Another niche weather tool, Dark Sky, has always been a staple for me due to its hyperlocal rain predictions.
Small anecdote: there's a unique satisfaction in creating something of your own. A few years back, I started a home hobby weather project using an Arduino to collect microclimate data right from my backyard. It quickly became an expansive endeavor. I integrated it with a backend running on Kubernetes, had a React Native interface for mobile updates, and to maximize performance, parts of data processing were done using Rust and OCaml. I was exploring the trade-offs of the CAP theorem in real-time. You'd be surprised at how often "weather inconsistencies" were just my nodes not agreeing on state. It was less about weather forecasting accuracy and more about the adventure in distributed systems.
It ended up getting pretty expensive so scrapped it but I learned a lot from that project and remember it fondly