|
|
|
|
|
6 February 2026
|
Using AI to help predict wildfires in Alaska
|
|
Alaska is one of the coldest states in the country so it may come as a surprise that it has experienced three of its four largest fire seasons in the last 25 years. But the combination of long summer days, high temperatures and global warming, along with dry vegetation, creates ideal conditions for destructive fires that threaten local communities and military training lands. Arizona State University researchers are addressing this growing risk through a US Army Engineer Research and Development Centre-funded plan that will use artificial intelligence to characterise a vast array of wildfire fuels and generate fuel maps at a much higher resolution than existing maps.
This will lead to improved forecasting and ultimately protect both civilian and military communities.
WenWen Li, the principal investigator for the effort, is an expert in environmental AI. The funder called the proposal the “Cadillac of solutions” that was needed. “We have extensive experience using AI for environmental mapping and monitoring,” said Li, a professor in the School of Geographical Sciences and Urban Planning. “This will be a new and important application that will spark new ideas. That’s what makes it so exciting.”
Li’s GeoAI team participants on the project include Chenyan Lu, a second year PhD student in geospacial information science, and Chia-Yu Hsu, a research professional who specializes in computer vision and image analysis.
ASU News spoke with Li, director of Spatial Analysis Research Centre or SPARC, about how AI and geospatial science will be used in this collaboration, why it matters and how it can “transform wildfire prediction and management — with insights that extend far beyond Alaska.”
Note: Answers have been edited for length and/or clarity.
Question: It makes sense that there would be wildfires in Arizona and California, areas that can be very hot and dry — but Alaska?
Answer: This is an interesting question. Alaska does burn and it can burn powerfully. In some years, such as 2022, Alaska accounted for nearly half of the total U.S. acreage burned by wildfires. Heat and drought are key drivers of wildfires in places like California and Arizona, but fuel is just as important. Alaska has vast forests, tundra regions, peatlands and other vegetation that provide abundant fuel. Although winters are cold, summers can be warm and very long — with nearly 24 hours of sunlight in some regions. This rapid drying of fuels makes the landscape highly flammable.
Q: What are high-resolution or AI-generated fuel maps, and what is their role in wildfire science?
A: Switching from traditional fuel mapping to AI is like going from a blurry road map to Google Earth in high definition. High-resolution fuel maps capture the details of vegetation and fuels — such as grasses, shrubs, fallen branches and tree canopy — at a much finer scale than current nationwide datasets. This level of detail is critical because fires spread based on small-scale fuel continuity and structure, and low-resolution maps often underestimate or misrepresent those risks. That makes the spatial resolution and quality of fuel maps one of the major gaps in wildfire science.
AI can help close this gap by combining data from satellites, aerial photos and topography to automatically map and describe fuels at high resolution, improve accuracy and apply them consistently from local to global scales. The new dataset will support better fire predictions, stronger planning and ultimately more resilient landscapes and communities.
Q: I understand how fuel maps identify fuel sources but how do scientists go from identifying these sources to using them to predict and hopefully prevent wildfires?
A: Scientists use AI-generated fuel maps along with weather and topography as inputs to fire behavior models. These models simulate how fast and in what direction a fire might spread. Fuel maps show what is actually burning, such as grass, shrubs or dense forest. Weather data add information about wind and humidity, and topography describes elevation and terrain changes. When combined, these inputs make it possible to forecast fire behavior and identify areas at higher risk.
But the accuracy of AI forecasts depends on the quality of the data.
Q: What makes this AI approach different from traditional mapping methods?
A: AI takes a smarter, data-driven approach that learns from large and diverse datasets, including satellites, aerial photos and topography to map fuels. A key advantage is that, because it can process so much information, cutting-edge AI models can extract generalizable knowledge that applies across different geographies and ecosystems. This makes AI more effective at producing accurate, high-resolution fuel maps, which in turn provide a clearer picture on wildfire risks.
Q: Given the accuracy with which the high-resolution mapping can identify vegetation that is fueling wildfires, can this approach also be used to stop fires that are already underway?
A: It can definitely help. The high-resolution fuel map provides detailed information to better model how fires spread in near real time. That said, stopping a fire once it’s underway is complex and requires close coordination with emergency response departments. Our data can support those teams in making faster, more informed decisions and can also guide long-term planning by identifying where vegetation management or firebreaks would most effectively reduce future risk.
Q: How will these maps help protect communities and military training lands in Alaska, and even beyond Alaska?
A: Wildfires are a growing concern in interior Alaska, especially on military training lands where managers are responsible for 1.6 million acres of fire-prone terrain. Right now, they lack detailed fuel maps that could help them anticipate risks and allocate resources more effectively. Our project uses AI to create high-resolution fuel maps to support fire-risk assessment, giving land managers the information they need to plan training activities more safely and to protect infrastructure and nearby communities.
