> "You seem also to speak to the concern your model might not be predictive "by chance", yet still be physically correct."
That's not what I said at all. I stated that even if my model made a correct prediction of the next El Nino, **an informed statistician** will claim that that can happen just by chance, because even a random guess can be correct a fraction of the time. So then I will be asked to make another prediction, and then another, ... until somebody deems that an adequate significance level is achieved.
If, as you say that *"There is a lot human life and assets riding on the best physics-based predictive models"* then we should not wait 20 years to get several correct El Nino predictions under our belt, and instead try to use all the best cross-validation tools available that will work on the historical data. Using your own words, that sounds like a much more socially responsible approach to take, right?