At Google I/O in May, Google revealed that it's working with the Earth Fire Alliance on FireSat, a program that combines new high-resolution satellites with AI analysis to pinpoint wildfires in their earliest stages and help responders knock them down before they grow. This week the alliance released the first images captured by the initial satellite, showing how fires as small as 5-by-5 meters -- about the size of a classroom -- can be detected from space.
FireSat identified this small roadside fire in Oregon in June 2025. Muon Space and Earth Fire Alliance
Existing satellite systems scan for fires, but at a coarser resolution. In one image from Oregon, using MWIR (Mid-Wave Infrared) heat-sensing imaging, a small roadside fire showed up as a bright speck. According to the alliance, it wasn't detected by other space-based systems.
Using up to six infrared channels, FireSat can detect new fires as well as hot burn scars from earlier fires, as shown in this June 15 image from Ontario, Canada. Muon Space and Earth Fire Alliance
This example from Ontario, Canada, on June 15, 2025, shows the Nipigon 6 fire, a new blaze detected using the MWIR spectrum, but it also shows how LWIR (Long-Wave Infrared) was used to identify areas left over from a previous burn in 2020, which are heated due to a lack of vegetation. At the bottom, a false-color composite of SWIR (Short-Wave Infrared), NIR (Near-Infrared) and visible Red channels helps track the life cycle of the fire.
Currently, the Earth Fire Alliance has one protoflight satellite, built by Muon Space, aloft for testing. With three satellites in orbit, FireSat will be able to scan locations globally twice a day. And when the program is fully operational, in 2030, a network of more than 50 satellites is expected to cut that time down to 20 minutes; for areas that are more prone to fires, that interval will be every 9 to 12 minutes.
One key reason for Google's involvement in the alliance is to sort through the massive amount of data that will be generated. Muon Space estimates that each satellite will cover 190 million square kilometers per day, and the multispectral instrument on each satellite records across six channels. With AI and software assistance from Google, the program should filter out false positives. AI is playing a larger role in fighting wildfires around the world -- NASA is using its vast trove of Landsat satellite data to build predictive models of where fires are likely to erupt next.
"There are millions of things that can be mistaken for a fire," said Chris Van Arsdale, Google Research climate and energy lead and chair of the Earth Fire Alliance board of directors. "Looking for fires becomes a game of looking for needles in a world of haystacks."
FireSat can locate wildfires in areas too remote to be detected by many other methods, as seen in this example showing Alaska on June 21. Muon Space and Earth Fire Alliance
It will also be important to prioritize fires that crews can respond to. A June 21, 2025, image of a remote area of Alaska shows a fire that wasn't observable by ground-based sources.
... continue reading