How the forecast works

Little Aurora answers one question for each night: is it worth heading out, and where? This is how that answer is made, what each number means, and where it can be wrong.

Published by Little Studio, the studio of Muaaz Saeed, which makes Little Aurora. Last reviewed 28 October 2026. Spotted a mistake? Write to hello@littlestudio.app.

The answer for tonight

Each place gets one of these for the night ahead, worked out hour by hour for every viewing spot.

Worth heading out
Moderate or stronger aurora activity is expected at the place, under a clear sky, for at least an hour of darkness. The page shows the best viewing window and the spot that suits it best.
Wait for a clearer sky
There’s activity, but the cloud is patchy; or the activity is weak under a clear sky. Check again later: these nights can turn either way.
Stay in
Cloud is likely to cover the sky, or the activity is too weak to see. Weak activity counts as none when the moon is up and more than half lit, or at a spot with strong light pollution.
Too light, or dark hours over
The sun doesn’t get far enough below the horizon, or tonight’s dark hours have passed. Aurora can’t be seen in daylight, whatever the activity.
Still waiting for a forecast
A source the answer needs is missing or out of date. Little Aurora says so rather than guessing, and never turns missing data into “stay in”.

Why strong activity can still mean stay in

Aurora activity and cloud are separate forecasts. A high activity forecast means the aurora is likely to be overhead; it can’t see through cloud. If the cloud forecast covers the dark hours, the answer is stay in, however strong the activity. A clear sky on a quiet night is stay in too, because there’s nothing to see.

An example: on 28 October 2026, Abisko’s page read stay in beside patchy cloud and a named spot. The activity forecast there was weak, and with a bright moon up, weak activity counts as none, so the cloud didn’t matter. Little Aurora ranks the spots every night, by cloud and darkness, even when there’s no window. A page now calls a spot recommended only on a night worth heading out, and “best chance if it clears” on a night to wait.

How each hour is judged

  1. Darkness. The sun’s position is calculated for each spot and hour. The night’s dark hours are when the sun is 12° or more below the horizon. Between 6° and 12°, in twilight, only strong displays count; above 6°, nothing does.
  2. Aurora activity. For the next few nights, from NOAA’s three-day Kp forecast, compared with how far the spot is from the magnetic pole. The aurora’s southern edge is taken to sit at about 66.5° magnetic latitude minus 2° for every point of Kp. A spot well inside that edge gets moderate activity, more as Kp rises; one near it gets weak activity, low on the horizon. Close to the time, NOAA’s 30-minute OVATION aurora model at the spot takes over. Live readings can raise it but never lower it: GFZ’s half-hourly Hp30 index, and in Finnish Lapland and around Fairbanks, the nearest magnetometer.
  3. Cloud. From MET Norway’s hourly forecast for each spot. The cloud that counts is the largest of low cloud, middle cloud and half of high cloud, because bright aurora shows through thin high cloud. Under 30% is clear, 30 to 70% patchy, over 70% cloudy. For the hour in progress and the next, a weather station within 40 km with a report under 75 minutes old can overrule the forecast: a clear report lifts a stay-in hour to wait, and a cloudy one holds a go hour back.
  4. The moon and lights. A moon above 5° and more than half lit, or a spot near bright lights, hides a weak display.

The hour is worth heading out when activity is moderate or stronger under a clear sky.

The viewing window and the spot

The window is the longest unbroken run of good hours, at least an hour long, with its edges moved to the half hour where the cloud forecast allows. Once the night has started, only the hours still to come count, so a window that has passed is never shown as tonight’s.

A night is named by the evening it starts. When the window opens after midnight, the page says so: a 00:00–03:00 window on the night of Wednesday is in the early hours of Thursday. Times are in the place’s own time zone.

The spot is chosen by the longest window, then the least cloud in it, then the darkest sky, then the most open view toward the pole. On a night without a window, spots are ranked by cloud over the dark hours still to come.

What the numbers on a page mean

Cloud: “Average cloud cover 24% in the window”
The average of each hour’s cloud, as defined above, across the viewing window, or across the dark hours when there isn’t one. It’s how much of the sky is forecast to be covered, not a chance of anything.
Aurora: “Moderate activity”, “Kp 3.7 forecast at its highest”
Activity is the strength expected at that place, from the rules above. Kp is a planetary index of geomagnetic activity from 0 to 9; the figure is the highest forecast for those hours. Neither is a probability of seeing the aurora.
Darkness and moon
When the sun reaches 12° below the horizon, and whether the moon is up during those hours and how much of it is lit.
Confidence
Starts high and drops a step for each of: a source running on a fallback, patchy cloud around the window, only one cloud model or models that disagree, a spot near the edge of the aurora’s usual reach, a weather station reporting cloud the forecast doesn’t have, and a night two or more nights ahead. Cloud comes from one forecast model per place at the moment, so a night worth heading out is at most medium confidence.
Up to date, and the forecast details
“Forecast data up to date” means the cloud and aurora forecasts behind the answer are within their freshness limits. If a source stops updating, the page says instead that the forecast hasn’t updated as it should. Under Forecast details: when the cloud forecast was last updated, when NOAA issued the Kp forecast (twice a day), when Little Aurora last recalculated the answer from them (about every ten minutes), and when the page itself was made (pages are kept for up to 15 minutes). Only the first two say how fresh the forecast is.

Clear dark nights by month

The chart on a place page shows the share of nights with at least two dark hours under 30% cloud cover, averaged by month. It comes from 30 years of Copernicus ERA5 reanalysis (1996–2025), checked hour by hour against the place’s own weather station. Where the two disagree badly, as in Abisko’s rain shadow, the station’s own recent record is used instead, scaled to match ERA5’s way of counting cloud; each chart names its years. Four places have no record to check against, so their pages leave the chart out.

It describes cloud, not aurora. A clear night still needs activity, and on any single night the forecast matters far more than the month.

Where it can be wrong

  • Cloud forecasts miss local cloud, especially along coasts, in valleys and with fog. Conditions a few kilometres apart can differ.
  • The Kp forecast is a planetary average in three-hour blocks. Displays come and go within minutes, and a quiet forecast can still bring a short burst.
  • The thresholds are starting points. Little Aurora compares its cloud forecasts with weather-station reports every hour, and asks travellers whether they saw the aurora, to check and adjust them.

No forecast can promise a display. Treat the answer as the best use of your evening, not a guarantee.

Sources, and what Little Aurora adds

Little Aurora brings them together for each viewing spot and hour, in the place’s own time, and turns them into one answer with a window and a spot. The places and spots are chosen and described by Little Studio from public sources; each spot says how it was checked.