Stochastic optimization provides key to combatting increasing frequency and intensity of wildfires

Ashley Kilgore:

When you take a look back, even just over the past couple of years, the devastation caused by the seemingly growing threat of large-scale wildfires has made headlines around the world: from the unprecedented brush fires in Australia in 2019 and 2020, to California’s record setting wildfire season later that same year, and most recently the Colorado prairie grass fires, which occurred well outside the traditional fire season. The increasing frequency and intensity of wildfire incidents are a growing global concern. Not only are these fires capable of causing billions of dollars in damage each year, but contribute to the deaths of thousands of people, not to mention, have a devastating impact on local wildlife and ecosystems. But are there steps that can be taken to better prepare and position first responders and others involved in combating and minimizing the impact of these fires?

Ashley Kilgore:

Joining me to discuss his work using stochastic optimization to support better decisions regarding research deployment, both in the stages before a fire occurs, as well as when multiple fires are underway, is Lewis Ntaimo with Texas A&M University.

Ashley Kilgore:

Lewis, it’s a pleasure to speak with you. Thank you for joining me to discuss your work.

Lewis Ntaimo:

Thank you, Ashley.

Ashley Kilgore:

Lewis, as I mentioned in my opening, we seem to be experiencing a significant increase in not only the number of wildfires, but in their severity as well. And on a global level. What are some of the contributing factors to this growing issue?

Lewis Ntaimo:

I think there are several contributing factors to this growing issue. Some of the factors that come to mind right away include climate change. I don’t know if you’re a believer in climate change but I am. Temperatures have been increasing globally, and so we have ended up with much dry vegetation. Also, I think the greenhouse gas emissions have also contributed to what we are seeing today. And when we look at the vegetation, we see a lot of build up of vegetation – we call that fuels. The other thing I can mention is the significant increase in the wide open interface across the globe due to open development. So there’s been a lot of homes being built near where you have a lot of vegetation or forests. And that is also one of the contributing factors.

Ashley Kilgore:

Now, Lewis, what have traditionally been some of the challenges or unique factors that first responders and other decision makers tasked with organizing and allocating resources face when they are attempting to both prevent and fight these fires?

Lewis Ntaimo:

That’s a very good question. Some of the challenges actually revolve around having limited firefighting resources, for example, ground crews, equipment like plows, dozers, air tankers. That is one of the challenges that they tell me they have. Another challenge, which you can appreciate is limited funding. So some of the state agencies have limited budgets to fight these fires. So they don’t have enough money to do it. From a technical point of view, one of the challenges that I’m aware of is predicting when the fire is going to happen. It’s very difficult to predict where a fire is going to occur because there are lots of causes. Some are natural like lightning, some are human causes like arson, electric power lines. For example, the Colorado prairie grass fire was believed to have been ignited by sparks from power lines and transformers.

Ashley Kilgore:

And now to build on that, how does the growing frequency and scale of wildfire events compound these already complex issues?

Lewis Ntaimo:

Oh yeah, it makes it worse. That’s why there is a dire need to optimize because frequency of these fires means more fires to respond to, which can lead to limitations in the resources that are available at any given time. Remember, I also mentioned limited budgets, right? So when you have more fires happening, it means that you’re not going to have enough funds available to support firefighting operations. Not to mention, of course, these large-scale fires require multiple resources at the same time.

Ashley Kilgore:

So, Lewis, in helping to address these challenges, your work has included the development of the explicit fire growth response model, or EFGRM, which helps with this complex process of resource allocation to combat wildfires. Before we jump into this, I’d love to know how you became involved in this type of research.

Lewis Ntaimo:

Oh, well, that’s a very interesting question. You’re taking me back to 2003, when I was a PhD student at the University of Arizona in Tucson. Back then, we had huge fires in Tucson, and I was just wondering, I was curious as to why and how they were fighting these large-scale fires. I wanted to understand how response planning is done. And so, when I moved to Texas in 2004, I found that they had fires in Texas as well. So I reached out to the Texas A&M Forest Service, which is a state agency tasked with responding to fires in the state of Texas. And I told them about my research interests and established collaboration with them. And that’s where I learned about their needs for what we ended up calling the explicit fire growth response model – EFGRM.

Ashley Kilgore:

So, now circling back, can you walk us through the development of the EFGRM? Was the use of stochastic optimization a new approach?

Lewis Ntaimo:

Oh yes. At the time, the Texas Forest Service wanted a model that would consider or take into account uncertainty in fire growth. As you know, fires are highly stochastic due to changing weather conditions at any given time. Of course, vegetation or fuels change as well. But when you have a fire happening, there’s rapid changing in the weather. And so, they wanted a model for what is known as initial attack. This is the first response after a fire is reported. So what do you do? So in order to do that, given my background in stochastic optimization, I decided to devise a stochastic programming model for this particular problem of the initial attack.

Lewis Ntaimo:

So, the next thing we did was, after we developed the mathematical model, meeting with them, talking about what decisions had to be made, what was the objective and so on, we collected historical data from them in order to calibrate and validate the model. The data actually range from 1986 to about 2003, included the vegetation fuels data, the rain, GIS data or landscape. And of course they keep weather data as well. So they define the center weather scenarios. And so all those things were included in the formulation of the EFGRM.

Ashley Kilgore:

And this model has already been applied in real world settings. Can you share the circumstances and the impact it’s had?

Lewis Ntaimo:

Oh yes. Actually this model was used by the Texas A&M Forest Service. At the time, after we completed the model, we were given a study area in East Texas. The idea was to figure out a way to preposition firefighting resources. In that part of the state, they mainly use plows and dozers. So they do what is known as indirect attack. So they wanted to know, to position themselves before fires actually took place. So we used the new EFGRM to actually do that.

Lewis Ntaimo:

The model also allowed to identify and prioritize areas that were prone to fire occurrence and needed more resources. Because once you use the model, it can tell you where more resources are needed based on the data that you feed it. Because the model takes into account the fire occurrence in each area, the vegetation at that particular location, the available resources at fire operations bases nearby, and then prepositions the resource. So you could basically identify priority areas.

Ashley Kilgore:

Lewis, I’d like to thank you again for joining me. It’s really been fascinating discussing your work and its impact. Before we go, are there plans for further applications of the EFGRM? What are next steps?

Lewis Ntaimo:

Oh yes. You know, since it’s development and publication, beginning I think around 2013, 14, 15, we have now developed other models for initial response planning. EFGRM is still useful for that purpose, but right now, Ashley, I’m working on the National Science Foundation funded project on fuel treatment planning. We have now moved on to strategic planning. This is before fires happen. What can you do with the vegetation? The EFGRM model is a tactical or operational planning model. So that can be used for that purpose, but we also need to think long term. So we are now developing models for more fuel treatment plans to help reduce or minimize these large-scale fires when they actually do occur. So that’s what we are doing now. Very exciting work.

Ashley Kilgore:

Want to learn more? Be sure to check out the show notes, posted in conjunction with ORMS Today magazine online to learn more about what was discussed in this episode. The podcast is also available for download from Apple Podcasts, Google Play, Stitcher and Spotify. Wherever you listen, please take the time to do a review as that helps other listeners find the podcast. Until next time, I’m Ashley Kilgore, and this is Resoundingly Human.