A student intern working with NASA shares what he’s learning about putting AI to work predicting wildfires.
Mason Lee was a high school graduate in Los Angeles preparing to study at Brown when, after experiencing California’s most severe recorded fire season, he opted instead to fight wildfires on the frontline. Now in his third year of college, Mason is studying artificial intelligence and machine learning. This, along with his interests in firefighting and environmentalism, led him to a summer internship at NASA, where he and his colleagues are developing creative solutions to integrate AI into wildfire spread forecasts. (He researches under Dr. Nancy Kiang, Dr. Robert Field, Dr. Keren Mezuman, and Dr. Marcus van Lier-Walqui.)
Mason recently sat down with his best friend (and rock climbing partner), Jasper Cerone, a student and writer for Bluedot Living. They talked about Mason’s work with NASA, how it complements his firefighting experience, and how this work prepares us for an otherwise unpredictable climate future. Oh, and they also talked about rock climbing.
Interview has been edited for brevity and clarity.
Jasper: I’m sitting with my best friend Mason. Mason, tell our readers where we are right now.
Mason: We’re up in the Shawangunk Mountains, in upstate New York. We just got done climbing an ultra-classic, multi-pitch route called High Exposure. It [requires] a style of rock climbing called trad, which means we had to place our own protection equipment in the rock while ascending the 300-foot route.
Jasper: It was my first real multi-pitch climb. Mason led me up it. I’ve never done anything like that – super sick, really scary. I’m still coming down from the adrenaline high of clinging to the face of an exposed cliff hundreds of feet above the tree canopy. … Before we get into your work with NASA, I want to talk about the applied science project that we started in ninth grade. Can you give a brief summary of the work we did together?
Mason: Sure. Our project explored the remediation capacities of methanotrophs, which are these bacteria with the ability to degrade methane, along with other environmental pollutants. We used them in a biofilter (a pollution control technique using living microorganisms — in our case, methanotrophs — to decontaminate water or air by breaking down pollutants into harmless byproducts) to detoxify contaminated ocean water. Our technology was a breakthrough. And our efforts received a lot of recognition. We won the Los Angeles County Science fair and were selected to participate in the 2020 International Science and Engineering Fair.
Jasper: Mason, let’s be honest: once the math for that project started to get crazy, you did most of the work.
Mason: You got lazy. I kinda carried you.
Jasper: Yeah, yeah; whatever. Still, I’m curious to know if our project motivated you to break into the field you’re in now, and if it continues to shape the way you think about land remediation and artificial intelligence.
Mason: That project, I think, was the first time that I initiated an independent study of science. From ninth grade onward, I knew I wanted to do something environmentally related. It was a really tough challenge over the years. There were many, many failures. The sheer difficulty of cultivating methanotrophs was one of the things that sparked my interest in AI. You know, you can look at nature and these large biological systems that are so complex and try to simplify their function down to protocols and simple equations, but these things aren’t always so simple. So we need computer assistants to speed up these processes, simplify them, and make them more efficient. Looking back on it, AI assistance would have made things easier for us.
Jasper: Tell me a bit about your current work at NASA.
Mason: I’m working in NASA’s climate science department. While most of the research they do is aeronautics- and space-oriented, they do have a climate center at Columbia University called the Goddard Institute for Space Studies. A good portion of the research is focused on a large climate model called ModelE. To put it very simply, we’re using ModelE to run wildfire simulations, hoping to understand the fundamental physics that drives wildfire spread.
Jasper: And what might we gain from a more comprehensive understanding of fire physics?
Mason: Well, in the long term, we can predict the complicated effects that fires have on land, air, and water. It’s about understanding the patterns and frequency of wildfire spread in an ecosystem over long periods of time, and being able to identify areas that are prone to fires so we can take proactive action.
Jasper: So how does AI come into play?
Mason: Essentially, we use AI to train the model, and extract physics from it. As a result we get a more accurate and physically correct fire model that tells us about the drivers of fire on a global scale.
Jasper: I’m really proud of you, bro. I can tell you love doing this.
Mason: Yeah, thanks so much man, I really do love it.
Jasper: Are you just researching broad-based trends or are you helping specific communities as well?
Mason: Right now, I’m analyzing historical fire trends over the years, looking at huge parts of the population. This NASA model does a lot to understand what places are going to be able to be habitable in the future, what places are going to be more prone to fires, and how we’re going to be affected by climate change at large. And I think that’s relevant to everyone on Earth.
Jasper: So you’re saying that, with the right inputs, the model you’re building can be applied anywhere around the world?
Mason: With the right tuning and physics, we can get semi-accurate predictions about what the state of the earth will look like decades into the future.
Jasper: That timeline is mindblowing. It makes me wonder what the future of this technology might look like? How can it improve from here?
