Abstract
Abstract. Here, we establish a spatiotemporal evolution of the sea-surface temperatures in the North Atlantic over Dansgaard–Oeschger (DO) events 5–8 (approximately 30–40 kyr) using the proxy surrogate reconstruction method. Proxy data suggest a large variability in North Atlantic sea-surface temperatures during the DO events of the last glacial period. However, proxy data availability is limited and cannot provide a full spatial picture of the oceanic changes. Therefore, we combine fully coupled, general circulation model simulations with planktic foraminifera based sea-surface temperature reconstructions to obtain a broader spatial picture of the ocean state during DO events 5–8. The resulting spatial sea-surface temperature patterns agree over a number of different general circulation models and simulations. We find that sea-surface temperature variability over the DO events is characterized by colder conditions in the subpolar North Atlantic during stadials than during interstadials, and the variability is linked to changes in the Atlantic Meridional Overturning circulation and in the sea-ice cover. Forced simulations are needed to capture the strength of the temperature variability and to reconstruct the variability in other climatic records not directly linked to the sea-surface temperature reconstructions. This is the first time the proxy surrogate reconstruction method has been applied to oceanic variability during MIS3. Our results remain robust, even when age uncertainties of proxy data, the number of available temperature reconstructions, and different climate models are considered. However, we also highlight shortcomings of the methodology that should be addressed in future implementations.
Highlights
The Dansgaard–Oeschger (DO) events of the last glacial are some of the most prominent climate variations known from the past
We transferred all ages of Greenland Ice Sheet Project 2 (GISP2) tie points to their equivalent Greenland Ice Core Chronology 2005 (GICC05) ages
Using the proxy surrogate reconstruction (PSR) method with these anomalies we find that the root mean square error (RMSE) between the proxy data and the model simulations increases, and the mean correlation between the surrogate and proxy reconstructions at the core locations drops by 35 %, and very few different model years are chosen for the analogs
Summary
The Dansgaard–Oeschger (DO) events of the last glacial are some of the most prominent climate variations known from the past. Regression models have been applied using climate models with MIS3 boundary conditions (Zhang et al, 2015); linearity between the model variables is still assumed and this assumption may be invalid during the abrupt non-linear DO events Another example is data assimilation and inversion techniques which are becoming more frequently used for paleoclimate studies. While data assimilation would be an ideal method for confining model simulations, it is not feasible in the transient case of MIS3 where the proxy data coverage is very sparse Due to their individual restrictions, both regression methods as well as data assimilation, together with long coupled climate model simulations, appear suboptimal for studying the non-linear changes over the long time period of MIS3. We use the SST variability in the North Atlantic and the Nordic Seas during the last glacial period as a test case for the PSR method to see whether the method can widen our knowledge of spatial patterns back in time.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
