A 12-Million-Year Temperature History of the Tropical Pacific Ocean

See allHide authors and affiliations

Science  04 Apr 2014:
Vol. 344, Issue 6179, pp. 84-87
DOI: 10.1126/science.1246172

Old Gradients

The surface ocean temperature gradient between the warmer Western Equatorial Pacific and the cooler Eastern Equatorial Pacific is smaller during El Niño episodes than during neutral periods or during La Niñas. Some reconstructions of Pacific Ocean sea surface temperatures (SST) covering periods before ∼3 million years ago have suggested a permanent El Niño–like state. Zhang et al. (p. 84; see the Perspective by Lea) present data from a biomarker-derived proxy for SST that indicate a sizable east-west gradient has existed for the past 12 million years, contradicting the concept of a permanent El Niño–like state existed.


The appearance of permanent El Niño–like conditions prior to 3 million years ago is founded on sea-surface temperature (SST) reconstructions that show invariant Pacific warm pool temperatures and negligible equatorial zonal temperature gradients. However, only a few SST records are available, and these are potentially compromised by changes in seawater chemistry, diagenesis, and calibration limitations. For this study, we establish new biomarker-SST records and show that the Pacific warm pool was ~4°C warmer 12 million years ago. Both the warm pool and cold tongue slowly cooled toward modern conditions while maintaining a zonal temperature gradient of ~3°C in the late Miocene, which increased during the Plio-Pleistocene. Our results contrast with previous temperature reconstructions that support the supposition of a permanent El Niño–like state.

Temperatures of the low-latitude Pacific Ocean substantially influence regional and global climates. In particular, the El Niño–Southern Oscillation (ENSO) dominates Earth’s interannual climate variability. El Niño is initiated by the eastward propagation of warm western Pacific equatorial waters that deepen thermocline depths and attenuate upwelling rates across the eastern equatorial Pacific (EEP) (1, 2). Regional sea-surface temperature (SST) change associated with El Niño affects the position and vigor of the Walker circulation and Hadley cell, leading to substantially reduced equatorial SST gradients, higher global mean temperatures, extratropical heat export, and regional hydrological impacts (1, 2).

The western equatorial Pacific warm pool is among the warmest surface water on Earth, whereas the EEP, known as the cold tongue, is characterized by cold, nutrient-rich waters that result from a shallow thermocline and intense upwelling rates. The modern SST gradient between the warm pool and cold tongue averages 4° to 5°C (3) and varies in response to ENSO. Accordingly, paleo-SST reconstructions from these regions and the resulting character of the east-west temperature gradient have been used to describe broad climate states over the past 5 million years, particularly during the Pliocene epoch [5.3 to 2.6 million years ago (Ma)], when atmospheric CO2 concentrations appear similar to today (4, 5) and global temperatures are simulated to be 3° to 4°C warmer than preindustrial conditions (6, 7).

Published temperature records based on magnesium-to-calcium ratios (Mg/Ca) of the planktonic foraminifera Globoritalia sacculifer, from Ocean Drilling Program (ODP) site 806 (0°N, 159°E) (Fig. 1) (8), suggest that warm pool temperatures remained relatively constant as Earth cooled over the past 5 million years. Curiously, existing Mg/Ca-based SST records also indicate that the Pacific warm pool of the Pliocene warm period (4.5 to 3 Ma) was ~0.5°C colder than the mean temperature of the late Quaternary (1.5 to 0 Ma), implying that warm pool temperatures were cooler or invariant during periods of global warmth. The appearance of invariant tropical temperatures during periods of global warmth implies a “tropical thermostat” (9) in which tropical warming is limited through evaporation or cloud feedbacks irrespective of the greenhouse gas forcing. However, this phenomenon is not supported by climate simulations (10), and much warmer tropical SSTs are evident during “super greenhouse” climate states in Earth history (e.g., the Cretaceous) (11).

Fig. 1 Pacific sites discussed in this study.

