COMPARISON OF ASTRONOMICALLY TUNED AGE MODELS

To further examine the quality of the astronomically calibrated timescales from Sites 1237 and 1241, we plotted the benthic ¹⁸O and ¹³C records from both sites vs. age (Fig. F8). This allows us to compare their age models for the time interval from 4.2 to 5.7 Ma. The oxygen isotope records from the two sites are well correlated and are almost in phase at the precession- and obliquity-related frequency bands (Table T2). Several eye-catching structures and prominent isotope stages are easily recognized and have almost identical ages (within a range of ±2 k.y.). Accordingly, we numbered the unambiguously correlated isotope stages in Figure F8 following the numbering scheme developed by Shackleton et al. (1995). In several intervals, however, patches are seen where the signal-to-noise ratio in the ¹⁸O records is so low that isotope stages bear no clear resemblance. The comparison of the carbon isotope records further strengthens the stratigraphic correlation between the two sites, as the ¹³C records are exceptionally well correlated, which is also indicated by cross-spectral analyses. Both ¹³C records are highly coherent with each other at the orbital periods of 41 and 23 k.y (Table T2). We used the excellent correlation to complete the numbering of isotope stages for the interval from 5.7 to 4.2 Ma (Fig. F8). In summary, the correlation of isotope stratigraphies between Sites 1237 and 1241 demonstrated that their astronomically calibrated age models are identical within an error range of a few thousand years.

The comparison of the benthic ¹⁸O and ¹³C records from Pacific Site 1241 and Atlantic Site 925/926 provides another excellent opportunity to test our age model for the late Miocene to Pliocene interval from 5.7 to 2.5 Ma (Fig. F9). The timescale at Atlantic Site 925/926 (Ceara Rise) (Tiedemann and Franz, 1997; Shackleton and Crowhurst, 1997) was developed by tuning precession-related variations in magnetic susceptibility to the same target record used for calibrating Sites 1241 and 1237. The benthic isotope records were established by Tiedemann and Franz (1997), Billups et al., (1997), and Shackleton and Hall (1997). For the time interval from 4 to 2.5 Ma, cross-spectral comparison between the ¹⁸O records from Sites 1241 and 925/926 indicates that the spectral distribution of variance is very similar and coherent for 41-k.y. cycles (Fig. F10). The dominant obliquity-related variations are highly coherent and in phase (Table T2). In the time interval from 5.7 to 4 Ma, the distribution of spectral variance for 41-k.y. cycles is less similar, but variations in ¹⁸O are highly coherent at the obliquity band. The visual comparison of the oxygen isotope records indicates a remarkably good correlation for the time intervals from 3.75 to 2.5 Ma and 5.1 to 4.5 Ma (Fig. F9), reflecting the global nature of the ¹⁸O ice volume signal and relatively uniform changes in deepwater temperatures and/or salinity at the two sites. The correlation is less pronounced in the intervals from 5.7 to 5.1 Ma and from 4.5 to 3.75 Ma because the ¹⁸O amplitude fluctuations are more different between both sites. These differences cannot be ascribed to differences in the time resolution, as the sampling resolution and the sedimentation rates are similar at both sites. Thus, we attribute these deviations to larger local differences in deepwater temperature/salinity fluctuations. The regional overprint of the global ¹⁸O signal is expected to have been larger at Site 925/926 than at Site 1241, at least for the interval from 4.5 to 3.75 Ma. This has been suggested by the isotope study of Billups et al. (1997). The bathymetric comparison of Pliocene benthic ¹⁸O records from Ceara Rise indicated anomalous high ¹⁸O values at Site 925 between 4.2 and 3.7 Ma, which were interpreted to represent a stronger flux of relatively warmer and more saline North Atlantic Deep Water (NADW) at 3000 m water depth. This is consistent with other studies that attribute early Pliocene warmth and enhanced NADW production to an increased northward heat and salt transport, which probably resulted from a critical threshold in the closure history of the Central American Isthmus (Haug and Tiedemann, 1998; Haug et al., 2001).

