FORECASTSEASON.COMRoute 13 · The QBOSEAS5 ExplorerProbabilityAnalogsSecond OpinionCitiesModelsDriversAbout
SEASON90Forecast SeasonSeasonal Outlooks · U.S. · Europe · Global
Forecast Season/Polar Vortex/QBO
Route 13 · The equatorial stratosphere

The QBO

Twenty to thirty kilometres above the equator, the wind reverses about every fourteen months — westerly, then easterly, then westerly again — and each new phase works its way down from the top. It is the most regular thing in the atmosphere and the one part of the stratosphere a seasonal model can genuinely forecast. This page puts everything the site measures about it in one place: ERA5 since 1959, every SEAS5 run on file, how much of the signal the model keeps, and what the phase does to the polar vortex.

QBO13

QBO at 30 hPa now

+15.2 m/s
ERA5 September 2026 · westerly since May 2026

SEAS5 this winter

+7.8 m/s
DJF 30 hPa · 51 of 51 members westerly

Amplitude SEAS5 kept

77%
of last winter's DJF easterly at 30 hPa, Aug 2022 start (+9.0 vs +11.7)

Vortex under easterly QBO

−5.7 m/s
DJF U60 vs westerly winters · ERA5, 67 winters · p = 0.02

What we know

ERA5 to Sep 2026 · SEAS5 30 starts

Where it is. At 30 hPa the equatorial wind turned westerly in May 2026 and stood at +15.2 m/s in September 2026, ERA5's latest month.

How long it lasts. The 32 completed westerly spells at 30 hPa since 1959 lasted a median 11 months (half of them between 10 and 13). Counted from May 2026, that puts the turn back to easterly in April 2027, most likely between Mar 2027 and Jun 2027. That is a persistence forecast, and the two disruptions of 2016 and 2020 show it can be wrong.

What SEAS5 says. The October 2026 run keeps the winter westerly: 51 of 51 members have a westerly DJF mean at 30 hPa, +7.8 m/s on average (+5.5 to +10.4). Its 30 hPa wind starts at +15.5 and falls to −0.1 by Apr 2027. That fall is most likely the model's bias, not a forecast of an early turn: see the next paragraph.

How much of the amplitude it keeps. Last winter's 27 SEAS5 runs got the 30 hPa phase right in every month, but carried only -1% to 99% of ERA5's December–February easterly (−25.9 m/s), more the later the start. They began 0–5 m/s too weak in their first month and weakened it further every month after. At 50 hPa the damping is milder, about a quarter.

What it does to the vortex. In 67 ERA5 winters the polar vortex averaged 27.7 m/s under an easterly QBO and 33.4 m/s under a westerly one — 5.7 m/s weaker, p = 0.02. Sudden warmings follow the same direction (62% of easterly winters had one, 43% of westerly) but a count of events is a blunter test, and that one reaches only p = 0.15. This is Holton–Tan, measured here rather than quoted.

This winter's combination. The forecast winter is El Niño with a westerly QBO. In ERA5 that pairing has occurred 8 times; 3 had a sudden warming, and the vortex averaged 33.8 m/s. El Niño with an easterly QBO is the most disturbed cell in the table (9 of 12, 22.0 m/s). With 8 winters in the cell, it is a lean, not a forecast.

Does SEAS5's vortex feel it? Only slightly. Split SEAS5's DJF vortex by the observed QBO phase and it differs by 0.7–1.3 m/s between easterly and westerly winters, against 3.6 m/s in ERA5 over the same 33 winters. That is roughly 19%–36% of the observed contrast. A model with a weak QBO would be expected to feel it weakly.

Inside one forecast, it does not pick members. The 51 members' QBO differs by only 0.6–1.9 m/s (one standard deviation), against a vortex spread of 4.9–9.3 m/s. The QBO is the same backdrop for every member. It shifts what the ensemble expects; it does not explain why one member warms and another does not.

