Atmospheric Chemistry and Physics (Jan 2017)

Revisiting the observed surface climate response to large volcanic eruptions

  • F. Wunderlich,
  • D. M. Mitchell

DOI
https://doi.org/10.5194/acp-17-485-2017
Journal volume & issue
Vol. 17, no. 1
pp. 485 – 499

Abstract

Read online

In light of the range in presently available observational, reanalysis and model data, we revisit the surface climate response to large tropical volcanic eruptions from the end of the 19th century until present. We focus on the dynamically driven response of the North Atlantic Oscillation (NAO) and the radiative-driven tropical temperature response. Using 10 different reanalysis products and the Hadley Centre Sea Level Pressure observational dataset (HadSLP2) we confirm a positive tendency in the phase of the NAO during boreal winters following large volcanic eruptions, although we conclude that it is not as clear cut as the current literature suggests. While different reanalyses agree well on the sign of the surface volcanic NAO response for individual volcanoes, the spread in the response is often large (∼ 1/2 standard deviation). This inter-reanalysis spread is actually larger for the more recent volcanic eruptions, and in one case does not encompass observations (El Chichón). These are all in the satellite era and therefore assimilate more atmospheric data that may lead to a more complex interaction for the surface response. The phase of the NAO leads to a dynamically driven warm anomaly over northern Europe in winter, which is present in all datasets considered. The general cooling of the surface temperature due to reduced incoming shortwave radiation is therefore disturbed by dynamical impacts. In the tropics, where less dynamically driven influences are present, we confirm a predominant cooling after most but not all eruptions. All datasets agree well on the strength of the tropical response, with the observed and reanalysis response being statistically significant but the modelled response not being significant due to the high variability across models.