Two climate scientists suggest they’ve come closer to resolving a critical debate about how quickly human activity will heat up the planet. The answer isn’t good news.

It’s almost universally understood that the Earth will continue to get warmer for the foreseeable future. The rate at which the planet warms, however, won’t remain the same, report Cristian Proistosescu and Peter Huybers of Harvard University. They say it’s likely to speed up.

Some parts of the planet heat up more slowly than others, they explain. But as more time passes, regions once less affected by global warming will get hotter. Thus, the bulk of planetary warming this century may actually be back-loaded onto its final decades.

The analysis, published in Science Advances, addresses the gap between two long-battling camps struggling to understand how quickly the world will warm. One group looks at the historical record and projects into the future all the warming that has already been known to occur, mainly through direct observations. Those studies have found that, once there is twice as much carbon dioxide in the atmosphere as there was prior to the Industrial Revolution, the temperature may rise between 1.6 degrees and 3 degrees Celsius (2.9 degrees to 5.4 degrees Fahrenheit).

Those estimates, while unnerving, are significantly lower than the projections generated by climate models held up by the other group. These are built from equations supplied by Earth physics and allow scientists to do something they can’t do in the real world: simulate how the planet behaves under various conditions over enormous periods of time. Computer models are the lab rats of climate science.

The contradictory evidence posited by the two camps was canonized in 2013, when the most authoritative climate-science group, the Intergovernmental Panel on Climate Change, broke with its own practice and declined to provide a single “best estimate” of warming. Instead it punted, providing only a range (from 1.5C to 4.5C) that extended the low end of the conflict below previous assessments.

Climate change may have two speeds, depending where you are on Earth.

Why does this fight over a few degrees really matter? It’s crucial to estimate the warming rate as precisely as possible because the answer will dictate how aggressively global policymakers respond to the problem.

However, each scientific camp relies on different analytical methods, making comparisons difficult. The new study seeks to address that problem by breaking down previous studies into smaller components—an attempt to provide an “apples to apples comparison,” Proistosescu said.

The authors conclude that it’s worth considering two speeds of climate change. So far, the world has been in “fast” mode, in which the regions likeliest to heat up fastest are showing the greatest increases in temperature. That includes land in the Northern Hemisphere. When it comes to analyzing global warming in the fast lane, both the observation camp and modeling camp agree, Proistosescu said.

There is also a “slow” mode. Places like the eastern tropical Pacific Ocean and the Southern Ocean, which are colder relative to much of the rest of the world, take longer to warm. But warm they will. As the atmosphere traps more heat, their temperatures will rise and the overall rate of planetary warming will accelerate. Warming projections based only on historical observations assume the pace of climate change will remain the same; the new study says that as time goes on, things may get worse faster.

Gavin Schmidt, director of NASA’s Goddard Institute for Space Studies, said the new paper provides independent backing for recent work from his unit and elsewhere. He concludes that projections of future warming derived from recorded temperature trends “are biased low.”

“Fixing that makes things line up much better,” Schmidt said.

Debate persists, however. Nicholas Lewis is an independent climate scientist who has published studies about climate sensitivity. He said he read the paper before publication and tried to replicate its findings. Among many technical criticisms, he said eight models analyzed in the new study varied in quality. A better-quality range for the observations-only camp reaches from 1.65C to 1.9C, he said.

Meanwhile, Proistosescu’s apples-to-apples analysis spans a more dangerous range, from 1.6C to 4C.

Part of the distance between the two camps may come from professional or philosophical differences over the utility of physics-based models to provide insight into the path of warming. There’s no substitute for direct observation, but models strongly suggest the past has limited predictive value in circumstances moving—geologically speaking—this quickly.