Published in the journal Science, their study is a first-of-its-kind global overview of heavy metal contamination in soils, based on a meta-analysis -- an approach drawing on data from many previous studies -- of almost 800,000 samples.

After ensuring the reliability and representativeness of the data, for example by ruling out samples taken purposefully at contaminated sites, the researchers used machine-learning algorithms to identify the worst-affected areas of the world.

The team led by environmental specialist Deyi Hou of China's Tsinghua University focused in on areas where concentrations of at least one of seven metals -- arsenic, cadmium, cobalt, chromium, copper, nickel and lead -- were above recommended safe limits for agriculture and human health.

Metals can be toxic to people, animals and plants at different doses, spreading into different ecosystems through food chains and water.

By running their samples through artificial intelligence-assisted analysis, the researchers found that between 14 and 17 percent of arable land is contaminated by at least one of the metals.

Between 900 million and 1.4 billion people live in "high-risk areas."

Heavy metal contamination can stem from both natural geological processes and human activity such as industrial waste, agriculture or mining.

The researchers noted that there was not enough data in some regions, especially Africa, to allow for designing targeted programs to pare down risks.

Rather, their study is intended as "a scientific alert for policymakers and farmers to take immediate and necessary measures," they wrote.

In fact, "the actual extent of global soil pollution may far exceed what is presented by the authors, due to limited data availability and likely underestimation," commented Wakene Negassa, a soil chemist at the James Hutton Institute.