Temperature Gradients Across the Pacific Ocean During the Middle Miocene

Sea surface temperatures (SSTs) of the tropical Pacific Ocean exert powerful controls on regional and global climates. Previous studies have suggested that during warm climate phases, the east‐west temperature gradient collapsed. To date, there has been no high‐resolution reconstruction of sea surface conditions in both the east and west Pacific Ocean during the Miocene Climate Optimum (MCO) and across the middle Miocene climate transition (MMCT); therefore, our understanding of the mean oceanographic state during this major global climatic shift is limited. Here, we present new SST reconstructions for the eastern Pacific Ocean (15.5–13.3 Ma) which show a clear east‐west temperature gradient of ∼4°C during the warmest interval of the Neogene, implying that the oceanographic processes that produce the modern gradient were present and active. There is no shift in the east‐west gradient across the MMCT indicating that the gradient was not impacted by global cooling and ice growth. We find a 2°C sea surface cooling in the eastern equatorial Pacific, that lags the benthic foraminiferal δ18O positive shift by 150 kyr, indicating that tropical temperature did not decrease synchronously with the expansion of the Antarctic ice sheet. Reconstructed variations in the δ18O composition of seawater, determined by combining our Mg/Ca and δ18O records, reveal a freshening in the eastern Pacific Ocean after 13.8 Ma, suggesting changes in the hydrological cycle and in tropical fronts in response to the new icehouse regime.

With average annual temperatures >28°C, the WPWP is the warmest water body on Earth and contributes to the heating of the tropical atmosphere, affecting both the Hadley and Walker Circulations (De Deckker, 2016;Nathan & Leckie, 2009). Investigations into the initiation of the WPWP and development of equatorial Pacific SST gradients during the middle Miocene have been hindered by a paucity of high resolution geochemical records extending beyond 13 Ma. However, Ocean Drilling Program (ODP) Site 1146 (116°16.37′E, 19°27.40′N) is located in the South China Sea (SCS) at the northern edge of the WPWP. High resolution (∼3-6 k.y.) middle Miocene upper ocean temperature and salinity estimates derived from planktonic foraminiferal δ 18 O and Mg/Ca reaching back to ∼15.7 Ma were previously generated at this site (Holbourn et al., 2010).
The seasonal and decadal variability of SSTs and the extent of warm water masses in the SCS are highly correlated with the development of the WPWP, suggesting that the WPWP strongly influences surface hydrology in the SCS (e.g., Lin et al., 2011). Although seasonality is more pronounced in the SCS, in comparison to the core area of the WPWP, the summer SST signal (mainly recorded in surface dwelling planktonic foraminifera) closely tracks long-term (decadal) variability in the WPWP (Lin et al., 2011). In the middle Miocene, when the SCS was more open to the western equatorial Pacific Ocean and when polar ice sheets may have been reduced and the latitudinal temperature gradient decreased, connections between the SCS and WPWP were presumably even stronger than they are at present. This notion is also supported by a comparison of tropical SST over the past 12 Myr, which shows comparable values and trends at ODP Site 1143 (South China Sea) and at ODP Site 806 (center of WPWP) during the Miocene (Zhang et al., 2014).
The present day EEP is characterized by cold nutrient-rich waters that result from a shallow thermocline and intense upwelling. The modern SST gradient between the EEP and WPWP averages 4°C-5°C (Karnauskas et al., 2009; Figure 1) and varies in response to ENSO (Zhang et al., 2014). The absence of an equatorial temperature gradient, caused by weak trade winds and the eastward propagation of warm western Pacific equatorial waters, is thought to reflect a "permanent El-Niño" state, which results in deeper thermocline depths and attenuated upwelling rates in the EEP (Wara et al., 2005). The establishment of "permanent El-Niño" like conditions has been suggested to have taken place during intervals of global warmth, for exam-FOX ET AL. Figure 1. Paleo-location of tropical Pacific sites discussed in this study. Top panel: Colors represent remotely sensed long-term mean sea surface temperature (SST); Bottom panel: Colors represent remotely sensed long-term mean sea surface salinity (SSS) (data from Tyberghein et al. (2012), Maps were generated using the rdgal and ggplot2 packages in R. ple during the early Pliocene when planktonic foraminiferal Mg/Ca data reveal that the E-W temperature gradient was nearly absent (Fedorov et al., 2006;Wara et al., 2005). More recently, this hypothesis has been challenged as new SST reconstructions from TEX 86 and U K 37 revealed a sustained temperature gradient between the east and west Pacific Ocean extending back 12 million years (O'Brien et al., 2014;Zhang et al., 2014). Given the absence of high resolution geochemical records from the WPWP, we compared the planktonic foraminiferal δ 18 O and Mg/Ca records from IODP Site U1338 (2°30.47′N, 117°58.18′W, 4200 m water depth) in the EEP with those from ODP Site 1146 (19°27.40′N, 116°16.37′E, 2092 m water depth) in the northern South China Sea (Holbourn et al., 2010; Figure 1).
Despite extensive studies of benthic foraminifera (e.g., Badger et al., 2013;Gasperi & Kennett, 1993;Holbourn et al., 2014;Shevenell et al., 2004;Tian et al., 2013) existing planktonic foraminiferal stable isotope records spanning the middle Miocene are scarce and of low resolution for the eastern equatorial Pacific Ocean. This is due to sedimentary successions spanning this interval having been strongly affected by carbonate dissolution or burial diagenesis, or proved incomplete due to hiatuses (Barron & Keller, 1982;Pearson, 1995). However, continuous sediment cores were recovered at both Site 1146 (Holbourn et al., 2010) and Site U1338, with abundant and diverse planktonic foraminifera in middle Miocene sediments (Fox & Wade, 2013). The Expedition 320/321 sites have been shown to be sensitive to dissolution (Kochhann et al., 2016), which could alter our Mg/Ca values. Yet scanning electron microscope studies of Site U1338 specimens reveal exceptional preservation of planktonic foraminifera with open pore spaces and little evidence of calcitic overgrowth on the wall surface (Fox & Wade, 2013). The interval of extensive carbonate loss at 15.6 Ma (Kochhann et al., 2016) is not part of our record, which begins at 15.5 Ma, and dissolution susceptible species are present throughout (Fox & Wade, 2013).