Colors represent modern mean annual SSTs. Circles represent sites analyzed in this study: ODP sites 769, 806, 850, and 1143. Triangles represent sites with previously published SSTs: ODP sites 846, 847, and 1208 (36°N, 128°E) and Integrated Ocean Drilling Program (IODP) site U1338 (3°N, 118°W). Map generated by Ocean Data View.

In contrast to the western Pacific warm pool, temperature records from the eastern Pacific sites 847 (0°N, 95°W; based on Mg/Ca and the alkenone unsaturation index Embedded Image) (8, 12) and 846 (3°S, 91°W; based on Embedded Image) (13) exhibit cooling for the past 5 million years. Consequently, warm pool and cold tongue SSTs result in a negligible equatorial temperature gradient during the early Pliocene (8, 14) similar to those expressed during brief modern El Niño events (8, 14). A nearly absent equatorial temperature gradient is argued to reflect a permanent El Niño–like state prior to 3 million years ago and has been further interpreted to reflect the establishment of a deep thermocline in the EEP, substantially reduced cold-water upwelling (14), an attenuated Walker circulation (7), and negligible ENSO variability (14). In contrast, high-resolution, coral δ18O records (15) and isotope records of individual planktonic foraminifera from site 846 (16) indicate that the frequency and intensity of Pliocene ENSO-like oscillations were similar to those of today.

The veracity of the permanent El Niño supposition rests on the accuracy of paleotemperature reconstructions. In particular, G. sacculifer Mg/Ca-based SST records used to infer invariant warm pool SSTs assume that foraminifera Mg/Ca compositions are uninfluenced by diagenetic alteration or temporal changes in seawater chemistry [e.g., seawater Mg/Ca ratio (Mg/Casw)] and/or carbonate ion effect; see supplementary materials). However, δ18O records on the identical samples (8) fail to capture the modern SST gradient (supplementary materials) and suggest the contribution of diagenetic carbonates. In addition, changes in Mg/Casw could have lowered Pliocene SST estimates because fluid inclusion (17) and carbonate veins (18) data indicate lower Mg/Casw during the Pliocene. Notably, other warm pool temperature records derived from Embedded Image values (19) are limited by the nature of the temperature proxy itself, which reaches a maximum calculable SST value of ~28.5°C (20) and thus compromises its capacity to record even higher SSTs (supplementary materials).

In summary, given possible diagenetic overprinting, seawater chemistry changes, and calibration limitations, available SST records likely underestimate maximum warm pool temperatures and potentially distort zonal SST gradient estimates.

For this study, we established new equatorial Pacific SST records using the TEX86 temperature proxy. TEX86 thermometry is founded on the distribution of the archaeal lipid membranes, glycerol dialkyl glycerol tetraethers, primarily produced by archaea from the phylum Thaumarchaeota. Thaumarchaeota are aerobic ammonia oxidizers (21), raising concerns that the TEX86 index can also reflect subsurface temperatures rather than a mixed-layer SST. However, available compound-specific carbon-isotope evidence indicates that the TEX86 temperature signal largely derives from the ocean mixed-layer (22). Using the current calibration [e.g., (23)], TEX86 tends to overestimate SSTs in the high latitudes and underestimate temperatures in the low latitudes (24). Indeed, TEX86 SSTs derived from surface sediments of the tropical Pacific show slightly lower temperature estimates than observed SSTs (see supplementary materials). Nevertheless, the late Pleistocene TEX86-derived zonal SST gradient agrees with the modern gradient between sites 806 and 850 (supplementary materials) and broadly captures the temperature characteristics of these distinct regions. Finally, TEX86 temperature reconstructions can record much higher SSTs compared to the alkenone proxy (25), making it a more suitable proxy to evaluate the evolution of warm pool temperatures.

