Illinois-led study uses machine learning to read 25,000 years of grass pollen
Researchers found grass diversity on Mount Kenya dropped sharply in the coldest stretch of the last ice age.
CHAMPAIGN - Researchers led by University of Illinois Urbana-Champaign plant biology professor Surangi Punyasena used machine learning and super-resolution microscopy to tell apart grass pollen grains that look nearly identical under a standard microscope, the university's News Bureau reported Oct. 5. The work appears in the Proceedings of the National Academy of Sciences.
Punyasena led the research with former doctoral student Marc-Élie Adaimé, now a postdoctoral researcher at the Smithsonian. Their model cannot identify individual grass species in a sample, but it can estimate how many species are present, the News Bureau said.
The team applied the method to a core holding 25,000 years of pollen from a lake bottom on Mount Kenya. Pollen diversity was substantially lower during the last ice age, especially between about 21,000 and 18,000 years ago, and rose afterward along with carbon dioxide levels and temperatures, Adaimé said.
"But the level at which we were able to analyze it before machine learning was limited by human ability," Punyasena said.
Source: University of Illinois News Bureau, Oct. 5, 2026.