A Statistical Modelling Approach Evaluating Explosive Volcanism as a Trigger of Millennial-Scale Climate Change

Abstract

Although millennial-scale climate change events are well documented across an array of globally distributed paleoclimate records, their driving mechanics remain ambiguous; a single coherent theory with concrete evidence providing a comprehensive explanation of the nature and origin of these phenomena remains elusive. Here a statistical model uses explosive volcanism as the sole trigger of millennial-scale climate change to reconstruct the NGRIP δ18O ice core chronology over the last 100 ka BP. The model takes numerous steps to address the inferred undercount in the volcanic catalogue by using a weighted probability Monte Carlo simulation approach to generate statistically grounded ‘missing’ eruptions. The results make a compelling case for the consideration of explosive volcanism as a potential trigger of millennial-scale climate
change. A model using known eruptions generated a modelled δ18O time-series significantly correlated with the NGRIP δ18O time-series (R2 = 0.678, p < 0.001). Another model using both known eruptions and statistically generated ‘missing’ eruptions generated the strongest reproduction of the NGRIP δ18O time-series (R2 = 0.681, p < 0.001), strongly supporting the hypothesis that unknown eruptions could have triggered millennial-scale climate change events that occurred during gaps in the known eruptions catalogue. The results also suggest it is highly likely that the abrupt climate transitions at the onset of the Younger Dryas and GS-20 had volcanic origins, using the recently dated rare supereruption doublet at the onset of GS-20 to underscore the importance of accurately dating eruptions to fully understand the potential climatic impacts of explosive volcanism. Future investigations should focus modelling attempts on shorter windows to adequately investigate the plausibility of explosive volcanism as a trigger of millennial-scale climate change at an individual event level while considering eruption dating uncertainties to address offset issues and maximise correlations between modelled and NGRIP δ18O time-series.