Discussion on cross-validation of extremes

Dear WG4-members,

at the Trieste meeting we have been thinking about how to do cross-validation of extremes. This is actually a very important issue, and seems not so easy to handle.

One idea involved stratified sampling (e.g. calibrate on wet years, validate on dry years), but this would involve a "handpicking" of specific years, which might not be appropriate.

Another option might be "leave-one-out cross-validation"

We should also keep in mind that with the validation of extremes there are generally few observations to validate on.

These are just a few thoughts from the Trieste meeting- it would be great if we can discuss this issue in this forum in more detail- looking forward to your ideas and suggestions.