Orbital Forcing of the Marine Isotope Stage 9 Interglacial

See allHide authors and affiliations

Science  12 Jan 2001:
Vol. 291, Issue 5502, pp. 290-293
DOI: 10.1126/science.291.5502.290


Milankovitch orbital forcing theory has been used to assign time scales to many paleoclimate records. However, the validity of this theory remains uncertain, and independent sea-level chronologies used to test its applicability have been restricted largely to the past ∼135,000 years. Here, we report U-series ages for coral reefs formed on Henderson Island during sea-level high-stands occurring at ∼630,000 and ∼330,000 years ago. These data are consistent with the hypothesis that interglacial climates are forced by Northern Hemisphere summer solar insolation centered at 65°N latitude, as predicted by Milankovitch theory.

Knowledge of the precise timing of past sea-level high-stands provides a crucial test of the Milankovitch model of climate change. This theory postulates that glacial-interglacial cycles are driven by periodic changes in July solar insolation at 65°N, caused by predictable variations in Earth's orbit (1). Previous U-series studies have focused almost exclusively on the last interglacial period, Marine Isotope Stage (MIS) 5.5, occurring at about 125,000 years ago (ka) (2–9). The timing of the MIS 5.5 sea-level high-stand appears consistent with Milankovitch forcing theory (8, 10). However, there is a growing body of evidence that factors other than 65°N summer solar insolation may have triggered the MIS 6-5 deglaciation, which appears to have begun before insolation started to increase (10–12). In contrast, the last deglaciation, which began at ∼21 ka, can be adequately explained by 65°N insolation forcing alone. This raises the possibility that a different combination of forcing mechanisms may have operated during previous glacial-interglacial cycles (13).

To help resolve this apparent contradiction, the Milankovitch climate model can be tested by dating coral reefs that formed during older interglacials. However, it is exceedingly difficult to obtain reliable chronologic information for older sea-level high-stands because of a lack of well-preserved, datable coral.

Here, we report 30 precise U-series ages for a set of reef terraces formed during MIS 9 and MIS 15 on Henderson Island near Pitcairn Island in the equatorial Pacific (Fig. 1). These data, obtained using thermal ionization mass spectrometry (TIMS) (7) and multiple-collector ICP sector mass spectrometry (MC-ICPMS) (14–16), provide reliable and independent radiometric constraints on the global sea level curve at ∼630 and ∼330 ka.

Figure 1

Location of Henderson Island which, together with Pitcairn, Ducie, and Oeno islands, compose the Pitcairn Island Group, located near the summit of the East Pacific Rise, ∼2100 km west of Easter Island. Henderson Island is tectonically stable, and its slow isostatic uplift [<0.1 m/ka (18)], in response to volcanic loading by the emplacement of Pitcairn Island at ∼0.8 million years ago (Ma), need not be considered when resolving MIS 9 sea-level change. (Inset) Composite geological cross-section of Henderson Island [modified from (18)]. The oldest exposed reefs occur in the center of the island and represent a major atoll construction phase; MIS 15 corals were sampled from the interior fossil lagoon. MIS 9 fringing reefs occur at the seaward margin of the early-formed atoll.

Henderson Island is located in the extreme east of the Indo-Pacific subtropical province, resulting in severe restrictions in ecological development (17, 18). Therefore, coral terraces form only during exceptionally long (18) or warm interglacials. Offshore fringing reefs grew prolifically on Henderson Island during the MIS 9 interglacial, but no coral terraces appear to have formed subsequently during the MIS 7.1, MIS 5.5, or mid-Holocene sea-level high-stands.

Growth-position corals were collected from reef terraces along seven lyphenate cliff-section transects and one central fossil lagoon transect during the 1991–92 Pitcairn Islands Scientific Expedition (18, 19). U-series results for corals selected for dating are displayed in Web table 1 (16). In the 330 ka samples, 230Th-age uncertainties (excluding the systematic contribution from the decay constants) can be smaller than ±2 ka, allowing the timing and duration of the MIS 9 interglacial to be well resolved. For completeness, all age uncertainties reported here include the decay constant contributions. Approximately one-third of the samples are considered to have acceptable 230Th-ages [Web table 1 (16)] (Fig. 2A) (20). This is a very large proportion, given the old age of the reefs. Even in MIS 5.5 samples, it is not unusual to reject more than half of the data on the basis of selection criteria for reliability (5, 9). Acceptable ages range between 334 ± 4 and 293 ± 5 ka. Three reliable samples were measured in duplicate, and the 230Th-ages are concordant [Web fig. 1 (16)], despite the expected heterogeneity in U-Th isotopic composition resulting from enhanced diagenesis in some parts of the coral skeleton over others. Even samples that do not pass reliability criteria show a high degree of consistency within a sampling transect: the highest levels of diagenesis appear to have occurred along Transect 3 and in the fossil lagoon, as indicated by very elevated δ234U(T) [a proxy for the U isotopic composition of the seawater in which the coral formed; Web table 1 (16)] in all samples.

