Paleoceanography

Coherent obliquity band and heterogeneous precession band responses in early Pleistocene tropical sea surface temperatures

Authors


Abstract

[1] The nature of the connection between high- and low-latitude climates during the early Pleistocene “41 kyr world” has important implications for our understanding of the feedbacks involved in translating insolation changes into global climate states. Here we focus on the tropical marine record, presenting alkenone-derived sea surface temperature (SST) and productivity records from the eastern equatorial Atlantic, eastern equatorial Pacific, the Arabian Sea, and the South China Sea for a time interval covering the heart of the 41 kyr world (1.2–2.0 Ma). All four SST records are dominated by variance in the obliquity band, suggesting that high-latitude dynamics and low-latitude climate were tightly coupled in the 41 kyr world, despite smaller ice volume variability during this interval as compared to the late Pleistocene. At the 41 kyr period, SST varied coherently and nearly synchronously between the four study regions, suggesting a tropic-wide feedback to high-latitude processes. Productivity variations at our equatorial Atlantic and Pacific sites were also coherent in the obliquity band, implying tropical trade wind variability at this frequency during the early Pleistocene. In contrast, we observe heterogeneous SST and productivity responses in the precession band between each of the tropical locations. Local atmospheric circulation patterns, rather than a globally coordinated response to precessional insolation forcing, apparently determined SSTs and productivity in the tropics at precessional frequencies during the early Pleistocene.

1. Introduction

[2] The early Pleistocene “41 kyr world” (∼1.2—1.8 Ma) represents an intriguing period in Earth’s climate history. We know from benthic oxygen isotope records that global ice volume varied in concert with the Earth’s orbital obliquity [Ruddiman et al., 1986], but our understanding of orbitally driven climate dynamics during the early Pleistocene remains incomplete. Most importantly, the lack of strong precessional power in the ice volume record is at odds with traditional Milankovitch theory, which states that summer insolation at high northern latitudes should control ice sheet size by modulating summer melting and ablation rates. Because northern high-latitude summer insolation curves are dominated by variance in the precession band, the lack of a significant ice volume response at this frequency requires explanation [Huybers, 2006; Raymo et al., 2006; Raymo and Nisancioglu, 2003]. Potentially, the contrast between the obliquity-dominated ice volume response of the early Pleistocene and the 100 kyr ice volume response of the late Pleistocene “100 kyr world” (∼0.0—0.6 Ma) holds critical clues to our understanding of the boundary conditions and/or critical feedbacks that dictate the climate’s response to orbital forcing.

[3] Early and late Pleistocene glacial cycles differ in their mean states, total variance, and dominant frequencies. During the early Pleistocene, benthic oxygen isotope values were on average 0.29‰ lighter than during the late Pleistocene [Lisiecki and Raymo, 2005]. Assuming a similar relationship between benthic δ18O and sea level for the Last Glacial Maximum and the early Pleistocene, this difference is equivalent to ∼20 m of global sea level change and indicates overall lower ice volume during the early Pleistocene. Additionally, the total amplitude of glacial-interglacial δ18O change was about 40% smaller during the early Pleistocene [Lisiecki and Raymo, 2005]. In the frequency domain, early Pleistocene glacial cycles oscillated nearly purely at the 41 kyr frequency, with very little concentration of spectral power in the precession band [i.e., Raymo and Nisancioglu, 2003]. In contrast, benthic oxygen isotope records from the late Pleistocene are dominated by 100 kyr variability and show much more significant variability in the precession band [Imbrie et al., 1993b]. Two important questions regarding these differences are (1) Did the smaller 41 kyr glacial cycles of the early Pleistocene impose a similar scale of global climatic change as the 100 kyr ice volume cycles of the late Pleistocene? and (2) To what degree did the switch from 41 kyr to 100 kyr glacial cycles during the mid-Pleistocene require the development of new feedbacks within the climate system?

[4] During the late Pleistocene, 100 kyr glacial cycles were closely tied to climatic changes throughout the globe, as evidenced by the strong, coherent 100 kyr responses observed in greenhouse gas concentrations [Petit et al., 1999] and in an array of local climate records from both high and low latitudes [i.e., Beaufort et al., 2001; de Garidel-Thoron et al., 2005; deMenocal et al., 1993; Holbourn et al., 2005; Imbrie et al., 1993a; Lea et al., 2000; Liu and Herbert, 2004; Medina-Elizalde and Lea, 2005]. Our understanding of the internal climate dynamics associated with the 41 kyr glacial cycles of the early Pleistocene is limited by our lack of adequately resolved and spatially distributed records from this interval. Potentially, the smaller glacial cycles of the early Pleistocene were accompanied by reduced feedbacks related to ice albedo, sea ice extent, atmospheric circulation and/or the carbon cycle. If these feedbacks were diminished, low-latitude climate may have been largely independent of high-latitude glacial cycles, responding instead to local precessional forcing, which models show should dominate tropical dynamics [Clement et al., 1999; Kutzbach and Liu, 1997; Prell and Kutzbach, 1992]. However, several existing records suggest a strong high-latitude influence on low-latitude climate during the early Pleistocene, as evidenced by their significant concentrations of spectral power in the obliquity band. These records include eastern equatorial Pacific upwelling SST and productivity [Liu and Herbert, 2004], western Pacific warm pool SST [de Garidel-Thoron et al., 2005; Medina-Elizalde and Lea, 2005], and dust flux associated with the African and Asian monsoons [Bloemendal and deMenocal, 1989]. While these records provide a preliminary indication that high- and low-latitude climate change were synchronized in the obliquity band during the early Pleistocene, a more detailed and geographically diverse characterization is necessary in order to determine to what extent the 41 kyr signal was a pervasive feature of tropical climate. Additionally, the relationship between high- and low-latitude climates in the precession band during the early Pleistocene, when ice volume variability lacked strong 23 kyr power, has not been thoroughly explored.

