PLATE · STORM
SEPTEMBER CONVOY
2026-09-08 → 2026-09-11 · Dst min -38 nT · Kp peak 5 (G1)
Daily minimum Dst (aggregated from Kyoto WDC hourly data)
Four CMEs, two arrivals, one G1
Sunspot region AR4524 fired at least four coronal mass ejections toward Earth over the weekend of September 5-6. Ensemble models hinted the structures could merge en route; forecasters warned of G1 with a chance of G2, and headlines promised auroras over twenty US states. What actually arrived: two distinct impacts, the first on September 8 triggering a G1 (minor) storm, the second about ten hours later retriggering G1. Kp peaked at 5. The third and fourth CMEs either missed or arrived too weak to matter, and the G1 watch was called off by September 10. By September 11, Kp sat at 3 and the NOAA three-day outlook showed nothing above Kp 4.
This is the normal fate of CME convoy forecasts. Whether separate ejections merge into one magnetic structure is decided mid-flight, in solar wind conditions no model can observe in advance. A headline written from the maximum scenario is not a lie; it is a forecast shown its error bars too late.
What actually decides your aurora: Bz, not the CME count
The convoy's headline number was the CME count. The number that decided the aurora was Bz: hours of negative (southward) interplanetary magnetic field. This storm had them, briefly. A single bigger CME with northward field would have produced less.
The Russell-McPherron September bonus
September is a structurally good aurora month: around the equinoxes, the tilt of Earth's dipole lets southward-pointing solar wind fields couple to the magnetosphere more efficiently, the Russell-McPherron effect. The same CME that would scrape past in July can ring the magnetosphere in September. That is why a G1 in this week pushed visible aurora further south than its raw Kp suggests: northern US states, Scotland, southern Scandinavia reported low-horizon glow on the night of September 8-9.
Method: this page's Dst curve aggregates Kyoto WDC hourly raw data, reproducible; missing hours are marked, never interpolated. Kp values are NOAA SWPC provisional planetary indices. The CME count and launch times are cross-checked against SOHO/LASCO catalog entries; where models and catalogs disagree, both are noted.