We reconstruct the history of equatorial Pacific temperature gradients for the past 12 million years using both TEX86 and Embedded Image temperature proxies at ODP sites 769 (9°N, 121°E), 806, and 1143 (9°N, 113°E) in the western warm pool region and site 850 (1°N, 111°W) in the eastern Pacific cold tongue (Fig. 1). Our results illustrate a distinctly different thermal history of the Pacific warm pool compared to previously published records. For example, in contrast to relatively invariant Mg/Ca temperatures from site 806 (Fig. 2), TEX86 records from three localities near the center and edge of the modern warm pool unambiguously show ~4°C of cooling since the late Miocene (Fig. 3), with an ~2°C decline since the early Pliocene (Fig. 2). TEX86 values from the late Miocene to Pliocene also indicate that warm pool temperatures were almost always higher than the calculable limit of Embedded Image temperatures until ~3-5 million years ago. Consequently, Embedded Image temperatures cannot be applied to interpret the temperature history of the warm pool older than ~3 Ma (Figs. 2 and 3).

Fig. 2 SST reconstructions of the Pacific warm pool for the past 5 Ma.

Mg/Ca temperatures do not account for changes in seawater Mg/Ca (8). Dashed line denotes the maximum calculable temperature (~28.5°C) of the Embedded Image calibration. Calibration errors for Embedded Image (20) and TEX86 (23) are 1.1° and 2.5°C, respectively, but vary spatially (24). Warm pool temperature trend is calculated with all TEX86-derived SSTs and Embedded Image temperatures after 3 Ma. Embedded Image- SST records from site 806 (5 to 0 Ma) (4) and 1143 (4 to 0 Ma) (34) are previously published.

Fig. 3 Temperature evolution of the western and eastern equatorial Pacific since 12 Ma.

All published data [Embedded Image from sites 806 (5 to 0 Ma) (4); 846 (5 to 0 Ma) (13); 847 (5 to 0 Ma) (12); 1143 (4 to 0 Ma) (34); and U1338 (12 to 0 Ma) (26)] are converted to temperature with the calibration of Conte et al. (20) for Embedded Image and Kim et al. (23) for TEX86. Western equatorial Pacific (WEP) temperature trend is calculated with all TEX86 from sites 769, 806, and 1143. Embedded Image records from ODP sites 806 and 1143 are also applied for the past 3 Ma. Embedded Image data before 3 Ma, shown in gray, are mostly maximized and are not included in trend calculations. Temperature trend in the eastern equatorial Pacific (EEP) is calculated with Embedded Image values from sites 846, 847, 850, and U1338 and with TEX86 results from site 850.

Cold tongue Embedded Image temperatures from site 850 in the western portion of the eastern equatorial upwelling region (Fig. 1) compare well with published alkenone records from sites 846 (13) and 847 (12) over the past 5 million years (Fig. 3). Site U1338 shows comparatively warmer temperatures (26), consistent with its location at the edge of the upwelling region (Fig. 1). Both TEX86 and Embedded Image temperature reconstructions in the EEP region indicate that the cold tongue slowly cooled by ~6°C since 12 Ma, with higher cooling rates during the Pliocene and Pleistocene. TEX86 and Embedded Image temperature estimates begin to deviate at about 6 Ma, with TEX86 SSTs showing consistently cooler temperatures, readily explained by differential changes in the depth and/or seasonality of production between haptophyte algae (alkenone producers) and archaea (supplementary materials). If changes in production depth are responsible for the temperature offset between these two proxies, production of archaea and haptophytes is still within the top 50 m of the mixed layer given the sharp vertical temperature gradient in the EEP (supplementary materials). Although sites 850 and U1338 represent the warmer edge of the EEP, alkenone values are still below the maximum temperature limit for 12 million years (Fig. 3)—in contrast to Embedded Image temperatures in the western warm pool—and thus can record the cooling history of the EEP (Fig. 2 and supplementary materials).

