SEAS5 Outlook
One system, read in full: the ENSO indices from the coastal Niño-1+2 to the central-Pacific Niño-4, each against the previous three issues; the Pacific Decadal, Atlantic and Indian Ocean indices from the same SST fields; calibrated tercile maps of temperature, rainfall and 500 hPa height over the Americas; and the polar stratosphere. Every anomaly is the member minus SEAS5’s own 1993–2016 hindcast for the same start month and lead, so model drift is removed before anything is drawn. The multi-model C3S page follows once the other centres publish.
ENSO indices — this issue against the last three
The shaded fan is this issue (P10–P90 and P25–P75 of the 51 members) with its mean in gold; the coloured lines are the ensemble means of the previous issues, oldest darkest, each drawn over its own valid months, so the run-to-run shift is read directly. The grey band is the hindcast’s year-to-year spread (±1σ), the model’s own “normal range”. Black is observed (CPC ERSSTv5, 1991–2020 base) where an observed series exists. East-based indices matter this year: a warm Niño-1+2 with a cooler Niño-4 is a different event from the reverse, and the Trans-Niño index puts that gradient on one axis.
PDO, AMO, Indian Ocean Dipole, Atlantic Niño
Same fields, same treatment. The PDO is the projection of each member’s North Pacific anomaly onto the ERSST EOF pattern the daily monitor uses, with the global mean removed and NCEI’s scale applied; the AMO is the relative form (North Atlantic minus the 60°S–60°N mean), so the warming trend does not masquerade as a basin mode. Because the anomaly base here is the model’s 1993–2016 hindcast rather than 1991–2020 observations, absolute levels can sit a tenth or two from the published indices; the shape and the run-to-run change are what to read.
Calibrated terciles over the Americas
For each grid point, season and lead, the tercile boundaries come from SEAS5’s own hindcast — 24 years × 25 members, 600 seasonal means — and the 51 forecast members are counted against them. That removes the model’s mean bias and its spread bias point by point, which is the calibration that matters most for a single-system tercile map. Shading is the most likely category and its probability; white means no category reaches 40 %. Near-normal is not drawn: with a realistic spread it is almost never the most likely tercile, and where it is, the signal is weak anyway. Three overlapping seasons follow the start month. One caveat for temperature and for 500 hPa height: the hindcast base is 1993–2016, so two decades of warming (and the thickness rise that goes with it) sit inside “above normal” before any El Niño signal is added — read those maps for their pattern and their below-normal exceptions, the subtropical jet trough across the southern United States above all, rather than the sheer extent of warm shading.
Polar stratosphere — cap-mean heights
The 60–90° polar-cap mean geopotential height at 100, 50 and 10 hPa, as an anomaly against the hindcast at each lead. A high cap at 100 hPa is a weak, warm vortex (the SSW-side reading); a low cap is a strong, cold one. The hindcast ±1σ band shows how far a seasonal model normally strays, which at these levels is not far — read the sign of the tilt and whether the members agree, not the amplitude. Both hemispheres; the southern vortex is in its spring breakdown season at a September start.
Data and method
Forecast: ECMWF SEAS5 (C3S originating_centre=ecmwf, system=51), 51 members, monthly means for lead months 1–6, from the Copernicus Climate Data Store; hindcast: the same system’s 1993–2016 re-forecasts, 25 members, for the same start month. Sea-surface temperature on a 2° grid between 70°S and 70°N; 2 m temperature, precipitation and 500 hPa geopotential on a 1° grid over 170°W–30°W, 60°S–75°N; geopotential at 10, 50 and 100 hPa poleward of 60° in each hemisphere.
Indices are cosine-weighted box means; anomalies are member minus hindcast mean at the same lead. Relative Niño-3.4 follows the CPC/ECMWF construction used elsewhere on this site (Niño-3.4 minus the 20°S–20°N tropical mean, rescaled per calendar month by σ(ONI)/σ(relative)). The Trans-Niño index is Trenberth & Stepaniak (2001): standardised Niño-1+2 minus standardised Niño-4, standardised here by the hindcast spread at each lead. Previous issues are processed identically against their own start-month hindcasts. Terciles: per grid point, per season, thresholds at the 33.3rd and 66.7th percentiles of the 600 hindcast seasonal means; probabilities are member fractions. Built by scripts/sst/seas5_outlook.py; the JSON behind the charts is seas5_outlook.json.
Contains modified Copernicus Climate Change Service information 2026. Neither the European Commission nor ECMWF is responsible for any use of this information.