Figure 2

(A) δ234U(T) and (C) sample elevation versus 230Th-age for “reliable” MIS 9 corals from Henderson Island. Duplicate analyses in (A) are indicated by gray ellipses. The lower precision result for HEN 2-7 is represented by a dashed ellipse. Shown also in (C) are the Devils Hole δ18O Nevada temperature proxy (11), the North Atlantic Site 980 δ18O sea-level record (32), and the Milankovitch solar insolation curve for July 65°N. Sea-level and 65°N July insolation curves for MIS 5 are shown in (B) for comparison.

Reefs also grew on Henderson Island during a major atoll construction phase occurring near 630 ka. Because MC-ICPMS can routinely achieve sub-permil analytical uncertainties for both U and Th measurements, U-series chronology can be extended beyond the 500 ka upper limit usually reported for TIMS. Two samples recovered from the fossil lagoon in the interior of the island (Fig. 1) have statistically identical230Th-ages of 642 +36/–32 ka and 628 ± 25 ka (weighted mean of three results), and probably record the earliest emergent reef growth on Henderson Island. The weighted mean result of 632 ± 21 ka provides a direct U-series constraint on the timing of the MIS 15 interglacial sea-level high-stand.

Despite their old age, the Henderson Island “330 ka” coral terraces are exceptionally well preserved (21). Some diagenetic alteration has occurred, but the effect on the238U-234U-230Th system appears predictable, allowing disturbed samples to be identified and discarded, and allowing important constraints to be placed on the behavior of U and Th during diagenesis.

Following earlier work (7, 8), we consider δ234U(T) to be the most reliable quantitative test of alteration of the coral skeleton. With only a few exceptions (5, 22), diagenesis appears to shift both δ234U(T) and the 230Th-age toward elevated values, and this general correlation has been well documented (35, 23). Gallup et al.(24) demonstrated that a roughly linear correlation exists between 230Th-age and δ234U(T) in ∼200 ka Caribbean reefs. This is consistent with a diagenetic mechanism that adds both foreign230Th and 234U continuously and linearly to the coral skeleton.

Aside from these Caribbean data, only the MIS 5.5 observations for Western Australia (7, 8) and these MIS 9 interglacial results for Henderson Island are of sufficient number and precision to test the validity of any given model of diagenesis. A striking result is that the 238U-234U-230Th system exhibits similar diagenetic behavior at all three reef localities. Although exceptions do occur (for example HEN 1-21, HEN 1-22, and HEN 4-12 in the MIS 9 Henderson Island data set), approximately the same broadly linear relationship between the 230Th-age and δ234U(T) can be found in the reefs of Henderson Island (∼330 ka), the Caribbean (∼200 ka), and Western Australia [∼125 ka; Web fig. 1 (16)]. The three data sets appear consistent with a common alteration process independent of local climate, whether it follows the combined230Th-234U linear uptake model proposed by (24) or a more elaborate open-system model (25, 26). Assuming continuous uptake of external234U and 230Th, the rates must be approximately linear, and the ratio of the rates of 234U to230Th addition approximately fixed, irrespective of location. Similar diagenetic shifts in U-Th appear to be occurring at a fine (millimeter to centimeter) scale in the Henderson Island reefs. Replicate results for separately processed pieces of samples HEN 2-7 and FH 328 show that older 230Th-ages are correlated with more elevated δ234U(T). The degree of alteration of the coral skeleton is variable, but is consistent with the diagenetic behavior exhibited by the reef system as a whole.

Corals with the lowest δ234U(T) are assumed to have undergone minimal diagenetic exchange of U and Th, providing a reliable constraint on the seawater234U/238U during the MIS 9 interglacial. In Web fig. 1 (16), no Henderson Island initial δ234U lies below the δ234U(T) = 149 ± 1‰ contour for present-day seawater, excluding the few erroneous data points that do not lie on the same diagenetic trend as shown by the rest of the data. Isotopic studies of ∼200 and ∼125 ka corals (6, 7, 24) indicate that the marine δ234U was within error of the present-day value during the previous two interglacial periods. There is some indication that the marine δ234U was also near 149‰ during the 83 ka interstadial event (24). Our new Henderson Island data show that even at ∼330 ka, the seawater δ234U appears to have been within error of 149‰, and we conclude that the marine 234U/238U has returned to essentially the same value during each of the past four interglacial periods.

