Solar Cycles in 150 Years of Global Sea Surface Temperature Data

Abstract

The purpose of the present work is to demonstrate that a solar cycle response exists in surface temperature using the longest global dataset available, which is in the form of 1854?2007 sea surface temperature (SST), with an emphasis on methods and procedures, data quality, and statistical tests and the removal of deterministic signals, such as volcano aerosol forcing and greenhouse gas warming. Using the method of composite-mean difference (CMD) projection, signals of warming during solar maximum and cooling during solar minimum years are found in the global SST over the 14 cycles, dispelling previous claims that the solar cycle response before 1920 is opposite to that of the modern era. The magnitude of the solar cycle response averaged over the oceans between 60°S and 60°N is about 0.1°C of warming for each W m?2 variation of the solar constant (but is slightly lower, at ~0.085°C, when periods of suspected bad data are averaged in, which is consistent with previous work). The signal is robust provided that the years near the Second World War are excluded, during which transitions from British ships to U.S. ships introduced warm bias in the SST, as recently pointed out by D. Thompson and his colleagues. Monte Carlo tests show that the extracted signal has less than 0.02% chance of being a random occurrence. This establishes the existence of a solar cycle response at the earth?s surface at high statistical confidence. Contamination of the signal by volcano aerosols is estimated using the multiple CMD inversion method and found to be small over this long record, although ENSO contamination varies depending on the period chosen but is also small. The multidecadal trend of response to solar forcing is found to account for no more than a quarter of the observed warming in SST during the past 150 yr, under a reasonable but unproven assumption that the climate response to secular solar forcing and to solar cycle forcing has the same spatial pattern.