The benthic ¹³C records from Sites 925/926 and 1241 are exceptionally well correlated between 5.7 and 2.5 Ma (Fig. F9) and surprisingly good in those intervals, where the correlation of the ¹⁸O signal is relatively weak (5.7–5.1 and 4.5–3.75 Ma). Cross-spectral analyses between the ¹³C records indicate that the spectral distribution of variance is very similar and coherent for 41-k.y. cycles, especially in the older time interval from 5.7 to 4 Ma (Fig. F10; Table T2). The remarkable similarity between the Pacific and Atlantic ¹³C deepwater signal (Fig. F9) with strongest concentration of variance at the 41-k.y. cycle indicates a strong response to global variations in the carbon reservoirs and associated isotope fractionations. The amplitudes and the clarity of the 41-k.y. cycles are more strongly developed in the ¹³C than in the ¹⁸O records, especially in the late Miocene and early Pliocene intervals. This demonstrates that the benthic ¹³C records are not only a powerful medium for orbital tuning, but also a valuable tool for chronostratigraphic correlations that could assist the late Miocene and early Pliocene benthic oxygen isotope stratigraphy, as the ¹³C signal lags relatively constantly behind ¹⁸O at the 41-k.y. period (2–3 k.y.). This opportunity may primarily become important when extending the isotope stratigraphy further back in time, particularly for those intervals where the ¹³C response to cyclic changes in global climate is more clearly developed than in the oxygen isotope signal. The comparison between the isotope records from Sites 1241 and 925/926 also suggests that their orbitally tuned age models are almost identical within an error range of a few thousand years.

During the review process of our publication, Lisiecki and Raymo (2005) presented an orbitally tuned 5.3-m.y. stack (the "LR04" stack) of globally distributed benthic ¹⁸O records. Our oxygen isotope nomenclature as well as the timing of oxygen isotope stages is consistent with that presented in Lisiecki and Raymo (2005), except for the interval prior to 4.8 Ma. The comparison between the LR04 ¹⁸O stack and the ¹⁸O records from Sites 1237 and 1241 is shown and discussed in Tiedemann and Mix (this volume).

Finally, the excellent paleomagnetic stratigraphy at Site 1237 with clear definitions of all Pliocene chrons allows an independent comparison of our age model to the actual ATNTS2004 polarity timescale that has been compiled by Lourens et al. (2004). Our age assignments for the Pliocene magnetic reversal boundaries agree within their error ranges (depth range of magnetic reversals) with the ATNTS2004, except for the top of Kaena and the base of Sidufjall (Table T4).

Our astronomical age for the top of Kaena is ~30 k.y. older. Apart from the general discrepancy in absolute age control, our age model suggests that Kaena Chron spans a time interval of 54 interval is 103 cm/k.y. (Fig. F3). Considering the depth range for the top and base of Kaena, the interval of the reversal would be ~30 k.y. shorter than suggested by the ATNTS2004 timescale (Table T4). Expanding the interval by ~30 k.y. would significantly drop the sedimentation rate and means that we misinterpreted one or two obliquity cycles as precession cycles. This is not very likely for the following reasons. First, the magnetic reversals of Kaena as well as the top of Mammoth are determined at the same core and the mcd in this interval is not affected by a switch point between holes. Therefore, we can exclude that we missed a sediment section over this interval. Second, the next older magnetic reversals, the base of Kaena and the top of Mammoth, are only ~1 m and ~2 m downcore and their assigned ages agree well with those from the ATNTS2004 (Table T4). Our age model suggests a sedimentation rate of ~1.4 cm/k.y. for the interval between the top of Mammoth and the base of Kaena, which is similar to that found for Kaena Chron. Third, spectral variation in both GRA density and color reflectance reveals a dominant precession-related 44-cm cycle for the interval from 54 to 59 mcd (2.92–3.23 Ma). In addition, the spectral variance of the color reflectance is also marked by a pronounced eccentricity-related 222-cm cycle. One may argue that this cycle is statistically not convincing, as it occurs only twice in this short interval. On the other hand, its appearance between 2.92 and 3.23 Ma is not very surprising because the precession-related amplitudes of the color reflectance record may just reflect the strong eccentricity modulation of the astronomical precession record within this interval (Fig. F3).

Our astronomical age for the base of Sidufjall is about one precession cycle younger than that provided by the ATNTS2004 timescale (Table T4). Our age for the top of Sidufjall, however, is in good agreement with that from the ATNTS2004 timescale. Accordingly, our tuning suggests a time span of ~75 k.y. for Sidufjall Chron instead of 97 k.y. as suggested by the ATNTS2004 timescale. Although the top and base of Sidufjall were determined at sediment intervals that belong to the composite depth, they were determined in different holes: in Hole 1237C and 1237D, respectively (Mix, Tiedemann, Blum, et al., 2003). Therefore, we verified the tie point between the two holes at 89.11 mcd. This tie point occurs within the interval that defines the depth range for the base of Sidufjall (88.77–89.27 mcd) (Table T4), whereas the lower limit of the base of Sidufjall occurs within the composite section. The correlation of GRA density records between Holes 1237C and 1237D reveals a small mismatch for the original tie point in the range of 7–10 cm. Since this small inaccuracy does not add an extra cycle, it does not influence our age assignments. Consequently, the assigned age of 4.888 Ma for the lower limit of the base of Sidufjall at 89.27 mcd will remain unaffected (the depth of the upper limit has been corrected for the small offset) (Table T4). We consider both the age range for the base of Sidufjall and the tuning around Sidufjall Chron as reliable.

