FORECASTSEASON.COMRoute 05 · openSEAS5 ExplorerModelsDriversAbout
SEASON90Forecast SeasonSeasonal Outlooks · U.S. · Europe · Global
Route 05 · Reference

Drivers

A seasonal forecast is only possible because parts of the climate system are slow. These are the slow parts — what each one is, how it is measured, and what it honestly does to a season.

The slow components

Ocean · stratosphere · land
ENSO
What it is. The coupled ocean–atmosphere oscillation of the tropical Pacific, and by a distance the largest single source of seasonal predictability on the planet. Warm phase El Niño, cool phase La Niña, with an episode declared when the ONI holds beyond ±0.5 °C for five consecutive overlapping seasons.
What it does. It shifts the Walker circulation, and with it the position of the Pacific jet. In an El Niño winter the southern tier of the United States tends wet and the Ohio Valley and Pacific Northwest dry; La Niña reverses it. The tropical response is strong and reliable; the mid-latitude response is a shift in the odds, never a guarantee. Skill collapses through the northern spring — the predictability barrier — which is why an outlook made in April for the following winter deserves less weight than the same outlook made in August.
Measured by · Niño-3.4 SST anomaly; the ONI is its three-month running mean
Indian Ocean Dipole
What it is. The Indian Ocean's own east–west seesaw, peaking in the northern autumn and usually decaying by December. Frequently, but not always, co-varying with ENSO.
What it does. A positive IOD loads rainfall onto East Africa and takes it away from Indonesia and southern Australia; a negative IOD does the reverse. It is the single most useful driver for the Australian and East African seasons, and it modulates the Indian summer monsoon alongside ENSO.
Measured by · The Dipole Mode Index — western minus south-eastern tropical Indian Ocean SST
NAO & the Arctic Oscillation
What it is. The dominant mode of North Atlantic winter variability, and the AO is essentially its hemispheric expression. It is a fast mode — it reorganises in days — which is precisely what makes forecasting its seasonal mean so hard.
What it does. Positive NAO means a strong, northward storm track: mild and wet in northwest Europe, dry in the Mediterranean, mild in the eastern United States. Negative NAO blocks, drives cold outbreaks into Europe and the eastern US, and wets the Mediterranean. Models predict the winter-mean NAO with real but modest skill, and — this is the famous part — with a signal that is too weak relative to its own noise, so the ensemble mean has to be scaled up to be believed.
Measured by · The pressure difference between the Icelandic low and the Azores high
The stratospheric polar vortex
What it is. The winter circumpolar westerly jet in the stratosphere. Around six times a decade in the north it collapses in a sudden stratospheric warming, and the anomaly propagates downward into the troposphere.
What it does. A vortex disruption is followed, within roughly two weeks, by a negative Arctic Oscillation at the surface that can persist four to eight weeks — a genuine source of week-3-to-8 predictability, and one of the few mechanisms that lets a subseasonal forecast beat climatology in mid-latitude winter. Our sibling site polarvortex.earth watches it directly.
Measured by · 10 hPa zonal-mean zonal wind at 60° N; a sudden stratospheric warming reverses it
QBO
What it is. The quasi-biennial oscillation: a downward-propagating reversal of the equatorial stratospheric winds on a roughly 28-month cycle. The most regular thing in the atmosphere, and still not perfectly predictable.
What it does. Through the Holton–Tan relationship, an easterly QBO is associated with a weaker, more disturbed polar vortex and so a greater chance of a negative-NAO winter. It also modulates the MJO. The effect is real in the composite and modest in any one year.
Measured by · Equatorial stratospheric zonal wind, usually at 30 or 50 hPa
MJO
What it is. An eastward-propagating envelope of tropical convection circling the globe every 30 to 60 days. Subseasonal rather than seasonal, but it is how the tropics talk to the mid-latitudes on the way to a season.
What it does. The MJO is the main lever for weeks 2–4 forecasts: particular phases load the odds for atmospheric rivers into the US West Coast, for cold-air outbreaks, and for tropical cyclogenesis. Over a whole season it mostly averages out — but the number of strong MJO events in a season does not, and that is where it re-enters.
Measured by · The RMM index — eight phases around the tropics
The slow ocean — PDO, AMV & friends
What it is. The Pacific Decadal Oscillation and the Atlantic Multidecadal Variability are patterns of ocean temperature that persist for years to decades. They are less predictors than context: the background state an ENSO event happens on top of.
What it does. AMV is a strong control on Atlantic hurricane activity and Sahel rainfall. PDO modulates how strongly a given ENSO event projects onto North American weather. Because they are slow, they are also the part of a seasonal forecast that persistence alone gets mostly right.
Measured by · Basin-scale SST pattern indices, on decadal timescales
Land memory — snow, soil and ice
What it is. The surface remembers. Soil moisture anomalies persist for months, snow cover changes the albedo and the surface energy balance, and sea ice does both while also moving the baroclinic zone.
What it does. Dry soil in spring amplifies summer heat — one of the clearer seasonal mechanisms there is, and central to European and Great Plains heatwave outlooks. Snow-cover and sea-ice pathways into the winter circulation are actively argued over in the literature and should be treated as open questions, not settled ones.
Measured by · Snow-cover extent, soil-moisture anomalies, sea-ice concentration
The trend
What it is. Not a mode of variability at all, but the largest single term in most temperature outlooks. A model's own drift is removed by taking anomalies against its hindcast climatology; the real-world trend is not removed, and should not be.
What it does. Over much of the globe the honest baseline for a seasonal temperature outlook is 'warmer than the 1991–2020 normal', before a single dynamical model is consulted. A forecast that beats climatology only because it captured the trend has still beaten climatology — but it is worth knowing which part of the skill came from where.
Measured by · The observed warming rate at each place and season

How they combine

The part nobody can put on a dial

The temptation with a list like this is to treat it as a set of dials: a bit of El Niño, an easterly QBO, a low solar cycle, and out comes a winter. It does not work, for two reasons.

The first is that the drivers are not independent. ENSO modulates the MJO, the MJO modulates the vortex, the vortex modulates the NAO, the IOD co-varies with ENSO, and the PDO sets how strongly any of it projects onto North America. Composites built by sorting years on one index quietly inherit whatever the other indices happened to be doing in those years.

The second is sample size. There have been roughly two dozen well-defined El Niño events in the reliable record. Split them by QBO phase and you have a handful. Split by anything else and you are describing individual winters, not a climatology.

This is exactly what the dynamical models are for: a coupled model run 51 times combines the drivers physically rather than statistically, and it does not need the analogue years to exist. The drivers on this page are for reading a model output — for understanding why it says what it says — not for beating it.

Next exits