[5] Here we examine tropical responses to local and high-latitude forcing during the 41 kyr world using new high-resolution SST and productivity records from the eastern equatorial Atlantic (EEAtl) and Arabian Sea upwelling regions, and from the South China Sea. We compare these records with an existing record from the tropical upwelling zone of the eastern equatorial Pacific (EEP) [Liu and Herbert, 2004] in order to evaluate whether the strong obliquity band connection between high-latitude glacial cycles and EEP dynamics is unique to this region or is representative of a broader tropical response, and to assess the coherence of the precession band SST and productivity responses at different tropical locations.

2. Site Locations

2.1. Tropical Upwelling Sites (ODP 846, 662, and 722)

[6] Ocean Drilling Program (ODP) Site 846 (3°5′S, 90°46′W, 3296 m water depth) is located in the heart of the EEP cold tongue (Figure 1). Trade winds drive open ocean divergence and year-round upwelling, which results in relatively cool SSTs and high biological production. Subsurface waters originate in the subantarctic and travel to the EEP via the Equatorial Undercurrent, also providing the main source of nutrients to the EEP [Harper, 2000; Sarmiento et al., 2003]. Modern mean annual SST at the location of Site 846 is ∼23.5°C, with a yearly range of ∼20 to 27°C [Levitus and Boyer, 1994]. Early Pleistocene alkenone SST and inferred productivity records for this site were originally published in Liu and Herbert [2004].

Figure 1.

Site locations for all of the tropical SST and productivity records discussed here. The locations of western Pacific warm pool SST records published by deGaridel-Thoron et al. [2005] (asterisk) and Medina-Elizalde and Lea [2005] (plus sign) are also indicated.

[7] ODP Site 662 (1°23′S, 11°44′W, 3824 m water depth) lies beneath the Southern Equatorial Current in the center of the Atlantic equatorial divergence zone (Figure 1). During boreal summer, northward advancement of the Southern Hemisphere trade winds leads to enhanced divergence and the upwelling of cool, nutrient-rich waters, which stimulates a pulse of biological productivity. On annual timescales, cooler SSTs are closely coupled with increased productivity [Pérez et al., 2005; Weingartner and Weisberg, 1991]. Upwelling dynamics in the EEAtl are also strongly influenced by the development of the African monsoon, which suppresses open ocean divergence by diverting the equatorial trade winds northward and reducing their zonal component [McIntyre et al., 1989]. Subsurface waters in the EEAtl originate in the subantarctic Atlantic, and also contain small amounts of warm Indian Ocean waters that are entrained at the Agulhas retroflection off the southern tip of Africa [Peterson and Stramma, 1991]. This addition of heat at high latitudes is unique to the South Atlantic eastern boundary current and provides an additional control on the temperature of thermocline waters in the EEAtl [Gordon, 1986]. Modern SST at Site 662 ranges between ∼23.5 and 28°C over the course of a year, with a mean annual SST of ∼26°C [Levitus and Boyer, 1994].

[8] ODP Site 722 (16°37′N, 59°48′E, 2028 m water depth) is located on the Owen Ridge in the northwestern Arabian Sea (Figure 1). Modern ocean dynamics are dominated by monsoon circulation, which drives coastal and open ocean upwelling during the summer monsoon as strong southwest winds develop in response to sensible heating and low atmospheric pressure over the Indian subcontinent. The upwelling of cool, nutrient-rich waters during the summer monsoon stimulates a pulse in biological productivity and also results in an ∼5°C drop in SST [Honjo et al., 1999]. A smaller pulse in export productivity is also observed during the winter monsoon [Honjo et al., 1999] when a reversal in the land-sea pressure gradient generates northeast winds which flow down across the northwestern Arabian Sea from the Tibetan Plateau. Thermocline waters in the Arabian Sea originate in the southern Indian Ocean, where winter cooling of surface waters causes the formation and subduction of subantarctic mode waters [Harper, 2000]. Modern SST at Site 722 ranges between ∼24 and 29.5°C, with an annual mean of ∼27°C [Levitus and Boyer, 1994].

2.2. A Tropical Nonupwelling Site (ODP 1146)

[9] ODP Site 1146 (19°27′N, 116°16′E, 2091 m water depth) is located in the South China Sea (Figure 1). In the modern annual cycle, minimum SSTs occur during the Asian winter monsoon, as cold northwest winds off the Asian continent drive sensible heat loss. A shallow nitricline and deep vertical mixing also result in enhanced productivity during the winter months [Chen, 2005]. Modern SST at Site 1146 ranges between ∼25.5 and 29.5°C, with an annual mean of ∼28°C [Levitus and Boyer, 1994].