Mason: You know, that’s a huge question … Without getting too deep into it, I think it comes down to advancements in machine learning. We’ll be able to increase accuracy and physical consistency, make predictions further and further into the future, and scale up the existing model to make these predictions exponentially faster. We’ll be able to plan farther ahead and know where certain catastrophes will happen.
Jasper: What about your own experiences working as a wildland firefighter? I know that fighting fires on the front lines ignited your passion for wildfire prevention.
Mason: Yeah, firefighting was an incredibly intense experience. In 2020, I worked at Cal Fire, an agency in the California Conservation Corps. We would work twenty-four- to forty-eight-hour shifts for months on end when there was a big fire.
Jasper: Literally fighting fires for two days straight.
Mason: Yeah. We would build a city in the middle of the forest to go back to when we needed rest. We’d wake up and immediately go out to cut the fire line, one foot in the blaze, the other on green. And the 2020 California wildfire season was record setting.
Jasper: What was it like to be in the field?
Mason: It was a very intense situation. Mentally, you’re always looking out for the people ahead of you, and on the ground it feels like war. You have planes flying overhead, dropping retardant on you. You have trees falling, or exploding with embers. You have houses that you’re trying to save. People are getting injured all around you, and you’re constantly having to evac them off the fire lines. You’re always considering your exit routes. And you have to work through hours of dehydration and sleep deprivation. And you’re carrying packs with anywhere from thirty to eighty pounds on your back. It’s definitely a tough job. You work throughout the day and the night.
Jasper: You found it rewarding, yeah?
Mason: Yeah, I really like being pushed to extremes. Growing up in California, fighting fire was always something that made sense, something that felt like the right thing to do. I remember going to the library as a kid and always checking out the wildfire DVDs. And I really enjoyed that physical challenge.
Jasper: The intensity of the 2020 season had a lot to do with anthropogenic factors, right?
Mason: It had a lot to do with increased tree mortality due to pine beetle migration patterns. They created a forest that was dense, dry, and ready to blow. Most of the fires were started by dry lightning strikes, which are boosted by climate change.
Jasper: Do you think your front line experience lends you a unique perspective that some of your co-workers might not have?
Mason: Absolutely. It’s one thing to interpret outputs of a simulation from an off-site lab, and something else entirely to be on the ground and see these fires up-close. You experience things that simulations will never be able to model. The most accurate simulations can’t predict split-second atmospheric changes, nor the erratic wind conditions that occur on the ground moments later. Not many researchers have a true sense of what it’s like to be fighting fires in the field. So being on the front lines and understanding how conditions change on a moment-to-moment basis is crucial for creating the best simulations.
Jasper: The other day you were telling me how you were rehired to do more work for NASA this upcoming semester.
Mason: Yeah, so I just accepted another offer at NASA, working under Peter Mehlitz and Sequoia Andrade, and our research will contribute to The ODIN [Open Data Integration Network] Project. We’re gonna be using California’s satellite data to create real-time fire maps and make predictions about specific fires. This network will allow more wildfire scientists and government agencies to input data into their own machine learning and physics models. We hope to provide local communities with real-time updates about wildfire spread and potential access route cutoffs.
Jasper: What’s a data integration network?
Mason: It’s software that combines data from a bunch of different sources and brings them into a unified, easily-understandable format. We take data from different sources like NASA’s satellites, UAV’s, planes, thermal cameras, and weather stations. These sources give us input variables like weather, wind direction, and moisture. The model will then output things like the rate of spread, the intensity of the crown fire (a fire reaching the canopy spreading between treetops), the spotting (the wind transporting burning pieces, risking the ignition of new fires), how much fuel is burned, stuff like that.
Jasper: And how else might this technology be used in service to the planet years down the line when it improves?
Mason: That’s a tough question. People are just now starting to have data integration platforms that can make predictions based on satellite data. I think that as we launch more satellites, collect more data, and develop better computers that can run these physics models faster, we’ll begin to come to a more accurate understanding of how the earth works.
Jasper: I can see why satellite data is so crucial.
Mason: Satellites give us images and weather forecasts. They tell us things about the wind, precipitation, and land cover; about the percentage of dead trees in the canopy, the type of fuel in the ground. They give us topographical information, like slope, terrain, and how wind flows through valleys. All the driving factors for fires. And it’s all relative to human settlements. We can get a map of where people live and quickly identify vulnerable communities and areas that will be impacted by fires. So much data can be sensed through satellites, and with it, we can get a broader picture of forest composition and factors that drive not only fires but climate change at large. The more satellite data we have, the more accurate our predictions will be. We might be able to make near-real time predictions about environmental conditions, instead of using historical guesses. And that’s big.