The oscillation

ten years observed, then SEAS5
Zonal wind, m/s← easterly≤ −25−25 … −15−15 … −5±55 … 1515 … 25≥ 25westerly →
10 hPa30 hPa50 hPa70 hPa100 hPaERA5 OBSERVEDSEAS52020 disruption201820192020202120222023202420252026202713FORECASTSEASON.COMERA5 (ECMWF) · SEAS5 (ECMWF) · contains modified Copernicus C3S information 2026
Zonal-mean zonal wind averaged 5°S–5°N, by pressure level (10 hPa is about 31 km up, 100 hPa about 16 km). Each phase appears at the top and works its way down over a year or so. Right of the line: the latest SEAS5 ensemble mean, which has no 70 hPa level. Its colours are weaker than ERA5's partly because SEAS5 damps the QBO, as the tables below measure. Hover for values.

The whole record, 1959 to now

10 hPa30 hPa50 hPa70 hPa100 hPa196019701980199020002010202013FORECASTSEASON.COMERA5 (ECMWF) · contains modified Copernicus C3S information 2026
The same section for every month since January 1959. The two brackets mark the only times the cycle has broken.

2016 disruption

Feb 2016 – Sep 2016

Easterlies formed at 40–50 hPa inside a westerly phase that should have descended undisturbed — the first break in the cycle since records began, driven by waves from the Northern Hemisphere.

Osprey et al. 2016, Science; Newman et al. 2016, GRL

2020 disruption

Dec 2019 – May 2020

A second disruption four years later, this time forced from the Southern Hemisphere; the westerly phase at 30 hPa lasted barely a year.

Anstey et al. 2021, GRL

The rhythm

every cycle's length since 1959
101520253035median 26.5 months1960197019801990200020102020Westerly onset June 1959: 20 months to the nextWesterly onset February 1961: 30 months to the nextWesterly onset August 1963: 32 months to the nextWesterly onset April 1966: 35 months to the nextWesterly onset March 1969: 23 months to the nextWesterly onset February 1971: 23 months to the nextWesterly onset January 1973: 28 months to the nextWesterly onset May 1975: 29 months to the nextWesterly onset October 1977: 16 months to the nextWesterly onset February 1979: 13 months to the nextWesterly onset March 1980: 29 months to the nextWesterly onset August 1982: 29 months to the nextWesterly onset January 1985: 31 months to the nextWesterly onset August 1987: 31 months to the nextWesterly onset March 1990: 30 months to the nextWesterly onset September 1992: 26 months to the nextWesterly onset November 1994: 27 months to the nextWesterly onset February 1997: 23 months to the nextWesterly onset January 1999: 35 months to the nextWesterly onset December 2001: 26 months to the nextWesterly onset February 2004: 25 months to the nextWesterly onset March 2006: 11 months to the nextWesterly onset February 2007: 13 months to the nextWesterly onset March 2008: 29 months to the nextWesterly onset August 2010: 31 months to the nextWesterly onset March 2013: 27 months to the nextWesterly onset June 2015: 11 months to the next — 2016 disruption2016 disruptionWesterly onset May 2016: 30 months to the nextWesterly onset November 2018: 19 months to the next — 2020 disruption2020 disruptionWesterly onset June 2020: 25 months to the nextWesterly onset July 2022: 23 months to the nextWesterly onset June 2024: 23 months to the next13FORECASTSEASON.COMERA5 (ECMWF) · contains modified Copernicus C3S information 2026
Months from each westerly onset at 30 hPa to the next, ERA5. Each point is one full cycle, placed at the year it began. Hover a point for its dates.
CompletedMedianMiddle halfShortestLongest
Full cyclewesterly onset to the next3226.5 mo23–301135
Westerly spellat 30 hPa3211 mo10–13116
Easterly spellat 30 hPa3215 mo12–16122

Spells are counted at 30 hPa from monthly means. A one-month sign flip counts as its own spell — the record has four — which is why the shortest entries are a single month. The spell under way now is excluded from every number because it has not ended.

This winter, from SEAS5

51 members against ERA5
Started
Level
ERA5 observedSEAS5 ensemble mean10–90 % of memberseach member
−30−20−100+10westerly ↑easterly ↓ERA5 OBSERVEDSEAS5 · STARTED OCT 2026Oct2025NovDecJan2026FebMarAprMayJunJulAugSepOctNovDecJan2027FebMarApr−0.1 SEAS5 mean13FORECASTSEASON.COMModel: ECMWF SEAS5 · ERA5 (ECMWF) · contains modified Copernicus C3S information 2026
30 hPa zonal-mean wind, 5°S–5°N, monthly means. SEAS5 started October 2026, 51 members; ERA5 in ink. The first forecast month is the initialisation month.

