Northern Lights
Heavy cloud covers the night sky above snow-covered hills, with no aurora visible

When the Sky Actually Opens

Which months have offered more usable sky in Tromsø, Reykjavík and Rovaniemi? Compare the published historical cloud figures, their measurement rules and the limits of the source.

Photo: Freepik

Almost every aurora guide tells you to travel in winter, because the nights are long. That advice looks at darkness and ignores cloud. The two do not move together. In Tromso they move in opposite directions, and December, the month with the most darkness and the highest prices, has the worst sky of the entire season.

We know this because the data has been sitting in public archives for decades. Nobody in this business publishes it, which is understandable: it is not a number that helps you sell a December tour.

How often the sky is actually usable

The table counts nights. The script checks hourly cloud-cover values at 18:00, 19:00 and each subsequent hour through 02:00, assigning the after-midnight samples to the preceding evening. A night qualifies when at least two samples are at or below 50% cloud cover. They need not be consecutive. This is a screening rule, not a measurement of two uninterrupted clear hours or a guarantee that a faint aurora could be seen.

Method correction — 20 October 2026. The cloud figures on this page are the previously published calculations from the Open-Meteo historical archive, covering September 2015 to April 2025. The September–March months in that interval form ten aurora seasons, not eleven. The archived calculation script did not select a weather model explicitly. Open-Meteo documents its default as a combination of models, so we cannot verify that the original sample used ERA5 alone or assign it a single grid resolution. The table has not been recalculated for this correction; use it as an indicative historical comparison, not a precise probability of seeing the aurora. Check the archive and model documentation.

Requested coordinates (latitude, longitude): Tromsø 69.65, 18.96; Reykjavík 64.15, −21.94; Rovaniemi 66.50, 25.73.

Nights with at least two qualifying hourly cloud samples, 18:00–02:00. Previously published Open-Meteo archive figures; ten seasons, 2015–2025.
Destination Sep Oct Nov Dec Jan Feb Mar
Tromso34.031.920.016.823.523.328.4
Reykjavik40.734.537.038.733.931.133.5
Rovaniemi41.733.220.022.322.930.450.0

December is the worst month in Tromso, and it is peak season

16.8%. Roughly one night in six. That is the December figure for Tromso, and it is the lowest number anywhere in the table.

Meanwhile the darkness is at its theoretical maximum. During the polar night the sun does not rise at all, and every one of the eight hours in our window counts as full astronomical darkness. So the hours are there. The sky is not.

This is worth sitting with for a moment, because it inverts the standard advice. You are being told to come at the exact time when your odds of a clear sky are at their annual worst, and to pay high-season prices for the privilege. The tour operators are not lying to you. They are answering a different question: December has long nights, Christmas holidays and the most flights.

Thick cloud over snowy hills at night, with no aurora visible through it
This is what a cloudy aurora night looks like. The lights may well have been active above that layer. You would not have known. · Freepik

September and March beat midwinter almost everywhere

Look at the two ends of the table instead of the middle. September gives Tromso 34.0% and Rovaniemi 41.7%. March gives Rovaniemi 50.0%, the highest figure in the whole dataset, against 22.3% in December.

Both are shoulder season, so they cost less. There is also a physical reason the equinox months do well, and it has nothing to do with cloud: the Russell-McPherron effect, described in 1973. Near the equinoxes the interplanetary magnetic field lines up more favourably with Earth's own field, and geomagnetic storms are close to twice as common as they are around the solstices. So the two months with the best odds of a clear sky are also the two months with the best odds of something happening in it.

The catch, and it is a real one, is darkness. We cover it properly on the moon and darkness page, but the short version is that September in Tromso gives you only about 1.8 hours of true astronomical darkness per night. The month with the best sky has the least dark in it. Nobody tells you that either.

Reykjavik is the boring one, and boring is a feature

Reykjavik never tops the table and never bottoms it. It sits between 31.1% and 40.7% all season, a spread of about nine points. Tromso swings from 16.8 to 34.0 and Rovaniemi from 20.0 to 50.0, which is more than double, end to end.

If you have a week and can chase the weather, that volatility is fine and occasionally works in your favour. If you have one night, booked months ago, it is not. Consistency is worth more than a high ceiling when you only get one roll of the dice, and Reykjavik is the only one of the three that offers it.

A number worth comparing against

A tour operator’s sighting rate and this table’s fixed-location cloud measure are not directly comparable. A guide may drive to another valley, count a brief gap as a successful night or use a different observation period. Ask what was counted before treating one percentage as better than another. Our own source uncertainty is disclosed above.

How we counted

Every choice behind the figures above. The thresholds are ours, and we would rather say so.
Choice Value Why
SourceOpen-Meteo historical archive, ten seasons 2015–2025; original model selection unverifiedOriginal request did not select a model
Samples18:00, 19:00, …, 02:00Local timestamps returned by the archive; nine hourly samples
Cloud threshold≤50%An editorial threshold, not a visibility guarantee
Qualifying nightAt least two qualifying samplesNot necessarily consecutive
Night attribution00:00–02:00 belongs to the previous eveningKeeps one night together across month boundaries
Period2015-09-01 → 2025-04-30September–March only: ten seasons

What these numbers do not tell you

  • Model and location: the original model selection is unverified. A grid value is not an observation at your exact viewing spot and cannot settle a local microclimate claim.
  • Thresholds: changing the cloud threshold or number of qualifying samples changes the result.
  • Visibility: passing the cloud rule does not mean an aurora was present, dark enough or visible.
  • Travel: tours can move. This fixed-location table does not measure an operator’s success rate.

