Data Sources & Update Cadence

Lumina stitches together data from half a dozen different spacecraft, ground stations, and models to build a real-time picture of what's happening between the Sun and your backyard. Here's where everything comes from and how often it updates.

Solar wind & IMF — SOLAR1, DSCOVR & ACE

The core data feeding Lumina's probability model comes from satellites parked at Lagrange Point 1 (L1) , about 1.5 million kilometres upstream of Earth — directly between us and the Sun. This is the perfect spot to measure the solar wind before it hits our magnetosphere.

SOLAR1 (Space Weather Observations at L1 to Advance Readiness – 1) — the primary source. This is NOAA's newest satellite, launched in 2025. Provides solar wind speed, density, temperature, and IMF vector data with high accuracy and reliability.
DSCOVR (Deep Space Climate Observatory) — the primary backup. Provides accurate solar wind speed, density, temperature, and IMF data.
ACE (Advanced Composition Explorer) — the legacy. Launched in 1997 and operating far beyond its original livespan, ACE still provides accurate measurements when both SOLAR1 and DSCOVR are unavailable.

Update cadence: Lumina polls the NOAA real-time solar wind feed every 1 minute . The L1-to-Earth propagation delay (typically 30–90 minutes depending on solar wind speed) is estimated per update so Lumina knows approximately when each measurement will actually reach Earth.

Hemispheric power — NOAA SWPC

NOAA's Space Weather Prediction Centre provides the global metrics that tell Lumina how energised the magnetosphere-thermosphere system is right now:

Hemispheric power (HP) — estimated total auroral particle energy deposition in gigawatts, separately for the northern and southern hemispheres - only the southern hemisphere value is displayed & used by Lumina. Derived from NOAA's OVATION model. Updated every ~5 minutes.
HP forecast — a 90-minute ahead prediction of hemispheric power, updated every few minutes. This is what powers the "HP forecast peaks near X GW" messages in Plan mode.

CME data — NASA DONKI

Coronal mass ejection information comes from NASA's DONKI database, which catalogues CMEs detected by solar observatories (SOHO's LASCO coronagraph, STEREO, and SDO):

DONKI CME API — pulled every 30 minutes with a 5-day look-back window. Includes analyst-measured CME parameters and WSA-ENLIL model runs predicting arrival time and intensity at Earth.
Linked solar flares — when a CME is associated with a flare, Lumina pulls the flare classification (C, M, or X-class) and displays it alongside the CME.

Magnetometers — GOES & Geoscience Australia

Substorm detection relies on two independent magnetometer networks:

GOES satellites (geosynchronous orbit, ~36,000 km) — carry boom-mounted magnetometers measuring the magnetic field at geosynchronous altitude. The Hp component (parallel to Earth's dipole) is the key substorm signature. Updated every ~1 minute.
Geoscience Australia ground stations — three Antarctic/sub-Antarctic observatories (Casey, Mawson, Macquarie Island) provide H-component magnetometer readings every minute. These are the most directly relevant stations for AU/NZ aurora watchers.

Cloud cover — Open-Meteo

Cloud forecasts come from Open-Meteo , a free weather API that provides hourly cloud cover percentages. Lumina fetches the forecast for your saved locations and uses it to determine whether there's a clearing expected, which direction has the clearest sky, and whether it's worth travelling to a different location.

Update cadence: every 15 minutes on the frontend, cached for 30 minutes per location.

How Lumina stores data

Live conditions are cached in Cloudflare Workers KV for fast access (sub-millisecond reads). Historical snapshots go into a D1 database (SQLite) for trend analysis, momentum detection, and the sparkline history you see on the dashboard.

Data health & degradation

Space weather data isn't always pristine. Satellites go into safe mode. Ground stations have outages. Lumina tracks data health for every source and surfaces degradation in the confidence score:

Stale data — if a source hasn't updated within its expected cadence, confidence drops.
Quality flags — SOLAR1, DSCOVR and ACE data include quality indicators; Lumina deprioritises low-quality readings.
Missing sources — if one magnetometer station is offline, substorm detection still works from the remaining stations, but confidence is lower.

When you see a data confidence warning on the dashboard, it means one or more sources are degraded and the probability estimate should be treated with a bit more caution than usual.