3. Methods

[10] Alkenone paleothermometry utilizes the temperature dependence of alkenone synthesis by marine-dwelling coccolithophorid algae to reconstruct past sea surface temperatures [Brassell et al., 1986; Prahl et al., 1988]. Alkenones were extracted from freeze-dried sediments using an Accelerated Solvent Extractor and were analyzed by gas chromatography, following procedures described in Herbert et al. [1998]. We correlated the alkenone unsaturation ratio (Uk37) to SST using the calibration of Prahl et al. [1988]. Analytical error is ±0.007 Uk37 units, or the equivalent of ±0.20°C. We also measured the total concentration of alkenones in each sample, [C37] total (mass/grams dry sediment), by comparison with an internal reference standard. We estimate a relative error of ±10% for [C37] total measurements, on the basis of replicate extractions of a laboratory standard. Alkenone abundance values provide a qualitative indicator of coccolith productivity, which has been shown to scale with total organic carbon export [Brassell, 1993; Budziak et al., 2000; Prahl et al., 1988; Schubert et al., 1998].

[11] To establish age control for Site 662, we extended an existing benthic δ18O record [Lisiecki and Raymo, 2005] by adding 174 new benthic δ18O values and correlating the full record to the LR04 stack [Lisiecki and Raymo, 2005] as shown in Figure 2. We generated oxygen isotope values for Cibicidoides spp. and Uvigerina spp. specimens (typically 2–3 individuals per analysis), which were hand-picked from the >150 μm fraction of wet-sieved bulk sediment. Oxygen isotope measurements, reported as per mil deviations from Vienna Pee Dee Belemnite isotopic values, were made on a Finnegan MAT 252 mass spectrometer. Dry lab standards (n = 34) were reproducible to ±0.03‰ for δ18O (long term lab reproducibility (1σ) = ±0.06‰), and replicate sample analyses had a reproducibility of ±0.07‰ (average half range, n = 23). The resulting oxygen isotope record is composed primarily of C. wuellerstorfi values, which have been adjusted by 0.64‰ to correct for oxygen isotope fractionation differences between Cibicidoides spp. and Uvigerina spp. [Shackleton, 1974]. For Site 722, age control follows the published age model of Clemens et al. [1996], with slight modifications made here on the basis of realignment of some features of the Site 722 benthic isotope record with the LR04 global benthic stack [Lisiecki and Raymo, 2005]. Age control for Site 1146 is based on benthic oxygen isotope stratigraphy [Clemens and Prell, 2003]. For Site 846, we use the age model of Lisiecki and Raymo [2005], which was generated by matching the Site 846 benthic oxygen isotope record [Mix et al., 1995; Shackleton et al., 1995] to the LR04 stack.

Figure 2.

Correlation between Site 662 benthic δ18O and the LR04 stack [Lisiecki and Raymo, 2005]. Below meter level 117.25 (∼1.8 Ma), oxygen isotope values for the Site 662 are from Lisiecki and Raymo [2005]. Values above this meter level are from this study. The average sedimentation rate implied by this age model is 6.7 ± 3.6 cm/kyr.

[12] Spectral and cross-spectral analyses were performed using the Arand software package [Howell, 2001]. All SST and productivity records were interpolated at an interval of 2 kyr for these analyses, following original sampling resolutions of 2.2 ± 1.7 kyr for Site 662, 2.0 ± 0.9 kyr for Site 722, and 1.7 ± 0.7 kyr for Site 1146 (corresponding to average sedimentation rates of 6.7 ± 3.6 cm/kyr, 2.9 ± 1.0 cm/kyr and 6.5 ± 1.4 cm/kyr, respectively). Because alkenone paleotemperature and paleoproductivity estimates can be directly compared to benthic δ18O values measured in the same sediments at each site, the relative lead/lag relationships we report are not highly sensitive to absolute errors in the isotope-based age models.

4. Results

4.1. Uk37–SST Relationship

[13] Haptophyte productivity is highly seasonal in the monsoonal and upwelling-dominated regimes of the EEAtl, Arabian Sea, and South China Sea, suggesting the possibility that the sedimentary Uk37 record may be biased toward a seasonal temperature signal, rather than representing mean annual temperature (MAT). However, core-top studies from all three study regions have shown that the correlation between Uk37 and MAT values is high and not statistically different from a regression against productivity-weighted temperature values [Müller et al., 1998; Pelejero and Grimalt, 1997; Sonzogni et al., 1997]. On the basis of these data, we interpret our alkenone-SST estimates as a representation of MAT variation. This suggests either that sedimentary alkenones integrate a mean annual SST signal, despite seasonal variations in coccolithophorid productivity, or that maximum alkenone production in the seasonal cycle coincides fortuitously with temperatures similar to the annual mean.

4.2. Early Pleistocene SST and Productivity Reconstructions: Eastern Equatorial Atlantic

[14] At Site 662 in the EEAtl, early Pleistocene SSTs range between 23.5 and 27.5°C, with a mean value of 25.7°C (Figure 3). These values are quite similar to the late Pleistocene alkenone SST estimates generated from a nearby core, which range between ∼23.0 and 27.3°C [Schneider et al., 1996]. Alkenone C37 concentrations at Site 662 range between 0.1 and 37.6 nmol/g, and average 3.6 nmol/g (Figure 3). Strong correlation (r = 0.76) between [C37] total and an opal record from Site 662 [Ruddiman and Janecek, 1989] (not shown) suggests that our [C37] total record represents variations in total productivity. Cool SSTs and high alkenone abundance are positively correlated (r = 0.58). Because of the non-Gaussian distribution of [C37] total values, we normalize the [C37] total distribution for spectral analysis by taking its natural logarithm.