Sample Materials
Our study focuses on an 80-m section of lower-middle Miocene sediments from 350 to 430 m composite depth (mcd) from Site U1338. Pälike et al. (2012) indicate that the paleo latitude of Site U1338 is within ±2° latitude of the equator. The sediments consist of white, pale yellow, and very pale brown nannofossil ooze, grading to nannofossil and calcareous chalk (Pälike et al., 2010). For planktonic foraminiferal geochemical analyses, we use the same samples used for benthic foraminiferal stable isotope analysis in Holbourn et al. (2014), ∼630 samples were analyzed at 10 cm intervals providing a ∼3 kyr temporal resolution.
To assess preservation, selected specimens of planktonic foraminifera were mounted on SEM stubs, coated with gold, and inspected using a FEI Quanta 650 SEM at the University of Leeds, UK. After imaging their external surfaces at low and high resolution, several tests were broken using moderate pressure under a glass slide, and the fragments were used for the investigation of internal surfaces and test walls in cross-section ( Figure 2).

Stable Isotopes
Oxygen and carbon isotope ratios (δ 18 O, δ 13 C) were measured on 10-12 shells of mixed-layer dwelling species Globigerinoides subquadratus from the 250 to 315 μm size fraction from 425 meters composite depth (mcd) until its extinction at 390 mcd. Analyses then continue with Globigerinoides spp. until 350 mcd. In a few samples, where foraminiferal density was low, only 5-7 specimens were analyzed. Analyses were made with a VG Optima mass spectrometer with multi prep device at the British Geological Survey, Keyworth, UK. When picking shells, care was taken to exclude individuals with broken or missing chambers, although preservation of specimens was generally excellent (Fox & Wade, 2013). The external reproducibility of our measurements is ±0.07‰ and ±0.05‰ for δ 18 O and δ 13 C, respectively. To examine the reproducibility of the results, duplicate measurements were made on 35 samples (5%), which indicate mean reproducibility better than ±0.12‰ and ±0.14‰ for δ 18 O and δ 13 C, respectively. Oxygen isotope data are reported as per mil on the VPDB scale calibrated through laboratory and international standards.
At ODP Site 1146, δ 18 O and δ 13 C were measured by Holbourn et al. (2010) on the mixed-layer dwelling planktonic foraminifera Globigerinoides obliquus or Gs. subquadratus, using 10-20 well-preserved tests from FOX ET AL.

Mg/Ca
For Mg/Ca analyses, we selected 25-35 specimens of Trilobatus quadrilobatus (140-550 μg) from the 250 to 315 μm size fraction; the same size fraction as used for δ 18 O analysis, to minimize size-related intraspecific elemental variation (Elderfield et al., 2002). Analyses were performed on 86 samples over the studied interval. The tests were gently crushed and subsequently cleaned according to the protocol of Martin and Lea (2002) to remove clays. Cleaning included a reductive step with hydrazine to remove Mn-(hydr)oxides. Samples were measured on an ICP-AES device at Christian-Albrechts-Universität zu Kiel, Germany. Analytical precision is ∼1.1%, based on measurements of an internal laboratory standard. Replicate Mg/Ca measurements revealed an average standard deviation of ∼0.08 mmol/mol. Adequate cleaning is indicated by very low Fe/Ca, Al/Ca and Mn/Ca ratios.
For ODP Site 1146 up to 40 tests of Gs. subquadratus or Gs. obliquus from the 250 to 350 μm size fraction were analyzed by Holbourn et al. (2010). Specimens were gently crushed and cleaned following methods outlined by Martin and Lea (2002), which include a reductive step with hydrazine to remove Mn-(hydr) oxides. Details of the procedure are outlined in Holbourn et al. (2010).

