Weather Sources

Naola queries multiple weather services at the same time and combines their forecasts into a single consensus prediction. Here's how it works and which sources are used.

Current Sources

The free tier queries 3 sources. Naola Advanced adds 3 more for a tighter consensus — 6 in total.

ProviderForecast ModelTierWebsite
Open-MeteoECMWF (European model)Freeopen-meteo.com
OpenWeatherGFS (US model)Freeopenweathermap.org
Apple WeatherBlended (Apple)FreeApple Weather data sources
WeatherAPI.comBlendedAdvancedweatherapi.com
Visual Crossing WeatherBlendedAdvancedvisualcrossing.com
GoogleBlended (Google Maps Platform)Advanceddevelopers.google.com/.../weather

Home-screen widgets show the sources your plan includes, except Google, which appears only in the app.

How the Consensus Works

  1. All sources are queried at once -- If one source is slow or down, the others still work. You always get a result.
  2. Forecasts are normalized -- All temperatures are converted to Celsius and all wind speeds to m/s, so they can be compared fairly.
  3. A weighted median is computed -- Sources powered by higher-quality forecast models (like the European ECMWF model) count more than blended models. This produces a more accurate consensus.
  4. Outliers are removed -- If one source gives a wildly different reading, it's excluded from the final result to avoid skewing the consensus.

Confidence Colors

The colored badge on the weather card tells you how much the sources agree:

ColorWhat It Means
GreenHigh confidence -- sources agree within 1.5 degrees C
YellowModerate confidence -- some disagreement (within 3 degrees C)
RedLow confidence -- sources disagree by more than 3 degrees C

Forecasts

The consensus approach also applies to forecasts:

How Source Accuracy Is Measured

The Source Accuracy screen ranks sources by how close their forecasts came to what was observed. Here is the full method, so anyone can repeat it:

  1. Snapshot -- Every day at 08:00 UTC, for up to 50 recently viewed locations (rounded to about 1 km), Naola records each source's forecast high and low for the next day. Apple Weather, Google and WeatherAPI.com are never recorded, because their terms do not allow it, so they are never ranked.
  2. Observation -- Two days after that day, its observed high and low for the location's local calendar day are taken from Open-Meteo's historical weather data.
  3. Score -- For each source and day, the error is the forecast minus the observation, for the high and for the low. The day's score is the average of the two errors, ignoring their sign, in °C.
  4. Ranking -- Sources are ranked by their average score over the chosen period (7, 14 or 30 days), lowest first.

Scoring is paused while Naola's Open-Meteo plan has no historical data, so the screen shows no new results.

Attribution and Licences

Copernicus citations