The Effects of Volcanic Eruptions on Climate Prediction Skill
Decadal climate prediction aims to use climate models initialized with observations to predict the time-evolving statistics of regional climate over the next ten years. It was previously thought that a large volcanic eruption would provide additional skill in the predictions after an eruption due to the well-known cooling effects on global climate. Though this may be the case for some aspects of global climate, it is more complicated for the pattern of sea surface temperatures (SSTs) in the Pacific that are associated with regional precipitation over south Asia, Australia, and North America.
Retrospective decadal climate predictions show a post-Mt. Pinatubo eruption sequence of Pacific SSTs that includes a La Niña-like pattern the third northern winter after an eruption, opposite in sign to what was observed after Pinatubo. This leads to a loss of hindcast skill for years in the 1990s affected by the Pinatubo eruption because the post-eruption internal variability of the climate system did not match the multi-model forced response. Agung (1963) and El Chichón (1982) happened to have post-eruption Pacific SST sequences more similar to the multi-model response, and thus do not degrade prediction skill. Thus, decadal prediction skill is reduced if the post-eruption randomly-occurring internal El Niño variability in the observations deviates from the multi-model forced response that, by definition, averages out internal variability in favor of the forced response.
The implications for decadal climate prediction are of critical relevance for scientists attempting to understand decadal climate prediction skill, and for stakeholders who would want to use credible decadal climate prediction information.