As artificial intelligence (AI) has become a powerful tool in combating the harms of natural disasters like hurricanes, researchers across the country are working to deploy the evolving technology in wildfire prediction/preparation, early detection and damage assessment with improved accuracy and speed. This important work is continuing to grow in urgency as the proliferation of extreme wildfires is expected to increase 50% by the turn of the century.
In terms of wildfire prediction, there are three key areas where AI tools can be leveraged: risk prediction, preparation and behavior/spread forecasting once a fire is detected.
Machine learning is uniquely positioned to aid in predicting future wildfires due to its ability to quickly analyze large amounts of data and identify patterns for accurate predictions. Utilizing historical data such as “weather, vegetation types, and previous fire incidents, machine learning algorithms can determine if the conditions are prime for disaster and identify potential wildfire risks with increasing accuracy.” This allows for more ample alerting time for surrounding communities to help with evacuation efforts and proper resource allocation.
AI has also been useful because of its ability to generate “digital twins,” or virtual models of different landscapes and fire-prone areas where researchers can use the tools to simulate different wildfire scenarios. This helps them better understand the varying behaviors wildfires could take on depending upon changes in different variables. This again provides another tool that communities can use to improve their response plans and make better and faster resource decisions.
Once a wildfire is detected, AI is also able to analyze data quicker than traditional wildfire models, updating responders on spread and growth more effectively. In 2024, University of Southern California researchers created a computer algorithm that combines generative AI with satellite imagery, and it was successful in predicting a “fire’s most likely path, intensity, and growth rate. The researchers’ data analysis [also] showed how fire patterns were impacted by factors such as weather, terrain, and fuel.”
Arguably, the most crucial part in fighting wildfires is early detection. Falko Kuester, Ph.D., an engineering professor at the University of California San Diego, notes that “With shorter detection times, fire agencies can shift from a defensive stance of protecting actively threatened lives and property to an offensive one, where they can proactively manage a wildfire’s path before it becomes uncontrollable.”
AI is incorporated into early detection by being fed high-resolution images from devices like cameras, drones and satellites to visually detect smoke plumes, while also being sent data from sensors in high-risk areas collecting information on early indicators of potential fires, “such as extreme dryness, heavy brush, and high heat in an area.” When combined, AI tools consolidate this information to identify fires faster and with more baseline information on size and spread than traditional methods of spotting, like eyewitness accounts and public reporting.
New systems like this are being created and implemented every day, like the FireSat project, a constellation of fire-detection satellites currently being built and tested. FireSat was created from a collaboration between Google, the Earth Fire Alliance, the Gordon & Betty Moore Foundation and Muon Space. Once it begins operating, the system “will provide global high-resolution imagery that is updated every 20 minutes, enabling the detection of wildfires that are roughly the size of a classroom,” or about “1/400th the size of what current early detection satellites are capable of.”
San Francisco-based startup Pano AI and public safety program ALERTCalifornia are other great examples of these systems that are already successfully in use. Pano AI co-founder and CEO, Sonia Kastner, recounts how in 2023, Pano AI was deployed in the Washington State Jackson Road Fire, “alerting agencies within minutes of onset and reducing the time to put resources on-site by 20-30 minutes,” helping to contain the fire at “23 acres with no loss of life or structure.”
Finally, AI is also being deployed post-fire to assess damages and assist with cleanup. AI can be used to do several things post-disaster, including identifying affected areas, analyzing “the compromised structure integrity of affected buildings,” estimating the potential death toll, and assessing air quality, pollutant dispersion, and high smoke residuals to note “areas with elevated risks to the population.”
After the 2023 Maui wildfire, “an AI-driven computer vision program was developed to identify burnt vehicles in Lahaina and expedite debris removal.” The model utilized high-resolution satellite imagery captured directly after the fire, and it took approximately 8.5 hours to train. Using methods like this can be safer, quicker and cheaper than on-the-ground survey crews. Post vehicle removal, the model was found to achieve a 99.6% accuracy rate, predicting just ten more burnt vehicles than were actually found and removed.
The power of AI offers a huge opportunity for public safety and recovery from wildfires, particularly because of the technology’s ability to quickly synthesize large amounts of data and translate this into predictive analyses. This is becoming more crucial as wildfires are increasing and evolving in size and spread. Assad Oberai, Hughes Professor and Professor of Aerospace and Mechanical Engineering at USC Viterbi, eloquently summarizes the strong need for AI integration into this field by saying: “Wildfires involve intricate processes…highly complex, chaotic and nonlinear processes. To model them accurately, you need to account for all these different factors. You need advanced computing.”
Image via Unsplash.