We count, rather than guess. For Abisko and Fairbanks we match real hourly cloud records to the geomagnetic record night by night. For Iceland we do the same with airport weather reports from Reykjavík and Keflavík, from 2012.
Key takeaways
- Weather comes in runs, so real odds are lower than the formula: a proper display within 3 March nights in Fairbanks is 52%, not 65%.
- A proper display means Kp 2 or more plus a clear hour; a faint aurora needs Kp 1 or more, a strong show Kp 3 or more.
- Reykjavík city airport reads about 10 points clearer than Keflavík, so Reykjavík is quoted for a city stay and Keflavík as the conservative figure.
- Driving to clear sky can't be measured by fixed weather stations, so any figure for it on the site is an estimate.
What “faint”, “proper” and “strong” mean here
In short Faint means Kp 1 or more, a proper display Kp 2 or more, and a strong show Kp 3 or more, each with at least one clear hour.
Kp is the standard 0–9 scale of geomagnetic activity, measured every three hours. We use it to sort nights by how strong the aurora was likely to be. The Icelandic Met Office says even grade 2 “can be beautiful” and grade 3 “can be dazzling”.
Faint
Kp 1 or more, plus a clear hour
- A pale arc, often low in the northern sky
- Used on the Abisko and Fairbanks pages
A proper display
Kp 2 or more, plus a clear hour
- A clearly visible arc or band that moves
- The main measure on our Abisko, Fairbanks and Iceland pages, including how many nights you need and the Fairbanks guide
A strong show
Kp 3 or more, plus a clear hour
- Brighter, higher, more likely to fill the sky
- Used the same way on the Abisko, Fairbanks and Iceland pages
A clear hour means 25% cloud or less at Abisko, and no cloud layer beyond “few” (two-eighths of the sky or less) at Fairbanks and at the two Iceland airports. A night counts if at least one hour in the evening window is clear while Kp is at or above the level.
The records we use, and the years they cover
In short Kp comes from GFZ Potsdam (1932–2026), and cloud comes from station and airport reports at Abisko (2019–2026), Fairbanks (2000–2026) and Reykjavík and Keflavík (2012–2026).
| What | Source | Period | Used for |
|---|---|---|---|
| Geomagnetic activity (Kp) | GFZ Potsdam, 3-hourly | 1932–2026 | Share of evening 3-hour periods at Kp 3+ by month |
| Kp per night | GFZ Potsdam | 1996–2026 | Share of nights whose most active three hours reached Kp 3 (activity only) |
| Cloud, Abisko | SMHI station 188790, hourly | 7 winters, 2019–2026 (1,435 nights) | Nights needed, clear nights; window 7pm–midnight (18:00–23:00 UTC) |
| Cloud, Fairbanks | Fairbanks airport reports, via Iowa Environmental Mesonet | 26 winters, 2000–2026 (6,057 nights) | Nights needed, clear nights; window 9pm–3am Alaska time |
| Cloud, Reykjavík and Keflavík airports | Airport weather reports, via Iowa Environmental Mesonet | 14 winters, 2012–2026 (3,271 nights at each) | Nights needed, clear nights; window 9pm–2am (21:00–02:00, Iceland keeps UTC) |
| Cloud, Reykjavík | Icelandic Met Office station 1 | 1996–2025 monthly averages | Sky covered by month |
| Cloud, Yellowknife | Environment and Climate Change Canada normals | 1981–2010 | Share of all hours with 0–2 tenths cloud |
| Cloud, Tromsø and Rovaniemi | WeatherSpark, from NASA's MERRA-2 model | Modelled | Month-to-month comparison only |
Two ways of counting Kp 3
You'll see two different Kp 3 figures on this site, and they are not the same measure.
- Evening 3-hour periods, 1932–2026: the share of 3-hour periods between 21:00 and 03:00 UT with Kp 3 or more. March is 39.2%, December 27.9%. This is the column on the best time to see the northern lights.
- Nights, 1996–2026: the share of nights whose most active three hours reached Kp 3. March is 53%, December 38%. This is the figure on the Iceland month-by-month page and the Iceland month guides.
A night counts if any of its periods reaches Kp 3, so the night figure comes out higher. Both put March top and December bottom.
How we worked out the nights needed
In short We matched each station's cloud reports to the Kp index, then for each month counted the share of real runs of consecutive nights with at least one good night.
- Cloud, Abisko: every hourly total-cloud reading from the Swedish Meteorological and Hydrological Institute's Abisko station (188790), 2019–2026. Clear means 25% cloud or less.
- Cloud, Fairbanks: every routine hourly report from Fairbanks International Airport, 2000–2026, via the Iowa Environmental Mesonet archive. Clear means no layer beyond “few”.
- Cloud, Iceland: every report from Reykjavík city airport and Keflavík airport, 2012–2026, from the same archive. Clear means no layer beyond “few”, or the report says no significant cloud.
- Activity: the Kp index from GFZ, the German Research Centre for Geosciences, matched to each hour's 3-hour Kp interval.
- Counting: for each month, every real run of n consecutive nights (1 to 14) inside the season, with no missing data. The share of runs with at least one good night is the chance shown. Ranges come from resampling whole winters 2,000 times.
Formula against real runs of nights
In short Real runs of nights come out lower than the formula 1 − (1 − p)ⁿ, because weather comes in runs rather than fresh rolls of the dice each night.
