MACHINE LEARNING MACHETE : GLOBALLY REMARKABLE ESSAY

 

NEW WAY of identifying wood runs rings around old approach : Using a unique combination of old-fashioned field work and sophisticated computer modeling, scientists in Sweden have found a way to trace a single beam of lumber to the forest in Europe where it originated.

The researchers said the new method, described in a recent paper in the Nature Plants journal, could significantly curb the sale of Russian timber, which is prohibited in the European Union because of the war in Ukraine.

But birch, oak, pine and other types of lumber from Russia are still finding European buyers and surging demand.

The novel approach was recently used to identify large shipments of illegal Russian lumber in Belgium.

The new study looked at the chemical composition of 900 wood samples collected from 11 countries in Eastern Europe. The data was fed into a model powered by machine learning, which found patterns that could predict the geographic origin of the samples.

OVERALL, the model caught 60 percent of the samples that had been intentionally labeled with the wrong country of origin.

The model could also narrow the wood's origin in a roughly 125-mile [ 200 kilometers] radius, a remarkable feat in a continent that's roughly 40 percent covered by forest.

The method is '' very, very solid from a technical point,'' said Naren Ramakrishnan, a data scientist at Virginia Tech who was not involved in the research.

Under the direction of Victor Deklerck, a lead author of the study, researchers with Preferred by Nature, a nonprofit organization based in Copenhagen, fanned out across Europe to collect tree samples by using a long, tubelike device that pulls out wood tissue.

A tree is not harmed when a sample is extracted from its trunk, Dr. Deklerck said, because the rest of the '' walls off '' the wounded tissue.

The samples were analyzed for the minerals they pulled in from the soil, as well as elements, like nitrogen and carbon, that they absorbed through rainfall.

The result was a '' chemical fingerprint '' for each free sample in the study, said Dr. Deklerck, who is also chief scientist at World Forest ID, a nonprofit organization in Washington that fights deforestation.

The World Students Society thanks author Alexander Nazaryan.