Scientists use machine learning to map how carbon sinks in the deep ocean

In short: Scientists combined a robotic underwater camera with machine learning to map how carbon-carrying particles sink from the ocean surface to deeper water.

What happened

Researchers deployed an autonomous underwater imaging system, basically a robotic camera that can take constant, high-resolution photos while submerged. It photographed tiny bits of material drifting and sinking through the water, including “marine snow” (small clumps of dead plankton, mucus, and other organic debris).

The robot produced millions of images. The team used machine learning (software that learns patterns from examples, like a spam filter that learns what junk mail looks like) to spot and label different particle types in those photos. This helped them measure particle size, how many were present, and how much material was moving downward over time.

Together, the camera and the software let the scientists build detailed, 3D maps for specific ocean areas. These maps showed where carbon-rich particles were sinking, including narrow “jets” and “filaments” (thin streams in the water that can act like fast lanes for sinking material).

Some headlines have described this as mapping Earth’s entire “carbon circulatory system.” That is not what the study did. It focused on one important piece, how the ocean’s “biological carbon pump” moves carbon from surface waters into the deep sea.

Why it matters

The ocean can store carbon for a long time if that carbon sinks deep enough before it breaks down. Better maps of where and when this sinking happens can improve climate forecasts, because it helps scientists estimate how much carbon the ocean is likely to keep out of the air.

Source: NYTimes