A recent paper in Nature Communications about a potential state-change in the Atlantic meridional overturning circulation (AMOC) has been getting some attention, and I wanted to dig into what the authors did.
First of all, what is the AMOC? It’s not the Gulf Stream current, which is a flow of water on the surface of the Atlantic Ocean. The Gulf Stream brings warm water north along the east coast of North America, all the way to Europe, and is the main reason why the climate is London is more temperate than the climate in, say, Calgary (which is at the same latitude). The AMOC, on the other hand, moves warm water and nutrients throughout the Atlantic Ocean in a huge loop, from north to south and back again. A complete journey takes about 1,000 years. And that time has been increasing over the last hundred years as the AMOC has slowed.
As with many major earth system processes, it would be pretty bad for a lot of people if the AMOC were to change substantially, let alone stop completely. Increasing meltwater from polar regions could shift the AMOC, causing major changes in where, when, and how much precipitation happens. Places that are great for agriculture now could become to cold or dry if the AMOC were to shift or stop. This has been a focus of research for a long time, so none of this is really news within the scientific community.
With that background, a new paper uses statistical modeling to determine when the AMOC might collapse. Their basic answer is that, using their statistics, the average time at which the AMOC will collapse is 2050, and the 95% confidence interval runs from 2025 to 2095. It’s hard to think intuitively about things like 95% confidence intervals. Accepting it at face value, it’s possible that the AMOC could collapse outside that interval, say, in 2100… or 2023!
But we shouldn’t accept that finding at face value. A statistical model like this makes sense when we have little information about the processes at play. We’ve never seen a major oceanic overturning circulation collapse, so it might make sense to claim ignorance and turn to a statistical model. But we actually know a lot about climate and planetary processes, and have a ton of measurements that aren’t a part of this statistical modeling approach. As the authors noted, when other scientists have modeled this phenomenon, either for the Intergovernmental Panel on Climate Change or the Climate Model Intercomparison Project, they have determined with moderate confidence that the AMOC will slow through the 21st century, but persist. Those projections are based on modeling projects that attempt to simulate how the oceans and atmosphere move around the planet. Provided that the people doing that work have some expertise (which they do), they will almost certainly come up with a better idea of what’s going to happen that scientists who are only relying on statistical correlations.
It’s overblown to move from statistical correlations to project a state change outside of anything represented in the data. This is an example of the statistical sin of projecting beyond the range of your regression. Nevertheless, the authors do show that there is a high degree of variation in key metrics associated with AMOC. As other authors have found, this variability itself will contribute to extreme weather events as climate changes. I’m glad that the authors did this work, if for no other reason than the increased attention it will bring to climate challenges we are all facing.