Excellent analysis! Both are very good fits over the time period in question. This spotlights the problem of Observability of the system in question. It is why I have always discounted the claims of successful hindcasting as support for a given model. The possible configurations of the system which could produce the observed behavior are not unique. It is unfortunate that reliable data is available for such a short time interval. Perhaps you could try fitting your models to half the data, and see how well it matches for the other half. It wouldn’t prove anything, but might give a feel for sensitivity of the results. I have, however, observed the ~60-65 year cycle in temperature reconstructions extending back over 1000 years. While that may also be happenstance, I suspect that the solution which accentuates this contribution more heavily will be closer to the truth. Reply
Very interesting — I hope you do more. First, I’ve read that the dependence of temperature (sensitivity) on CO2 is proportionate to the log of CO2 so I hope you try that correlation. Second, I’ve wondered about Hansen’s claim about the stability of temperature anomaly correlation with distance on which he bases his averaging scheme for sensors within 1200 km of the center of a subbox. I hope you look at that issue also. Reply