A groundbreaking study reveals an AI that can track melting icebergs 10,000 times faster than humans. The AI identifies icebergs in satellite images in less than 0.01 seconds, a task that takes humans several minutes.
AI's role in monitoring iceberg shrinkage
With the looming threat of climate change, it's more crucial than ever to keep a close eye on our planet's icy giants. Scientists are tapping into the potential of artificial intelligence to scan through satellite images, identifying and tracking the changes in icebergs over time. Unlike the typical approach that requires a human to painstakingly outline these structures in an image - a process that takes a few minutes - this AI system can do the job in less than 0.01 seconds. That's a whopping 10,000 times faster.
To achieve such efficiency, researchers trained the AI on images from the European Space Agency's Sentinel-1 satellite, which is capable of capturing Earth's surface regardless of cloud cover or lack of light. Despite some minor misses with icebergs larger than the AI's training examples, the system demonstrated a staggering 99 percent accuracy. It even correctly identified icebergs ranging from 54 square kilometers (roughly the size of Switzerland's city of Bern) to 1052 square kilometers (about the size of Hong Kong).
Operational potential of AI in iceberg tracking
The advantages of using AI for iceberg tracking are significant. Not only does it offer greater speed and accuracy, but it also eliminates the common mistakes made by other automated approaches, like miscategorizing individual ice bits as one collective iceberg. According to Anne Braakmann-Folgmann, the lead author of the study, this advance 'paves the way for an operational application,' meaning it could soon become a standard tool for climate scientists.