In 2014, forecasters expected a strong El Niño, but it arrived only a year later. The ocean seemed to change its mind. Researchers studied a simple model of the tropical Pacific and found that the system has two distinct "personalities": one regularly generates powerful El Niños, while the other oscillates chaotically without them.

This model resembles a pendulum that sometimes jumps to a different swing. In one regime, the pendulum swings almost steadily, with a period of about twelve years, and each cycle includes one strong El Niño. In the other regime, the same pendulum jerks unpredictably, and strong events are absent altogether. Both regimes coexist, like two hollows on the same surface, and the system can roll from one into the other.

In a simulation spanning one hundred thousand years, chaotic epochs lasted on average several hundred years, with the longest about 540 years. The durations of adjacent epochs are uncorrelated: knowing how long the current one lasts tells you nothing about the next. It is as if a coin were tossed every few centuries, and the outcome determined whether strong El Niños would occur in the coming decades.

Previously, the unpredictability of strong El Niños was attributed to chaos, that is, sensitivity to tiny differences in initial conditions. The new work shows there is a second mechanism: the system does not just diverge gradually but switches regime abruptly. The boundaries between regimes are intertwined like lace, so even a minuscule perturbation can push the system over the edge.

This perspective changes the approach to forecasting. If similar coexisting regimes exist in the real ocean, then long-term prediction becomes a question of which swing the pendulum is currently on. Consequently, the disappearance of strong El Niños for decades is not a model failure but a natural behavior of a system that sometimes forgets how to swing steadily.