El Nino oscillation

Why does the Pacific keep swinging between warm and cool states instead of settling into one of them?

Most large natural systems settle. Given constant forcing — the same sunlight, the same rotating planet, the same ocean basin — you would expect the tropical Pacific to find a state and stay in it. Instead it swings every few years between an unusually warm eastern Pacific and an unusually cool one, and has apparently done so for millennia. What would a system have to be like for not settling to be its natural behaviour?

Start with the ordinary state, because the anomaly is a departure from it. Trade winds blow from east to west along the equator. Wind dragging on water for thousands of kilometres piles warm surface water in the western Pacific — enough that sea level near Indonesia stands appreciably higher than off Ecuador — and with it pushes down the thermocline, the sharp boundary between the warm surface layer and the cold water beneath. In the east, the opposite: warm water is drawn away, the thermocline rises close to the surface, and cold water wells up along the South American coast.

Now ask what maintains those winds. The trades are driven partly by the temperature contrast they themselves produce: warm water in the west feeds rising air and deep convection, cool water in the east suppresses it, and that pressure difference along the equator is what keeps air flowing westward. So the wind sustains the temperature pattern and the temperature pattern sustains the wind.

That is a positive feedback, and it changes the question entirely. Weaken the trades slightly — and the tropics supply plenty of noise to do so, from bursts of westerly wind associated with tropical convection — and less warm water is held in the west. The eastern thermocline deepens, so the water welling up is drawn from a warmer layer and arrives less cold. The east-west contrast shrinks. Weaker contrast means weaker winds. The disturbance amplifies itself. This is the Bjerknes feedback, and it explains why the system does not sit still: the neutral state is not a valley the system rolls back into but a ridge it rolls off.

But a runaway does not oscillate either — it would simply stay warm. So there must be a restoring force, and crucially one that arrives late. Here the ocean's memory does the work. The same wind anomalies that warm the east also drive warm water away from the equator, so the total volume of warm water stored in the equatorial band is drained during a warm event. The surface can look warm while the reservoir beneath is emptying. When the reservoir is low enough, the shallow thermocline lets cold upwelling reassert itself, the east cools, the contrast steepens, the trades strengthen, and the whole feedback runs in reverse into a cool phase — which then spends years recharging the reservoir, setting up the next warm one. The delay is not incidental; it is what converts an unstable state into a cycle.

There is a second, closely related account in which the delay is carried by waves: wind anomalies launch equatorial waves that cross the basin, reflect off the western boundary, and return months later carrying the opposite signal, arriving after the event that launched them has already matured. The two descriptions are not rivals so much as different bookkeeping on the same delayed ocean adjustment, and the balance between them is still argued.

Why so irregular, then — anywhere from about two to seven years between events, rather than a clean beat? Because the trigger is weather, which is random, while the response is phase-locked to the seasonal cycle: events overwhelmingly peak around the end of the calendar year, which is how the phenomenon got its name from Peruvian fishermen. A noisy trigger with a seasonally gated response gives you recurrence without periodicity. Whether the system is a self-sustaining oscillator that noise merely perturbs, or a damped mode that would die out and is kept alive by noise, is genuinely unresolved.

Two corrections are worth making explicit. First, El Nino is not warmth arriving from somewhere; it is warm water that was already there redistributed eastward, together with a change in how much cold water is allowed to reach the surface. Second, this is not a system with a controller. In an engineered feedback loop a delay turns a stabilising correction into hunting around a setpoint. Here the near-instant feedback is the destabilising one and the delayed feedback is the restoring one, and there is no setpoint at all — the "target" is itself a state of the same fluid. That inversion is why the tropical Pacific oscillates rather than merely overshoots.

It also explains the reach. A shift in where the warmest water sits moves where the atmosphere's deepest convection happens, and that convection is a heat source large enough to bend the upper-level winds that steer weather systems far outside the tropics. The effects on distant regions are statistical tendencies, not schedules — a genuinely important caveat, since no two events have the same pattern and forecasting skill beyond about a season, and especially across the northern spring, remains limited.

• Why forecasts made in the northern spring are markedly worse than those made in summer, and what that says about where the predictability lives.

Every term the collection defines is gathered in the glossary.