Scientists use AI to track icebergs from birth to break up for the first time
Researchers at the British Antarctic Survey (BAS) have developed a new artificial intelligence tool that can, for the first time, automatically track the lifecycles of icebergs as they drift, melt and fragment across the world’s oceans – reconstructing detailed “family trees” of these massive ice formations.
Using satellite imagery, the system identifies individual icebergs, assigns each a unique identity and follows it over time. Crucially, when a large iceberg fractures, the AI can link the scattered “child” fragments back to their original “parent” iceberg – something that has not previously been possible at scale.
This matters because when icebergs melt, they release huge volumes of fresh water into the ocean. Where that fresh water enters the sea can reshape ocean currents, influence marine ecosystems and affect global climate patterns. Until now, scientists have struggled to track icebergs once they split into smaller pieces, leaving a major blind spot in climate models.
Until recently, scientists have tracked only the largest named icebergs by visually following them one by one via satellites. As a result, the fate and impact of the thousands of smaller fragments that break away across the Southern Ocean has largely remained unknown.
Ben Evans, lead author and machine learning expert at the British Antarctic Survey said:
“What’s exciting is that this finally gives us the observations we’ve been missing. We’ve gone from tracking a few famous icebergs to building full family trees. For the first time, we can see where each fragment came from, where it goes and why that matters for the climate.”
The system works by analysing the distinctive geometric shapes of icebergs captured in satellite images. When an iceberg fractures, the AI effectively performs a digital “jigsaw puzzle,” piecing together which fragments once belonged together. The system was tested on real satellite observations of Arctic icebergs calved from Petermann Glacier and other parts of north-west Greenland.
By revealing where iceberg meltwater is released into the ocean, the method provides vital new information that can be fed into global climate and ocean models, improving predictions at a time when ice loss from polar ice sheets is expected to increase as the climate warms. Data will feed into the NEMO ocean model that is part of the UK Earth System Model.
The approach lends itself to a range of applications including navigation, helping ships operating in polar waters better understand where hazardous ice is coming from and how it is likely to move.
The research was funded by EPSRC Grant EP/Y028880/1, The Alan Turing Institute, and the Polar Science for a Sustainable Planet programme at the British Antarctic Survey.