You should still use the original title, some subset of it or representative language from the article. Rewriting them makes sense if they are misleading or clickbait but this one is neither - 'We mapped weather forecast accuracy across the U.S.'.

I agree it seems like flimsy justification. But it is also likely harder to assess and communicate. Temperature they get a point prediction for the high and you can easily calculate the mean absolute error.

For precipitation you will be getting percent chance often with an interval, 10% chance of 0.1-0.25 inches with higher likely in thunderstorms. Also precipitation patterns tend to be much more irregular within small spatial extents. You can asses things like calibration and perhaps take a mean value for there intervals to get point errors. But all of this will make it harder to communicate actual performance.

Microsoft did weather forecasts?

There was another I tried to dig up. It listed how one day forecast was better at x, during the day at y, 2-3 day at z, etc.

Windy.com allows you to compare models. Apparently weather forecasts accuracy are a popular topic and I didn’t bookmark it.

I think you are looking for ForecastAdvisor. However depending on your geo, different weather providers will be more accurate than others.

Specifically, for the contiguous United States how many days into the future the Weather Service’s forecast of the daily high temperature is accurate within 3 degrees Fahrenheit.

Something I learned only recently: Weather forecasts don't actually directly use sensor data. Instead, a physically consistent model is first fitted to all the available sensor data, and then the forecast is made based on the values that model produces. Doing it this way has the benefit that physically implausible sensor readings are given less importance, and the fact that this model can be sampled in regular intervals, whereas the sensors are all over the place (and often moving, e.g. in aircraft, which contribute crucial data).

Of course, higher density of sensors would lead to a better fit of the model to the real world, but there would still be no guarantee that the model would reflect the measured values exactly. I found that pretty interesting.

(And it's kind of funny to think about our own consciousness in this way, which seems to work somewhat similarly: we don't experience the actual 'sensor values', but instead we experience the output of a model our brain fits to those inputs.)