Hello Bert,
you mentioned that "I think that in many cases these seiches could indeed be routinely predicted IF you have an accurate high-resolution wind model, and IF you have a sufficiently detailed hydrodynamic model (possibly including waves)."
In this it would make sense to use a coupled model such WRF-ROMS, or it is improper from a physical and technological standpoint to try such avenue to improve the prediction of these seiches.
thanks
All the best
Antonio