Why Tromso is the cloudy one

The numbers make more sense once you know where the cloud comes from, and the answer is the same thing that makes Tromso habitable in the first place.

Tromso sits at 69 degrees north, further north than most of Alaska, and its winters are far milder than that latitude has any right to produce. The North Atlantic Current carries warm water up the Norwegian coast and keeps the fjords ice-free through winter. That warm water sits under a constant procession of low pressure systems tracking in off the Atlantic, and those systems arrive loaded with moisture that has nowhere to go but up over the coastal mountains.

So the mildness and the cloud are the same phenomenon. You cannot have an ice-free port at 69 degrees north and a desert sky above it.

Move inland and the trade reverses. Rovaniemi is roughly 300 kilometres from the coast, behind the Scandinavian mountains, with a continental climate: colder, drier, and progressively clearer as the season goes on. That is why its March figure is the highest in our dataset while its November figure is among the lowest.

Reykjavik is a third case. It sits in the middle of the North Atlantic, so it is cloudy by default, but it has no mountain range upwind and no continental interior to swing between. The weather is unsettled and consistently unsettled, which is exactly what its narrow 31 to 41 percent band describes.

This is also the reason serious operators drive inland when the coast closes. From Tromso the standard move is south and east towards Skibotn and the Finnish border, where the mountains have already wrung the moisture out of the air. Guides do this routinely, and it is common enough that they will tell you to bring a passport.

Checking this against the national weather services

Historical archive data and forecasts answer different questions. Use the historical comparison to understand the planning trade-offs, then check local cloud forecasts for your actual dates. MET Norway and the Icelandic Met Office are primary sources for local conditions. A forecast for a particular night cannot be inferred from a monthly historical percentage.

Using this when you actually book

Three practical consequences follow from the table, and they are worth more than the individual percentages.

  1. Allow more than one night, without treating it as a guarantee. As an illustration, if each night had a 25% chance of usable sky and nights were independent, three nights would give a 57.8% chance of at least one usable night: 1 − 0.75³. The chance of none would still be 42.2%. Weather is not independent from night to night: cloudy spells can persist. This example is not a forecast for a destination, and usable sky is not the same as seeing an aurora.
  2. Let the destination follow the month, not the other way round. If you can only travel in March, Rovaniemi is the strongest option on this table. If you can only travel in December, Reykjavik is. Picking the place first and the date second is the more common approach and the weaker one.
  3. Treat any single fixed night as a gamble, and price it accordingly. That does not mean do not do it. It means do not pay a premium for a guarantee that will not fire, and consider putting the same money into a second attempt instead.

Use these comparisons for your next decision: moon phases and darkness hours · Tromsø, Reykjavík and Rovaniemi compared · choosing the month for your trip.

Questions this raises

Should I avoid December entirely?

No. One night in six is still a real chance, and a December trip usually has other reasons behind it. But go in knowing the odds, book more than one night if you can, and treat any single-night December booking as a gamble rather than a plan.

Is September really better than January?

For cloud, clearly: 34.0% against 23.5% in Tromso. For darkness, no, September is much worse. The honest answer is that September suits people who want decent odds of a clear sky and are happy with a short window each night, and January suits people who want long dark hours and will accept more cloud.

Why not just use a forecast?

Because you are booking months ahead. Weather forecasts are useful three to five days out. Climatology is what you have when the decision is made in August for a trip in February. On the night itself, use the national met service instead.

Do these figures apply to the whole country?

No, and not even to the whole region. They are calculated at the coordinates of each city, and cloud varies sharply over short distances in mountainous terrain. We measured a 20 point difference between Abisko and Kiruna, 90 km apart. Numbers for one place do not transfer to another.

What about Abisko and its blue hole?

A gridded cloud estimate cannot establish or disprove Abisko’s local clear-sky effect. The original model selection has not been verified, so we do not assign this calculation a specific spatial resolution.

Why use a reanalysis instead of weather station records?

The archive provides hourly values for each requested location. That makes a consistent counting rule possible, but it does not remove uncertainty about the source model or local weather. Station observations are useful independent checks where available.

Could I reproduce these numbers myself?

The counting rule and coordinates can be checked, but we cannot currently promise exact reproduction of the published table: the original request did not select a model explicitly. A reproducible release requires a fixed model, saved input data and the exact calculation version. This correction does not claim that such a release has been completed.

Sources

  • Historical cloud figures: Open-Meteo historical archive, September 2015 to April 2025, covering ten September–March seasons. These are previously published values; the original model selection has not been verified. See the cloud method and correction and the Open-Meteo documentation.
  • MET Norway — national observations and forecasts for Norway: met.no.
  • Icelandic Met Office — cloud cover and aurora forecast for Iceland: vedur.is.
  • NOAA Space Weather Prediction Center — geomagnetic activity, which this page deliberately does not cover: swpc.noaa.gov.
  • Russell, C. T. and McPherron, R. L. (1973), on the semiannual variation of geomagnetic activity, Journal of Geophysical Research, doi 10.1029/JA078i001p00092.

Data verified on . Source description corrected on 20 October 2026. Cloud values are the previously published Open-Meteo archive calculations for September 2015–April 2025, ten September–March seasons. No new calculation is claimed; the original weather-model selection is unverified. Darkness and moon tables are separate astronomical calculations.

Cloud is only half the calculation

The other half is darkness and moonlight, and unlike the weather those can be worked out exactly, years in advance.