Heat of the Moment

Small white plan flies over fire and smoke dropping red fire retardant.

Fighting wildfire with data science and supercomputing.

When a wildfire starts in the western U.S., it sets off a chain reaction of responses. Much of it you can imagine: alarm bells in a fire station, firefighters gearing up and rushing to the scene. But in recent years, teams at UC San Diego have leapt into action too—not with trucks and hoses, but with state-of-the-art technology and computing power.

At the first signs of smoke or reports of a possible fire, first responders, dispatchers and UC San Diego researchers alike zoom in on video feeds and triangulate the fire’s location from a vast network of mountaintop cameras. Firefighters scale up their response and decide on deployment of resources. Supercomputers at UC San Diego run simulations to predict the fire’s path—how far the winds will take it, how much vegetation will fuel its flames. Maps are sent to command centers, calls come in and researchers convene with battalion chiefs—how severe could this fire become? Do they send out a plane to gather more intel?

You might not expect a university to be so hands-on. But with increasing frequency, severity and conditions ripe for wildfire, these are the times that call for it, and this is how we answer.

In 2020, California experienced the most devastating wildfire season in modern history. Fires torched 4.2 million acres throughout the state, resulting in 33 deaths and more than 10,000 structures destroyed. The outlook for the rest of 2021 isn’t much better.

Meager amounts of rain last winter mean more parched chaparral in wildlands across the state, which means a greater chance for larger and more destructive fires this season. “We’re watching the planet respond to extreme climate,” says Neal Driscoll, professor at UC San Diego’s Scripps Institution of Oceanography. “This weather whiplash caused by climate change is leading to cascading disasters. In this extreme climate, we need to stop thinking about these disasters as unprecedented—they are only going to be longer and stronger in the future.”

Driscoll himself has an unprecedented view of the state’s fires. He is the co-director of ALERTWildfire, a rapidly expanding network of cameras that has become an essential tool for fire officials. Having started in 2013 with just a handful of cameras in the Lake Tahoe area, by the end of this year, the network will have grown to 800+ high-tech infrared cameras, monitoring many wildlands and communities at risk all across the west.

The high-definition cameras are able to pan, tilt, zoom and perform 360-degree sweeps approximately every three minutes. Each camera can view as far as 60 miles on a clear day and 120 miles on a clear night.

Dry brush atop a mountaintop in California with scientists and camera technologies.
A new ALERTWildfire camera is installed on the hills above Santa Monica, Calif. by Scripps Institution of Oceanography personnel Ernest Aaron, Colby Nicholson and Bryan Hoban. There are now 800+ ALERTWildfire cameras across the west.

Fire agencies, public utilities and local governments have invested in the network given how indispensable it has become for confirming locations of fires in those critical first moments and providing intel that helps officials sequence evacuations and manage containment.

“The cameras have been a game changer for us in San Diego,” says CAL FIRE San Diego County Unit Chief Tony Mecham. “They have given us real-time situational awareness and allowed us to make informed decisions much quicker than we used to. It used to take 20 to 30 minutes for our fireground commanders to get to fires and make decisions, and now with the cameras, we are reacting within seconds of the first report. That extra time is significant when it comes to moving resources or starting evacuations. It’s making a difference. I can’t even put into words how important those first few minutes are.”

Driscoll runs ALERTWildfire with Graham Kent, PhD ’92, a fellow seismologist who he has previously collaborated with on earthquake research. The program itself evolved from networking technology created for an earthquake detection system, part of the AlertTahoe pilot project Kent developed as the head of the Nevada Seismological Laboratory at the University of Nevada, Reno (UNR). But within 30 minutes of the first cameras going live, they happened to capture a fire and immediately knew they were onto something major.

Seeing a new utility for the infrastructure they developed, Kent and Driscoll ramped up ALERTWildfire starting in 2017, overseeing the logistics of installation, network support and resilience, user training and more. The program—now a collaboration between UC San Diego, UNR, and the University of Oregon—grew by dozens and then hundreds of cameras each year as demand skyrocketed due to increasing fire season severity and disasters up and down the West Coast.

“It’s one of the most important endeavors I’ve undertaken as a professor at UC San Diego—developing this network to help first responders save lives, conserve ecosystems, and protect infrastructure,” says Driscoll, who still teaches undergraduate and graduate courses on tectonics and sedimentation. “My research has always focused on hazards, and this has become the most severe hazard.”

Given the remote location of many of these cameras, installation can be challenging. Kent, Driscoll and members of UNR, UC San Diego and the University of Oregon’s highly skilled team often drive for hours along backroads in order to scale towers up to 150 feet, sometimes even employing helicopter operations to get to remote mountaintops. But the human impact of keeping people safe is what keeps the team motivated.

“There is a sense that fighting wildfires out west is like a war,” says Kent. “I say no, it is a war—there is no ‘like.’ The danger is there, and operationally, from how they deploy air assets to how they deploy hand crews, sometimes it is life safety triage.”

ALERTWildfire isn’t the only UC San Diego technology being used by fire officials. The WIFIRE Lab, founded by Chief Data Science Officer Ilkay Altintas at the San Diego Supercomputer Center (SDSC), is using the power of data science and supercomputing to provide officials with real-time fire models that help strategize the best approach to containment.

