I’ve used SSA analysis to extract the first mode from the northern hemisphere SST record available from http://www.metoffice.gov.uk/hadobs/hadsst3/data/HadSST. _ts.txt”


The plot above shows the dominate eigen mode plotted with the detrended SST data. The plot below show the mode plotted against a sine wave fit.


The best fit period is about 64 years and shows signs of slight phase modulation. At first I thought difference in frequency between this mode and the similar mode extracted from the Hadcrut3 surface temperature data pointed away from a celestial forcing. Then I decided to look at correlations between the two.

Here’s the auto-correlation of the Hadcrut3 climate signal reconstructed from the first eigenpair,

ACF of Hadcrut3 S[43,1-2]

ACF of Hadcrut3 S[43,1-2]

 It shows cyclostationarity with a period of ~57 years, similar to what was found by harmonic decomposition. Here’s the ACF of the same mode extracted from the SST northern hemisphere data.

ACF of SST NH S[55,1-2]

ACF of SST NH S[55,1-2]

Here’s both on the same plot

Comparison of SST and Hadcrut3 ACF

Comparison of SST and Hadcrut3 ACF.

This is pretty convincing evidence that the mode in both datasets have the same period (56.8 years).

The cross correlation shows global surface temperature 4.2 years lags the sea surface temperature by 4.2 years.

Cross correlation

Cross correlation

The cross spectrum is pretty amazing

Cross spectrum being dominant mode of Hadcrut3 and SST

Cross spectrum between the dominant mode of Hadcrut3 and SST

And then there is this: