Climate change is going to get a lot worse if we don’t act soon. While that’s not breaking news, it was the thrust of a landmark report by the Intergovernmental Panel on Climate Change, released in August. The IPCC’s report warned that every region of the planet will be affected by rising temperatures.

A new paper, published in the journal Nature Climate Change on Oct. 11, adds some specificity to that forecast. Using machine learning technology to analyze over 60,000 climate change-related studies, researchers in Germany estimate that 85% of the population is affected by human-induced climate change. The study was led by Max Callaghan of Berlin’s Mercator Research Institute on Global Commons and Climate Change.

«There is overwhelming evidence that the impacts of climate change are already being observed in human and natural systems,» the paper reads. «We infer that attributable anthropogenic impacts may be occurring across 80% of the world’s land area, where 85% of the population reside.»

The study comes ahead of COP26, the UN Climate Change Conference in Glasgow that runs from Oct. 31 to Nov. 12. COP26 will bring together world leaders, including US President Joe Biden and UK Prime Minister Boris Johnson but notably not China’s Xi Jinping, in hopes they’ll make new commitments to lowering carbon emissions. It was at COP21 in 2015 that the Paris Accords were struck, and observers hope that more ambitious commitments to carbon neutrality can be agreed to in Glasgow.

Machine learning is a type of artificial intelligence that gets smarter as it’s fed more information: think speech-to-text software, which gets more accurate the more hours of voices it’s able to hear. Callaghan and the team aimed not just to highlight the planet’s plight as climate change’s impacts become more known, but also to use machine learning to reveal gaps within scientific study.

The researchers fed machine learning software called BERT (or «bidirectional encoder representations from transformers») 2,373 abstracts on papers related to climate change. Having digested the information on climate change, the algorithm then identified studies that may show the impacts of climate change even if those studies didn’t attribute their findings to climate change. The paper referenced one such study on the relationship between the timing of snowmelt and the population growth of mammals.

«Our objective is to map all possibly relevant studies on climate-related changes, rather than a list of studies where the relationship between an observed climate trend and specific impacts has been demonstrated with high confidence,» the paper reads. «While traditional assessments can offer relatively precise but incomplete pictures of the evidence, our machine-learning-assisted approach generates an expansive preliminary but quantifiably uncertain map.»