The models, for instance, suggest that if the carbon dioxide concentration in the atmosphere were to reach double its pre-industrial level, the planet would warm by anywhere from about 1.5C to 4.5C. But the warming patterns we've actually observed over the past 200 years or so would suggest that a doubling in carbon dioxide should only lead to about 3C of warming at the most.
The discrepancy has become a major cause for concern among climate scientists in recent years, according to the new study's lead author, Cristian Proistosescu, a research associate at the University of Washington who conducted the research while completing a PhD at Harvard. Even the United Nations' Intergovernmental Panel on Climate Change acknowledged the issue in its most recent report, he said, stating that it could no longer provide a best estimate for the climate's sensitivity.
"It worried a lot of us," Proistosescu said. "We needed to understand why our different estimates didn't work."
The new study helps to reconcile the models with the historical record. It suggests that global warming occurs in different phases or "modes" throughout the planet, some of which happen more quickly than others. Scientists now increasingly believe that certain slow-developing climate processes will amplify warming to a greater extent in the future, putting the models in the right after all. But these processes take time, even up to several hundred years, to really take effect - and because not enough time has passed since the industrial revolution for their signal to really develop, the historical record is what's actually misleading at the moment.
This conclusion is supported by a growing body of research, which suggests that warming estimates made from the historical record alone are "potentially biased low, for reasons we are now just beginning to understand", said Timothy Andrews, a climate scientist with the Met Office, Britain's national weather service. While Andrews was not involved with the new study, he's one of multiple scientists whose recent research has tackled the same issue.
The new study uses a statistical method to separate "fast" and "slow" climate modes in the models. According to Proistosescu, when greenhouse gases are emitted into the atmosphere, a "fast" warming effect begins to take place almost immediately in certain parts of the planet, mainly over the land masses in the Northern Hemisphere. Indeed, these are the parts of Earth where the most rapid warming has been observed since the industrial revolution.
On the other hand, he said, other parts of the planet - namely, the Southern Ocean and the eastern Pacific - respond much more slowly, in part because they're just so deep and cold to begin with. But as they absorb more heat and finally start to warm up, they may produce a variety of climate feedback effects that enhance the global warming that's already occurring. For instance, changes in ocean temperatures can alter atmospheric patterns around the world. The models suggest that these warming ocean regions may lead to a decrease in reflective cloud cover in the future, allowing more solar radiation to make it through the atmosphere to Earth's surface.
While the impact of these processes may be profound, they can also take long periods of time to unfold, Proistosescu said - potentially up to 300 years or so. Through their statistical parsing, the researchers found that the models suggest the effects of this slow climate mode only account for about 3 per cent of the human-caused warming we've seen so far. But in the future, under a scenario in which atmospheric carbon dioxide concentrations double, it could account for up to half of all warming that occurs.
Since they were able to separate the fast and slow climate modes, the researchers also conducted a test to see what would happen if they only applied fast climate warming in the models. When they did so, the model predictions fell in line with the historical record - suggesting that the still-developing slow mode is the reason the two have failed to match so far.
That said, the study does beg the question of how we can be so sure the slow climate mode will actually develop as it does in the model, since humans have yet to really observe it in nature. According to Piers Forster, a University of Leeds climate scientist who has also studied climate sensitivity, the models tend to rely on certain assumptions that have not yet unfolded in real life - for instance, that the eastern Pacific will eventually warm to a greater extent than the western Pacific.
"This doesn't necessarily mean the models are wrong," he added. But, he said, it does suggest that "we do not yet fully understand these long-term changes in the Pacific and therefore these climate change amplification effects".
According to Proistosescu, some studies of Earth's ancient climate - which scientists can conduct using information from sources like ice cores and preserved sediments - do suggest that a slow climate mode does exist and has occurred in the past. And he added that the climate models rely on basic physical processes to a great extent, and "we trust that they do the basic physics correctly". But he agreed that some caution - or at least a great deal more research on how these slow climate responses will continue to develop - is called for. In fact, identifying the uncertainties that remain about Earth's climate sensitivity is another significant outcome of the study, he suggested.