Deglacial Indian monsoon failure and North Atlantic stadials linked by Indian Ocean surface cooling

The Indian monsoon, the largest monsoon system on Earth, responds to remote climatic forcings, including temperature changes in the North Atlantic1,2. The monsoon was weak during two cool periods that punctuated the last deglaciation— HeinrichStadial 1 and theYoungerDryas. It hasbeensuggested that sea surface cooling in the IndianOceanwas the critical link between these North Atlantic stadials and monsoon failure3; however, based on existing proxy records4 it is unclearwhether surface temperatures in the Indian Ocean and Arabian Sea dropped during these intervals. Here we compile new and existing temperature proxy data4–7 from the Arabian Sea, and find that surface temperatures cooled whereas subsurface temperatures warmed during both Heinrich Stadial 1 and the Younger Dryas. Our analysis of model simulations shows that surface cooling weakens the monsoon winds and leads to destratificationof thewater columnandsubstantial subsurface warming. We thus conclude that sea surface temperatures in the Indian Ocean are indeed the link between North Atlantic climate and the strength of the Indian monsoon. During the transition from the Last Glacial Maximum (LGM) to the Holocene (about 20,000–10,000 yr BP), two abrupt, millennialscale coolings interrupted a gradually warming climate: Heinrich Stadial 1 (H1; 17,500–14,500 yr BP) and the Younger Dryas (YD; 12,800–11,500 yr BP). These events are associated with the sudden release of freshwater into the North Atlantic, which slowed thermohaline circulation8, cooled the Northern Hemisphere, and had far-reaching effects on global climate9,10. In the tropics, the Indian monsoon system weakened, causing dry conditions throughout the Indian Ocean rim11. Although these events are specific to a glaciated climate state, they highlight the teleconnection between the North Atlantic and the tropical monsoons. Understanding howwarmings and coolings in theNorth Atlantic impact the world’s largest monsoon system may improve our ability to predict future changes in the Indianmonsoon domain. Previous modelling work suggests that sea surface temperatures (SSTs) in the Indian Ocean are the crucial link between North Atlantic cooling and monsoon failure during Heinrich Events3. Specifically, cooling in the IndianOcean, which is most pronounced in the western part of the basin, causes a delayed onset of the monsoon and weakened convection3. If SSTs do not cool, the Indian monsoon is largely unaffected in these simulations, even though the cooling in the Northern Hemisphere causes a southward shift in the tropical convergence zones3. This implicates cool Indian Ocean SSTs—and not just a change in the hemispheric temperature gradients—as a necessary precursor for a weakening of the Indian monsoon during these millennial-scale events. However, some proxy data suggest that the Arabian Sea—where the model shows a strong decrease in temerature—may not have cooled during H1 (ref. 4), questioningwhether this proposedmechanism formonsoon failure is valid. Here, we analyse temperature proxy evidence from the Arabian Sea during the deglaciation to explicitly test the hypothesis that the surface ocean cooled during H1 and the YD. We then compare the proxy data to both timeslice and transient climate simulations to understand more generally the regional oceanic response to these millennial-scale events. The temperature proxy data include Mg/Ca and δ18O measured on the planktonic foraminifera Globigerinoides ruber (white), and the lipid-based alkenone (UK ′ 37 ), and glycerol dialkyl glycerol tetraether (GDGT) (TEX86) indices. We use published, publicly available proxy time series that have adequate chronology and time resolution to resolve H1 and the YD, and also present new proxy data from core P178-15P in the Gulf of Aden, which has a well-defined chronology12,13. In addition, we include data from a timeslice study5 that provide a spatial array of paired Mg/Ca and δ18O measurements at 8,000, 15,000 (H1) and 20,000 (LGM) yr BP, respectively (Supplementary Table 1). For consistency, we re-calculated all agemodels in the samemanner and transformed the proxy data into palaeotemperature estimates using the same set of calibrations (see Methods and Supplementary Information for details). The transient behaviour of the proxy data across the deglaciation demonstrates that substantial and consistent changes in the thermal structure of the western Indian Ocean occurred during H1 and the YD. The twoMg/Ca–δ18O-derived temperature records indicate that SSTs were comparable to those of the LGM during the early part of H1, but then warmed slightly during the latest stage of the event (16–15 kyr BP) (Fig. 1a). After H1, there was an abrupt warming at 14.5 kyr BP associated with the Bølling/Allerød, then a cooling during the YD of about 0.5 C (Fig. 1a). Similar to the Mg/Ca–δ18O data, the UK ′ 37 -derived temperature records show, with the exception of site NIOP 905, cooling during early H1 followed by a slight warming before the end of the event at 14.5 kyr BP, and a small cooling (about 0.3 C) during the YD (Fig. 1b). To assess whether the observed trends in the Mg/Ca, δ18O and UK ′ 37 data agree with expected climatic responses during the deglaciation, we compare the proxy data to output from the TraCE-21ka (Simulation of Transient Climate Evolution over the past 21,000 yr) fully coupled climate model simulation,