Figure 3.

Early Pleistocene alkenone SST and [C37] total records for Site 662 in the eastern equatorial Atlantic show very regular cyclicity. Gray shading indicates the strong correlation between SST and [C37] total, with SST minima occurring during inferred productivity maxima and vice versa.

[15] Early Pleistocene records of both SST and ln[C37] total show strong concentrations of spectral power in the obliquity and precession bands (Figure 4a). The tight coupling observed between cool SSTs and increased productivity in the time domain is also evident from cross-spectral analysis, which indicates significant coherence (>95% confidence) between the records in both the 41 kyr and 23 kyr bands (Figure 4a). Cross-spectral analysis between SST and benthic δ18O from Site 622 also shows high coherence (>95% confidence) between the two records in the 41 kyr and 23 kyr bands (Figure 4b).

Figure 4.

Cross-spectral analyses between SST, alkenone abundance, and benthic δ18O at Site 622. (a) SST and alkenone abundance are strongly coherent in the obliquity and precession bands. (b) SST and benthic δ18O are also coherent at periods of 41 and 23 kyr.

4.3. Early Pleistocene SST and Productivity Reconstructions: Arabian Sea

[16] Early Pleistocene alkenone SST estimates for Site 722 in the Arabian Sea vary between 24.5 and 27.8°C and average 26.5°C (Figure 5). Early Pleistocene minimum and maximum SST values are only slightly higher than late Pleistocene alkenone SST minima and maxima (∼23 and 27°C) reported for two nearby cores from the Oman Margin [Emeis et al., 1995; Rostek et al., 1997]. Alkenone C37 concentrations at Site 722 range between 1.3 and 43.4 nmol/g, with a mean value of 12.2 nmol/g (Figure 5). Comparison between our [C37] total record and an opal percentage record from the same site [Clemens et al., 1996] shows general agreement between the two proxies, although the higher-resolution alkenone record captures additional variation that is perhaps aliased in the percent opal record (Figure 6). The similarity between alkenone and opal concentrations suggests that both records reliably capture overall productivity variations, despite the fact that they originate from carbonate and silica production, respectively. Notably, there is not a clear correlation between cooler SST and inferred productivity highs, suggesting that water temperature is not directly coupled to productivity levels (Figure 5).

Figure 5.

Early Pleistocene alkenone SST for Site 722 in the Arabian Sea is dominated by 41 kyr cyclicity and is not tightly coupled to [C37] total; note inferred productivity maxima (and minima) which occur during both SST highs and SST lows (gray shading).

Figure 6.

Comparison between alkenone abundance (this work) and percent opal [Clemens et al., 1996] records for Site 722. Good correspondence between the two records suggests that the [C37] total record reliably captures changes in total export productivity at this site.

[17] Early Pleistocene SST variability at Site 722 is dominated by 41 kyr cyclicity (Figure 7a). Additional spectral power is observed near periods of 23 and 19 kyr, and at the nonorbital periods of 62 and 25 kyr. Spectral analysis of the [C37] total record from Site 722 reveals a distinct lack of spectral power in the obliquity band (Figure 7a). Spectral peaks at around 21 and 18 kyr could reflect precessional forcing, while dominant spectral peaks have periods of 79 and 28 kyr. Cross-spectral analysis indicates significant coherence between SST and [C37] total in the precession band, though maximum coherence is not well centered at the 23 kyr frequency (Figure 7a). Cross-spectral analysis between benthic δ18O from Site 722 [Clemens et al., 1996] and our SST record indicates strong coherence (>95% confidence) between ice volume and SST in the obliquity band, as well as a degree of coherence (>80% confidence) in the precession band (Figure 7b).

Figure 7.

Cross-spectral analyses of SST, [C37] total, and benthic δ18O at Site 722 in the Arabian Sea. (a) SST variability is strongly concentrated in the 41 kyr band, with additional variance at precessional frequencies. Alkenone abundance contains spectral power at precessional frequencies, but lacks 41 kyr power. Coherence between SST and alkenone abundance is significant at precessional frequencies, though not well centered on the 23 kyr band. (b) SST is coherent with benthic δ18O in both the obliquity and precession bands.

4.4. Early Pleistocene SST and Productivity Reconstructions: South China Sea

[18] For Site 1146 in the South China Sea, early Pleistocene SSTs range between 23.7 and 27.6°C, with an average of 26.1°C, while alkenone abundance varies between 0.05 and 2.00 nmol/g, with an average of 0.35 nmol/g (Figure 8). In the spectral domain, SSTs vary strongly in the obliquity band, and also show a significant concentration of spectral power in the precession band (Figure 9a). As observed for the Arabian Sea, alkenone abundance in the South China Sea lacks 41 kyr variability, but does contain spectral power in the precession band, which is coherent with SST variation at the 80% confidence level (Figure 9a). Site 1146 SST and benthic δ18O [Clemens and Prell, 2003] are also coherent in both the obliquity (>95% confidence) and precession (>80% confidence) bands (Figure 9b).

