> "is appropriate for dealing with possible regime changes eg. 1935; 1880-1920,1920-1960,1960-2000; 1975 etc.?"

Possible, look at the residual in #44 . Corresponding to the year after the Pacific climate shift of 1976/1977, there is a significant transient cooling. These again are the features that are not captured by the strong periodic forcings of QBO etc, and so may benefit from masking out.


I wrote earlier :
>"I will post later, but the correlation coefficient is a bit above 0.4 for the model to IOD. It doesn’t look bad as far as it goes."

this is what it looks like using the model for SOI on IOD. I adjusted the initial conditions so they would align as much as possible, and I tweaked the characteristic frequency slightly, from 4.25 years to 4.2 years.

![iod model](http://imageshack.com/a/img674/2362/jyuUL7.gif)

The IOD isn't as striking a dipole as SOI as it has somewhat weak and non-emphatic excursions. It's possible that if I adjust the phases of the forcing functions such as QBO, the fit could improve substantially. This could provide a better way to look at the phase relationship between the SOI and IOD dipoles.