Fig. 1. Locations of ODP Site 1089 and Core PS2821-1 (large dot) and of the core-top samples used for the radiolarian reference database, in comparison to the average position of the oceanographic fronts (Peterson and Stramma, 1991), and the mean winter sea-ice distribution ( Naval Oceanography Command Detachment, 1985). Different symbols have been assigned to stations according to which radiolarian assemblage reaches the higher factor component value at that station (also see Table 2 and Table 3): factor 1=filled circle, factor 2=open circle, factor 3=square, factor 4=asterisk.

Plate I. Scale BAR=100 μm for all figures, except figs. 1 and 15 (scale BAR=200 μm). All samples are core tops, unless otherwise indicated.

Plate II. Scale BAR=100 μm for all figures, except figs. 16, 17 and 22 (scale BAR=200 μm). All samples are core tops, unless otherwise indicated.

Fig. 2. Observed versus estimated mean summer SST (December to March) for all core tops included in the reference database. Indicated are the linear regression line, equation and standard error of estimate. Observed summer SST versus residual for SST with ±σ (one standard deviation).

Fig. 3. (A) Down core SSST over the past 160 ka at ODP Site 1089 and PS2821, and δ¹⁸O PDB variations for the planktic foraminifer Globigerina bulloides (Mortyn et al., submitted) at ODP Site 1089. The age model for ODP Site 1089 and PS2821-1 is the one developed in this paper (see 2.4. Age model). The dashed line represents today’s temperature (15.8°C) at the core location. The topmost bar includes the MIS 1–6, with their age according to Martinson et al. (1987). (B) Changes over the past 40 ka in CH₄ concentrations (in parts per billion in volume) for gas trapped in air bubbles in the GRIP ice core (Greenland). CH₄ data and the gas-age model for this core are from Chappellaz et al. (1993). Vertical dashed lines mark the position of quasi-synchronous climatic changes. Interstadials 5–8 and 10–11 are the same as those recognized when comparing the CH₄ and the δ¹⁸O record in the GRIP ice core (Chappellaz et al., 1993). (C) Paleotemperature record at PS2821-1. (D) Paleotemperatures above the atmospheric inversion layer at Vostok Station (Antarctica), given as a difference from today’s value ( Petit et al., 1999). The age model for the Vostok ice core is from Petit et al. (1999). The age assignment of some of the most prominent climatic changes during MIS 5 ( Fig. 3A) discussed in the text, as well as the approximate location of the YD-I (‘YD-type’) and the ACR-I (‘ACR-type’) cooling events ( Fig. 3C) are reported for clarity. Note the change in age scale between A and B–D.

Fig. 4. Comparison of paleotemperature changes at PS2821/ODP Site 1089 and at Vostok Station, Antarctica. The age model and temperature data for Vostok are from Petit et al. (1999). (A) SSST record for PS2821/ODP Site 1089. (B) Atmospheric paleotemperatures at the inversion layer reconstructed from the Vostok ice core (Antarctica), given as a difference from today’s value ( Petit et al., 1999). (C) Blow-up of A and B for a 30-kyr interval including Termination I. The positions of the cooling events recognized at this termination in the SSST record of PS2821 are reported as YD-I and ACR-I. For the Vostok ice core, the position of the ACR and of an event similar to YD-I are also shown. (D) Same as above, for Termination II. The positions of the cooling events recognized at this termination in the SSST record of Site 1089 are reported as YD-II and ACR-II. For the Vostok ice core, the position of an event similar to YD-II is also shown.

Fig. 5. Physical properties, isotopic and paleotemperature records for ODP Site 1089 and Core PS2821-1 over the past 160 ka. Magnetic susceptibility and color reflectance measurements from Shipboard Scientific Party (1999). Age assignment of MIS 1–6 after Martinson et al. (1987). The arrows and dashed line in E and F identify the portions of these records belonging to Core PS2821-1 and to ODP Site 1089. (A) Smoothed color reflectance (percentage of reflected to incoming radiation) in the 550–650-nm band for ODP Site 1089. (B) Color reflectance (percentage of reflected to incoming radiation) in the 550–650-nm band for Core PS2821-1. (C) Magnetic susceptibility (SI×10⁻⁵ units) for ODP Site 1089. (D) Magnetic susceptibility (SI×10⁻⁵ units) for Core PS2821. (E) Changes in the relative abundance of Cycladophora davisiana along the PS2821-1/ODP Site 1089 record. (F) Estimated IKM-radiolarian SSST for the PS2821-1/ODP Site 1089 record. (G) δ¹⁸O PDB for the planktic foraminifer Globigerina bulloides (Mortyn et al., submitted).

Fig. 6. (A) The paleotemperature record of ODP Site 1089 and Core PS2821-1 for the last 160 ka compared to its precessional component (thin line). (B,C) The results of MTM spectral analysis for the paleotemperature record, together with the frequency, amplitude and F-test values (thin line) for the spectra. The periods for spectral peaks with an F-test value higher than 0.90 (see Table 6) are also shown. Note the change in scale for the frequency axis between B and C.

Table 1. Tie-points between the different holes (A–D) at ODP Site 1089 used to produce the composite section (modified after Shipboard Scientific Party, 1999)

Table 2. Geographic location (latitude, longitude, water depth), sampling device (LBC=large box corer, MUC=multi-corer, MIC=micro-corer) and estimated SSST for the core tops in the reference dataset

For each sample the communality, the varimax factor component and the percent of the total information explained (along with its corresponding cumulative value) for the first four factors are also shown.

Table 3. Overview of the radiolarian assemblages recognized in the core-top dataset

The factor names, the temperature and latitudinal ranges, as well as the taxa having varimax factor scores higher than 1, are indicated for each factor. The symbols below the assemblage names correspond to those shown in Fig. 1.

Table 4. Scaled varimax factor scores matrix (species vs. factor scores) for the 24 species used in the reference dataset

Absolute values higher than 1.000 have been highlighted (bold).

Table 5. Statistical summary of the first seven factors extracted by the program CABFAC: eigenvalue and cumulative variance explained (top)

Regression terms (radiolarian assemblages/factors) and coefficients for the paleotemperature regression equation (bottom).

Table 6. Statistically significant (F-test value >0.90) spectral peaks obtained by an MTM applied to the paleotemperature record

The frequency (in cycles/kyr), period (in kyr), amplitude and F-test value are reported for each peak. The bold values correspond to periodicities close to those recognized in the Vostok ice-core record (Yiou et al., 1991).