True, but fails to explain why the longer - term mean about which those cycles fluctuate is trending up other than
curve - fitting an «approximation by three sinusoids of periods 1000 years, 210 years and 60 years,» ANSWER: The
curve fitting exercise is labeled as
such «heuristic»; the lengths of the cycles are from other observations, some displayed on figures 5 - B & C; only the amplitudes and
phase of the 215 and 60 years sinusoids are subject to optimization; Singular Spectrum Analysis has been applied by Diego Macias et al (note 18) to the HadCRUT series with equivalent results, and among many others by Liu Yu et al..