Weather data discrepancies for SFO between two sources, despite being part of the same system

[janluong]

I’m currently using api.weather.gov to pull weather data and noticed some discrepancies for San Francisco International Airport (SFO) compared to forecast.weather.gov. api.weather.gov appears to have less data overall, but it provides historical data beyond three days, whereas forecast.weather.gov seems more complete, but doesn’t go as far back.

I was hoping to rely on api.weather.gov for its extended historical data, but the missing information compared to forecast.weather.gov really concerns me about its accuracy and completeness.

Does anyone know why these two sources, despite being part of the same system, contain different data? Any insight into how they are sourced or updated would be greatly appreciated!

Here is a comparison that I did for SFO at 3/12/2025 @ 13:56 (20:56 UTC – api.weather.gov):

[dwhitemv25]

The issues you've pointed out are documented on the API webpage on Updates tab under "Known Issues."

• The observations endpoints always show missing (null) wind gust and may show null values for other weather properties (temperature, wind, precipitation, etc.) at stations when the MADIS source has data.
• Precipitation values in the observations endpoints are rounded down to the nearest centimeter (less than 0.4" may be improperly rounded down to 0).

Wind reporting as null is a bug with decoding gust indications.

Honestly, if you want better METAR decoding, use aviationweather.gov's API. In addition, aviationweather posts METARs a LOT faster than api.weather.gov does. Hopefully this API will get better observations support in the future.