A cooler south, and a tropics that models miss

Over the last three decades, the tropical eastern Pacific has stubbornly refused to warm as fast as the rest of the tropics. The slow trend looks like a long, gentle La Niña, and it has been blamed on everything from trade winds to aerosol pollution. A paper published on 5 June 2026 in Science Advances argues that part of the explanation lies much farther south: the Southern Ocean, and specifically the way climate models misrepresent the tropical rain belt, has been hiding how strongly a cooler south reaches into the tropics.

The starting point is a known but awkward fact. Observed sea-surface temperatures show cooling both around Antarctica and in the eastern tropical Pacific, even as the planet as a whole warms. A growing body of work suggests these two trends are connected: meltwater and cloud changes around the Southern Ocean alter winds and rainfall, and those changes propagate north across the equator. The question has been how strong that teleconnection really is, because models disagree wildly about it.

The team, working across a hierarchy of simulations from idealized slab-ocean runs to multi-model ensembles, found that the disagreement tracks a single old model flaw. Nearly every climate model places the tropical rain belt, the intertropical convergence zone, too far south in the Atlantic and eastern Pacific, producing a double band of rainfall where observations show one. When the team corrected that mean-state bias, the simulated tropical response to Southern Ocean cooling roughly doubled. The mechanism is mechanical: a more realistic rain belt means stronger mean trade winds, and stronger mean winds let the wind-evaporation feedback amplify a distant cooling signal into a tropical temperature response.

The numbers make the point. Across seven models, the position of the rain belt explained about three quarters of the spread in how strongly the eastern Pacific reacted to Southern Ocean cooling. In a separate set of meltwater experiments, the same bias explained about four fifths of the spread. When the researchers took a model that had been run with observationally constrained Antarctic meltwater from 1990 to 2021 and corrected its rain-belt bias, the simulated tropical cooling moved into much better agreement with what thermometers have actually recorded.

The implication is uncomfortable. The models that are used to project the future also carry this bias, which means they have been systematically understating how much a changing Southern Ocean can reach into the tropics. Antarctic meltwater, in other words, may already be doing more to shape tropical temperatures than the current generation of forecasts credits. That matters for predicting regional rainfall, for the timing of monsoons, and for the apparent pause in tropical warming that has puzzled modellers for years.

The paper is also a quiet argument about where to spend modelling effort. The usual instinct is to add better physics, finer grids or new forcings. This result says that cleaning up the mean state, the average climate the model returns to when left alone, may matter as much as any of those. A model that cannot reproduce where the rain falls on a quiet Tuesday cannot be trusted to tell you how the tropics will answer a meltwater pulse.

None of this settles the future, and the authors are explicit that the work is about diagnosing a bias, not issuing a new forecast. But it reframes a long-running argument. The tropics have not been warming as expected partly because the south has been quietly pulling the strings, and the tools meant to reveal that have been looking at the wrong tropical map.