I'm working a model of one of the fastest ocean oscillations, the North Atlantic Oscillation (NAO). The NAO can show up to 2 strong cycles per year in comparison to ENSO, which has El Nino peaks every 2 to 7 years
It appears that the model for NAO is not as sensitive to annual or semi-annual impulses like all the other behaviors (ENSO, QBO, PDO, AMO, IOP) require for modeling. So instead of an impulse, it appears to more directly correspond to monthly tidal variations.
This is promising. The time-series is dense enough that cross-validation may work well here.
This is a cross-validation, trying to overfit an interval from 1980 to 2000 and observing how the out-of-band intervals respond -- quite stable.
The non-impulsed lunisolar tidal forcing is identical for ENSO and NAO, which removes many degrees of freedom from the fitting process.
The LTE modulation for NAO is quite strong, approximately that of used in PDO in the prior comment #309. Perhaps this is expected as both NAO and PDO are northern/higher latitude behaviors