A forecast you can read. And question.

1. The northern lights model

Our map uses NOAA’s OVATION model. It is a short-term model of the location and intensity of the northern lights. Its lead time varies with solar-wind conditions; NOAA’s fallback mode can have no lead time. We display the actual observation and forecast times, rather than promising a fixed forecast horizon.

The overlay retains the source’s one-degree grid. The colours are the native model intensity values, not a calibrated personal probability of seeing the northern lights. A city’s activity category uses its nearest model cell: low overhead at 0–5, present above 5, and elevated above 30. These are explanatory display categories, not scientifically validated visibility thresholds.

2. The weather where you are

Cloud coverage, temperature and wind come from MET Norway Locationforecast. We show total cloud coverage for the destination’s coordinates. This is forecast spatial coverage, not the probability of seeing northern lights. Lofoten is a region represented by a point in Svolvær. There is no continuous cloud map in this version.

Weather data is attributed to MET Norway under CC BY 4.0. Forecasts are refreshed according to the provider’s expiry information. A late weather forecast is labelled stale.

3. A useful definition of darkness

We calculate the sun’s altitude with SunCalc 2. A practical dark window begins when the sun is at least 12° below the horizon, sampled in five-minute steps. This is the end of nautical twilight; it is not a guarantee of astronomical darkness. A bright display may still be visible before this threshold.

“Tonight” runs from 18:00 to 08:00 at the destination. Before 08:00, it refers to the evening that began the previous day. Times use the destination’s IANA time zone, including daylight-saving changes. The map’s night shading uses the horizon at the selected model’s forecast time; it is a separate day/night context layer.

The seven-night calendar

The calendar combines MET Norway weather with calculated darkness for seven local nights. Cloud ranges use actual forecast samples during dark hours, not a continuous prediction of the whole night. Later weather forecasts may have six-hour gaps between samples; we show their original times without inventing hourly values. For an ongoing night, the calendar shows only remaining darkness. After dawn, it marks the dark window as ended instead of presenting past darkness or daylight Kp as a future viewing opportunity.

Activity comes from NOAA’s issued three-day forecast. We show the highest forecast Kp among three-hour intervals that overlap the dark window. This is a planetary index, not local northern lights intensity or a sightings probability. UTC forecast boundaries can cover only part of a local night; those nights are labelled partial. Later nights say “Too early to forecast”.

The activity bulletin is checked every 30 minutes and becomes stale after 24 hours or a failed refresh. An expired bulletin is not used for future nights. Weather, activity and darkness retain separate availability indicators. No seven-day northern lights prediction is inferred from the short-term OVATION map.

How we describe the outlook

Our deterministic rules, version separate-conditions-v1, first check brightness, then data availability, clouds and modelled activity. Cloud coverage of 75% or more is described as limiting; 35–74% suggests watching for gaps. We don’t combine the inputs into a visibility percentage.

Kp does not veto a viewing opportunity. Low Kp can still coincide with northern lights at high latitudes. Moonlight is not treated as a universal blocker. These rules explain conditions; they have not been validated as a sighting prediction model.

Measurements, timestamps and stale data

Planetary Kp is NOAA’s estimated geomagnetic index. Solar-wind speed and Bz come from NOAA’s current real-time solar-wind feeds using records marked active. The original spacecraft name and overall quality are retained. Invalid or missing measurements create gaps rather than fabricated zeroes.

Solar-wind measurements become stale after 15 minutes, Kp after four hours, and weather after six hours. Northern lights model data becomes stale when its observation is older than 90 minutes or its valid time is more than 20 minutes past. Refresh failures preserve and label the last available data. A valid zero remains a zero. Each source has its own time in UTC; the hourly destination timeline uses local time.

What the history player shows

Playback contains only model snapshots captured by Northern Forecast. It is neither a record of confirmed sightings nor an extended future forecast. The first snapshot starts the archive; gaps are possible. This version refreshes shared data when requested. Continuous unattended collection requires an external scheduler.

Map and photography credits

Base-map boundaries are public-domain Natural Earth data, rendered with MapLibre GL JS. They are for geographic context and are not suitable for road navigation. The photograph of the northern lights above the mountains is by Eugen on Unsplash, used under the Unsplash License. It is an illustrative photograph, not a current observation.