How A.I. Can Help Humans Battle Wildfires, From Advanced Camera Systems to Forecasting Models

May Be Interested In:Inside the ‘lawfare’ plot to sabotage DOGE


Earlier this month, relentless 90-mile-an-hour Santa Ana winds sent wildfires tearing through Greater Los Angeles, taking more than two dozen lives, flattening neighborhoods and decimating biodiversity. Urban sprawl, a lack of winter rain, excessive vegetation around homes, insufficient water systems and, of course, climate change all contributed to the massive blazes.

Human-induced climate change has led to a 172 percent increase in burned areas since the 1970s, and resulting intensifying heat and dry air has left Southern California in one of the most severe droughts since the 1500s.

“Wildfire is natural in many temperate forests, yet human-caused climate change is intensifying the heat that drives fire, tripling burned area in California mountain forests above natural levels,” says Patrick Gonzalez, former National Park Service principal climate change scientist and former assistant director for climate and biodiversity of the White House Office of Science and Technology Policy.

After what’s been called one of the worst natural disasters in recent U.S. history, researchers are beginning to use new methods of wildfire detection and prediction—including artificial intelligence. This field enables machines to learn from experience by processing massive amounts of data to make predictions and recommendations. One sophisticated A.I. capability called machine learning could be used to alert firefighters to the source and likely path of a wildfire. And these computer models could train themselves to forecast wildfires with greater speed and complexity than humans could ever imagine. But the rapidly evolving nature of A.I. has also made it difficult to establish comprehensive federal regulations, leading to concerns about ethics, trustworthiness and accuracy, with many experts emphasizing the importance of leaving the decision making about wildfire response up to a human, rather than a machine.

The use of A.I. to detect and predict wildfires is still in its infancy. While A.I. has been used successfully in some climate and weather models, for now, most wildfire machine learning algorithms only calculate factors like terrain, temperature, humidity, wind and human activity to evaluate risk at a given location or time. But scientists are building more sophisticated A.I. models with up-to-date climate data that can detect wildfires quicker and map out their spread. As publicly available weather, deforestation and other data is amassed through increasing networks of A.I. cameras, drones, satellites and sensors, A.I. technologies could revolutionize firefighting.

Mapping wildfires

Assad Oberai, an aerospace and mechanical engineer at the University of Southern California, is building an A.I. forecasting model that measures and predicts the spread of a wildfire—and aims to eventually feed more accurate data into other forecasting models. Very shortly after the deadly Eaton wildfire broke out, he used his model to forecast the wildfire’s intensity and growth by tracking it back to the place of ignition—in the vicinity of Eaton Canyon, where, it turns out, an electrical tower is now being investigated as the cause of the fire.

While he refines the part of his algorithm that updates the current state of a wildfire, Oberai says it can already map out the likely path of a wildfire with 85 percent accuracy. Once his models are more reliable, he hopes to work with wildfire agencies in real-world scenarios to improve forecasting so officials can gain earlier control of wildfires, like Eaton, as they become more common due to climate change.

“These were densely populated neighborhoods with homes and businesses,” says Oberai. “Usually, we see this type of spread in wildlands.”

A map from Assad Oberai showing the time it took the fire to spread by color. Darker colors indicate earlier fire, while lighter colors indicate later spread. According to this image, the ignition likely took place somewhere in Altadena in the vicinity of Eaton Canyon and East Altadena Drive, then spread outward into Altadena and the surrounding mountains. 

Courtesy of Assad Oberai

The algorithms Oberai used in his model to predict the source of the Eaton Fire, which were published in this past July’s issue of the American Meteorological Society’s Artificial Intelligence for the Earth Systems, pull NASA satellite imagery into wildfire spread models to forecast the evolution of the wildfire. With improvements, he believes his novel wildfire modeling approach could be a game changer in wildfire forecasting.

Using ultra-high-definition A.I. camera systems

Since 2023, the California Department of Forestry and Fire Protection and other agencies have used an A.I. timelapse video camera system to detect and monitor wildfires and their causes. The system, called ALERTCalifornia, was built and is run by the University of California San Diego, and deploys a public network of more than 1,140 cameras that use A.I. to help alert firefighters and other agencies to wildfires—so they can respond earlier.

“The camera has been an important tool to help emergency managers across California spot fires at the incipient phase,” says Caitlin Scully, who works for ALERTCalifornia at the university’s Scripps Institution of Oceanography.

