California’s fire department wants to put out forest fires before they get too big. To do this, she is testing a system that works with artificial intelligence – a model for other regions?
To fight a fire, you first have to know that it is burning. In California this can be a challenge. The US state is slightly larger than Germany and has half as many inhabitants.
Until now, fire departments had to rely on a fire being reported by telephone, for example. In remote areas there are observation towers with security guards to view the fires.
Cameras instead of fire towers
But: “The fire towers that were common, similar to lighthouses in Germany, are becoming fewer and fewer,” says Falko Küster from the University of California, San Diego. The Berlin native is a big data specialist and is involved in a “ALERTCalifornia” project that aims to detect fires more quickly with the help of artificial intelligence. Instead of watchful eyes, hundreds of cameras now help.
“We currently have over 1,030 recording cameras in California. The cameras are designed so that they can record virtually continuously with a 360-degree viewing angle – with up to 1,000 images per second,” explains Küster. “If you take that a step further: Each of the images has two million pixels in it, so that’s billions of data points. And that’s where artificial intelligence comes in – we’ve trained algorithms to look for corresponding signals.”
Fast and smart AI
AI can detect smoke faster and better. If she discovers smoke, the fire department is alerted.
When it comes to fires, every minute counts, Fire Inspector Zachary Wells of the Kern County Fire Department told NPR: “The more time we give firefighters, the greater our chance of containing these incidents before they become large, complex fires to contain Hundreds, if not thousands, of firefighters will be required.”
The AI learns with every use. When a fire is detected, the system receives feedback and continues to improve.
“Fires that no one ever hears about”
In addition to the cameras, there are also sensors – for example thermal imaging cameras. And the system can not only detect fires better, but also try to predict where a fire will move.
Sexton explains: “If you can calculate in advance how quickly a fire will spread, you can prepare. The firefighters say: ‘Don’t tell me where the fire is, I can see that. Describe to me where it will be in an hour will be.’ Because then I can send my people there, create a breach and, if necessary, evacuate people.”
Dozens of fires were contained at an early stage as a result. “The biggest success story we can tell about are the fires that no one ever hears about,” says Küster proudly.
A model for other regions?
The wildfires in California have become more severe in recent years due to climate change, climate researchers say. The state is spending millions of dollars fighting fires and trying to protect itself with the most modern means.
The AI firefighting project is said to be unique in the world – but not exclusive. The data is open source, meaning it can be viewed by everyone: “The goal is to create a framework model that can be copied elsewhere,” says Küster.
Because, as this summer shows with violent forest fires in Greece, Canada and Hawaii: the topic is omnipresent.