The most conspicuous feature within Sidufjall Chron is the occurrence of pronounced obliquity-related GRA density fluctuations and relatively low precession-related variability (Fig. F3). The filtered 22-k.y. component of the GRA density record unfortunately overestimates the amplitudes within this interval, as indicated by visual inspection of the GRA density record. The relatively large amplitudes of the filtered 22-k.y. component most likely resulted from interferences with strong obliquity-related amplitudes. The observed strong response to obliquity is also expected from orbital forcing as amplitude variations in orbital obliquity are strong and those in orbital precession are very small during Sidufjall Chron. Accordingly, the tuning was relatively straightforward in this interval.

The stratigraphic comparison of benthic ¹⁸O and ¹³C records from Site 1237 and 1241 corroborates our age assignments along Sidufjall Chron (Fig. F8). The isotope records consistently provide a clear 41-k.y. signal over this interval as well as a good match between sites as documented by the simultaneous occurrence of conspicuous isotope stages Si4 and Si6. Any tuning attempt to stretch Sidufjall Chron at Site 1237 by one precession cycle either distorted the sedimentation rates within this interval or led to an out of phase relationship with orbital obliquity.

Apart from the age discrepancies associated with the top of Kaena and base of Sidufjall, our orbitally derived ages for the other geomagnetic reversal boundaries at Site 1237 are in good agreement with those from the ATNTS2004 timescale and suggest that our age model is very similar to that of Lourens et al. (2004) given the error involved in reversal identification and orbital tuning. This result demonstrates the potential for developing an APTS at Site 1237, probably for the entire Neogene.

The records from Sites 1236 and 1239 are not directly tuned to variations in Earth's orbital parameters. At Site 1236, the main drawback is its poor time resolution. Sedimentation rates vary from 0.5 to 1 cm/k.y. and thus are not suited for orbital tuning. Although Site 1239 has high sedimentation rates of up to 10 cm/k.y., it still lacks a composite depth for the Miocene–Pliocene interval, a major precondition for orbital tuning. However, the opportunity for reconstructing such a composite section is excellent. High-resolution core logging data from two holes as well as high-resolution borehole logging data cover the Miocene–Pliocene interval. Density and natural gamma ray intensity records from borehole and core logging data exhibit strong correlation of meter-scale variability and allow the construction of an equivalent logging depth (eld) scale for the extended core barrel–cored intervals (Mix, Tiedemann, Blum et al., 2003). Such an eld-based composite depth is under construction. The advantage of this depth scale is that it corrects for stretching and squeezing of cored sediment sections. Furthermore, it provides the best estimate of in situ depth and is ideal for estimating mass accumulation rates.

Instead of orbital tuning, we established an initial age model based on magnetostratigraphic (Site 1236) and biostratigraphic information (Site 1239) (Mix, Tiedemann, Blum, et al., 2003). In a second step, we matched the benthic ¹⁸O and ¹³C isotope records from Site 1236 and Hole 1239A with those from Site 1241 by visual identification of oxygen and carbon isotope stages. This procedure indirectly resulted in orbitally tuned age models for Sites 1236 and 1239, spanning the intervals from 2.5 to 5.3 Ma and 2.7 to 5 Ma, respectively. The comparison between benthic ¹⁸O and ¹³C records is shown in Figures F11 and F12 and the age-depth control points for Site 1236 and Hole 1239A are given in Tables T5 and T6. The age model for Hole 1239A is regarded as very preliminary, as our stratigraphic correlation to Site 1241 suggests significant gaps at core breaks. The detailed age model for Site 1236 could have never been achieved without using its ¹³C record as a tool for chronostratigraphic correlation with Site 1241. Although the ¹³C record from Site 1236 is indicative of isotope changes at the Pacific intermediate water level, it clearly resembles the general ¹³C structure of the Pacific central and deepwater sites. In contrast, the intermediate water ¹⁸O signal shows a relatively weak correlation to the "globally correlative" oxygen isotope stratigraphy, represented by Site 1241 in Figure F11. This again demonstrates the utility of early Pliocene benthic ¹³C records for stratigraphic correlations.