Last winter, checked

SEAS5 against what happened
Level
Monthly wind, m/s← easterly≤ −25−25 … −15−15 … −5±55 … 1515 … 25≥ 25westerly →
Aug 2017Sep 2017Oct 2017Nov 2017Dec 2017Jan 2018Feb 2018Mar 2018Apr 2018Aug 2018Sep 2018Oct 2018Nov 2018Dec 2018Jan 2019Feb 2019Mar 2019Apr 2019Aug 2019Sep 2019Oct 2019Nov 2019Dec 2019Jan 2020Feb 2020Mar 2020Apr 2020Aug 2020Sep 2020Oct 2020Nov 2020Dec 2020Jan 2021Feb 2021Mar 2021Apr 2021Aug 2021Sep 2021Oct 2021Nov 2021Dec 2021Jan 2022Feb 2022Mar 2022Apr 2022Aug 2022Sep 2022Oct 2022Nov 2022Dec 2022Jan 2023Feb 2023Mar 2023Apr 2023Aug 2023Sep 2023Oct 2023Nov 2023Dec 2023Jan 2024Feb 2024Mar 2024Apr 2024Aug 2024Sep 2024Oct 2024Nov 2024Dec 2024Jan 2025Feb 2025Mar 2025Apr 2025Aug 2025Sep 2025Oct 2025Nov 2025Dec 2025Jan 2026Feb 2026Mar 2026Apr 2026DJFShare of
ERA5 DJF
ERA5what happened−20.9−21.2−22.7−24.3−25.6−26.4−25.6−26.5−28.9−27.4−14.3−1.6+8.0+12.1+10.4+10.9+12.5+14.1+8.4+6.9+5.9+2.5−0.6−3.1−5.1−6.6−7.4+9.2+10.2+10.5+10.3+9.3+7.8+7.0+7.7+4.7−19.6−22.1−24.5−26.5−27.8−28.3−28.3−28.6−29.0+10.6+13.4+13.7+13.3+12.3+12.1+10.8+10.9+12.0−14.1−18.6−21.8−25.9−27.4−27.7−27.9−29.0−25.7+10.4+11.6+11.9+11.8+10.3+9.8+9.0+8.4+3.7−28.5−29.3−29.7−29.6−29.9−30.8−30.6−26.5−12.0−25.9
SEAS5 · Aug 2017ensemble mean, 51 members−15.8−11.8−11.0−12.1−13.9−14.8−14.3−14.455%
SEAS5 · Sep 2017ensemble mean, 51 members−17.0−13.5−12.8−13.9−15.0−15.2−13.9−14.757%
SEAS5 · Oct 2017ensemble mean, 51 members−18.1−15.4−15.2−15.7−15.9−15.1−10.9−15.660%
SEAS5 · Aug 2018ensemble mean, 51 members−22.8−8.3+2.2+7.4+8.6+7.5+6.8+7.668%
SEAS5 · Sep 2018ensemble mean, 51 members−11.6+0.2+6.4+7.2+5.4+4.6+5.8+5.751%
SEAS5 · Oct 2018ensemble mean, 51 members−1.4+6.1+7.6+5.6+4.5+5.4+8.3+5.953%
SEAS5 · Aug 2019ensemble mean, 51 members+8.7+8.2+7.3+4.1+0.8−1.1−2.5−0.9—
SEAS5 · Sep 2019ensemble mean, 51 members+6.5+5.7+3.9+0.9−1.3−2.8−4.1−1.1—
SEAS5 · Oct 2019ensemble mean, 51 members+4.3+0.8−1.7−3.7−4.9−6.2−8.2−3.4—
SEAS5 · Aug 2020ensemble mean, 51 members+7.9+7.8+6.8+4.7+2.1−0.3−2.1−0.1-1%
SEAS5 · Sep 2020ensemble mean, 51 members+10.6+9.8+7.8+5.3+3.0+1.3−0.1+3.240%
SEAS5 · Oct 2020ensemble mean, 51 members+11.6+9.5+6.6+4.0+2.1+0.6−1.4+4.253%
SEAS5 · Aug 2021ensemble mean, 51 members−15.5−13.4−13.6−14.7−13.9−11.6−9.0−11.541%
SEAS5 · Sep 2021ensemble mean, 51 members−17.1−14.7−14.9−15.6−15.2−14.1−11.0−15.053%
SEAS5 · Oct 2021ensemble mean, 51 members−19.6−17.3−17.4−16.9−15.6−12.3−6.1−16.659%
SEAS5 · Aug 2022ensemble mean, 51 members+9.4+10.8+11.9+12.3+11.2+8.9+7.0+9.077%
SEAS5 · Sep 2022ensemble mean, 51 members+13.2+13.3+13.6+12.6+10.2+8.4+7.8+10.489%
SEAS5 · Oct 2022ensemble mean, 51 members+14.9+14.5+13.6+11.5+9.8+9.5+9.1+11.699%
SEAS5 · Aug 2023ensemble mean, 51 members−10.6−9.0−9.4−10.9−12.8−13.8−13.2−13.248%
SEAS5 · Sep 2023ensemble mean, 51 members−13.8−11.8−11.9−13.6−14.6−14.1−11.3−14.151%
SEAS5 · Oct 2023ensemble mean, 51 members−16.7−14.4−14.8−15.6−15.4−13.3−7.6−15.355%
SEAS5 · Aug 2024ensemble mean, 51 members+11.1+11.2+11.3+11.1+9.5+7.0+5.1+7.274%
SEAS5 · Sep 2024ensemble mean, 51 members+12.4+12.1+11.0+8.9+6.5+4.6+3.4+6.769%
SEAS5 · Oct 2024ensemble mean, 51 members+13.1+11.9+9.7+7.2+5.3+4.0+2.3+7.476%
SEAS5 · Aug 2025ensemble mean, 51 members−24.6−20.9−20.0−19.0−14.2−9.9−6.7−10.334%
SEAS5 · Sep 2025ensemble mean, 51 members−25.8−22.3−21.5−17.9−13.2−8.9−5.0−13.444%
SEAS5 · Oct 2025ensemble mean, 51 members−26.5−24.5−21.5−16.5−11.8−6.1−0.2−16.654%

