Methodology
How the Aurora Hunt forecast works
Aurora Hunt forecast pages combine public space-weather data with a local cloud forecast, the position of the Sun and the location’s magnetic latitude. The result is an estimate that helps you decide whether it is worth checking the sky. It is not a measurement of aurora at your location and it cannot guarantee a sighting.
Published by Aurora Hunt (Leveling PRO). Last reviewed 2026-11-11.
Data sources
- Planetary Kp index from the NOAA Space Weather Prediction Center (SWPC). City pages use the most recent observed or estimated value in the SWPC planetary K-index forecast product, never a future value. Kp describes global geomagnetic activity, not conditions over one town.
- NOAA OVATION aurora model (SWPC “ovation_aurora_latest”), a short-term model of where the auroral oval is expected. We use the value of the grid point nearest to the location.
- Cloud cover from the MET Norway Locationforecast API (cloud area fraction). A scheduled job refreshes these values for our locations in rolling batches and stores them; pages read the stored value and its refresh time.
- Magnetic latitude computed with the AACGM-v2 model (version 2.7.1, IGRF-14 coefficients, epoch 4 September 2026, ground level) from each location’s coordinates. The values are precomputed and stored with the site.
- Solar wind (Bz), the observed Kp product, NOAA alerts/watches and the NOAA 27-day outlook from SWPC are used on the live aurora map, the geomagnetic storm page and our curated country and destination pages.
- Darkness and Moon are calculated by us: the Sun’s altitude from the date, time and coordinates, and the Moon’s phase and illumination from the average lunar cycle. Local time zones come from the coordinates.
Attribution and terms for each provider are listed on the data sources page.
How the local estimate is calculated
- Activity estimate. We start from the NOAA OVATION value nearest to the location, add a small amount for locations at high magnetic latitude given the current Kp, and add more during geomagnetic storms (Kp 5 or higher). When OVATION shows very little aurora nearby and Kp is below 6, the estimate is capped low (at most 15–35% depending on the OVATION value) so a quiet night cannot look “excellent”. If OVATION is unavailable, a conservative rule based only on Kp and magnetic latitude is used and the estimate is capped at 20%.
- Data checks. An estimate needs a verified magnetic latitude for the location, a stored MET Norway cloud reading refreshed within the last 6 hours, a Kp observation no older than 6 hours and, when OVATION is available, an OVATION observation no older than 1 hour. If any of these is missing or stale, the page shows no percentage or viewing window and says that current data is insufficient. If NOAA’s Kp feed cannot be read, no Kp value is shown either. No substitute Kp or cloud value is used.
- Darkness and cloud. For each of the next 24 hours the activity estimate is reduced by the cloud cover (for example 60% cloud leaves 40% of the estimate) and set to zero unless the Sun is at least 12° below the horizon.
- Headline number and verdict. The percentage on a city page is the highest hourly estimate in the next 24 hours, never more than 95%. The verdict sentence follows fixed bands: 76% and above, 51–75%, 21–50%, and below 21%.
- Viewing window. The window is the stretch of consecutive dark hours with an estimate above zero that has the highest average estimate (ties go to the longer, then the earlier stretch). It is shown in the location’s local time. If there is no such stretch, the page says so.
The percentage is an uncalibrated estimate. It has not been statistically validated against sightings, so it should be read as “more or less promising”, not as a measured probability. On our curated country and destination pages we do not show a percentage at all; those pages show the current Kp and NOAA G-scale, the solar-wind Bz direction, a Kp range for each location’s magnetic latitude, darkness and the local cloud reading when a recent one is available.
Magnetic latitude and Kp ranges
Aurora follows the Earth’s magnetic field, so magnetic latitude says more than geographic latitude about how strong activity has to be. Where we show a Kp range for a location, it comes from these bands of AACGM-v2 magnetic latitude (north or south):
- 67° and above: Kp 0–2
- 62° to below 67°: Kp 1–3
- 57° to below 62°: Kp 3–5
- 52° to below 57°: Kp 5–7
- below 52°: Kp 7–9
These ranges are screening guidance for planning, not a threshold at which aurora becomes visible. If a location’s coordinates do not match our magnetic-coordinate registry, the magnetic latitude is shown as unavailable instead of being guessed.