Figure 8.

Early Pleistocene alkenone SST estimates and [C37] total record for Site 1146 in the South China Sea. SST variations show pronounced 41 kyr cyclicity, which is not observed in the alkenone abundance record.

Figure 9.

Cross-spectral analyses of SST, [C37] total, and benthic δ18O at Site 1146 in the South China Sea. (a) SST is dominated by 41 kyr variability, which is lacking in the [C37] total record. Precessional variability in SST and [C37] total are coherent at the 80% confidence interval. (b) SST and benthic δ18O are coherent in both the obliquity and the precession bands.

4.5. Comparison With the Eastern Equatorial Pacific

[19] Comparison between early Pleistocene SST records from Site 662, Site 722, Site 1146, and Site 846 in the EEP [Liu and Herbert, 2004] reveals marked similarities between all four tropical SST records, despite considerable differences in local dynamics in the modern world. In the time domain, SST patterns in the Arabian Sea, EEP, and (to a lesser degree) the South China Sea are similar not only in the shapes and relative amplitudes of individual SST cycles, but also in long-term (∼400 kyr) cyclicity that gives rise to periods of increased interglacial SST values around 1.5 and 1.95 Ma and pronounced SST minima at ∼1.25 and 1.65 Ma (Figure 10). Some of these similarities are also apparent in the EEAtl record, which additionally contains an interval of decreased SST variability between 1.8 and 1.9 Ma that is also common to the EEP record. Alkenone abundance records are compared in Figure 11. Though sharing some broad similarities, inferred productivity records from all four sites show a large degree of independent variation.

Figure 10.

Comparison of our early Pleistocene alkenone SST records from Sites 1146, 722, and 662 with an alkenone SST record from the eastern equatorial Pacific [Liu and Herbert, 2004] and with a global benthic oxygen isotope stack [Lisiecki and Raymo, 2005]. All four tropical SST records are dominated by 41 kyr cyclicity, which is coherent and nearly in phase in each of the locations. In particular, the structure and relative amplitudes of individual SST cycles at Sites 1146, 722, and 846 are remarkably similar. All four tropical SST records also share similar long-wavelength (∼400 kyr) cyclicity. Comparison between the SST records and the oxygen isotope record shows that coolest SSTs occur during ice volume maxima and vice versa.

Figure 11.

Comparison between alkenone abundance records from Sites 1146, 722, 662, and 846 [Liu and Herbert, 2004]. Inferred productivity at Sites 662 and 846 is coherent and similarly phased in the obliquity band. While sharing some broadly similar features, all four records contain a significant amount of independent variation.

[20] In the frequency domain, all four SST records are mutually coherent above the 95% confidence interval in the obliquity band. Notably, 41 kyr variability at all four locations is nearly in phase (relative to the common benthic δ18O signal), with small differences more likely due to age model uncertainties than to real differences in timing (Figure 12a). Our tropical SST records lead the benthic oxygen isotope record by an average of 19° (2 kyr) in the obliquity band. Inferred productivity maxima at Sites 662 and 846 are also coherent (>95% confidence) and closely phased in the obliquity band (Figure 12a); such a coordinated glacial-interglacial response between EEAtl and EEP productivity has also been observed for the late Pleistocene [Lyle, 1988].

Figure 12.

Phase relationships for the early Pleistocene records discussed here, relative to insolation in the obliquity (a) and precession (b) bands. The similar phasing of the SST responses at Sites 662, 846, 722, and 1146 in the obliquity band contrasts sharply with the nonuniform responses in the precession band. Phase relationships for SST are calculated relative to benthic δ18O records for each site; phase relationships for [C37] total are calculated relative to SST. For ice volume, phases are calculated for the LR04 stack from 1.2–2.0 Ma [Lisiecki and Raymo, 2005] relative to insolation over this interval [Laskar et al., 2004]. SST and [C37] total records from Site 846 are not coherent with ice volume in the precession band and are not depicted on the phase wheel.

[21] In contrast to the close phasing observed between our SST and [C37] total records from different sites in the obliquity band, none of the SST or productivity records are mutually coherent in the precession band, and they do not show uniform phasing with respect to ice volume (Figure 12b). At this frequency, maximum SST leads minimum ice volume at Sites 722 and 662, while SST maxima significantly lag ice volume minima at Site 1146. Though similarly phased, [C37] total records at Sites 722, 662, and 1146 are not coherent (cross-spectra not shown). At Site 846, neither SST nor [C37] total is coherent with ice volume above the 80% confidence interval in the precession band.

5. Discussion

[22] Our results confirm that the strong 41 kyr SST and productivity signals observed in the early Pleistocene at Site 846 [Liu and Herbert, 2004] are not unique to the eastern equatorial Pacific but instead are a widespread feature of the tropical oceans. The data we present here illustrate that the phrase “41 kyr world” applies to more than ice volume and other aspects of high-latitude climate, and join a growing body of evidence demonstrating that the 41 kyr beat was a pervasive feature of tropical climate as well. Significantly, the SST response at all of these tropical locations is nearly 180° out of phase with obliquity band insolation forcing in the tropics, implying a remote forcing mechanism originating in the high latitudes. Because early Pleistocene tropical SSTs led the benthic δ18O record by a few thousand years in the obliquity band, this signal cannot be initiated directly by processes related to the ice sheets themselves and therefore require additional high-latitude processes capable of being exported to the tropics. At the same time, the lack of a coordinated tropical SST response in the precession band implies forcing by local dynamics. Below we discuss potential forcing mechanisms for first the obliquity and then the precession band SST and inferred productivity responses we observe in the EEAtl, EEP, Arabian Sea, and South China Sea, and their implications for early Pleistocene climate dynamics.

