Storm Nights, Counted
The mean says the sky reaches geomagnetic storm level — the threshold aurora alerts fire on — 59 nights a year. The mean describes almost no actual year: the spread runs from 3 nights to 141, April doubles December, storms echo 27 nights later — and 2024, the year everyone finally saw the aurora, was a below-average storm year with an exceptional tail.
One question, one instrument
The one question an aurora chaser actually has — how many nights a year is it worth going out? — is usually answered with folklore. Solar maximum is good. Equinoxes are good. Last year was amazing. Every one of those claims is checkable against a single instrument record, so we checked.
The record is the planetary Kp index: every three hours since January 1932, a fixed network of ground magnetometer stations has been condensed into one number, 0 to 9, saying how disturbed Earth’s magnetic field is. It is the exact quantity aurora alerts are built on — NOAA’s storm scale is defined directly on it, from G1 (Kp 5, aurora pushed toward the northern-tier states) to G5 (Kp 9, aurora over the deep south). We counted a storm night every day whose peak Kp reached 5.
The record is remarkably complete: 274,672 readings across 34,334 days, 94 complete years, and we verified every single day carries all eight readings. Nothing here is modelled, forecast, or interpolated — every number on this page is a count over measured readings.
Ninety-four years of storm nights
Nights per year reaching G1 or stronger (peak Kp ≥ 5). The heartbeat is the 11-year solar cycle, visible in raw counts.
Highlighted: 1951 (141 nights), 2009 and 2020 (3 nights each), and 2024 (36). The mean is 59, but the average year barely exists — the record swings by a factor of 47. The longest wait between two storm nights was 334 days, October 2019 to September 2020. And the 2010s averaged just 26.8 storm nights a year against the 1950s’ 91.3: anyone who took up aurora photography in the 2010s learned the sky in the quietest decade of the entire record.
2024: the year of few but great storms
Here is the finding we did not expect. 2024 — the year aurora photographs saturated social media — had 36 storm nights, ranking 71st of 94. By how often the sky reached storm level, it was a well below average year.
But it ranks 7th of 94 by nights at G4 or stronger, matched 1960- and 1989-class years with 13 nights at G3+, and produced the record’s first Kp 9.0 in twenty-one years, on May 11. No other year in the record pairs so few storm nights with so many great ones. Both public impressions of 2024 are correct — it just wasn’t a year of many storms. It was a year of great ones. 2025, for comparison, has had nearly double the storm nights but nothing above G4, and far fewer headlines.
Every night at the top of the scale
The 27 days in 94 years on which Kp reached 9.0 — one every 3.5 years on average, arriving in clusters separated by droughts of up to two decades.
- 1938 Apr 16
- 1940 Mar 24
- 1941 Mar 1
- 1941 Jul 5
- 1941 Sep 19
- 1946 Jul 27
- 1946 Sep 22
- 1957 Sep 4
- 1958 Feb 11
- 1958 Jul 8
- 1959 Jul 15
- 1960 Apr 30
- 1960 Oct 7
- 1960 Nov 13
- 1967 May 25
- 1967 May 26
- 1970 Mar 8
- 1972 Aug 4
- 1982 Jul 13
- 1982 Jul 14
- 1986 Feb 8
- 1989 Mar 13the Québec blackout storm
- 1989 Mar 14
- 2000 Jul 15the Bastille Day storm
- 2003 Oct 29the Halloween storms
- 2003 Oct 30
- 2024 May 11the Gannon storm
The list reads as the history of famous space weather, which is a check on the counting itself. The gap before the most recent entry — October 2003 to May 2024, twenty and a half years — is the longest in the record.
April doubles December
Percent of nights reaching storm level, by calendar month, all 94 years pooled. Equinox months highlighted.
A given night reaches storm level 21.5% of the time in April and 9.4% in December — a factor of 2.3. This is the century-old “equinox effect” sitting plainly in raw counts, and it compounds with darkness: the storm-rich months are also months that actually have night at high latitudes.
The 27-night echo
Chance of a storm night, N nights after a storm night. The dashed line is the 16.1% base rate — any night at all.
Storms cluster, then echo. After a storm night, the next night storms 46% of the time — nearly three times the base rate. The probability decays within a few nights, then rises again to 30% exactly 27 nights later: one rotation of the sun, when the same coronal hole or active region swings back to face Earth. Julius Bartels invented this index in part to see that pattern; ninety-four years later it is still there in nothing more than conditional counting. The practical rule: after a storm night, go out again the next night — and mark the calendar 27 nights ahead.
What this counts, and what it doesn’t
A storm night here means the planet reached storm level that day — not that you, personally, saw the aurora. Whether a given storm was visible from your driveway depends on your latitude (G1 reaches the northern-tier states; it takes G3 or G4 for the mid-latitudes), on whether the storm peaked while you had darkness, and on cloud and moon. These counts are the opportunity ceiling: the number of times a year the sky gives you a chance. Local weather then takes its cut.
One more honesty note: Kp comes in thirds of a unit, and we count a threshold as crossed only at the whole number — a night peaking at Kp 4⅔ does not count as a storm night. Counting those near-misses would raise the average from 59 to 79 nights a year without changing a single comparison, ranking, or pattern above. We publish the stricter number.
Reproducibility
Every number on this page comes from the results.json file in the aurora-nights workspace, and a checker script re-derives all 34 quoted figures independently of the analysis code. Pipeline: the GFZ Kp record is read from the TerraPulse kp_index dex — one slot per year, 95 slots, each carrying the complete 3-hourly series; coverage is verified exhaustively (every day, all eight readings, no nulls); and every result is a count at NOAA’s published thresholds, with the threshold-rule sensitivity reported. Because storm nights cluster, the 27-night echo also carries a year-block bootstrap interval alongside the naive one. No model, forecast or reanalysis field is used anywhere.
TerraPulse Data Lab, 2026-09-01. Data: Kp and ap index by GFZ German Research Centre for Geosciences (DOI 10.5880/Kp.0001, CC BY 4.0). G-scale thresholds and visibility descriptions: NOAA Space Weather Prediction Center. Storm nights are UTC days. All times UTC.
Published paper
The full scientific paper, with methods, tables, and references.