Was millennial scale climate change during the Last Glacial triggered by explosive volcanism?

The mechanisms responsible for millennial scale climate change within glacial time intervals are equivocal. Here we show that all eight known radiometrically-dated Tambora-sized or larger NH eruptions over the interval 30 to 80 ka BP are associated with abrupt Greenland cooling (>95% confidence). Additionally, previous research reported a strong statistical correlation between the timing of Southern Hemisphere volcanism and Dansgaard-Oeschger (DO) events (>99% confidence), but did not identify a causative mechanism. Volcanic aerosol-induced asymmetrical hemispheric cooling over the last few hundred years restructured atmospheric circulation in a similar fashion as that associated with Last Glacial millennial-scale shifts (albeit on a smaller scale). We hypothesise that following both recent and Last Glacial NH eruptions, volcanogenic sulphate injections into the stratosphere cooled the NH preferentially, inducing a hemispheric temperature asymmetry that shifted atmospheric circulation cells southward. This resulted in Greenland cooling, Antarctic warming, and a southward shifted ITCZ. However, during the Last Glacial, the initial eruption-induced climate response was prolonged by NH glacier and sea ice expansion, increased NH albedo, AMOC weakening, more NH cooling, and a consequent positive feedback. Conversely, preferential SH cooling following large SH eruptions shifted atmospheric circulation to the north, resulting in the characteristic features of DO events.

which is the most recent high-precision date available. However, this date is slightly older 48 than another recently published YTT 40 Ar/ 39 Ar date of 73.88 ± 0.32 ka BP by Storey et al. 49 (2012) 5 . Mark et al. suggest that this discrepancy is partly due to differences in astronomical 50 tuning and calibration of 40 Ar/ 39 Ar between the two dates. They suggest that if the same 51 (recently updated) calibration techniques they used are applied to the Storey et al (2012)   Where appropriate, we have used radiometric data from very recent literature that has not yet 81 been incorporated into LaMEVE. ii) For NH eruptions, we chose to use only Magnitude 7 82 eruptions because these are the best-dated, but also because errors exist associated with the 83 estimation of magnitude of all eruptions, and these are more important for smaller eruptions.

84
For example, if we had included Magnitude 6 eruptions, not only would the age uncertainties 85 been greater, but their magnitude estimate might have been inaccurate as well, and the 86 eruption might actually have been a Magnitude 5 with few climatic repercussions. On the 87 other hand, the magnitude of Magnitude 7 eruptions is better constrained (they are larger 88 eruptions and have attracted more research) and even if the estimate were inaccurate for any 89 given eruption, it is very unlikely that the real magnitude would be less than Magnitude 6 (it 90 is an exponential scale), which would probably still have climatic effects. iii) For Southern

91
Hemisphere eruptions, we chose eruptions that were Magnitude 6 or above for the simple 92 reason that there were far fewer eruptions in the SH than in the NH, and if we had used a 93 threshold of M7 we would have eliminated all eruptions. iv) For both NH and SH eruptions,94 the eruptions needed to have a radiometric date. This is critical, because we do not wish to 95 include eruptions that are not well-dated. The selection of the largest, most well-dated eruptions from an independent database seems to us to be the best way to objectively seleft 97 eruptions for our analysis. of NH eruptions) AMOC, which then continues to affect climate over longer timescales.

137
Atmospheric reorganisation following volcanic eruptions might directly affect AMOC, or 138 may first trigger glacier and sea ice shifts that subsequently affect AMOC.

139
The magnitude of an eruption ( i.e., the volume of ejecta injected into the atmosphere) is 140 clearly linked to the potential of that eruption to affect climate 11 . However, erupted volume is   Conversely, during the coldest intervals of the last 100 ka (e.g., from ~20-30 ka BP and ~60-195 70 ka BP) a combination of substantial ice volume, reduced insolation, and low CO 2 may 196 have mitigated the climatic effects of SH eruptions, explaining the lack of substantial DO 197 events. Only after integrated summer insolation (65°N) 26 began to increase were SH 198 eruptions again able to force DO events. 199 We therefore restrict our analysis to the interval 30-80 ka BP, which is characterised by   where N e is the number of eruptions, T k e are the eruption dates, and T j c are the dates of abrupt 290 climate change events (e.g., DO events or abrupt stadial onsets). The probability distribution 291 function and the cumulative distribution function of this statistic were generated by using 10 292 million random samples (for each hemisphere) (Fig. 4, main text). In other words, 10 million 293 simulations with random eruption dates selected from a uniform distribution of dates from 30 294 to 80 ka BP were conducted per hemisphere, and the probability that the root-mean-squared 295 best match statistic for the distribution of the actual ages was significantly less than for the 296 randomly generated age distributions was determined. The RMS statistic provides a 297 summation of the distances between individual eruption ages and the nearest abrupt climate 298 change event. If no correlation exists between eruptions and climate shifts, the sum of all the 299 root-mean-squared best match statistics should be no different than that produced using the 300 randomly-generated eruption dates.