5.1. Obliquity Band Responses

5.1.1. Insolation Forcing

[23] The dominant 41 kyr SST variability we observe in the EEAtl, the EEP, the Arabian Sea, and the South China Sea cannot have a local origin in low-latitude insolation forcing, for two reasons. First, low-latitude seasonal insolation spectra are dominated by precession, with less than 1% of total insolation variance contained in the obliquity frequency band. Secondly, variations in annual insolation at low latitudes are minor (accounting for less than 1% of the total insolation budget), and are furthermore antiphased to our SST measurements (i.e., warm SSTs occur during minima in the local mean annual insolation curve, and vice versa). Instead, the 41 kyr pattern in our early Pleistocene tropical SST records very closely tracks high-latitude glacial cycles, as displayed in the time domain in Figure 10 and as revealed in the spectral domain by the strong coherence in the obliquity band between SST and benthic δ18O in all of our tropical records (Figures 4b, 7b, and 9b). This close relationship between tropical SST changes and high-latitude processes during the early Pleistocene requires strong physical feedbacks linking tropical and high-latitude climate dynamics, as previously discussed for the eastern equatorial Pacific by Liu and Herbert [2004].

5.1.2. Thermocline Ventilation

[24] One mechanism by which a high-latitude obliquity signal could be transported to the tropics is through the ventilation of the thermocline. According to the model proposed by Philander and Fedorov [2003], heat loss in the high latitudes during periods of low axial tilt is balanced by heat gain in low-latitude upwelling zones, which is achieved through a shoaling of the tropical thermocline. This implies a direct link between the high latitudes and the upwelling zones of the EEAtl, EEP, and Arabian Sea, which are primarily sourced by mode waters formed in the subantarctic region [Harper, 2000]. Because seasonal temperature fluctuations are rapid relative to the timescale of thermocline circulation and do not carry a net signal when integrated over a year, thermocline ventilation should not vary in the precession band [Philander and Fedorov, 2003], a hypothesis consistent with the lack of coherence we observe in our early Pleistocene tropical SST records at this frequency.

[25] While high-latitude obliquity forcing may play an important role in determining the temperature of subsurface source waters in the upwelling regions of the EEAtl, EEP, and Arabian Sea, this mechanism alone cannot account for all of the 41 kyr SST variability we observe in the tropics, for several reasons. First, a recent modeling study concluded that although high-latitude insolation forcing of subantarctic mode water temperatures over the course of an obliquity cycle results in temperature differences in EEP subsurface waters, much of this temperature change is mitigated at the surface by the opposing effect of local insolation [Lee and Poulsen, 2005]. According to this study, obliquity-driven changes in high-latitude mode water temperatures result in a difference of about 0.6°C in EEP surface waters between conditions of high and low obliquity, which is considerably smaller than the 1–4°C amplitude of temperature change we observe in the EEAtl, EEP, and Arabian Sea in the obliquity band. Additionally, high-latitude insolation forcing of mode water temperatures cannot account for the significant phase lag of several thousand years between high-latitude insolation maxima and tropical SST maxima. Most importantly, thermocline ventilation cannot explain the dominant 41 kyr SST signal we observe in the nonupwelling regions of the South China Sea (this study) and the western Pacific warm pool [de Garidel-Thoron et al., 2005; Medina-Elizalde and Lea, 2005], where the greater depth of the thermocline prevents a direct connection between surface ocean temperatures and high-latitude mode waters. Thus the amplitude, phasing, and ubiquity of the 41 kyr tropical SST signal all require forcing mechanisms additional to thermocline ventilation.

5.1.3. Lateral Heat Transport

[26] In the case of Site 662 in the EEAtl, we note that unique controls on subsurface circulation may have played a role in determining SSTs at Site 662, through variations in the entrainment of warm Agulhas waters into Atlantic subsurface waters [McIntyre et al., 1989; Schefuß et al., 2004]. The north-south migration of the subtropical convergence zone at 100, 41, and 23 kyr periodicities during the late Pleistocene [Howard and Prell, 1992; Labeyrie et al., 1996] has been associated with variations in Agulhas leakage at the same frequencies [Peeters et al., 2004]. The impact of subantarctic front migration on Agulhas leakage during the early Pleistocene, when ice sheets were smaller, is unknown. Our data neither provide evidence for nor rule out sensitivity of Agulhas leakage to orbital change during the early Pleistocene, which may be better monitored at other locations (i.e., in the South Atlantic, closer to the source of the original signal) and/or using other means, such as faunal assemblage data.

