Data sources and methodology
Aurora Tonight doesn't measure space weather itself -- it reads public data products published by the NOAA Space Weather Prediction Center(SWPC), the US government agency responsible for space weather forecasting, and combines them with local computation to answer one specific question: is the aurora likely to be visible from a given place, right now. This page lists exactly which NOAA products are used, how often each one is expected to refresh, and exactly how the verdict shown throughout this site is calculated -- including its known limitations.
The NOAA data products this site uses
All of the following are public JSON products served fromservices.swpc.noaa.gov, fetched server-side by this site's Cloudflare Pages Functions (and, in local development only, directly from the browser as a fallback):
| Product | What it provides | Typical update cadence |
|---|---|---|
/products/noaa-planetary-k-index.json | Official (definitive) 3-hour planetary Kp index | Every 3 hours |
/products/noaa-planetary-k-index-forecast.json | 1-3 day Kp forecast, in 3-hour bins | Several times a day |
/json/planetary_k_index_1m.json | Near-real-time estimated Kp | Roughly every minute |
/json/rtsw/rtsw_mag_1m.json | Real-time solar wind magnetic field (Bt, Bz, By) | Roughly every minute |
/json/rtsw/rtsw_wind_1m.json | Real-time solar wind speed, density, temperature | Roughly every minute |
/json/ovation_aurora_latest.json | OVATION aurora oval probability grid (global, 1° resolution) | Roughly every 5 minutes |
/products/alerts.json | Space weather watches, warnings, alerts and summaries | As issued (event-driven) |
/json/goes/primary/xray-flares-latest.json | Recent solar flare (X-ray) events from the GOES satellites | As flares occur / periodically through the day |
Cadence figures above describe NOAA's typical publishing pattern for each product, not a guarantee -- any of these feeds can occasionally lag or be temporarily unavailable, and this site's status indicators reflect that (a reading older than about 30 minutes is flagged as stale rather than presented as current).
How the live verdict is actually computed
The verdict shown on city pages and the live dashboard runs entirely from four inputs, combined with a specific, fixed logic (implemented once, in src/client/verdict.ts, and used identically everywhere on this site):
- Geomagnetic latitude. Converted from a location's geographic coordinates using a centred-dipole approximation of Earth's magnetic field (see the accuracy caveat below). This determines the Kp index at which the auroral oval typically reaches that location, using a linear fit to NOAA's published Kp-to-latitude table.
- Current and forecast Kp. The current estimated Kp (or, if unavailable, the last observed value) is compared against that location's threshold, alongside the forecast Kp for the dark hours ahead.
- NOAA's OVATION model, where available. Kp is a single, planetary-scale number, and the real auroral oval is not a perfect circle -- it's lumpier, and shifts in shape with solar wind conditions. Where OVATION data is available, its higher-resolution local probability is used as the primary signal for the strongest verdict tier, with the Kp-based threshold used as the baseline and fallback.
- Darkness. None of the above matters if the sky is too bright to see anything against. The verdict is gated on solar elevation dropping below -6° (civil twilight) at the location and moment in question; above that, the verdict reports "daylight" or, at very high latitudes in summer, "midnight sun" rather than a probability.
Moon phase is also factored in as a caveat (a bright moon, above roughly 70% illumination in this site's logic, is flagged as likely to wash out a faint display) rather than as a hard gate, since a genuinely active aurora remains visible under a bright moon.
The verdict never claims certainty. Every result includes its underlying reasons (geomagnetic latitude, Kp thresholds, OVATION probability if available, darkness state, moon phase) and explicit caveats -- including that cloud cover is not accounted for at all, since no data source used here measures it, and checking a local weather forecast is left to the visitor. See how aurora forecasts workfor the fuller explanation of why no forecast, including this one, can promise a sighting more than roughly 30-90 minutes ahead with real confidence.
The geomagnetic latitude accuracy caveat
Geomagnetic latitude is computed here using a centred-dipole approximation: Earth's magnetic field is modelled as a single bar magnet through the planet's centre, tilted relative to the geographic poles. This is a standard, well-established simplification, and it is accurate to within about a degree for Scandinavia, continental North America, and the United Kingdom -- comfortably precise enough for the Kp thresholds this site quotes.
It is measurably less accurate in two specific regions, because Earth's real magnetic field is not perfectly centred on the planet's core (an "eccentric dipole" effect the simplified model doesn't capture): the North Atlantic, most notably Iceland, and the Australia/Tasmania sector. Cross-referencing multiple published aurora-forecasting services' "magnetic latitude" figures against this site's centred-dipole calculation found:
- Reykjavík, Iceland -- centred-dipole calculation: roughly 68.9°N. Commonly cited published values: roughly 64.7-65.6°N -- a gap of 3-4°.
- Hobart, Australia -- centred-dipole calculation: roughly 49.7°S. A published value cited by aurora-forecasting services: roughly 53.7°S -- a gap of about 4°.
A 3-4° gap is not a rounding error for this purpose -- run through this site's Kp conversion, it shifts the resulting Kp threshold by roughly one to two whole steps, which can be the difference between a city page reading as a "regular" versus a "storm-only" destination. Where this site has a specific, sourced corrected value for a city, it is set as an mlatOverride in that city's page data and used in place of the raw calculation (Reykjavík and Hobart, at the time of writing). For other cities in the same affected regions (for example Akureyri, Vík, Tórshavn, and Launceston) no separately sourced correction was available with enough confidence to publish as a specific number, so those pages state the uncertainty in their own text rather than presenting an unverified figure.
The live verdict computation cross-checks against NOAA's OVATION model wherever it's available specifically because OVATION works from real geographic coordinates and actual solar wind physics, not a simplified magnetic model -- when both signals are present, OVATION is treated as the more reliable one.
What this site does not do
It does not forecast cloud cover, does not have access to any live camera or reported sighting data, and cannot promise a specific sighting at a specific time -- see theDisclaimer for the complete list of limitations. Every number on this site is computed transparently from the data listed above, using logic that is identical across every page rather than tuned per location.
Frequently asked questions
Is Aurora Tonight affiliated with NOAA?
No. Aurora Tonight is an independent site that reads NOAA Space Weather Prediction Center's public data products and presents them with additional local computation (geomagnetic latitude, darkness, and a combined verdict). NOAA does not endorse or operate this site.
Why does the live verdict sometimes disagree with a plain Kp reading?
Because it isn't based on Kp alone -- it also weighs NOAA's OVATION aurora oval model where available, whether it is actually dark at the location in question, and the moon phase. Two locations at a similar Kp threshold can show different verdicts if OVATION shows the oval genuinely shaped differently on one side of the planet that night.
How accurate is the geomagnetic latitude shown for my city?
Generally accurate to within about a degree for Scandinavia, continental North America, and the UK. In the North Atlantic (notably Iceland) and the Australia/Tasmania sector, the simplified model used here can run 3-4 degrees off from better, cross-referenced values -- see the caveat below for the full explanation and which city pages carry a corrected value.
How often is the underlying NOAA data refreshed?
It varies by product, from roughly every minute (real-time solar wind and estimated Kp) to roughly every five minutes (the OVATION aurora model) to event-driven (space weather alerts, issued as NOAA publishes them). See the endpoint table below for specifics.