Jasper: So the ODIN Project is gonna be a massively useful tool for local citizens and firefighters, right?
Mason: Yeah, the hope is that agencies like Cal Fire and the Forest Service will use it. It’s open source and public access. We hope to make it as easy as possible to use.
Jasper: So if you define the work you’re doing now as large-scale problem solving, the work you’re doing next semester is more locally targeted.
Mason: Exactly. We’re gonna be using algorithms that integrate data as fast as possible to predict how a specific fire will spread. All of this data will be quickly synthesized into a simple user interface that anyone can understand. Firefighters will receive this information so they can take action more effectively and efficiently.
Jasper: What do you and your colleagues think about the fires in Maui?
Mason: We’re following closely what’s happening there. The rate at which the fire spread is terrifying — who ever thought a wildfire would collide with a hurricane? And the increasing death toll saddens me deeply.
Jasper: Residents said they weren’t warned before they fled for their lives. Could the technology you’re working on have been of use here?
Mason: It’s tough because the Maui fires spread so fast and were so wildly
unpredictable. Time will tell if these systems can decrease response times, but they haven’t been deployed yet. So I can’t say for sure.
Jasper: Okay, so now we’re walking back to the car with crash pads and ropes on our backs. We’re exhausted. My phone is about to die, but I have one more important subject I want to talk about. Can you share your perspective on how the AI revolution might facilitate long-term sustainability and multi-species coexistence?
Mason: That’s a great question. I think a lot of people are really excited about this. It’s exciting for people interested in planetary science. I think AI is going to be able to resolve some of the earth’s complex processes that we’ve never been able to understand. As we continue to collect so much Earth data, we’ll be able to understand these huge global interactions on a large scale in much shorter time frames.
Jasper: How so?
Mason: If we can run simulations thousands of times faster than we can now, we can simulate different scenarios about the future in ways that we were never able to before. Like, would the desert in Africa change relative to the deforestation of the Amazon? Things across the world.
Jasper: So it’s about being able to identify these large scale planetary interconnections that were previously impossible because we just didn’t have the technological capacity. We’ll move towards a more comprehensive understanding of earth systems and take action accordingly.
Mason: Right, and we’re at a point where our connection to the environment is becoming inextricable. We can no longer envision ourselves as separate from the Earth. We will need to become stewards of the land in a way that we never have been before. We have to anticipate the repercussions of human actions. We’re going to have to decide what type of ecosystems we want and where.
Jasper: What do you mean?
Mason: Like, when a fire happens, what type of trees should be replanted in the forest? Should we go back 200 years to when there were manzanitas growing there? Or do we keep pine trees? Should we plant perennial wildflowers? Things like that – these decisions can definitely be assisted by AI. I think building upon our traditional knowledge, using climate models and neural networks, we can create accurate simulations that will lead us to sound decisions. Of course it’s not as simple as that, but I do think we can use AI to work towards something like a planetary control system.
Jasper: Really fascinating. To your point about becoming stewards of the land, I think it’s so important to reconsider what it means to be human beings living on earth in the age of the Anthropocene. Technology can aid the fight against climate change, but a radical shift in perspective is necessary now more than ever. We’ve both lived our lives in big cities, so it’s easy for us to think about nature as some abstract thing in the distance. But as you said, we’re inextricably connected to it, and we need to recognize that we are a part of nature, no matter where we are.
Mason: Yeah, I think that, at the end of the day, it’s about listening to the planet. And I think, again, the shift towards data-driven learning algorithms is a good way to develop and hone our ability to listen. The better we can listen, the better we can understand what’s really going on. Then we can fight.
Jasper: It’s reassuring to hear about projects using AI for the betterment of the planet. But what do you have to say to people who are scared about the future of AI? One of the many things I fear is corporations harnessing the power of AI to streamline their methods of extraction and land degradation. Do you see this as an issue?
Mason: Yeah, I think it’s a real issue. We’re getting more and more efficient in the ways we extract. AI is already being used by gas and mining companies. It’s the dark side of this geology work.
Jasper: Is planetary destruction not an inevitable byproduct of their extractive practices?
Mason: Yes. I don’t think that the power of AI ever voids the personal responsibility that people have to protect their land and speak up. AI isn’t going to save us; it’s going to take a paradigm shift in the way that we think about the planet. Everyone has to be on board.
Jasper: People need to wake up. Maybe use AI to combat corporate extraction.
Mason: Yeah, I think it’s like most things. Big tools like AI can definitely be used to reinforce the way that we already think. Not enough people think about AI in terms of Earth science. And there are too many people who think about it in terms of making money. There’s so much potential for artificial intelligence to be used to make great discoveries in the realm of earth science, discoveries that I can’t even begin to comprehend. I think the big challenge moving forward will be getting a lot more people putting energy into that.