[27] Heat advection may also play a role in determining SSTs in the South China Sea. During the late Pleistocene, South China Sea SSTs were strongly influenced by sea level changes, which controlled the inflow of warm Indian Ocean surface waters through the shallow Sunda Shelf (sill depth 30–50 m) [Wang et al., 1999; Wang and Wang, 1990; Zhao et al., 2006]. If this were also the case for the smaller glaciations of the early Pleistocene, we would expect South China Sea SST to occur in phase with or slightly lag the δ18O record. While South China Sea SST instead shows a slight lead over δ18O, this lead is not statistically significant and does not preclude sea level changes as a control on Site 1146 SST.

5.1.4. Upwelling Intensity

[28] At Site 722 in the Arabian Sea, the phasing of SST maxima precludes monsoon-driven upwelling as the major control on SST in the 41 kyr band, since monsoon maxima were closely phased with obliquity maxima during the early Pleistocene [Clemens et al., 1996] (Figure 12a). If upwelling intensity directly drove SST changes at Site 722, we would expect SST minima to slightly lag obliquity maxima, and would also anticipate a strong correlation between SST and inferred productivity in the obliquity band, neither of which we observe. Such a decoupling between Arabian Sea SST and monsoon strength has also been observed for the late Pleistocene [Emeis et al., 1995; Rostek et al., 1997].

[29] Given the similarity of SST changes in the trade wind–dominated upwelling systems of the EEAtl and the EEP and the monsoon-driven regions of the Arabian and South China seas, it also seems unlikely that changes in upwelling strength directly drove the tropical SST variability we observe in the EEAtl and EEP. While the strong 41 kyr inferred productivity signals in the EEAtl and EEP are likely related to changes in upwelling intensity (discussed below), the similar temperature responses observed in the EEAtl, EEP, Arabian Sea, and South China Sea rule out changes in upwelling strength as a dominant control on SST in these locations.

5.1.5. Greenhouse Gases

[30] The dominance and near synchrony of the 41 kyr SST signal across the tropics together imply a coordinated tropical response that cannot readily be explained by local insolation, ventilation of the thermocline, lateral heat transport, or changes in upwelling intensity. These mechanisms fail to fully account not only for the magnitude, phasing, and high degree of similarity we observe in SST records from the EEAtl, EEP, Arabian Sea, and South China Sea, but also for the dominant 41 kyr SST variability observed in the western Pacific warm pool [de Garidel-Thoron et al., 2005; Medina-Elizalde and Lea, 2005], which varies in phase with our tropical SST records. Together these records imply a unified tropical response to high-latitude processes during the early Pleistocene, despite the smaller amplitude of glacial-interglacial cycles during this interval.

[31] Because the low-latitude SST signal significantly leads the benthic δ18O record, high-latitude forcing of tropical climate could not have originated solely from feedbacks related directly to the early Pleistocene ice sheets themselves, such as ice albedo or changes in atmospheric circulation patterns. We therefore favor the involvement of additional high-latitude processes, and suggest that biogeochemical responses driving atmospheric greenhouse gas concentrations [c.f. Archer et al., 2000; Sarmiento and Toggweiler, 1984; Sigman and Boyle, 2000] played a dominant role in linking the high and low latitudes during the early Pleistocene. The relationship we observe between tropical SST and early Pleistocene high-latitude glacial cycles is analogous to the tropical Pacific SST lead over the 100 kyr glacial cycles of the late Pleistocene, which has been attributed to greenhouse gas forcing [Lea, 2004]. We propose that greenhouse gas forcing also acted as a major driver of tropical SST variability during the early Pleistocene, coordinating the coherent, synchronous, temperature responses we observe at the 41 kyr frequency at locations across the tropics. This would imply that the glacial-interglacial greenhouse gas feedback observed for the 100 kyr glacial cycles of the late Pleistocene [Petit et al., 1999] was also an important factor in the smaller amplitude glacial cycles of the 41 kyr world [Medina-Elizalde and Lea, 2005].

5.1.6. Productivity Variability

[32] Because the EEAtl, EEP, and Arabian Sea upwelling zones are all sourced by mode waters formed in the subantarctic, we may expect to observe coherent, synchronous productivity responses in these locations if productivity variability were forced by high-latitude controls on the subsurface nutrient budget. While the EEAtl and EEP alkenone abundance records are coherent in the obliquity band, our Arabian Sea inferred productivity record distinctly lacks 41 kyr cyclicity. This decoupling between productivity records suggests localized atmospheric controls on upwelling intensity and/or iron fertilization, rather than global nutrient cycling, as more likely controls on productivity variability.

[33] In the EEAtl and EEP the tight correlation between increased productivity and cooler SSTs in the obliquity band suggests that changes in upwelling strength drove the productivity variations and implies significant trade wind variability at the 41 kyr rhythm during the early Pleistocene. Potentially, enhanced trade winds contributed to productivity not only by increasing upwelling strength but also by delivering greater quantities of iron from continental sources areas to the EEAtl and EEP [Martin et al., 1994]. The phasing of productivity maxima between obliquity minima and ice volume maxima (Figure 12a) suggests that the meridional temperature and pressure gradients driving the tropical trade winds [Rind, 1998] were sensitive to both insolation forcing and high-latitude glacial cycles during the early Pleistocene. Interestingly, productivity maxima in the EEAtl are phased more closely to ice volume maxima than in the EEP, which may suggest a greater sensitivity of the Atlantic basin to early Pleistocene ice sheet dynamics. In contrast, productivity variations in the EEAtl and EEP are nearly in phase with 100 kyr ice volume cycles during the late Pleistocene [Beaufort et al., 2001; Lyle, 1988], perhaps indicating an evolution toward increased ice sheet influence on equatorial trade wind dynamics over the course of the Pleistocene.

