GraphCast: Learning skillful medium-range global weather forecasting

[@chiponline123]

Is there a weather app that is using GraphCast technology that i can download? Thx

[@superkaran20]

Thanks for the talk, it was very helpful, but i didn't get what is opreational analysis ground truth at 31:50 time stamp.

[@nkhgs3392]

Is there any guidance on how to interpret the output it provides? It would be a cool project to build a site which uses the data.

[@simonpedley9729]

does anyone know what software he used to create the spilt screen with presentation on one side and video on the other?

[@anonymous-q2b5s]

I have a question about the surface temperature bias visualization. You can see some areas are "blinking" because the bias seems to have significant differences in magnitude depending on the point in time. Does this again have to do with the assimilation windows? Why is specifically Angola, India and the australian outback blinking so much? It seems to correlate with magnitude as one can see north america and europe starting to blink as magnitude increases as well but antarctica for example doesn't blink at all despite having the highest bias.

[@simonpedley9729]

if i understand correctly, for most of your results you are comparing your in-sample predictions with the hres out-of-sample predictions. what i mean is, you trained your model once on 40 years of era, and then tested it using the same 40 years of era. if thats right, shouldnt you be testing in a cross-validation or out-of-sample framework? ie training on the first 35 years, and testing on the last 5 years? or did i misunderstand something. you very briefly seemed to address this at 44m, but mostly as an after-thought

[@ajitabhkumar5449]

Great talk. Thanks!

[@simonpedley9729]

the comments about ibtracs seemed slightly wrong to me…ibtracs includes various versions, but no gold standard. its a huge pain in the neck