Applying the singular spectrum analysis (SSA) R package to the global temperature record is potentially insightful. SSA is a type of principal components analysis for time series data, recovering orthogonal components, EOF’s, of different period. Just the ticket for decomposing climate data into potential sources. Although, it must be remembered that temperature series have high enough autocorrelation to produce spurious results in any such method — its not ‘reliable’ in a strict sense.
Above are the results of the first seven EOF’s from the HadCRU3gl monthly global temperature dataset, and immediately a number of possible attributions to the EOF’s pop out. Not on the y axis is the anomaly degrees associated with that EOF.
1. A broad peak – solar forcing? – 0.1 C
2. An exponential increase – ? – 0.6 C
3. A 50+ year period fluctuation – AMO? – 0.06
4. A ~20 year fluctuation – PDO? – 0.1
5. A ~10 year fluctuation – ENSO? – 0.06
These attributions to EOFs roughly match the profiles of the variables, and the approximate intensities in the literature. But what of Series 2 the exponential?
2.1 Increase in CO2?
2.2 Or artifact of adjustments?
A neat summarizing of the climate dilemma by SSA, don’t you think?
You might want to read How the US Temperature Record is Adjusted by Jennifer Marohasy.