What I particularly value about this project is that, in addition to advancing AI research for wildfire science, we are also working closely with land managers to translate our research outcomes into practice. And while our focus is Alaska, the AI methods we are developing can be adapted anywhere, helping protect communities and landscapes well beyond Alaska.
Source: Arizona State University
This will lead to improved forecasting and ultimately protect both civilian and military communities.
WenWen Li, the principal investigator for the effort, is an expert in environmental AI. The funder called the proposal the “Cadillac of solutions” that was needed. “We have extensive experience using AI for environmental mapping and monitoring,” said Li, a professor in the School of Geographical Sciences and Urban Planning. “This will be a new and important application that will spark new ideas. That’s what makes it so exciting.”
Li’s GeoAI team participants on the project include Chenyan Lu, a second year PhD student in geospacial information science, and Chia-Yu Hsu, a research professional who specializes in computer vision and image analysis.
ASU News spoke with Li, director of Spatial Analysis Research Centre or SPARC, about how AI and geospatial science will be used in this collaboration, why it matters and how it can “transform wildfire prediction and management — with insights that extend far beyond Alaska.”
Note: Answers have been edited for length and/or clarity.
Question: It makes sense that there would be wildfires in Arizona and California, areas that can be very hot and dry — but Alaska?
Answer: This is an interesting question. Alaska does burn and it can burn powerfully. In some years, such as 2022, Alaska accounted for nearly half of the total U.S. acreage burned by wildfires. Heat and drought are key drivers of wildfires in places like California and Arizona, but fuel is just as important. Alaska has vast forests, tundra regions, peatlands and other vegetation that provide abundant fuel. Although winters are cold, summers can be warm and very long — with nearly 24 hours of sunlight in some regions. This rapid drying of fuels makes the landscape highly flammable.
Q: What are high-resolution or AI-generated fuel maps, and what is their role in wildfire science?
A: Switching from traditional fuel mapping to AI is like going from a blurry road map to Google Earth in high definition. High-resolution fuel maps capture the details of vegetation and fuels — such as grasses, shrubs, fallen branches and tree canopy — at a much finer scale than current nationwide datasets. This level of detail is critical because fires spread based on small-scale fuel continuity and structure, and low-resolution maps often underestimate or misrepresent those risks. That makes the spatial resolution and quality of fuel maps one of the major gaps in wildfire science.
AI can help close this gap by combining data from satellites, aerial photos and topography to automatically map and describe fuels at high resolution, improve accuracy and apply them consistently from local to global scales. The new dataset will support better fire predictions, stronger planning and ultimately more resilient landscapes and communities.
Q: I understand how fuel maps identify fuel sources but how do scientists go from identifying these sources to using them to predict and hopefully prevent wildfires?
A: Scientists use AI-generated fuel maps along with weather and topography as inputs to fire behavior models. These models simulate how fast and in what direction a fire might spread. Fuel maps show what is actually burning, such as grass, shrubs or dense forest. Weather data add information about wind and humidity, and topography describes elevation and terrain changes. When combined, these inputs make it possible to forecast fire behavior and identify areas at higher risk.
But the accuracy of AI forecasts depends on the quality of the data.
Q: What makes this AI approach different from traditional mapping methods?
A: AI takes a smarter, data-driven approach that learns from large and diverse datasets, including satellites, aerial photos and topography to map fuels. A key advantage is that, because it can process so much information, cutting-edge AI models can extract generalizable knowledge that applies across different geographies and ecosystems. This makes AI more effective at producing accurate, high-resolution fuel maps, which in turn provide a clearer picture on wildfire risks.
Q: Given the accuracy with which the high-resolution mapping can identify vegetation that is fueling wildfires, can this approach also be used to stop fires that are already underway?
A: It can definitely help. The high-resolution fuel map provides detailed information to better model how fires spread in near real time. That said, stopping a fire once it’s underway is complex and requires close coordination with emergency response departments. Our data can support those teams in making faster, more informed decisions and can also guide long-term planning by identifying where vegetation management or firebreaks would most effectively reduce future risk.
Q: How will these maps help protect communities and military training lands in Alaska, and even beyond Alaska?
A: Wildfires are a growing concern in interior Alaska, especially on military training lands where managers are responsible for 1.6 million acres of fire-prone terrain. Right now, they lack detailed fuel maps that could help them anticipate risks and allocate resources more effectively. Our project uses AI to create high-resolution fuel maps to support fire-risk assessment, giving land managers the information they need to plan training activities more safely and to protect infrastructure and nearby communities.
What I particularly value about this project is that, in addition to advancing AI research for wildfire science, we are also working closely with land managers to translate our research outcomes into practice. And while our focus is Alaska, the AI methods we are developing can be adapted anywhere, helping protect communities and landscapes well beyond Alaska.
Source: Arizona State University