Where does Weatherlow get their forecast data from? It’s 10 degrees off today from their forecast. The actual matches up exactly with Apple and local news.
Other people noticed the bad forecasts as well. In a reaction weatherflow said that the use of your own local data to improve the forecast was not yet implemented but it soon will and they were confident that it would work. Personally I would be surprised if it would improve that much. Start with an already good prediction and make that one better using your local data, is an approach that I could understand.
Thanks for your feedback. The forecast in the app is produced by WeatherFlow, originating from a variety of third-party sources, including global and regional numerical models from US and European weather agencies, as well as WeatherFlow’s in house regional numerical models.
A 10-degree temperature discrepancy seems large and may be evidence of a bug or the product of growth pains as we implement and tune our forecast systems. With our app and scalable forecast system just launched, a few bugs are inevitable.
This could be an error or temporary bug in our forecast presentation. If you can provide your station ID, a screenshot of what you’re seeing, as well as the date & time of your observation, that will help us investigate the problem.
Hello, this is for ST-00002253
Attached are a few screen grabs from today. It was less off today, but yesterday was a full 10 degrees different. The rest of the week’s forecast is also much lower than any other service forecast. The actual readings from the unit seem to be exactly in line with what the temperature is reported from other nearby stations, Apple Weather and Weather.com. So the unit is good, just not the forecast.
Also, the option to reply to my email notification from you guys returns an error: We’re sorry, but your email message to [“email@example.com”] (titled Re: [WeatherFlow Smart Weather] Email issue – Posting error) didn’t work.
None of the destination email addresses are recognized, or the Message-ID header in the email has been modified. Please make sure that you are sending to the correct email address provided by staff.
Here’s another. 9:43am it was 76 which was the forecasted high for the day. Obviously it was looking like a bad start to the day when it was already wrong.
Thanks very much @allenmaris we are investigating and will have improvements in place soon.
Just want to chime in that I have the same problem. Newly installed Tempest, the actual weather readings seem quite consistent with other stations nearby (from Wunderground), but the forecast is pretty consistently 10 degrees off. Tomorrow is forecast in Tempest to be 85F, Wunderground forecasts 94F. Based on the past week, my money is on the Wunderground forecast…
Let me know if you want me to start a new thread with details of my forecast issues…
Hi Chris, here is fine. Thank you for your report. The temperature discrepancies will be addressed.
I’ve got the same issues around temperature forecasts. I found myself reinstalling other weather apps because the forecast makes no sense and are nowhere near reality. Station ID 21253.
Thanks, Joe. If you can share specific observations where your forecast showed something that was clearly wrong, please do!
Sorry for the delay, but documenting issues are a pain, I have to screenshot in advance, and then compare after the fact. Here is an example.
Thanks. One more piece of info: Is that for yesterday, 18 Nov?
That was for Tuesday the 17th Nov.
Thanks, Joe. We’ve had a look at your data and confirmed the discrepancy. As I’m sure you know, you see a lot of temperature inversions in your neck of the woods during winter and that can make it tricky to forecast the temperature at the surface. We compared our forecast with several other sources that also missed the inversion on this particular day. Some probably got it right, but miss it on other days. Hopefully we’re getting it right more often than they are!
Note, because you own a Tempest your forecast will get better over time as the data from your Tempest trains our machine learning algorithms on the local patterns. It can take several months before you see any significant improvement, but we guarantee it will happen.