Comment on “A 12-million-year temperature history of the tropical Pacific Ocean”

Even small changes (≤1°C) in absolute tropical sea surface temperatures (SSTs) have implications for climate conditions both regionally (1) and globally, including climate sensitivity to greenhouse gases (2). Thus, it is crucial to obtain accurate tropical SST estimates during the warm early Pliocene [5.0 to 3.5 million years ago (Ma)], when atmospheric CO₂ concentrations and extratropical temperatures were, on average, higher than pre-industrial values. SST reconstructions, based on planktonic foraminifera Mg/Ca and biomarker proxy ${\mathrm{U}}_{37}^{K\text{'}}$, indicate two key features of the tropical Pacific (3): average Western Equatorial Pacific (WEP) warm-pool SSTs (4, 5) similar to today and a reduced (but not absent) zonal equatorial SST gradient (4–6). We compile tropical Pacific SST data (4, 6, 7, 8–12) (Table 1) and show, contrary to claims made by Zhang et al. (7), that their low-resolution (1 sample per >100,000 years) records, only appropriate for a cursory characterization of average Pliocene conditions, are largely consistent with previously published data.

The only published data representing the central WEP warm pool is from site 806 (Table 1 and Fig. 1) and indicate average early Pliocene SSTs cooler than today (13): –0.03°C for Mg/Ca, –1.10°C for TEX₈₆, and –0.66°C for ${\mathrm{U}}_{37}^{K\text{'}}$. Because absolute SST estimates are calibration dependent (14–17), a better approach compares proxy measurements and shows that average Pliocene values (0.70 for TEX₈₆, 3.45 mmol/mol for Mg/Ca, and 0.99 for ${\mathrm{U}}_{37}^{K\text{'}}$) are similar to core-top values (0.70 for TEX₈₆, 3.43 mmol/mol for Mg/Ca, and 0.98 for ${\mathrm{U}}_{37}^{K\text{'}}$). Within calibration errors of at least ±1°C (14–17), all proxies concur that the differences between Pliocene and core-top SSTs are negligible (+0.08°C for Mg/Ca, +0.04°C for TEX₈₆, and +0.26°C for ${\mathrm{U}}_{37}^{K\text{'}}$). ${\mathrm{U}}_{37}^{K\text{'}}$ is saturated at 29°C (17), but Mg/Ca and TEX₈₆ should detect warmer SSTs if they existed; although potential biases (e.g., changing seawater Mg/Ca and subsurface glycerol dialkyl glycerol tetraether (GDGT) lipid production) need further study, at this point neither Mg/Ca nor TEX₈₆ data provide solid evidence of average warm-pool SSTs substantially warmer than today. Average Pliocene TEX₈₆ SSTs at site 1143 (9°N) of 29.7°C can be explained by an expanded, but not warmer, central warm pool (3); expansion of the tropics during global warming is documented by 21st-century observations (18) and future (19) and Pliocene model simulations (20).

Focusing on long-term smoothed trends, Zhang et al. (7) argue that the Pliocene warm pool was substantially warmer relative to today but in fact show that it was warmer relative to the Quaternary (which includes Pleistocene glacial intervals). Their approach contrasts with a large body of published work, including modeling (20, 21) and observational (5, 8) studies, that compare Pliocene conditions to a modern baseline. This difference in approach is a source of misinterpretation of the TEX₈₆ data (7, 22). If one considers TEX₈₆ variability, instead of the mean, their approach still misses the profound importance of the data. From 3.5 to 5.0 Ma, TEX₈₆ variability is within the ±2.5°C calibration error with minima and maxima straddling the core-top value; this is in stark contrast to CO₂ and climate records from outside the warm pool in which even discrete glacial minima are generally higher than core-top and Pleistocene interglacial values (3).

Unlike the WEP warm pool, the East Equatorial Pacific (EEP) cold tongue was substantially warmer than today during the Pliocene (Fig. 1). This mean state with a reduced zonal SST gradient is referred to as “permanent El Niño-like conditions” and is supported by many data types (5, 23). This term does not imply a lack of variability nor that mean state changes are due to the dynamics that generate interannual El Niño events (3). Rather, it highlights that, like an El Niño event, small changes in the zonal gradient induce major changes in atmospheric circulation with far-field effects. The term “El Padre” is also used to differentiate this Pliocene state from the dynamical El Niño–Southern Oscillation pattern. The Zhang et al. (7) data support previous work showing a reduced zonal gradient during the Pliocene.

At EEP site 850, TEX₈₆ SSTs are cooler than ${\mathrm{U}}_{37}^{K\text{'}}$ SSTs (7); a similar cold bias occurs at site 1241 (Fig. 2A), due to subsurface production of crenarchaeota GDGTs (11). A shallow cool thermocline can explain this cold bias: TEX₈₆ apparently records depth-integrated temperature rather than SST. The cold bias likely changed as the thermocline shoaled (24) or as ecosystem structure changed (11) since 5 Ma. At WEP site 806 (Fig. 2B), higher Pleistocene ${\mathrm{U}}_{37}^{K\text{'}}$ and Mg/Ca compared to TEX₈₆ SSTs may be due to problems with Mg/Ca (7) and/or a cold bias in TEX₈₆ that was diminished in the Pliocene when the tropical thermocline was deeper (4, 23–25). Even if regional calibrations can statistically circumvent the TEX₈₆ cold bias in the modern ocean (14), they cannot be expected to provide accurate SST estimates as subsurface temperatures changed through the Pliocene (4, 23–25). Zhang et al.’s TEX₈₆ data reflect the WEP-EEP depth-integrated temperature difference, but their ability to quantify the surface temperature gradient is tenuous.

To quantify changes in EEP SSTs, we exclude TEX₈₆ data (at sites 850 and 1241) because of the clearly documented severe cold bias (11) (Fig. 2A) compared with other available EEP records (e.g., ${\mathrm{U}}_{37}^{K\text{'}}$). All EEP records show Pliocene SSTs warmer than today (Figs. 1B and 2B). The Zhang et al. (7) equatorial data indicate a mean Pliocene WEP-EEP difference of 1.5°C, in agreement with previously published data (Fig. 2C). Given the profound importance of tropical SST, and the fact that biases of all proxy types require additional study, a conservative approach that identifies convergence among data records, and does not overinterpret the importance of divergent records, is warranted.

The new data (7) confirm previous Pliocene data but provide evidence of Miocene warm-pool temperatures warmer than today (Table 1). Lea (22) states that “the TEX₈₆ data support previous evidence that the warm pool responds strongly to radiative forcing,” but in fact, the warm pool was warmer in the late Miocene than in the Pliocene, whereas partial pressure of CO₂ was not (26). The new TEX₈₆ data highlight the importance of future research aimed at solving this paleoclimate enigma.