Constraints on the magnitude and patterns of ocean cooling at the Last Glacial Maximum

Observation-based reconstructions of sea surface temperature from relatively stable periods in the past, such as the Last Glacial Maximum, represent an important means of constraining climate sensitivity and evaluating model simulations1. The first quantitative global reconstruction of sea surface temperatures during the Last Glacial Maximum was developed by the Climate Long-Range Investigation, Mapping and Prediction (CLIMAP) project in the 1970s and 1980s (refs 2,3). Since that time, several shortcomings of that earlier effort have become apparent4. Here we present an updated synthesis of sea surface temperatures during the Last Glacial Maximum, rigorously defined as the period between 23 and 19 thousand years before present, from the Multiproxy Approach for the Reconstruction of the Glacial Ocean Surface (MARGO) project5. We integrate microfossil and geochemical reconstructions of surface temperatures and include assessments of the reliability of individual records. Our reconstruction reveals the presence of large longitudinal gradients in sea surface temperature in all of the ocean basins, in contrast to the simulations of the Last Glacial Maximum climate available at present6,7. Studies following the Climate Long-Range Investigation, Mapping and Prediction (CLIMAP) project mainly reanalysed old primary data, applied single new proxies or concentrated on specific ocean basins8–11. The Multiproxy Approach for the Reconstruction of the Glacial Ocean Surface (MARGO) project’s objective has been to compile and analyse available estimates of Last Glacial Maximum (LGM) sea surface temperatures (SSTs) based on all prevalent microfossil-based (transfer functions based on planktonic foraminifera, diatom, dinoflagellate cyst and radiolarian abundances) and geochemical (alkenones and planktonic foraminifera Mg/Ca) palaeothermometers. The MARGOproject approach is to argue that no current proxymethod is objectively better than another to provide an accurate picture of past SST, and that consequently the multiproxy approach yields the least biased representation of past reality. By using a rigorous definition of the LGM time interval (19–23 cal kyr bp; ref. 4), many records used previously had to be discarded. Other key features are the use of a common data set of ambient temperatures for the calibration of all proxies and the assessment of the reliability of individual SST estimates5. The MARGO compilation combines 696 individual SST reconstructions (Fig. 1). The coverage is especially dense in the North Atlantic, the SouthernOcean and the tropics. Several oceanic regions remain undersampled, such as the subtropical gyres, especially in the Pacific Ocean. Individual proxies have different spatial coverage and analyses of multiple palaeothermometers on the same material remain rare. To address this spatial bias, we have projected the entire SST data set onto a regular grid of