Show HN: Weather app to show actual performance of weather models | Hacker News Reader

Show HN: Weather app to show actual performance of weather models

apps.apple.com

5 pointsnickoteeno5 comments

Got tired of forecasts being wrong and having no way to check. So I built an app that overlays what models predicted days ago vs. what actually happened, so you can literally see the accuracy on a timeline.

You can compare all open-meteo models (such as IFS, AIFS, AIGFS...) side by side. Choose how many days back you want to check (it uses open-meteo's Previous Model Runs API) and the app shows the previous prediction as a "ghost line" against observed data. There's also a scoring system that calculates an accuracy by temp, wind, rain and sun hours.

Features: - Compares previous predictions vs. observed data - Built-in scoring system that calculates accuracy by temp, wind, rain, and sun hours - Separate storm accuracy page to choose the best model for predicting storms. - Free and no ads!

Would love to hear what fellow weather nerds think. Any feedback is welcome!

If you're interested: iOS: https://apps.apple.com/nl/app/verisky/id6759251875 Android: https://play.google.com/store/apps/details?id=com.verisky.ap...

Nice! Even for a non weather nerd like me it makes for a helpful, and privacy friendly, forecast app, especially using the ensemble.

The scoring on rain seems a bit binary though. In my area, one got 100%, the rest all scored 0.

Yeah when there's been very little rain, it's just hit or miss ;) Does it still look like that when you increase the Score tab lookback window?

I will add some minimum amount of rain though, otherwise the score feels weird indeed

I can't see you are retrieving any actual observations (open-meteo only provides model data). So you are really measuring model consistency or confidence since you aren't comparing to actual observations.

I’m getting a “Forecast API error: 429 - Too many concurrent requests” error.

Reply on news.ycombinator.com