A season-long observational study. A graph-neural-network weather model runs on this site's own server twice a day; NOAA ocean indices and National Hurricane Center storm data are pulled alongside it. Every figure below is evidence for or against one hypothesis — collected automatically and archived for the whole season.
The 2026 event against the two strongest El Niños on record, aligned by calendar week. Horizontal lines mark NOAA's intensity thresholds.
"Super" is not an official category — the strongest NOAA label is very strong, Niño 3.4 running ≥ +2.0 °C. The relative index RONI subtracts the tropics-wide warming trend and is the stricter yardstick in a warming ocean: when RONI and raw Niño 3.4 agree, the signal is real, not background warmth.
El Niño suppresses the Atlantic and fuels the East Pacific. Bars show 2026 counts against NOAA's outlook ranges (shaded) and the 1991–2020 average (tick).
Vertical wind shear over the hurricane Main Development Region (10–20°N, 20–80°W), measured twice daily by this site's own model. The dashed trace is the current 10-day forecast.
Shear is the mechanism El Niño uses to suppress hurricanes — and it is directly measurable. Sustained values above ~10 m/s are hostile to developing storms. This record grows for the rest of the season; watch for multi-day windows where shear relaxes below the line.
1992 was a suppressed season: seven named storms. One of them was Hurricane Andrew — Category 5, ~$27B damage, 65 dead. A quiet season delivered one of the most catastrophic landfalls in U.S. history.
El Niño shear doesn't blow at full blast every day. When it relaxes for even a few days over a Gulf that stays bathtub-warm, storms can spin up close to land with far less warning than a long-track Cape Verde system. Below-normal is not the same thing as safe.
This site is a research exercise — consult the NHC and NWS for any real decision.
Campaign start. The seesaw is already visible in the raw best-track files: the East Pacific has produced seven systems — two of them hurricanes spinning right now — while the Atlantic has managed two. Exactly the split a strengthening El Niño predicts.
CPC's July 9 discussion raised the odds of a very strong event (Niño 3.4 ≥ +2.0 °C, Oct–Dec) to 81% — it was ~33% in spring and 63% in June. If the +3 °C model solutions verify, this becomes the strongest event in the modern record. Watch whether RONI keeps pace with the raw index.
NOAA's seasonal outlooks frame the bet: Atlantic 55% below-normal (8–14 named / 3–6 hurricanes / 1–3 major) versus an East Pacific at 70% above-normal (15–22 named). The scoreboard above tracks both against reality through November 30.
Everything on this page is precomputed on a single small server that this site runs on. A pipeline fires twice daily, at 08:30 and 20:30 UTC: it runs a graph-neural-network weather model (Keisler 2022) out to ten days from ECMWF open analysis data, renders the global fields shown on the model bench, refreshes the ocean indices and storm records below, and publishes the result as a static site. No numbers are computed in your browser; each panel renders a JSON file stamped with its generation time.
ENSO indices. Weekly Niño 3.4 sea-surface temperature anomalies come from NOAA CPC's weekly OISST file (1990-base climatology), including the 1997 and 2015 analog years drawn from the same file. The seasonal ONI and the trend-corrected RONI come from the CPC ocean indices. Note the weekly file trails real time by up to ~11 days, and RONI/ONI are 3-month averages that update monthly.
Storms and ACE. Storm counts and accumulated cyclone energy are computed from NHC's ATCF best-track files: best-track lines at synoptic hours only, systems at tropical/subtropical-storm status or stronger (≥ 34 kt), ACE = Σ v² × 10⁻⁴ over 6-hourly fixes. Invests are excluded. Active-storm positions come from NHC's CurrentStorms feed.
Wind shear. The model bench instrument is a Keisler-2022 graph neural network at 1° resolution. At each run, 200–850 hPa vertical wind shear is computed pointwise as |V₂₀₀ − V₈₅₀|, then averaged (cosine-latitude weighted) over the Atlantic Main Development Region, 10–20°N and 20–80°W. The analysis-time value is appended to the season-long record in Figure 3; the model's own 10-day trace is kept so its forecasts can be scored against later analyses.
Limitations. The model is experimental: 1° resolution, atmosphere-only — no precipitation, no surface variables, no storm-intensity skill. Storm counts can shift in post-season best-track reanalysis. Nothing here is an official forecast.
This site is a research exercise. For watches, warnings, and anything that affects real decisions, consult the National Hurricane Center and the National Weather Service.