Top row: ERA5. Below it, each 2025 start's ensemble mean for the same months; hover a cell for the member range and how many of the 51 members were as strong as what happened. The last column is the forecast DJF mean as a share of ERA5's. It is left blank at 100 hPa, where the observed wind was too close to zero for a ratio to mean anything.

What it does to the vortex

Holton–Tan, measured
westerly wintereasterly winterone sudden warmingSEAS5, this winter
−30−20−100+10+2019601970198019902000201020202026/27SEAS5westerly winters ↑easterly winters ↓13FORECASTSEASON.COMERA5 (ECMWF) · SEAS5 (ECMWF) · contains modified Copernicus C3S information 2026
DJF-mean 30 hPa QBO wind for every winter since 1959/60, with each confirmed major sudden stratospheric warming (Charlton–Polvani, ERA5) as a dot. Hover a bar for the vortex strength and Niño 3.4 that winter.

67 winters, split by phase

WintersWith a warmingWarmingsVortex, DJF U60 (m/s)
Easterly QBODJF 30 hPa wind < 03924 (62%)3027.7
Westerly QBODJF 30 hPa wind > 02812 (43%)1533.4
Differenceeasterly minus westerly+19 pts p = 0.15−5.7 p = 0.02

…and by ENSO

Easterly QBOWesterly QBO
El NiñoDJF Niño 3.4 ≥ +0.59 of 1275% had a warming · vortex 22.0 m/s3 of 838% had a warming · vortex 33.8 m/s▲ this winter, forecast
Neutralbetween5 of 1050% had a warming · vortex 29.4 m/s3 of 838% had a warming · vortex 31.9 m/s
La NiñaDJF Niño 3.4 ≤ −0.510 of 1759% had a warming · vortex 30.8 m/s6 of 1250% had a warming · vortex 34.1 m/s

Niño 3.4 is COBE2's DJF mean against 1991–2020. Each cell is coloured by the share of winters that had a warming; hover it for the winters. The "this winter" mark uses the Copernicus consensus on /nino/ for ENSO and SEAS5 for the QBO phase.

