Apple Weather's forecasts are so bad my meteorologist made a post about it
Apple Weather's forecasts are so bad my meteorologist made a post about it
Good morning. Since the forecast is pretty much unchanged—expect very hot and sunny weather for the foreseeable future—I want to briefly address a question I’ve received several times. And th…
Apple Weather forecasts for my area have shown a 30-50% chance of precipitation for eternity, even when there isn't a cloud in sight. My meteorologist has told his readers to ignore Apple's forecasts.
This started around the time the Apple Weather outages a few months ago and haven't improved. YSK that you can report inaccurate forecasts from within the app, but it hasn't improved for me.
I'm currently using Foreca on iPhone and the free version of Carrot on Apple Watch.
You're viewing a single thread.
How can Apple got this wrong? Pretty much all meteorological data are sourced from the same set of weather forecasting data produced from running weather simulations on weather agencies supercomputers using satellites and ground observation data. The forecasting models will pretty much produce almost similar results when fed with the same data, which is why weather reports are usually pretty similar even though they're produced by different weather agencies.
Does Apple actually run their own weather simulations here? Probably for their hyper local next-hour weather prediction feature? Or could it be that Apple is using some "AI" for their weather forecasting and the AI conjured some number out of its ass instead of running full physics simulation?
That's not true - I'm seeing 6 different weather sources in Carrot, and while they probably share raw data, they use different models. Dark Sky (which apple bought and re-branded) wasn't just an app, it was its own hyper local (and accurate) weather model. Apple broke it.
It started to go to shit right around the end of 2020 for me. Dark Sky went to complete dog shit in my part of the Bay Area. It became comically inaccurate.
The different models definitely do not produce the same output and some weather conditions are extremely difficult to predict with any accuracy (especially rain). Even the "current" conditions are often very different depending on the model since they rarely use actual observations... and even if they did the next city block over could be totally different).
Which one is the most accurate depends on your location (in particular, your latitude, proximity to the ocean, nearby mountains, etc).
Even little things like standing near (not even under) a tree can drastically alter your air temperature - because trees raise the humidity of the air around them and a slight humidity change has a drastic effect on temperature.
Apple most likely uses multiple models.
That's why when they take air temperature samples they take multiple samples from multiple locations and they explicitly don't put them under trees, behind shade, unusually high up.
If you're standing under a tree resulting in a lower air temperature that won't affect the weather prediction because the temperature isn't taken from your phone.
Per apple it looks like for the US they use the weather channel.
Apple sources it us data from the weather channel. Which is a pretty shitty source. Would be nice if they had options for the other sources.