5.2. Precession Band Responses

[34] The precession band SST and inferred productivity responses we observe in the EEAtl, Arabian Sea, and South China Sea can be attributed to local atmospheric dynamics at each of the locations. In the EEAtl, strengthened monsoons decrease equatorial divergence by diverting the trade winds northward and reducing their zonal component. Modeling studies and proxy data from the late Pleistocene both suggest that African monsoon maxima occur nearly in phase with Northern Hemisphere summer insolation maxima [deMenocal et al., 1993; McIntyre et al., 1989; Prell and Kutzbach, 1987]. Assuming that this phase relationship also applies to the early Pleistocene, the phasing of the EEAtl SST and inferred productivity responses in the precession band is consistent with forcing by monsoon modulation of upwelling strength, perhaps with some additional sensitivity to glacial-interglacial states, which would account for the later phasing of these signals.

[35] In the Arabian Sea, monsoon winds drive upwelling strength, such that monsoon maxima are associated with SST minima and productivity maxima over an annual cycle. The relationship between SST minima and inferred productivity maxima persists in our Arabian Sea records in the precession band, suggesting that the modern annual cycle provides a good analog for orbital-scale changes at this frequency. Our data show that minimum SST and maximum productivity occurred ∼180° out of phase with Northern Hemisphere summer insolation forcing during the early Pleistocene, which is consistent with the phase estimate for Asian monsoon maxima over this interval [Clemens et al., 1996].

[36] In the South China Sea, early Pleistocene 23 kyr SST variability appears to be sensitive to both ice volume and winter monsoon strength. SST maxima are phased between ice volume minima and Northern Hemisphere winter insolation maxima (Figure 12b), both of which should contribute to a weaker winter monsoon and warmer SSTs in the South China Sea [Wang et al., 1999]. The productivity signal, however, remains a puzzle. Productivity maxima are phased between Northern Hemisphere winter insolation maxima and ice volume maxima, which is opposite to the expectation of increased productivity during periods of enhanced winter monsoon strength (i.e., Northern Hemisphere insolation minima). Precession-driven productivity dynamics for the early Pleistocene South China Sea thus may bear further investigation.

[37] The coherent, near in-phase SST responses at all four sites in the obliquity band contrast sharply with the noncoherent, out-of-phase SST responses in the precession band (Figure 12b). While it is possible that greenhouse gas forcing in the precession band existed but was overridden by more dominant local controls on SST, the noncoherence of our tropical SST responses, combined with the near absence of precessional variability in the benthic oxygen isotope record, may suggest a lack of coherent global climate variation in the precession band during the early Pleistocene.

6. Conclusions

[38] We find that early Pleistocene SST records from the eastern equatorial Atlantic, the Arabian Sea, and the South China Sea (this study), in addition to the eastern equatorial Pacific [Liu and Herbert, 2004] are nearly in phase and universally dominated by 41 kyr variability. We attribute this strong 41 kyr signal to greenhouse gas forcing, which implies glacial-interglacial greenhouse gas feedbacks during the early Pleistocene 41 kyr world that were analogous to the 100 kyr greenhouse gas cycles of the late Pleistocene. High-latitude controls on thermocline ventilation likely also contributed to the 41 kyr SST signals observed in the upwelling regions of the EEAtl, the EEP, and the Arabian Sea, which are connected to the high latitudes via subsurface circulation. We also find evidence for 41 kyr variability in tropical trade wind strength during the early Pleistocene, which was sensitive to both high-latitude insolation and ice volume. In the precession band, heterogeneous SST responses imply forcing by regional atmospheric dynamics related to the African and Asian monsoon systems.

[39] Our data show that the high and low latitudes were tightly linked during the early Pleistocene, despite the smaller size of ice sheets and the reduced amplitude of glacial-interglacial cycles. The apparent connectivity between high- and low-latitude climates during the early Pleistocene implies that many of the internal climate feedbacks associated with the 100 kyr glacial cycles of the late Pleistocene also operated during the early Pleistocene, and that the shift from 41 to 100 kyr glacial cycles was not associated with a major revolution in low-latitude climate dynamics. At the same time, the relative weakness of the precessional response in our tropical SST and productivity records raises a puzzle for the early Pleistocene world. While the smaller early Pleistocene ice volume variations should have enhanced the impact of local precessional control on tropical dynamics, this expectation is directly contradicted by our records, which are clearly dominated by 41 kyr power. The relationship between the tropical response to precession forcing, the late Pleistocene emergence of precessional power in the benthic δ18O record, and the evolution of precession band greenhouse gas variability may be a key mystery to unravel as we advance our understanding of ice age climate dynamics.

Acknowledgments

[40] Samples for this study were provided by the International Ocean Drilling Program. Funding was provided by NSF grants OCE 0351599 to T.D.H. and OCE 0352215 to Steven Clemens and Warren Prell. We thank Steve Clemens, Zhonghui Liu, and Kira Lawrence for valuable discussions on this manuscript and Mitch Lyle and two anonymous reviewers for providing helpful reviews.

Ancillary