Algorithm predicts location of meteorites

It looked like a football on the centre spot just waiting to be kicked. It lay in the open, bar a few sprinkles of snow, on the immense plain of ice. “We saw it from a couple of hundred metres away,” says glaciologist Harry Zekollari, who until recently worked at the Faculty of Civil Engineering and Geosciences and now works at ETH Zurich. “We were as excited as children who had just discovered that the Easter bunny had left a huge surprise for them in the garden.”

Nine years ago during an expedition to Antarctica, Zekollari and his colleagues from the Laboratoire de Glaciologie at the Université Libre de Bruxelles (where Zekollari also has an appointment) found a meteorite weighing an incredible 18 kilogrammes. The space rock, a chondrite from the asteroid belt between Mars and Jupiter, is now on display at the Museum of Natural Sciences in Brussels. It crashed into the continent about 16,000 years ago. After a period of millenia during which it was being moved around deep in the ice, it ultimately surfaced 100 kilometres south of the Belgian Antarctic research station, the Princess Elisabeth Base.

‘Everything has to come together precisely’

By studying meteorites, scientists hope to learn more about the origins of the solar system. Antarctica is a gold mine as there are still between 300,000 and 900,000 samples to collect. At least, they may lie on the surface but you still need to know where to find them. The researchers were lucky in 2013 when they formed a line and combed a large area on snow scooters.

Artificial intelligence
Up to now, researchers have had to rely on their gut feeling when they went looking for meteorites. But artificial intelligence will change this. In the journal Science Advances, a team of researchers, including Zekollari, Stef Lhermitte (CEG) and first author Veronica Tollenaar (a TU Delft alumna who is now doing a PhD at the Université Libre de Bruxelles) reported on a calculation programme that is said to predict where meteorites can be found with an 80% accuracy rate. “We have actually made a kind of treasure map,” says Zekollari.

• That treasure map is accessible to everyone and is available here.

Meteorites are always found on the blue ice sheets that cover about 1% of the continent. The ground beneath the ice is marked by protruding rocks that have the effect of pushing the ice upwards. The top layer of ice is continuously pushed and polished by the strong wind and the meteorites pile up there.

Nevertheless, in an area like this finding a meteorite is like looking for a needle in a haystack. “Everything has to come together precisely,” adds Stef Lhermitte. “The wind needs to be strong enough to erode the ice at the right speed and the ice needs to have the right flow velocity. The topography of the underlying ground and the ice cap and the temperature are also relevant factors.”

The researchers have input all these factors into a self-learning algorithm and validated it by checking if the calculation programme could identify areas – that were not used in the algorithm itself – which field studies already know are rich in meteorites. It transpired that this was the case. They now have to test it on the continent itself. A Belgian expedition was ready to go last month but was cancelled at the last minute because of a Covid outbreak in the research station. “We hope that a mission can depart next year,” says Zekollari. “And researchers from Aerospace Engineering may be able to work on it too.” 

‘This is a pleasant trip outside our own field of work’

“The idea is to also deploy drones”, confirms meteorite expert Sebastiaan de Vet, of the Astrodynamics and Space Missions group and head of the Delft Meteorite Lab. “We want to see if they will speed up the search process.” But there are major challenges ahead. Strong winds make flying difficult and the batteries run out quickly in freezing temperatures. The researchers also need to train an algorithm to help them recognise meteorites on drone images. As the sun is low on the horizon, even the smallest protrusion on the ice casts a shadow. The algorithm needs to be trained to distinguish meteorites from shadows.

Lhermitte and Zekollari are not meteorite experts. “Our field is glaciology,” says Lhermitte. “But this will be a very pleasant trip outside our own field of work.”