Where the data comes from

DONKI is operated by NASA's Community Coordinated Modeling Center (CCMC) at Goddard Space Flight Center. It is the primary public database for space weather events - solar flares, coronal mass ejections, and geomagnetic storms - issued and annotated by space weather forecasters.

The forecast updates every 30 minutes by querying the DONKI API. NASA's involvement means the underlying data has been reviewed by professional space weather analysts, not generated automatically.

NOAA - the National Oceanic and Atmospheric Administration - is the US federal agency responsible for weather forecasting, including space weather. Its Space Weather Prediction Center (SWPC) in Boulder, Colorado operates 24 hours a day and issues the official geomagnetic storm watches, warnings, and alerts used by governments and utilities worldwide.

The real-time solar wind measurements come from the DSCOVR satellite (Deep Space Climate Observatory), which sits at the L1 Lagrange point - around 1.5 million kilometres between the Earth and Sun. From this position it can measure the incoming solar wind roughly 15-60 minutes before it reaches Earth, giving just enough warning to be useful.

These readings update every minute at the source. Aurora Tonight fetches a 2-hour rolling window and displays the most recent valid reading, refreshing every five minutes on the forecast page, the homepage, and every location page.

Knowing that a Kp 7 storm is in progress is only half the picture - if there is 100% cloud cover overhead, you won't see anything. Open-Meteo is an open-source weather API that provides hourly and daily cloud cover forecasts, drawing on data from national meteorological services including ECMWF, GFS, and others depending on the region.

On individual location pages, Aurora Tonight fetches cloud cover data directly from Open-Meteo in your browser using the page's coordinates. This means the cloud forecast is always pulled fresh for the specific latitude and longitude of that location - not a generic regional figure.

Open-Meteo is free to use and does not require an API key. It is widely used in weather and astronomy applications. The underlying forecast models are the same ones used by professional meteorologists - Open-Meteo simply makes them accessible via a clean API without the licensing restrictions of commercial weather services.

The live feeds above answer what is happening tonight. They cannot answer whether February is a better bet than October, which is the question most people actually have when booking a trip. That comes from a separate archive: every three-hourly Kp reading from 2010 to 2024, published by the GFZ German Research Centre for Geosciences in Potsdam.

The archive is processed once at build time, not fetched by your browser. For every location on the site, the build counts the nights when Kp reached that location's threshold and the sun was more than 18 degrees below the horizon at solar midnight - real darkness, not just night. That gives an average number of aurora nights per month for each place, which is what the "how often does the aurora appear" charts and the trip planners show. The result is stored in aurora-history.json and currently covers 358 locations.

Two limits are worth stating plainly. These counts are geomagnetic only - cloud cover is not in them, and cloud is the single most common reason a statistically good night produces no sighting. And they are historical averages across fifteen years spanning a solar minimum and a solar maximum, so any individual year can sit well above or below them.

The current dataset is built from measured GFZ readings throughout. Every entry records which of the two it is, and the wording on each chart follows that record rather than assuming.

Kp data: GFZ German Research Centre for Geosciences, CC BY 4.0