We assess the evolution of the east-west equatorial Pacific temperature gradient (ΔTzonal) by interpolating and averaging temperature results using TEX86 and Embedded Image temperature records, and solely using TEX86 data (Fig. 4). Integrating all of the available SST data expands the spatial assessment for each region. However, because Embedded Image is mostly maximized in warm pool samples older than 3 Ma, warm pool temperature reconstructions largely rest on TEX86 SST records prior to 3 Ma (Figs. 2 and 3). In contrast, the composite temperature of the cold tongue is dominated by Embedded Image records (Figs. 2 and 3). At our localities, TEX86 values tend to yield lower temperature estimates than Embedded Image when both are available (Figs. 2 and 3 and supplementary materials). As a result, our SST trends from the integrated data set often yield lower average temperatures in the warm pool, higher temperatures in the cold tongue, and a smaller ΔTzonal relative to estimates that only use our TEX86 records (Fig. 4 and supplementary materials). For example, the Pliocene ΔTzonal shows a range of 3.2°C (total data set) to 4.1°C (based solely on TEX86 data; see supplementary materials).

Fig. 4 Zonal (ΔTzonal) and meridional (ΔTmeridional) temperature gradients of the Pacific Ocean for the past 12 million years.

Zonal gradient is computed with both Embedded Image and TEX86 data, and with TEX86 data only. Meridional gradient is calculated between sites 806 and 1208 (33). Fifty-point Savitzky-Golay smoothing has been applied to the raw data. Note the different scales for ΔTzonal and ΔTmeridional.

Averaging ΔTzonal data over the Quaternary (2.6 to 0 Ma), Pliocene (5.3 to 2.6 Ma), and late Miocene (12 to 5.3 Ma) indicates that about 60 to 70% of the Quaternary ΔTzonal is expressed in the late Miocene, whereas ~80% of the Quaternary ΔTzonal characterizes the Pliocene and thus closely reflects modern oceanographic conditions (supplementary materials). Notably, our records indicate that both the warm pool and the cold tongue were warmer in the past, but that a cold EEP (relative to the warm pool) was present. Thus, the oceanographic processes that produce the modern cold tongue, including a shallow EEP thermocline and active upwelling, were likely in play—consistent with high rates of biogenic opal accumulation in the EEP during the late Miocene and Pliocene (27).

A late Miocene–Pliocene climate state characterized by strong temperature asymmetry across the equatorial Pacific provides the necessary conditions for robust ENSO-type interannual climate variability but does not directly prove its existence. Regardless, our temperature reconstructions support other proxy records (15, 16) and climate simulations (28, 29) that indicate ENSO-like behavior during the Pliocene and beyond (30).

Today, waters of the Equatorial Undercurrent that source upwelled waters in the eastern equatorial Pacific derive from the extratropics and higher latitudes, including the Subantarctic mode water (31). Geochemical tracers, drifter experiments, and climate simulations (32) indicate that waters from the eastern, subtropical Pacific subduct and resurface in the EEP thermocline within two decades. Indeed, the linkage between the eastern equatorial Pacific and anomalously warm waters from higher latitudes was used to explain a period of prolonged warming of the EEP in the early 1990s (32). Similarly, during the late Miocene to Pliocene, warmer extratropical waters (7, 33) (Fig. 4) that sourced the EEP likely contributed a reduction in ΔTzonal. As meridional temperature gradients increased during global cooling, colder extratropical water contributed to EEP cooling and an increase in ΔTzonal (Fig. 4).

Given the areal extent of tropical warm pools and their importance in regulating global temperatures, our new warm pool temperature records substantially revise the character and nature of global warming in the recent past.

Supplementary Materials

Materials and Methods

Figs. S1 to S7

Tables S1 to S4


Database S1

References and Notes

  1. Acknowledgments: We thank M. Huber, H. Dijkstra, G. Foster, A. von der Heydt, M. Leckie, P. Hull, H. Spero, J. Zachos, A. Ravelo, A. Fedorov, and S. Hu for helpful discussions and four anonymous reviewers for their thoughtful reviews. This research used samples and data provided by the Integrated Ocean Drilling Program (IODP). Funding for this research was provided by NSF AGS 1203163 (to M.P.) and a Schlanger Ocean Drilling Fellowship (to Y.G.Z.), which is part of the NSF-sponsored U.S. Science Support Program for IODP that is administered by the Consortium for Ocean Leadership, Inc.
View Abstract

Stay Connected to Science

Navigate This Article