The usual shortcut treats every night as a fresh roll of the dice. Weather comes in runs, so the real odds are lower. Some examples from the same records:
| Example | Formula says | Real runs of nights |
|---|---|---|
| Abisko, January, proper display, 3 nights | 73% | 62% |
| Abisko, November, faint aurora, 3 nights | 81% | 72% |
| Fairbanks, March, proper display, 3 nights | 65% | 52% |
| Fairbanks, December, proper display, 3 nights | 43% | 36% |
| Fairbanks, March, proper display: nights for 80% | 5 | 9 |
| Abisko, January, proper display: nights for 80% | 4 | 6 |
So we report the real runs, not the formula. They are lower in almost every month and at every level. That is also why we say to add a night or two to whatever simple odds suggest.
Faint aurora and strong shows at Abisko
In short At Abisko a faint aurora usually takes about 4 nights for an 80% chance, while an 80% chance of a strong show takes 12 nights to more than two weeks.
| Month | Faint: nights for 80% | Strong: chance within 3 nights | Strong: nights for 50% |
|---|---|---|---|
| September | 4 | 39% | 5 |
| October | 4 | 32% | 6 |
| November | 5 | 31% | 7 |
| December | 4 | 26% | 7 |
| January | 4 | 40% | 5 |
| February | 3 | 33% | 6 |
| March | 4 | 32% | 7 |
Iceland, measured
The Iceland odds use the same method as Fairbanks, on the weather reports from two airports in the southwest: Reykjavík city airport and Keflavík. We use the 14 winters from 2012 to 2026. The archive goes back to 2000, but before 2012 it is mostly a later reconstruction of the reports, which shows implausibly few clear hours (7–10% of evening hours, against 16–32% in the years since), so we left those years out.
- Window: 9pm to 2am (21:00–02:00). Iceland keeps UTC all year, so local time and UTC are the same.
- Proper display: at least one clear hour in the window while Kp was 2 or more. Strong show: the same at Kp 3 or more.
- Two stations, two figures: Reykjavík reads about 10 points clearer than Keflavík. Reykjavík may be sheltered by the mountains to its north, or the two stations may simply report differently; we can't tell which. We quote Reykjavík for a stay in the city and Keflavík as the conservative figure.
Both airports are on the southwest coast, so these are odds for staying in the Reykjavík area. Whether driving to clear sky does better, and by how much, is not something fixed weather stations can measure; where a page talks about it, it is our estimate.
Limits
- Abisko has only seven winters, so its ranges are wide.
- Fairbanks sits slightly further from the oval than Abisko, so the same Kp gives a slightly lower aurora there; we didn't correct for that.
- Both are single weather stations; hills nearby can be clearer or cloudier.
- Fairbanks' cloud sensor can't see cloud above 12,000 ft, so its clear-night figures are, if anything, slightly generous.
- The Iceland odds come from two airports on the southwest coast, over 14 winters. Airport reports can't see thin high cloud well, so they are slightly optimistic, and they say nothing about the north.
- Cloud figures for Reykjavík, Yellowknife, Tromsø and Rovaniemi were gathered in different ways and can't be ranked against each other or against Abisko and Fairbanks.
Tours, stays and prices
Tour, hotel and activity details on this site are compiled from the operators' own published terms and prices, and from GetYourGuide's and Vio.com's listings and review counts, on the date given on each page. We have not taken these tours or stayed at these places. Prices shown in Indian rupees by the booking sites are converted at 95.54 per US dollar, the rate the sites themselves applied on 23 September 2026, and rounded to the nearest $5. Prices change by date, so check your own.
Questions about our aurora odds
What does Kp mean?
Kp is the standard 0–9 scale of geomagnetic activity, measured every three hours and published by GFZ in Germany. We use it to sort nights by how strong the aurora was likely to be. The Icelandic Met Office says even grade 2 “can be beautiful” and grade 3 “can be dazzling”.
Why are your odds lower than a simple calculation gives?
Because cloudy nights come in runs. We count real runs of consecutive nights instead of multiplying single-night odds. For three nights in Fairbanks in March, the formula says 65% and the real runs say 52%.
Are the Iceland odds measured?
Yes, from 2012. We match the weather reports from Reykjavík and Keflavík airports, 2012–2026, to the Kp record night by night, the same way as Fairbanks. Earlier years are left out because the archive before 2012 is a reconstruction with implausibly few clear hours. The one thing not measured is how much driving to clear sky adds.
Where do the tour and hotel prices come from?
From the operators' own published terms and prices, and from GetYourGuide's and Vio.com's listings, on the date given on each page. We have not taken these tours or stayed at these places. Prices change by date, so check your own.
Where these numbers are used
Sources (10)
- GFZ Potsdam: Kp index since 1932 (checked 2026-09-23)
- SMHI: hourly cloud, station 188790 Abisko Aut, corrected archive (checked 2026-09-23)
- Iowa Environmental Mesonet: Fairbanks airport (PAFA) weather reports, 2000–2026 (checked 2026-09-23)
- Iowa Environmental Mesonet: Reykjavík (BIRK) and Keflavík (BIKF) airport weather reports, 2012–2026 (checked 2026-09-24)
- Icelandic Met Office: Reykjavík station monthly normals (checked 2026-09-17)
- Icelandic Met Office: aurora forecast and how to read it (checked 2026-09-17)
- Environment and Climate Change Canada: climate normals 1981–2010, Yellowknife A (checked 2026-09-17)
- WeatherSpark: average weather in Tromsø (modelled) (checked 2026-09-17)
- WeatherSpark: average weather in Rovaniemi (modelled) (checked 2026-09-17)
- Our calculation: nights-needed method and results, Northern Lights Planner, 23 September 2026 (Iceland added 24 September 2026)