The 2023 wildfire season in Canada burned more than 18 million hectares. The Maui fire that August killed 101 people in Lahaina, Hawaii, spreading so fast that many residents had minutes, not hours, to evacuate. In both cases, detection came too late. By the time satellites confirmed the fires and alerts reached the ground, the blazes had already grown beyond the point where early intervention could have contained them.
FireSat is designed to close that gap. A joint effort between Google Research and the Earth Fire Alliance (EFA), the project is building a constellation of satellites that can detect a fire as small as 5 meters by 5 meters, roughly the size of a classroom, anywhere on Earth. The system will scan the entire planet every 20 minutes, feeding data through AI models that distinguish genuine fires from false positives like industrial flares, sunlight reflections, or hot rooftops.
Why Current Detection Falls Short
Today's satellite fire detection relies primarily on two NASA instruments: MODIS and VIIRS. Both are aging. MODIS, launched in 1999 and 2002, detects fires at a resolution of about one kilometer, meaning a fire must be large before it registers. VIIRS, launched in 2011, improved this to 375 meters but still revisits any given location only twice a day. A fire that starts at noon in a remote area might not be detected until the satellite passes overhead hours later, by which point it could have consumed thousands of acres.
Geostationary satellites like GOES offer near-continuous coverage but sit so far from Earth that their spatial resolution is measured in kilometers. They can see large fire fronts but not the critical first ignition point. Ground-based camera networks and human lookout towers cover only a fraction of fire-prone land.
How FireSat Works
FireSat uses a constellation of small satellites in low Earth orbit, each equipped with infrared sensors tuned to detect the thermal signatures of fire. The constellation is designed so that any point on Earth is within view of at least one satellite at all times, enabling the 20-minute revisit rate. Google contributed $13 million through Google.org to help fund the development and launch.
The raw thermal data streams down to ground stations where AI models process it in near real time. These models are trained on thousands of confirmed wildfire events, satellite imagery, and environmental data including vegetation dryness, wind patterns, and historical fire behavior. The AI doesn't just detect that something is hot. It classifies whether the heat signature is consistent with an active wildfire, estimates its size, and predicts its likely spread direction based on terrain and wind.
The first satellite in the constellation launched in 2025, and initial data from the engineering phase confirmed the sensor sensitivity targets. The full constellation is expected to be operational by 2027.
From Detection to Action
Early detection matters because wildfires grow exponentially. A fire that covers one acre in its first ten minutes can cover ten acres thirty minutes later and a hundred acres within a few hours under dry, windy conditions. Fire agencies describe the first hour as the "golden hour" because suppression efforts during this window are dramatically more effective and less costly than fighting an established blaze.
The Earth Fire Alliance is working with fire agencies in the United States, Australia, Brazil, and several Mediterranean countries to integrate FireSat alerts into existing dispatch systems. The goal is for a detection event to trigger an alert on a fire commander's screen within minutes, accompanied by the fire's precise coordinates and an AI-generated estimate of its current size and trajectory.
The Economics of Prevention
Wildfire suppression in the United States alone cost over $3.1 billion in 2023. The economic toll including property damage, health impacts from smoke, and lost productivity pushes the annual figure past $100 billion. Catching fires when they are small doesn't eliminate these costs, but modeling by EFA suggests that reducing average detection time from hours to minutes could cut total burned area by 25 to 40 percent in fire-prone regions.
The insurance industry is watching closely. As fire risk makes homeowner coverage unaffordable or unavailable in parts of California and Australia, better detection technology could help rebuild the actuarial models that underpin the market.
What Comes Next
FireSat is not the only player in the space. Planet Labs, Muon Space, and OroraTech are all building or operating fire-detection satellites. But FireSat's combination of resolution (5 meters), revisit time (20 minutes), and global coverage makes it the most ambitious effort to date. The fact that it is backed by a philanthropic coalition rather than a single commercial entity also means the data is intended to be shared freely with fire agencies worldwide, including those in developing nations where satellite data access is limited.
Five meters by five meters. Twenty minutes. Everywhere on Earth. If FireSat delivers on these specifications, it will fundamentally change how humanity fights wildfire.
Sources: Google Research, "Introducing FireSat" (2024); Earth Fire Alliance official documentation; Google.org announcement of $13M commitment; NASA FIRMS (Fire Information for Resource Management System); National Interagency Fire Center suppression cost data (2023).
