Anxiety-inducing images of skies hazy from wildfires and hillsides set ablaze are increasingly popping up on my Twitter feed and Instagram feed. The photos are posted by friends and colleagues worried about their homes, loved ones and dashed vacation dreams.
As hundreds of wildfires have sparked across the Western U.S. this summer, there’s a new interactive mapping tool available for firefighters and residents alike to track and respond to the fires.
The RADRFIRE tool uses infrared satellite imagery and artificial intelligence to create detailed wildfire maps to track and forecast fires. It was developed at the Pacific Northwest National Laboratory (PNNL) in Eastern Washington, in consultation with numerous agencies responsible for battling the fires — a job that keeps getting harder with worsening droughts and climate change. The Bootleg Fire currently burning in Southern Oregon is so fierce that it’s generating its own weather.
“We’re seeing much larger fires these days. They’re a lot more dynamic, faster moving. We can’t always use the traditional methods that we’ve used,” said RADRFIRE project lead Andre Coleman. “We need to up the game a little bit.”
Drawing on their experience in other areas of natural disaster response, the 14-member PNNL team began actively building the tool less than a year ago, in September 2020. It was released in May and updates are coming in August.
RADRFIRE received input from the U.S. Forest Service, the U.S. Bureau of Land Management, Washington’s Department of Natural Resources and Cal Fire.
Here are some details of the project, as recently described to GeekWire by Coleman:
Why build this tool? In the past, in order to track active fires, an incident commander would request an aircraft to fly over a site after dark to collect images — an exercise that’s limited by the availability of aircraft, safety concerns, and the ability to penetrate the smoke cover. The images were passed to an analyst to draw the maps by hand to be ready for use in the morning.
Firefighters have previously used satellite images, but employing infrared imaging and higher resolution sensors makes it much more powerful. (In earlier satellite images, a pixel could span 2 kilometers; now that’s down to 30 meters.)
One of the new tool’s capabilities is to identify small spot fires that are sparked by lightning strikes before they become larger and harder to fight.
What’s the information used for? The images help commanders make decisions on how to best deploy their resources, keep firefighters safe, predict a fire’s spread and identify structures in its path. It can aid municipalities and others who are calling for evacuations of people in harm’s way.
It also helps utilities understand what infrastructure is at risk. Power companies have limited access to areas with active fires, so the images can keep them better informed and help with quicker responses once they get access again. The Bootleg fire, for example, has threatened essential power transmission lines carrying electricity to California.
What about the non-headline grabbing fires? Because it’s automated, the RADRFIRE system can pull and analyze information on small and large fires alike. Firefighting resources are concentrated on the largest, most complex type 1 fires, but there are type 2, 3 and 4 fires that also impact communities and might be more remote. Making smart use of the few resources available to smaller blazes is crucial.
“This is really a huge benefit to folks working on these type 3 and type 4 fires,” Coleman said. “They get a whole lot less attention, a lot less resources to work on them.”
Whose satellites are these? Who owns the system? The tool uses information from open-access domestic and international government satellites. The team worried that using commercial satellites would quickly become too costly and limit who could use the tool.
“We just feel like this should be open-access information,” Coleman said. “Somebody shouldn’t have to have the purchase it.”
Because the team is piggybacking on others’ satellites, they take what images they can get so some areas are photographed more or less than others and at random times. The project’s goal is to capture a couple of images a day, which is added to information gathered through other means to create a more complete picture.
Coleman expects that the program will continue to be managed by a government agency. The U.S. Department of Energy’s Artificial Intelligence Technology Office has supported it in collaboration with the U.S. Department of Defense’s Joint Artificial Intelligence Center.
How does RADRFIRE contribute to the science of firefighting? Working with the Forest Service, the team has developed an algorithm to better plan and monitor the use of fire retardants that are dropped on blazes by aircraft. The tool can quickly determine where past drops were planned, where they actually landed and spread, and their effectiveness. In the past commanders were able to gather some of this information, but it took more time and resources.
Could RADRFIRE help with fire prevention? One area of investigation is into the use of LIDAR imaging, which can capture information about the nature and volume of vegetation in an area. The data could highlight areas with dense, at-risk trees and shrubs that might be well suited for targeted controlled burns or thinning to reduce the fuel available for wildfires.
This is modeling, AI, very mathematical stuff. But what about the human and environmental impacts of this work? “We’re all absolutely driven by this. It is not unusual during these events where we’re all working 17, 18 hours a day. Nobody complains because we’re all just driven by trying to do better, to do social good,” Coleman said. “For everybody on the team, this is personally important.”
