Machine learning points to prime places in Antarctica to find meteorites

The hunt for meteorites may have just gotten some new leads. A powerful new mac

The hunt for meteorites may have just gotten some new leads. A powerful new machine learning algorithm has identified over 600 hot spots in Antarctica where scientists are likely to find a bounty of the fallen alien rocks, researchers report January 26 in Science Advances.  

Antarctica isn’t necessarily the No. 1 landing spot for meteorites, bits of extraterrestrial rock that offer a window into the birth and evolution of the solar system. Previous estimates suggest more meteorites probably land closer to the equator (SN: 5/29/20). But the southern continent is still the best place to find them, says Veronica Tollenaar, a glaciologist at the Université libre de Bruxelles in Belgium. Not only are the dark specks at the surface starkly visible against the white background, but quirks of the ice sheet’s flow can also concentrate meteorites in “stranding zones.”

The trouble is that so far, meteorite stranding zones have been found by luck. Satellites help, but poring through the images is time-consuming, and field reconnaissance is costly. So Tollenaar and her colleagues trained computers to find these zones more quickly.

Space rocks’ road

This diagram shows what happens when a slowly creeping ice sheet, with meteorites (black dots) embedded in deeper layers, encounters a topographic rise such as a mountain. That obstacle bends the ice sheet’s layers upward, concentrating the space rocks embedded in them into a meteorite stranding zone. In regions where snow turns into water vapor (red arrows) faster than it accumulates, called blue ice areas, these stranding zones are particularly visible.

How meteorite hot spots form
diagram showing how meteorite hot spots form in Antarctica
Veronica Tollenaar

Such stranding zones form when the slow creep of the ice sheet over the land encounters a mountain or hidden rise in the ground. That barrier shifts the flow upward, carrying any embedded space rocks toward the surface.

Combining a machine learning algorithm with data on the ice’s velocity and thickness, surface temperatures, the shape of the bedrock and known stranding zones, Tollenaar and colleagues created a map of 613 probable meteorite hot spots, including some near existing Antarctic research stations.

Veronica TollenaarVeronica Tollenaar

Tags:to,Machine,learning,