Differing Temperatures From API vs Web

[jkreski]

I am doing a fetch of most current temperature observation from: https://api.weather.gov/stations/KMBL/observations/latest and I am checking the accuracy by comparing it to what I see at: https://forecast.weather.gov/MapClick.php?CityName=Beulah&state=MI&site=APX&textField1=44.6296&textField2=-86.0955&e=0

I am seeing significant differences at times and am wondering why this is or if I am missing something in my approach. Does anyone else notice this? Thanks!

[brewin]

In my experience, the /observations/latest endpoint is usually delayed by an hour or more, which makes it wildly inaccurate much of the time. There are other issues with it too. #726

[scadergit]

Unlike mapclick that has limited stations, the API uses MADIS to provide observation data. MADIS can take up to 20m from the observation time to run quality checks. For a station that makes an observation on the hour, that can be 80m from the previous observation. That tends to be the extreme unless another issue is present.

[brewin]

I think that aligns with what I'm seeing. Typically the /observations/latest timestamp is 30 to 120 minutes old. For example, at 10:45 AM CDT, /KMCI/observations/latest timestamp was 8:53, almost 2 hours old. At 10:47, it was updated to 9:53. So, the new "latest" was already almost an hour old. My expectation is that /latest would always be updated within a few minutes of the observation time, like the MapClick API has done over the years. But I know MapClick has had its own issues.

[jkreski]

What time zone are the time stamps in? UTC, EST, CST, etc?

[scadergit]

I believe they are all UTC, as indicated by +00:00

[brewin]

They are UTC. I converted them to the local time in my comment. If you use the forecast endpoint, it has timestamps with UTC offsets.

[jkreski]

Is there a better link to use? Or use a different service's API?

…

On Sun, Jun 16, 2024, 8:02 AM brewin ***@***.***> wrote: In my experience, the /observations/latest endpoint is usually delayed by an hour or more, which makes it wildly inaccurate much of the time. There are other issues with it too. #726 <#726> — Reply to this email directly, view it on GitHub <#729 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AF4E5WKOURWC5YBI3JE3OK3ZHV5GRAVCNFSM6AAAAABJLWDGL2VHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM4TOOBWGQYDQ> . You are receiving this because you authored the thread.Message ID: ***@***.***>

[brewin]

There is the aviationweather.gov metar endpoint. You have to calculate a bounding box and parse the metar. I have not used it, but I'm considering it.

[scadergit]

Will note here as well that the beta weather.gov 2.0 site uses the API, so a lot of these issues are getting attention internally. Won't be fixed tomorrow, but it's recognized this isn't to standard.