How to read the NOAA aurora forecast

The map on SWPC's aurora page is drawn from the OVATION Prime model. OVATION takes the solar wind speed and the interplanetary magnetic field measured at the L1 point, 1.6 million km upstream of Earth, and computes how much electron and proton energy should rain into the upper atmosphere. The forecast lead is the time the solar wind takes to cross from L1 to Earth, which SWPC puts at 30 to 90 minutes.

SWPC switched operations to the current version, which it calls OVATION 2020, in 2020. The same product page notes that the original model, built on DMSP satellite data alone, was only reliable to Kp 7.

Each cell carries a number from 0 to 100. SWPC describes it as an estimate of aurora viewing probability, derived by assuming a linear relationship to modeled intensity. So a cell at 40 is not a promise that 4 nights in 10 will show aurora. It says the model puts more auroral energy over that cell than over a cell at 10.

Glow reads the raw JSON file behind the map, where each cell is a longitude, a latitude and a value. It rounds values to whole percent and drops malformed cells. The oval map draws every cell at 1 percent or more, and the southern edge tool finds the lowest latitude at 10, 30 or 50 percent for each longitude.

The grid has 1 degree spacing, with longitudes from 0 to 359 east. To find a town's cell, Glow rounds its latitude to the nearest whole degree and converts its longitude to that 0 to 359 axis. Minneapolis at 45.0 north reads latitude 45. Its longitude of 93.3 west becomes 267.

A 1 degree cell is about 111 km north to south, and narrower east to west at high latitude. The value covers the whole cell, so a town near a cell border shares its neighbor's weather in space as much as its own. The chance tool does this lookup for any coordinates.

Check the two timestamps in the file. The observation time is when the solar wind data were measured. The forecast time is when the model says that air will arrive. When solar wind data are missing or contaminated, SWPC drives OVATION from the current Kp instead, and then there is no forecast lead time. Glow labels a file as a Kp-based fallback when the forecast time sits less than 20 minutes after the observation time.

Check the age as well. Glow rejects a file whose observation or forecast time is more than 2 hours old, and shows no value rather than an old oval. A stale map on another site can look like a quiet night when it is simply out of date.

Then check the sky. SWPC says the aurora is not visible during daylight, and the model does not know about your cloud. A high cell under full overcast, or in twilight, shows you nothing. Glow puts dark hours and NWS cloud beside the cell for that reason and never merges them into one score.

Finally, look north of your cell, not only at it. SWPC notes that bright aurora can be seen from as far as 1000 km away. The distance tool measures how far you are from the nearest cell at 10, 30 and 50 percent.