The cameras were originally intended for geologic monitoring—helping to spot natural disasters. But ALERTCalifornia cameras now employ a separate A.I. capability called image recognition, which is used to analyze and detect objects, recognize faces, and identify behavior patterns. Now by searching the landscape for signs of smoke, they help agencies track fire incidents in real time across thousands of miles. Day and night through near-infrared night-vision capabilities, the ultra-high-definition cameras pan, tilt, zoom and capture 360-degree views as far as 60 miles on a clear day, updating every two minutes. By delivering a better understanding of the environment and its elements, they help humans detect fires and make more informed decisions about containing them, scaling resources and supporting evacuations.

Sunset Fire

A shot of the Sunset Fire taken by an ALERTCalifornia camera on January 8.

ALERTCalifornia | UC San Diego

The cameras are currently located in Los Angeles County, Kern County, Orange County, Riverside County, San Bernardino County, Ventura County and San Diego County. These high-tech cameras can be viewed by anyone 24/7 and are currently aimed at the sites of the Palisades Fire, the Eaton Fire, the Hurst Fire and others.

Detecting red flags

Last July, an algorithm used by Pano A.I. detected smoke from a lightning strike that caused a remote high-risk fire on Bennett Mountain in Douglas County, Colorado. After Pano A.I. staff confirmed the wildfire, they dispatched the exact location of the blaze with a video of the smoke to alert fire agencies, which sent 40 firefighters two and a half hours into the mountains, where they successfully extinguished the three-acre blaze.

Pano A.I. is constantly monitoring forests, grasslands and other fire-risk areas across the western United States, Australia and Canada. The company’s algorithms combine satellite imagery, field sensors and other data with ultra-high-definition infrared video to spot temperature fluctuations at night as far as ten miles away. A human always reviews fires detected by the A.I. to eliminate false positives, before sending out alerts to fire agencies. The Pano A.I. system has been deployed in over a thousand locations, sending out alerts in 2024 about more than 110 wildfires in the U.S., from Warden, Washington, to Wellington, Colorado, while learning from the data it collects to get better at detecting fires.

“Every minute matters in wildfire response,” says Sonia Kastner, CEO and co-founder of Pano A.I. “Mountaintop camera networks paired with satellites and A.I. algorithms shave precious minutes and hours off of each stage in the process of detecting, confirming, pinpointing and responding to fires.”

AI and climate change

While becoming more efficient, large data centers—many of which store computing machines and hardware for A.I.—used as much electricity in 2022 as the entire country of France.

“The coal, oil, and methane burned for artificial intelligence generates the pollution that causes climate change,” says Gonzalez. “When artificial intelligence draws electricity from renewable sources, it displaces sustainable solar and wind energy from potentially more essential uses of lighting, heating, food production and health care.”

Gonzalez, now a climate change scientist and forest ecologist at the University of California, Berkeley, points out that in many areas measures like prescribed burning, allowing naturally occurring wildland fires to burn under management, clearing vegetation around buildings and preventing human ignitions is ultimately the best fire prevention. He adds that Southern California shrublands, where prescribed burning is not ecologically appropriate, have a natural regime of high-severity fires every 50 to 100 years. And the reality, he says, is that wildfires will continue to burn through counties and cities as long as developers continue building in fire-prone areas—especially as the Earth warms. “Cutting carbon pollution from cars, power plants, and other human sources offers the essential solution to reduce climate change,” he says, “and help protect people and nature from unnatural wildfires.”

Editors’ note, January 30, 2025: A previous version of this article misstated the location where Pano A.I. detected a remote fire in July 2024; it was Douglas County, Colorado. The article has been updated to correct the error.

Get the latest Science stories in your inbox.

share Share facebook pinterest whatsapp x print

Similar Content

Several killed in Russian missile attack on Ukraine's Zaporizhzhia
Several killed in Russian missile attack on Ukraine's Zaporizhzhia
Director Nabil Ayouch pays tribute to artistry of Morocco's Sheikha singers
Director Nabil Ayouch pays tribute to artistry of Morocco's Sheikha singers
Meghan Markle Set To Make A Comeback With Netflix Show And Brand Launch In 2025 - News18
Meghan Markle Set To Make A Comeback With Netflix Show And Brand Launch In 2025 – News18
COVID-19, cold, allergies and the flu: What are the differences?
The good news is that you don’t have to go completely grain-free.
Robyn Hynes DND Ottawa Citizen
Watchdog pushes government to help Afghan war advisers
How do you get better at social games like The Traitors?
How do you get better at social games like The Traitors?
Global Focus: Events that Define Our World | © 2025 | Daily News