Does SEAS5's own vortex feel the QBO?

−4−3−2−10ERA5, same winters−3.6 m/s p = 0.36SEAS5 · started August−0.7 m/s p = 0.35SEAS5 · started September−1.2 m/s p = 0.11SEAS5 · started October−1.3 m/s p = 0.0813FORECASTSEASON.COMModel: ECMWF SEAS5 · ERA5 (ECMWF) · contains modified Copernicus C3S information 2026
DJF-mean U60 at 10 hPa in easterly-QBO winters minus westerly-QBO winters, 1993/94 to 2025/26 (19 easterly, 14 westerly). Negative means a weaker vortex under the easterly phase. The phase is ERA5's for each winter: the reforecast's own QBO was not downloaded.

Inside one forecast

ForecastQBO 30 hPa, DJFQBO spread (sd)Vortex spread (sd)CorrelationVortex spread explained
Aug 201751 members−14.40.96.9-0.121%
Sep 201751 members−14.70.66.3-0.000%
Oct 201751 members−15.60.66.6-0.183%
Aug 201851 members+7.61.28.1+0.3311%
Sep 201851 members+5.71.17.5+0.3512%
Oct 201851 members+6.01.37.5+0.6238%
Aug 201951 members−0.91.48.2-0.060%
Sep 201951 members−1.00.98.8-0.246%
Oct 201951 members−3.40.88.4-0.050%
Aug 202051 members0.01.27.3+0.152%
Sep 202051 members+3.31.08.4+0.091%
Oct 202051 members+4.30.89.3+0.050%
Aug 202151 members−11.61.47.8+0.132%
Sep 202151 members−15.00.96.9+0.173%
Oct 202151 members−16.70.88.4-0.050%
Aug 202251 members+9.10.86.6+0.3714%
Sep 202251 members+10.50.77.7+0.4823%
Oct 202251 members+11.70.77.9+0.3915%
Aug 202351 members−13.30.86.6-0.193%
Sep 202351 members−14.10.66.8+0.071%
Oct 202351 members−15.30.68.6-0.183%
Aug 202451 members+7.31.27.1-0.070%
Sep 202451 members+6.71.18.0+0.020%
Oct 202451 members+7.50.98.2-0.070%
Aug 202551 members−10.41.96.7+0.163%
Sep 202551 members−13.51.88.5+0.163%
Oct 202551 members−16.71.37.2+0.4823%
Aug 202651 members+6.21.37.5+0.071%
Sep 202651 members+7.71.06.3+0.010%
Oct 202651 members+7.81.04.9+0.163%

Across one forecast's 51 members: the spread of the DJF 30 hPa QBO and of the DJF 10 hPa U60 vortex, and how much of the second the first explains (r²). The QBO barely varies between members, so it cannot sort them.

How to read this

four things to know

Index

5°S–5°N, zonal mean

The QBO here is the area-weighted mean of the zonal-mean zonal wind between 5°S and 5°N, the same definition for ERA5 and SEAS5. The Berlin station series (FU Berlin) and CPC's index use Singapore or a narrower band and run a few m/s stronger. The levels are never averaged together, because 30 and 50 hPa are months out of phase.

Amplitude

The phase is right; the strength is not

Every verifiable SEAS5 run got the phase right and the strength too weak, by more the further ahead it looked. Read SEAS5's QBO for its sign and timing; do not read its size against ERA5's percentiles.

Reforecast

Not yet downloaded

The 1993–2016 reforecast was fetched for the polar vortex only. Until its QBO is downloaded, the damping rests on three real-time runs and the Holton–Tan split of SEAS5's vortex uses ERA5's phase for each winter. The model starts from the observed QBO and has held its phase through every winter that can be checked.

Where next

The vortex it steers

This winter's warming probability, the model's own warming rate and how well it has scored are on the polar vortex page. The ENSO half of the combination is on /nino/, and the general background on Drivers.