.

Wojciech Wegrzynski 59:02

And and then with that message, I'm very happy to to conclude this. Uh Arno, uh, thank you so much for for coming here. Congratulations on being recognized uh by the combustion institute and given a plenary lecture. You're representing all of us, and we're proud of you. Like uh I'm I'm really happy that fire is such a strong part of this symposium. There's a I I saw on the program there's a fire track that goes through the entire symposium, I think, from the day one till the end, and there's like a mini symposium at the end.

Arnaud Trouvé 59:32

Yes. I mean, when I started in 2001 and the combustion symposium, you would have 25 presentations on fire, and it will be done in you know a day and a half. Now we are more than 100 papers, and it goes from day one to the last day. So that's very impressive.

Wojciech Wegrzynski 59:49

Fantastic. Have a have a great time in Kyoto. And and uh well, all the best, Arno. Thank you very much. Thank you very much for Jim. And that's

Future Conflagrations And Final Thoughts

Wojciech Wegrzynski 59:58

it. Thank you for listening. I have not been to the Arno's lecture in Kyoto, but my spy network tells me it was nice. And uh, thanks Arno for representing the fire community out there. Thanks to Combustion Institute for recognizing the importance of fire problem by giving planner lecture, a whole track that span the whole conference, and a workshop symposium uh at your conference. That's really great. And I'm I'm happy that also this uh this part of fundamental research on combustion in the fire problem is growing. From this interview, I've learned the new class of models, I've learned uh how Arnaud is thinking about the wildland urban interface problem. And you know, we we had a lot of episodes on Wii already in the Farm Science Show, and I think the image is clearly emerging on how they are different from a pure wildland fire problem, how the specific challenges come to those fires, you know, like the firebs, like the rapid fire spread, acceleration, surface-to-surface, uh ignition problem, the long-lasting fire exposures, the uncertainty in what the the heat release rates in the area will be. And all of those, you know, reflect on our ability to design prevention measures, which are critical if we want to have resilient and safe uh communities. I mean, it's great to know how far spread in those communities, but this knowledge ultimately has to let us build defense mechanisms against those fires, and I think that's the ultimate goal of building models like the ones that Arnaud and his team at UMD is building. And I hope this will continue and I hope this will grow, and I hope we will get better and better at modeling those fires, we will get access to the better data. And if you are a CIA engine listening to this podcast, please give us your satellite data. We would really, really uh do something good with that. I'm pretty sure they can sign a lot of NDAs. You can close them in an underground bunker, just give them the data they need, and they will develop something that will serve the humanity. Anyway, on that note, I think that would be it for this podcast episode. A series of heat waves hitting Europe. And uh yeah, a lot of what we were talking about with Arno uh in this podcast episode already materialized in France and Spain. So yeah, stay safe, guys, and uh I hope this wildfire season will not be the worst in the history. And uh yeah, I I hope we will get better for the next one in terms of our preparedness. Thanks for listening to the Fire Science Show. And the next week I'm gonna bring you more fire science, so stay with me. See you here next Wednesday, cheers. Bye.