Update to #141. Forcing with a 1st-order lag and an opposing feedback from one year prior. This serves to smooth the ENSO model fit (in contrast to the blocky one-year window integration) and to reinforce the biennial modulation in the lunar tidal forcing.

![enso](http://imagizer.imageshack.us/a/img923/8048/jKp07B.png)

Like the QBO model, this is a remarkable result showing the ability of a very short interval of the time series to recover the rest of the behavior. And like the QBO model, there are so few degrees of freedom in the fitting parameters that over-fitting is not an issue. No remaining kludges in the model and the only ansatz is the assumption of a metastable biennial modulation. Yet, this is by all accounts a result of a year-to-year compensating feedback leading to a period doubling (and which Rayleigh discovered in his [maintained vibrations paper](http://www.tandfonline.com/doi/abs/10.1080/14786448308627342?journalCode=tphm16) )

http://contextearth.com/2017/04/18/shortest-training-fit-for-enso/