Jim, All I have to do is find a list of identified ElNino and LaNina events prior to 1950. Asking a question, I am not sure if anyone has done this?

I also have an additional insight into what goes into pattern matching and whether a model is a good fit or not.

This is a stretched analogy but consider the case of Robert Durst that is in the news. There is an infamous comparison of his authenticated hand-writing sample with the hand-writing of someone writing a threatening note.


There are various elements to comparing the two samples. One can look at the overall style (block capitals), the stroke, and the spelling. At a finer detail, one can look at the L's and notice that the base differs, but not entirely.

So you ask yourself is there any doubt that the two samples are written by the same person?

I would say no, that there is no doubt that the same author wrote both..

But with ENSO, the comparison is not so cut-and-dried. For one, what I am doing isn't exactly performing a pattern match, but coming up with the best representative model of the underlying physics. For example, it is entirely possible that I am missing some important factor. This could mean that the model can be easily made better, but at the risk of over-fitting. Further, some inconsistency between the model and data comparison may be a dead giveway that the model is wrong.

But even with the Durst letter, consider that what would happen if the comparison was with a sample where the word Beverly was spelled correctly and one where it wasn't. Would that have invalidated the match? Not necessarily, as the model would just need to be changed to that the author was an inconsistent speller.

That's kind of similar to the noise in the data. We actually don't know the true signal that we are comparing the model to. There is enough noise in the underlying data that deviations may be simply spurious.

These are issues that I am always chewing on.