Temporal resolution of GFS and GEM models

I am located in the US. Both the GFS and GEM models are shown with a temporal resolution of 3 hrs. I believe both models are calculated with 1 hour resolution. The GFS should be freely available to you at that resolution and I think the GEM is as well. Several other free apps use this and I don't think they are interpolating.

Is there a reason you don't have the higher temporal resolution for these? I'm an astronomer and the GEM seems to have the best cloud predictions. I'd love to have the 1 hr resolution there.

Thanks,
Bill

btschumy
We are not considering the 1-hour step from these models. They are more suitable for long-term forecasts, where the 1-hour step is not important. For short-term forecasts, I rather recommend regional models (like HRRR or NAM) or ICON, which do include a 1-hour step.

That's disappointing. As I mentioned, the GEM in particular is extremely useful for short term cloud forecasts. I also think you are the only app that uses 3 hour intervals for GFS.

Please give this some more thought.

btschumy
In general, we haven't had good experiences with the GEM model. For which region and what forecast range does it perform better?

Thanks for keeping the conversation going.

The GEM model seems to do well in the US, especially for the cloud forecast. Astronomers are always interested in whether there will be clouds for the next couple of evenings. GEM is widely acknowledged in the Astro community to be one of the best for this.

I haven’t paid a lot of attention to how it does for other aspects of the forecast but I haven’t seen anything particularly wrong.

I do have other apps/sites I use for GEM, but it would be nice to be able to stay in Ventusky.

btschumy
Let's try a forecast for tomorrow at 15 UTC. GEM shows completely clear skies, HRRR predicts full overcast (with high clouds - the sun might shine through), and NAM is somewhere in between. So, in 24 hours we'll see who nailed it best

As I hope you know, this won’t prove anything (even if GEM is more correct tomorrow). You need to access the accuracy over a long period of time for it to have any validity. Nevertheless, we can see what happens.

btschumy
Sure, it is just for interest. But in our long-term tests, GEM has always come out poorly.

And here is the reality from the satellite. I think NAM did the best. It predicted well that it would be completely overcast in the east and scatter mostly high clouds elsewhere. HRRR overdid it with the high cloud cover – there is some, but not as much as it forecast. GEM was off – it expected almost no cloud cover in the entire area, yet it's completely overcast in the east and not entirely clear in the west either.

Agree that NAM probably one this one. I was out looking at the sky at the appropriate time and GEM wasn’t far behind. Both HRRR and ICON predicted much more cloud cover than was there.

I’m just curious what the reasoning is for not providing higher temporal resolution when it is available? Does it strain the servers more to provide this? More bandwidth needed? Perhaps you could do what Windy does and have the hourly forecasts only for Premium subscribers.

In either case, I still think Ventusky does an excellent job as a weather app.

btschumy
It's about processing time — we prefer to allocate computing resources to speed up the download of regional models first, and only then the global ones. For regional models, timely updates are crucial, whereas for the global model, not as much.