Super El Niño Winters and the Arctic Polar Vortex: Response Heterogeneity, QBO Modulation, and a Prospectively Defined Test for Winter 2026/27

Mehmet Can TANYERI 

In the three strongest El Niño winters of the instrumental record (1982/83, 1997/98, 2015/16) the Arctic polar vortex did not collapse in a major midwinter sudden stratospheric warming (SSW). This study starts from the premise that the observation is not a paradox: 58% of El Niño winters contain at least one major SSW, so the chance probability of three consecutive SSW-free winters is 0.42³≈0.07, and the canonical El Niño signature in the literature is not an SSW but a vortex weakened on average in late winter. By that measure the three are not homogeneous: the DJF vortex of 1997/98 is below average (~21st percentile) and conforms to the canonical chain, 1982/83 is intermediate (~76th), and the true exception is 2015/16 (~97th), the winter in which the QBO was disrupted for the first time on record. We re-examine this heterogeneous record, on the occasion of the El Niño developing in 2026 with record-exceeding values forecast for October–December, using public index series (ONI, RONI, NAO, AO, QBO; 1951–2026), the NOAA SSW Compendium catalog (1958–2024) and ERA5 fields. Four results stand out. First, the intraseasonal sign asymmetry of the ENSO–NAO relationship (November–December r=+0.21, January–March r=−0.12) is an independent reproduction of published findings (Moron and Gouirand 2003; Ayarzagüena et al. 2018); our contribution is that the paired difference over the same winters excludes zero (Δr=+0.33; 95% CI +0.08/+0.56) and survives a moving-block bootstrap. Second, while the seasonal-mean surface trace of the QBO is near zero (QBO50×AO r=−0.01), the winter-based probability of an SSW appears higher in the easterly phase: the risk ratio is 1.26 (95% CI 0.75–2.12) under the pre-winter November 50-hPa definition and 1.48 (0.94–2.32) in the 67-winter 30-hPa record. Detecting a difference of this size with 80% power requires roughly 450 winters; the observational record cannot settle the question on its own, and a single winter moves the likelihood ratio by only ~1.2. In the continuous vortex regression ONI alone gives R²=0.002 while adding wQBO raises R² to 0.126; this result rests on NCEP/NCAR Reanalysis-1. Third, the three winters defined by ONI≥+2.0 are westerly-QBO under the main phase definition (50 hPa, November), but the match is definition-dependent: at 30 hPa in DJF 1997/98 shifts to the easterly phase, and in 2015/16 the QBO was disrupted in February; the RONI≥+2.0 ruler enlarges the set to four with 1991/92 (4/4 SSW-free, 3/4 wQBO). The QBO is therefore not a deterministic switch but a possible conditioner of uncertain magnitude; the ONI–U10 relationship (r=−0.17) and the segmented fit (ΔAIC=−0.3) give no support to the vortex-level signature of a simple saturation scenario, and the mixed Central/Eastern Pacific structure of 2015/16 is added as a fourth candidate independent of the QBO. Fourth, the eastern Mediterranean end of the chain is measured directly with ERA5 (1979/80–2021/22, n=43): a +1 standard deviation stronger vortex is associated with a positive-NAM-like Z500 pattern in which 36.6% of grid cells in the Atlantic–European sector pass the BH-FDR (α=0.10) threshold and with a DJF precipitation anomaly of −5.5 mm month⁻¹ [95% CI −10.2/−0.6] in the Türkiye box; a land-masked repetition preserves the DJF result, the fifteen regional coefficients do not pass BH-FDR as a single family, and the January–March link is erased when the NAO is controlled. The decision rules for 2026/27 (phase, event, ruler and evidence-updating criteria) are fixed in Appendix D before the outcome is known; the winter is not a verdict but a small Bayesian update of prospectively defined expectations.