Google Invests in AI-Powered Satellites for Swift Wildfire Detection

Google Research has collaborated with fire community leaders to introduce a groundbreaking satellite constellation called FireSat. This innovative system utilizes AI technology and satellite imagery to detect and track small wildfires within a mere 20 minutes after ignition. FireSat is capable of identifying fires as small as a classroom, approximately 16 by 16 feet in size. Once a fire is spotted, the system provides real-time information on the fire's location, size, and intensity, enabling firefighters to respond promptly and effectively.

The increasing prevalence of wildfires due to hotter and drier climates underscores the necessity for improved detection mechanisms. Currently, firefighters rely on low-resolution satellite imagery that is updated infrequently, making it challenging to identify fires until they have expanded significantly. The development of FireSat involved close collaboration between the Google Research team, Muon Space, and the Environmental Defense Fund. Custom infrared sensors for the satellites were developed to enhance the detection capabilities for small-scale fires.

Utilizing AI technology, FireSat quickly analyzes specific areas on Earth, comparing them to past imagery and considering variables like nearby infrastructure and local weather conditions to identify potential fires. The detection model for smaller fires was validated by flying sensors over controlled burns to establish a baseline dataset for the AI. The first satellite in the FireSat constellation is scheduled for launch early next year by Muon Space, with the full constellation to follow in subsequent years.

The Earth Fire Alliance, a nonprofit organization dedicated to launching FireSat, will lead this initiative with support from the Moore Foundation and $13 million in funding from Google.org. In addition to enhancing emergency response efforts, the data collected by FireSat will contribute to creating a global historical record of fire spread. This historical data will aid Google and scientists in developing comprehensive models to better understand wildfire behavior, expanding upon Google's existing work on fire simulation in collaboration with the U.S. Forest Service.

Google's commitment to wildfire detection is apparent in its efforts since 2020, including the use of AI models for wildfire boundary tracking to provide alerts and crucial information to communities in close proximity to wildfires in over 20 countries. Recently, Google released FireBench, an open-source machine learning benchmark dataset for wildfire research, as part of its ongoing dedication to supporting emergency responders and scientists in addressing the escalating threat of wildfires.