This is part of my critical review of the palaeoenvironmental evidence for the influence of solar activity on climate.
Patterson et al (2013) have a new paper that reports evidence of the 11-year sunspot cycle on diatoms and sediment characteristics in a freeze-core spanning the second half of the 20th Century from a British Columbian fjord.
The fjord is anoxic at depth, preserving varved sediments and the fragile, weakly silicified diatom Skeletonema costatum. Skeletonema costatum is a distinctive planktonic marine diatom, with long spines holding the cells together into long chains.
The chronology for the core is based on the varve counts supported by 210Pb dates and the 1963 137Cs spike from atmospheric bomb tests. These three dating methods agree beautifully – the chronology has a much lower uncertainty (one or two years at most) than most cores, so it is not absurd to look for 11-year cycles in the data as it has been in some of the other studies I have examined.
As this region is affected by both El Niño and the Pacific Decadal Oscillation, it does not seem to be an obvious place to look for a solar signal, but Patterson et al (2013) are supported in their quest by a couple of papers reporting a relationship between sunspots and Pacific climate. Christoforou & Hameed (1997) look at how the location of Pacific “centres of action” (equivalent to the Atlantic Icelandic low and the Azores high) move in relation to sunspot number. Curiously, this paper does not simply test the correlation between sunspot number and location, but only uses the data from the years of with solar maxima or minima, throwing out most of the data. Looking at their graphs, I wonder how robust their analysis is, for example 1957 (solar maxima) had the Hawaiian high far north and west, the following year, with very similar sunspot number, it was far south and east, but this is ignored by Christoforou & Hameed’s (1997) analysis. The second supporting paper, Hameed and Lee (2005) concerns stratospheric-tropospheric interactions; on a quick skim, I didn’t understand what was going on – mea maxima cupla.
Patterson et al (2013) use redfit to test for periodicity in their data, and anova to test whether there are differences between the years with sunspot numbers greater or less than 82.5. I cannot help wondering whether a non-parametric test would be more appropriate given the non-normality of the data.
Redfit finds a nice 9-12 year peak in the spectrum of S. costatum abundance. If Patterson et al are only looking for this one frequency, then we don’t need to worry about the Redfit multiple testing problem, but we probably ought to worry about the non-normal distribution of the data. The anova for high vs low sunspot years is significant (at p=0.05) for varve thickness and for diatom abundance when the data are lagged by one year (no correction for multiple testing), but not for temperature, precipitation or the Aleutian low index.
The issues with the statistics in Patterson et al (2013) may be no more than niggles. The physical link between sunspot number and S. costatum abundance is unclear as it somehow bypasses all the climate data Patterson et al (2013) show. But even if I ignore these issues, I do not find the sun-Skeletonema relationship convincing.
- If noisy proxies like Skeletonema abundance or varve thickness have strong relationships with solar variability, then relationships between instrumental climate records and solar forcing should be strong, obvious and unambiguous, but they are not.
- If authors who find an 11-year cycle publish claims of a sun-climate relationship, but other authors (or even the same) analysing other data find no such cycle and move on to other analyses, there is a potential for a huge publication bias. As any single study has a 1 in 20 chance of being statistically significant at the p=0.05 level (more if multiple testing and other biases are not accounted for), it is difficult to determine whether the few papers reporting 11-year cycles in proxy data are just flukes.
What is needed is a meta-analysis of all available high-resolution palaeoecological data testing whether significant 11-year cycles are more prevalent than might be expected by chance. I have such an analysis in progress – preliminary results are not encouraging – about 5% of records I used had a significant correlation with solar activity. More news when I get the code to work again.
Christoforou, P., Hameed, S., (1997) Solar cycle and the Paciﬁc ‘centers of action’.
Geophysical Research Letters 24, 293e296.
Hameed, S., Lee, J.N. (2005) A mechanism for sun-climate connection. Geophysical
Research Letters 32, L23817. http://dx.doi.org/10.1029/2005GL024393.
Patterson, R.T., Chang, A.S., Prokoph, A., Roe, H.M. & Swindles, G.T. (2013) Inﬂuence of the Paciﬁc Decadal Oscillation, El Niño-Southern Oscillation and solar forcing on climate and primary productivity changes in the northeast Paciﬁc. Quaternary International