It has been quite the week for climate change news: We’ve learned that scientists can now quantify the United States’ expected levels of inundation by rising seas, that droughts in the Amazon could triple, and much more.

But the most troubling research — depending, that is, on how you interpret it — may have appeared in a less-noticed, first-of-its-kind study just published in the influential Proceedings of the National Academy of Sciences. In it, the researchers attempted something that seems never to have been successfully done before. Namely, they mined the data from a large suite of computerized climate change simulations, or models, to determine how often they produced abrupt and disruptive changes in a few decades or even less — surely the most feared impact of climate change.

The result — that out of 37 abrupt changes detected in these climate simulations, fully 18 of them occurred at temperature levels less than 2 degrees Celsius of warming — is simultaneously dramatic and yet also difficult to assess. Models, after all, are mathematically sophisticated simulacra that embed scientists’ best current physical understanding of how the Earth system and its components work, but still should not be confused with reality.

Nonetheless, the authors — led by Sybren Drijfhout, a professor at the Royal Netherlands Meteorological Institute — assert that their results represent a sign of how unstable the future could really be, even before we reach warming levels of 2 degrees Celsius above pre-industrial levels (often thought of as a kind of guardrail in international climate negotiations). “It is likely that the Earth system will experience sharp regional transitions at moderate warming,” they write, “although the prediction of any particular event has a very high uncertainty.”

So what’s going on here — and should we really believe this?

Mining models for major disruptions. The study — which Drijfhout undertook with a large team of researchers at institutes in Britain, the Netherlands, Germany and France — is in effect a massive “big data” inquiry into an urgent mystery about climate change. That mystery is this: When, precisely, can a relatively slow and steady rate of global warming trigger abrupt or sudden shifts in particular regions or Earth systems?

The topic has been much discussed — but also remains very murky. So to examine it in a new way, the researchers looked at the results of no less than 37 separate computerized climate change simulations, or models, which were used in the U.N. Intergovernmental Panel on Climate Change’s 2013 assessment report of the state of climate science. Each model’s results were examined out to the year 2100 — or farther, in cases where that was possible — under different assumptions about levels of greenhouse gas emissions, and resulting planetary warming.

In doing this, the work is entering new territory. “There has been no systematic study of the potential for abrupt shifts in state-of-the-art Earth System Models,” the research noted, calling the paper “a first step toward a robust assessment of abrupt change.”

And sure enough, the models did produce many abrupt changes — but they were also rarely in agreement with one another. Some changes — for instance, a massive oceanic algal bloom in the Indian Ocean in the next century, which only appeared in one model — are dismissed by the researchers as a possible fluke. “That’s the one we are really most unsure about,” says Drijfhout.

On the other hand, other changes that showed up more frequently are precisely the types of things that scientists have long forecast might result from a warming of the climate. For instance, multiple models showed rapid collapses of Arctic sea ice, particularly in extreme global warming scenarios. And multiple models also produced partial or full shutdowns of circulation in the North Atlantic— a change that sometimes occurred for only moderate levels of warming (less than 2 degrees Celsius above pre-industrial levels).

“A striking feature is that the majority of abrupt transitions occur in the ocean-sea ice system, implying that this Earth system component is more prone to abrupt change than other components,” the research added. “These are very nonlinear processes that are reasonably well resolved in the models,” says Drijfhout — suggesting that if they turn up again and again, they may be something that can really happen in the real world.

Other experts react. But here, perhaps, we should pause. Models are not predictions of the future — they’re more about understanding than about forecasting. And when an abrupt change shows up in just one model but not others, that could be simply due to the equations embedded in that particular simulation. Indeed, “no type of abrupt shifts occurs in all models,” the authors say.

Researchers asked to look at the study by The Washington Post offered some criticism, while also noting that the research certainly has consistencies with other evidence about abrupt climate changes. Until now, such work has largely been based on studies of the Earth’s past using so-called “proxy” evidence like ice cores or ocean sediments.

“As the authors note, it’s unclear whether these events are related to specific simplifications or perhaps even bugs in the codes, and without some consistency across models it’s hard to make any useful predictions,” said Gavin Schmidt, director of the Goddard Institute for Space Studies at NASA — which runs one of the models in question — by e-mail. “The places where most of these events occur are not surprising of course — the North Atlantic stands out.”

A similar take came from Richard Alley, a glaciologist at Penn State University who chaired a National Academy of Sciences panel on abrupt climate change. “Many questions exist about the ability of models to simulate ‘tipping points’ or abrupt changes accurately,” Alley said. “But the paleoclimate record shows clearly that such abrupt jumps have occurred, and this new paper shows that they are fairly common and widespread in the modeled climate system.”

The most critical take came from Kevin Trenberth, a climate researcher at the National Center for Atmospheric Research in Boulder, Colo. “I don’t find anything in this paper surprising or very illuminating,” he said by e-mail. “The paper does not validate the models to be able to say that any of them are realistic or likely. It is likely that some of these are more the result of model flaws. But it is a start,” Trenberth added.

Drijfhout said by e-mail that he agrees that climate models have not been validated based on their ability to capture abrupt changes — rather, he argues, they are validated based on their ability to capture the present climate and even “tuned” to be good at this. “There is consensus that climate models tend to be more stable than the real climate,” he said in his e-mail. For this reason, Drijfhout believes that models may underestimate its real instability and capability for abrupt shifts.

“In general there seem to be more missed cases than false alarms,” he said.

It’s also important to note that the simulations were not capable of detecting one possible abrupt shift that worries many climate scientists right now — the potential for a collapse of the West Antarctic ice sheet. “They are clearly correct in noting that some potential abrupt events aren’t possible in this class of models — no ice sheet components, or sub-surface methane hydrate routines,” noted Gavin Schmidt, director of the Goddard Institute for Space Studies at NASA, by e-mail.

In the end, the most striking finding from the study remains how many abrupt shifts occur — at least in the models — with relatively modest levels of warming. Granted, the paper also acknowledges that abrupt changes were most likely to occur at the highest warming scenarios.

“There is of course a certain tendency for the whole climate system to become more unstable when the warming gets larger,” said Drijfhout, “but we cannot say, ‘as long as it’s this and this much, nothing will happen.’ Every .1 or .2 degrees in temperature is as dangerous as any other, I would say. And that’s the main message of this exercise, or this paper.”

It remains to be seen how many other scientists agree with this assessment — and whether through future research, they can alter or improve on our abilities to detect truly abrupt climate change scenarios with modern computer simulations.