The linear feedback response with a time constant of 3.5 years (commonly found in climate models) is calculated for a trivial three cosine “forcing” signal.
Cosines of period 9, 22 and 60 years were aligned to 2005. All three were given equal weighting , no attempt being made to optimise or regress this to climate data.
The choice is not random but a rough guess at what is ‘climate-like’ variation.
The correlation with AMO (a detrended version of North Atlantic SST) is quite remarkable. Key features like the steep drops around 1905, 1945 and 1960 and the post 1974 warming are closely matched.
It is notable that it fully reproduces the early 1990’s dip without the need for volcanic forcing.
The shift introduced by the feedback response with tau=3.5 years seems to correctly align many of the main features.
This points out the futility of attempting to account for the post 2000 ‘plateau’ without any representation of the 60 year variability in the model.