The company said in a blog post that it is teaming up with a satellite maker and a group of nonprofits to launch satellites next year that can pinpoint small blazes all over the world before they have gobbled up acres of land.
"Until now, firefighters have had to rely on satellite imagery that's either low resolution or only updated a few times a day, making it difficult to detect fires until they've grown larger than a soccer field," said the post, written by Google Research Climate & Energy Lead Christopher Van Arsdale.
Using low earth orbit satellites from Mountain View manufacturer Muon Space equipped with Google-made custom infrared sensors, the constellation of satellites dubbed FireSat will give fire authorities a picture of fire zones updated every 20 minutes, Google said.
Van Arsdale said the technology will allow fire spotters to catch blazes that are only 5 meters across, about the dimensions of a school classroom, before they grow to the size of a few football fields, which early detection satellites can currently see.
Google said in an email the satellites will use infrared sensors along with other imaging bands to detect wildfires and filter out false positives. The satellites will also monitor burn scars and estimate other metrics like how much heat energy a fire is putting out.
The plan is to send up the first satellite next year, with the aim of eventually putting more than 50 satellites in orbit. Google said in an email the first satellite is scheduled to go up on a SpaceX rocket early next year.
Google is contributing $13 million to the effort, with additional funds from the Gordon and Betty Moore Foundation. The initiative will be led by the Earth Fire Alliance, a nonprofit established in May to launch the grouping of satellites.
The plan is to create a "consistent, accurate, and comprehensive view of fire activity for every fire on Earth," according to the alliance's website.
Having frequently updated, high-quality satellite imagery then allows for the use of AI to automatically compare any spot on the planet with what it looked like previously to detect small conflagrations.
With the satellites taking so many images, processing them and comparing them with previous pictures usually means there is a lag in detecting fires in near real time. Google said some of the image processing will be done on the satellite itself, and that the AI model for doing so will build on previous programs it has developed for programs including wildfire boundary tracking, FireSim and FireBench.
Google said it has already flown its sensors over controlled burns to establish a baseline of images for fire detection.
Other Google technology is already in use to model wildfire behavior, and the company has partnered with the U.S. Forest Service to predict fire spread based on conditions.
Other companies, including the European company OroraTech, already have satellites detecting wildfires, combining their images with those of government satellites. Their scan areas tend to be at the kilometer level, far bigger than the 5-meter areas that Google is promising to zoom in on.
For its part, Cal Fire has used AI-powered fire modeling software for a half decade or more. Last year the state agency upgraded its ability to predict a fire's behavior with technology built by San Jose State University researchers called the Weather Research Forecasting System.
That technology can predict where burning material from fires will travel and ultimately land depending on fire behavior and weather conditions. Such predictions permit firefighters to move into position and potentially get ahead of blazes before they can spread.
© 2024 the San Francisco Chronicle. Distributed by Tribune Content Agency, LLC.
