You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For years 2022 and beyond, you could run GraphCast_operational, which was trained up to 2021. For years 2018-2021 you could run GraphCast, which was trained up to 2017.
However for 2017 However, all of the models we provide train on data up to 2017, so unfortunately it is not possible to evaluate them on those years, as they may be overfitting to the training data, and it would not be a fair evaluation.
I want to conduct a study comparing the results of Graphcast with Copernicus from the year 2008-2023.
Is it feasible to conduct such a study?
The text was updated successfully, but these errors were encountered: