Alkenone SST: Sediment samples (2.5 to 2.57 g) were freeze-dried and extracted by sonication using dichloromethane. The extracts were dried under nitrogen and hydrolysed with 7% potassium hydroxide in methanol. Extraction with hexane yielded a fraction enriched in neutral compounds, which was cleaned with ultra-pure water. The purified extracts were derivatised with BSTFA. Samples were injected diluted in toluene. They were then analysed with a Varian gas chromatograph Model 3400 equipped with a septum programmable injector and a flame ionisation detector. Absolute concentration errors were below 10%. Reproducibility tests showed that the uncertainty in the Uk'37 determinations was lower than 0.015 (ca. ±0.5ºC). The SST reconstruction (Uk'37-SST) was obtained from the relative composition of C37 unsaturated alkenones [Uk'37 = C37:2/(C37:2+C37:3)], using the calibration equation Uk'37 = 0.033*SST+0.044. The record was generated by B. Martart and J.O. Grimalt.Age model: The age model was developed by aligning the abrupt transitions of d18O record of G. bulloides in core MD01-2444 to warming and cooling events in the NGRIP d18Oice record, presented on: (i) the GICC05 timescale [Svensson, A. et al. A 60,000 year Greenland stratigraphic ice core chronology. Clim. Past 4, 47–57 (2008)]; and (ii) the WD2014: modified timescale, whereby NGRIP GICC05 ice ages 31.2–67.2 ka are multiplied by 1.0063. [Buizert, C. et al. The WAIS Divide deep ice core WD2014 chronology – Part 1: Methane synchronization (68–31 ka BP) and the gas age–ice age difference. Clim. Past 11, 153-173 (2015)].