I've been interested in weather for as long as I can remember, but often even more so in the weather models behind it.

They can behave quite differently from one season to another, have their own specific biases, and sometimes you can see models shift significantly in their view of the synoptic situation from one run to the next. That has always fascinated me.

You can also see trends developing. Everyone gets that feeling sometimes: "It's warming up again." Being able to actually see that change in the data is something I find particularly intriguing.

Not just the final forecast, but why a model arrives at that forecast, how it changes over time, and how it actually gets things right.

At some point, I wanted to do more than just follow those changes. I wanted to keep track of them and see how the models actually perform and especially, evolve over time.

What started as making GIFs from synoptic charts gradually grew into something bigger.

I started collecting model data to look for connections and patterns that are difficult to see from a single forecast. After all, it's easy to forget what all the models showed yesterday, let alone a few days ago.

Along the way, I started building tools to explore those changes more easily, from comparing individual model runs to looking back at how forecasts evolved over time.

Meteoproof is essentially the result of that curiosity.

Forecasts are compared with observed weather to measure things like accuracy, bias and stability over time.

The goal isn't to declare one model the winner. Different models have different strengths, and their performance can change depending on the situation, season and forecast range.

Instead, Meteoproof tries to make those differences visible.

I'm constantly testing new models, ideas and ways of looking at forecast data. Some experiments work. Others don't. Some eventually become part of the site, while others stay experiments.

That's part of what makes this project interesting to me.

I'm building and maintaining Meteoproof so I can better understand how weather models gradually change their forecasts, and share those changes so others can follow them too.