I’ve thought before that I’d like error ranges in weather graphs. Alternative predictions aren’t quite how I’d imagined it. Showing some detail/personality/confidence level for the alternative prediction lines might help. This might also be solved by time with the app to learn how surprising the weather turns out to be when predictions were divergent.
Cumulative precipitation could use another dimension to the data to show either soon-ness in the next 24 hours or how short of a time period is predicted to deliver most of the rainfall. That’s a hard data visualization problem. Maybe color shifting plus color saturation? It would add a lot to quick glances at the app.
The design and widgets are nice. Should scale well to iPadOS when they launch it. Rainbow notifications will be awesome if they can get them right. (Seems hard to get enough testing data on those.)
Disclaimers: I’m not serious about weather apps in general, and the obligatory “I loved Dark Sky back in the day.”