Monthly, from the Copernicus C3S multi-system seasonal ensemble
Seasonal outlook by model
Eight seasonal systems from seven centres, one at a time or averaged: global anomaly maps for temperature, precipitation, pressure and sea-surface temperature, beside the Niño-3.4 plume that carries every one of their members. Each model is shown against its own 1993–2016 hindcast, so the systems are comparable even though their biases are not. ECMWF SEAS5 is one of the models here; its member-level products — calibrated terciles, threshold days, precipitation minus evaporation — live on the SEAS5 page, which is the only system whose full ensemble is held locally.
← → step through the options · ↑ ↓ change figure · click a still to enlarge
The map for each system is its ensemble-mean anomaly, taken by C3S against that system's own 1993–2016 hindcast and published in the postprocessed seasonal collection. That matters twice over: it is the comparison you want, because a seasonal model's raw field is dominated by its own drift and bias, and it is what makes eight systems affordable — the members themselves are tens of gigabytes per system per issue, the ensemble-mean anomalies are about a megabyte.
The multi-model mean weights every system equally rather than by member count, so NCEP's 124 members and ECCC's 20 count the same. That is the usual convention and it stops one large ensemble speaking for the group. Where systems disagree the mean is the wrong thing to read on its own; step through the models and see whether a feature survives.
Periods. Lead month 1 is the issue month itself, already partly observed by the time you read it. Months 2–4 and 5–7 are the seasonal ranges the systems are actually run for.
Copernicus Climate Change Service, seasonal-postprocessed-single-levels · anomalies vs each model's 1993–2016 hindcast · equal weight per system in the multi-model mean
Every member of every system, on one axis. This is the same feed the ENSO forecasts page draws, kept here so the maps and the index sit together: the Niño-3.4 box is small enough that the full ensemble is cheap to hold, which is why the plume has hundreds of members while the maps are ensemble means.
Anomalies are against each model's own 1993–2016 hindcast, while the observed lead-in uses ERSSTv5 on the fixed 1991–2020 base, so the two can differ by a tenth or two of a degree.
C3S seasonal multi-system, seasonal-monthly-single-levels · observed: NOAA ERSSTv5