Two women and two men gather in a dark room to view fire mapping imagery.
(L-R) Ilkay Altintas, Dan Crawl, John Graham and Jessica Block of the WIFIRE Lab at the San Diego Supercomputer Center review firemaps made by data and AI technology.

During periods of high fire risk, the WIFIRE Lab is staffed around the clock. When the ignition point of a fire has been confirmed, often by using  ALERTWildfire cameras, Altintas’s team moves into action to deliver continuous, predictive fire models using a number of data types, including the point of origin, nearby topography, weather and wind conditions, vegetation available as fuel and more. The ability of the SDSC to compute all these data together is key to predicting a fire’s path, especially when dry brush in a narrow canyon on a windy day can amplify flames in an instant. WIFIRE’s software pulls in vast government data sets as well as ground-level sensors and high-resolution satellite imagery then uses artificial intelligence and fire simulations to turn this information into dynamic predictions.

These models, known as Firemaps, are made accessible to the Wildfire Fusion Center, a command center housed in a downtown Los Angeles fire station, and then relayed to commanders in the field.

“For incident commanders, the WIFIRE Firemap is one of the most progressive decision-making tools developed in the last decade,” says L.A. Fire Department Fire Chief Ralph Terrazas. “Firemap gives the commanders accurate and real-time data to help make command decisions when prioritizing resource allocation or which communities to evacuate. This has never been available during the initial action phase of brush firefighting.”

Altintas and WIFIRE also have resources they can deploy if there is a need for better data in the moment. Two aircraft stand at the ready—one in Northern California and one in Southern California—to fly above fires for the sole purpose of data collection, as part of the Fire Integrated Realtime Intelligence System, a public/private partnership program funded through the California Office of Emergency Services and led by the Orange County Fire Authority. Within minutes of a fire, the WIFIRE team assists fire chiefs to make go or no-go decisions on deploying aircraft to collect data about the fire’s perimeter. In 2020, they flew more than 45 fires throughout the state.

“Only in a public university like UC San Diego can we turn this science funded by federal and state resources into solutions for society,” says Altintas, who initially grew the WIFIRE Lab thanks to funding from the National Science Foundation. “This program has seen a dynamic transition of science into the public space thanks to the openness of researchers and the trust we’ve built with agencies.”

Beyond the reactive fire modeling for use in the moment, Altintas and her team also proactively use their data to minimize fire risk in the first place.

“There are a lot of reasons why wildfires have become so severe, one of which is successful fire suppression. Putting fires out as soon as they ignite means vegetation and fuel is allowed to build up,” says Altintas. “This has created a fire deficit. California needs to proactively burn an order of magnitude more to treat vegetation every year to get us out of this deficit.”

The same data the WIFIRE Lab uses in the heat of the moment are also put to use outside of fire season, modeling the highest-risk areas of overgrown, dry vegetation to target areas for controlled burns.  Developed by Altintas and her team, the program known as BurnPro3D uses artificial intelligence and three-dimensional fire modeling to estimate the least risky days for land managers and burn bosses to do the most controlled and efficient burns. The models can even predict where smoke will go, to better minimize the health and emissions impacts of these necessary burns.

“It can be very risky to do controlled burns if you don’t have good fire models,” she said. “Our models can help predict the unpredictable by determining the brush fuels available down to one meter of resolution to make the process less risky.”

The uses ofthe ALERTWildfire cameras are also expanding beyond their most immediate utility in fire season. Efforts are underway to make the data collected from the cameras freely available to everyone, including researchers around the world, by giving them open-source accessibility. 

UC San Diego’s Falko Kuester, a professor of structural engineering and computer science and also a part of UC San Diego’s Qualcomm Institute, is leading the effort to bring this data network to bear for research. But with the network taking 12 million images per day, there’s a lot of data.

“We’re making sure this vast amount of data is processed and ingested for a deep time analysis to understand the environmental data better,” says Kuester. “Even though ALERTWildfire is mainly a fire confirmation system, it does so much more—it’s a very rich data record that allows the wider study of an array of environmental issues.”

More accessibility would allow for innovation on those issues from others well beyond campus. Companies are now testing artificial intelligence using ALERTWildfire images with the hopes of being able to quickly detect fire smoke and distinguish its appearance from clouds and marine layer fog. Other researchers are analyzing different facets of the data: the health impacts of wildfire smoke, how burn scars weaken soil and cause debris flows, and the effects of fire on plant regrowth.

Students at UC San Diego also get hands-on opportunities to work with the data. Both the ALERTWildfire and WIFIRE Lab programs offer students the chance to work behind the scenes on technology development—making an impact on public preparedness, data management and staffing during times of high fire risk.

“We’ve really built a village,” says Driscoll. “It’s been incredible to see the university and our partners bring this cutting-edge technology and applied science to save lives and habitat.”

Altintas echoes this, as the WIFIRE Data Commons also shares data with firefighting and artificial intelligence communities with the intention of fostering next-generation fire science. Kuester likewise agrees on the need for such information sharing.

“In industry and a lot of academia, there is typically an instinct to protect your data,” he says. “At UC San Diego, however, it’s more like, ‘How as a team do we use the data to benefit others?’ It is potentially lifesaving, and it should be out in the public domain.”

From engaging fire professionals and industry partners to training the next generation to build smarter tools to serve the public, these programs are a testament to how collaboration and partnership are necessary when it comes to saving lives and livelihoods.

 

Read stories from Tritons fighting fire: tritonmag.com/fire