Best months: long-term climatology
Indexed city pages show a 12-month table of long-term averages for 2001–2025. It describes past years and is not a forecast. The values are precomputed for each place:
- Clear night sky: 100% minus the mean night-time cloud amount (
CLOUD_AMT_NIGHT) from the NASA POWER climatology, based on CERES SYN1deg on a 1° grid, January 2001 to December 2025. It is the average cloud-free share of the night sky over that grid cell, not a count of clear nights. When the night value is missing because the sun does not set (polar day), the all-day cloud amount is used and the month is marked with an asterisk. - Darkness: for every night of the year, the interval during which the sun is at least 12° below the horizon, scanned from local noon to the next local noon in 5-minute steps with the NOAA solar-position equations (reference year 2025; the time zone is the place’s own). The table shows the hours of darkness in the night that starts on the 15th.
- Nights with enough activity: the share of all nights from 2001 to 2025 that had at least 2 hours of darkness and on which the highest 3-hour Kp value overlapping that night’s dark interval (in UTC) reached the lower bound of the place’s Kp range above. The window follows the season: short summer nights cover fewer Kp slots, and nights without 2 hours of darkness never count. Kp values come from the GFZ Helmholtz Centre for Geosciences (Matzka et al., 2021); a value of 5− (4.67) does not count as 5. Where the range starts at Kp 0, the column is not shown and only darkness counts.
- Aurora-friendly nights: days in the month × share of nights with enough activity during darkness × clear-sky share. The best months are the two months with the most such nights, plus a third when it reaches 80% of the second, ranked before rounding. Below one night a year no best months are named.
Limits: the cloud amount is a grid-cell average and is treated as independent of geomagnetic activity; local weather, terrain, light pollution and moonlight are not included; the Kp ranges are screening bands rather than visibility thresholds. A place without verified data for any input shows no table.
Places whose Kp range is 7–9 (the lowest band) show no table, best months or yearly count: there the aurora needs a strong geomagnetic storm and is a rare event, so the page says that instead and keeps the direction to look.
Where to look
The direction shown on indexed city pages is the initial great-circle bearing from the place to the geomagnetic pole of its magnetic hemisphere, rounded to the nearest degree and to one of eight compass points. The pole is the axis of the centred dipole of the IGRF-14 field for 2026.0, from the degree-1 Gauss coefficients for 2025.0 plus one year of secular variation (g10 = −29,337.4 nT, g11 = −1,400.3 nT, h11 = 4,524.0 nT): about 80.8° N, 72.8° W in the north and 80.8° S, 107.2° E in the south.
The aurora usually appears first low in the sky in that direction when you are well away from the auroral oval. At an AACGM-v2 magnetic latitude of 64° or more, the oval can be overhead, so during activity the aurora can appear anywhere in the sky; the page then says so. The bearing is a starting point, not a prediction of where a display will be.
Updates, freshness and caching
- Forecast pages are generated on request and may be served from a cache; a cached copy is normally refreshed after about 15 minutes, and an older copy can be shown while a refresh is running or if it fails.
- Kp is re-read from NOAA at most every 15 minutes and OVATION at most every 5 minutes; other space-weather inputs are cached for about 5 minutes.
- A Kp observation older than 6 hours, an OVATION observation older than 1 hour or a cloud value refreshed more than 6 hours ago is treated as stale. Stale or missing inputs are labelled on the page, and the affected numbers are hidden or marked as unavailable rather than filled in.
- Each city page shows when it was generated, and the state and time of its inputs under “How to read this forecast”.
Known limitations
- The activity estimate is held constant across the 24 hours; it does not forecast when a substorm will happen.
- The stored cloud value is a single reading applied to every hour unless an hourly cloud forecast is available, so changing cloud during the night is not captured.
- Moonlight, light pollution, terrain and the view of the horizon are not part of the estimate. Bright moonlight is flagged on the page when the Moon is more than 75% illuminated.
- The OVATION model has a coarse grid and a short lead time; Kp is a three-hour global index. Neither describes your exact sky.
- Faint aurora is often easier to see in a camera image than with the eye.
- Space weather changes quickly. The forecast is probabilistic: a high estimate can pass without a display, and aurora can appear when the estimate is low.
Corrections
If something on a forecast page looks wrong, email hello@aurorahunt.app with the page URL, the time and what you saw. More about who publishes Aurora Hunt is on the about page.