Two interpretations of the reliable U-series data for the MIS 9 Henderson Island reefs are possible, representing limiting cases of a range of intermediate models. In the first case, it is assumed that all corals formed during a continuous episode of reef growth, within a single and prolonged sea-level high-stand correlating with MIS 9.3. This interpretation assumes the different sample elevations [between <2 and 29 m above the present mean sea level (MSL) datum] reflect the variable water depths in which corals can grow. Today at Henderson Island, most corals live in an 18-m depth range, between 2 and 20 m below MSL (27). In this case, our230Th-ages suggest that high interglacial sea levels persisted from at least 334 ± 4 to 306 ± 4 ka (weighted mean of two results for HEN 2-2), and possibly later, until 293 ± 5 ka if the less precise result for HEN 2-7 is also considered (Fig. 2A). Considering, for the moment, precise results only, the data give a minimum duration of 20 ka and a maximum duration of 36 ka for the MIS 9 sea-level high-stand, taking into account the 2σ age uncertainties. The latter interval is very long compared with the duration of the Last Interglacial (Fig. 2B), but, given the extremely old age of these samples, it is possible that even “reliable” data points have still been disturbed by minor diagenesis, shifting their230Th-ages toward older values despite our screening efforts.

The second, and preferable, explanation assumes that all “reliable” samples are pristine and several discrete reef-building episodes occurred during multiple sea-level high-stands. Figure 2C considers an alternative sea-level curve for MIS 9, assuming all corals formed near the sea surface, at MSL. The data show several distinct oscillations in sea level, which may be correlated with the MIS 9.3 interglacial and MIS 9.1 and MIS 8.5 interstadials, respectively. Our results suggest that sea levels approached peak interglacial values near 324 ± 3 ka and high sea levels persisted for up to 8 ka. Sea levels then began to fall toward glacial values near 318 ± 3 ka (weighted mean of two results for HEN 5-6). Subsequent “interstadial” oscillations in sea level attained their maximum values near 306 ± 4 and 293 ± 5 ka, in-phase with Milankovitch insolation peaks at ∼311 and ∼290 ka, respectively (28). Two critical observations support this latter scenario. First, Montastrea sample HEN 4-1, dated at 334 ± 4 ka, probably correlates with the MIS 10/9 deglacial sea-level rise, while the sea surface was still significantly lower than at the interglacial maximum. As a comparison, livingMontastrea corals on the Great Barrier Reef, Australia, reside only in the upper 15 m of the water column. If the same ecological conditions can be assumed for the Montastreaspecies at Henderson Island, then the implication is that at least a 5 m rise in sea level must have occurred between 334 and 324 ka. This oscillation in sea level would have exceeded 20 m if HEN 1-26 grew in deeper water than HEN 4-1. Similarly, Montastreasample HEN 1-10 occurs ∼28 m stratigraphically lower than the clustering of samples correlating with the MIS 9.3 interglacial. This implies a sea-level fall of at least 13 m between 318 ± 2 and 317 ± 5 ka. Second, independent field observations suggest that the “MIS 9.3” samples (comprising massive corals within well-lithified reef complexes) formed during a prolonged episode of high and stable sea level, whereas the younger “interstadial” samples (composed of smaller corals within reef units that drape the lower part of the MIS 9.3 complexes) formed subsequently, during one or more lower and shorter lived sea-level still-stands.

The true sea-level curve for Henderson Island probably lies somewhere between the two end-member cases we present here. However, our results cannot resolve this. Importantly, the earliest phase of MIS 9 reef growth on Henderson Island, dated at 334 ± 4 ka, is consistent with the only other reliably dated TIMS U-series observation for reefs that formed near the MIS 9 interglacial: a coral sample recovered from a drowned carbonate platform in the Huon Gulf, Papua New Guinea, and correlated with the preceding MIS 10 sea-level low-stand, has a slightly earlier 230Th-age of 348 ± 10 ka (29). The Henderson Island sea-level curve in Fig. 2C also shows a very good correspondence with the Devils Hole (Nevada, USA) DH-11 paleoclimate record (11, 30), which is the only other independently dated TIMS U-series record spanning the MIS 9 interglacial interval. The DH-11 record shows that air temperatures over Nevada attained their maximum values near 335 ± 8 ka. This is coeval with the earliest recorded episode of reef growth (at 334 ± 4 ka) on Henderson Island, although the combined U-series age uncertainties preclude an exact comparison of the two records. The Devils Hole chronology implies that interglacial climates persisted for ∼20 ka in the Nevada region during MIS 9.3. This is in good agreement with the duration that we observe from the Henderson Island coral reef results.

Our Henderson Island results are consistent with the idea that interglacial climates are forced by orbitally induced Northern Hemisphere summer solar insolation changes centered on 65°N. The earliest high sea levels within the MIS 9 interglacial, dated between 334 ± 4 and 324 ± 3 ka, are either coincident with or slightly postdate the timing of peak insolation at 333 ka (31).

  • * To whom correspondence should be addressed. E-mail: stirling{at}


View Abstract

Navigate This Article