Planktonic Foraminiferal Taxonomy and Ecology
Trilobatus quadrilobatus is a multi-chambered and photosymbiont bearing species, which forms part of the T. sacculifer plexus (Poole & Wade, 2019). This species first appeared in the early Miocene (Subzone M1a), and T. quadrilobatus, sacculifer, and trilobus morphotypes are all found in the modern ocean produced by the same species (Andre et al., 2013). Many studies point to a depth habitat within the upper ∼80 m of the water column for T. sacculifer (Fairbanks et al., 1982;Steph et al., 2009), which make this an ideal species to reconstruct SSTs in the equatorial Pacific Ocean.
Globigerinoides obliquus is a long-ranging, tropical to subtropical species, which first appeared in the early Miocene (Kennett and Srinivasan, 1983). Duplicate measurements of Gs. obliquus and Gs. subquadratus at Site 1146 indicate no offsets in δ 18 O and δ 13 C between these two species (Holbourn et al., 2010). This suggests that these two species had a similar depth habitat within the mixed layer.

Mg/Ca Paleotemperature Calculation
To investigate the influence of temperature on the upper ocean oxygen isotope gradient between the eastern and western Pacific Oceans, Mg/Ca analyses were performed on selected samples at Site U1338 to compare with the Site 1146 records (Holbourn et al., 2010). The planktonic species T. quadrilobatus forms part of the sacculifer plexus (Poole & Wade, 2019;Spezzaferri et al., 2018). Therefore, the conversion of T. quadrilobatus Mg/Ca ratios to SSTs was carried out using the T. sacculifer calibration (Anand et al., 2003): (1) The lengthy residence times of Ca (∼1.1 Myr) and Mg (∼13 Myr) (Broecker et al., 1982) suggests that Mg/ Ca of seawater (Mg/Ca sw ) is unlikely to have varied significantly over the duration of our 2.3 Myr record. However, Mg/Ca sw may have been considerably different in the Miocene compared to the present day values of 5.2 mmol/mol (Evans Mueller, 2012). Attempts have been made to reconstruct Mg/Ca sw over time using geochemical models (Fantle & DePaolo, 2005Stanley & Hardie, 1999), echinoderm skeletons (Dickson, 2002(Dickson, , 2004, low-resolution analyses of fluid inclusions in marine evaporites (Horita et al., 2002;Lowenstein et al., 2001;Sime et al., 2007) and carbonate veins precipitated in oceanic basalt (Coggon et al., 2010). Whilst these various reconstructions agree that Mg/Ca sw was lower during the Miocene than today, its precise value remains uncertain and the long-term evolution of Mg/Ca sw is still unclear. For this study the fluid inclusion value of 3.43 mmol/mol is used in the paleotemperature equation (at both sites) for consistency with the recent Miocene study by ; following the calculations outlined in their methods, we used the T. sacculifer calibration of Anand et al. (2003) (A = 0.09, B = 0.347) for T. quadrilobatus and applied a power constant of 0.41 for T. sacculifer based upon the available data for this species (Delaney et al., 1985;Evans & Mueller, 2012). T. quadrilobatus SST estimates were therefore calculated from the following equation : (2) In addition to changes in seawater Mg/Ca, it has been suggested that a pH correction is necessary for estimating SST (Evans et al., 2016;Gray & Evans, 2019). However, as Mg/Ca ratios in the T. sacculifer plexus are unaffected by changes in ocean pH (Allen et al., 2016;Drury et al., 2018), no correction was applied in this study.

Site Location Water Depth
The direct comparison of two records from sites with contrasting paleo-water depths (∼2,000 m at Site 1146 and 3,000-3,500 m at Site U1338; Holbourn et al., 2010;Pälike et al., 2012) and thus saturation state, is a potential source of uncertainty in Mg/Ca paleotemperature calculations. In Figures 3 and 5 SST estimates are presented that correct for the potential effects of dissolution on Mg/Ca ratios due to decreasing [CO 3 2− ] with increasing water depth (Dekens et al., 2002;Elderfield et al., 2006;Rosenthal et al., 2006). This is achieved by using the Pacific Ocean depth calibration of Dekens et al. (2002) (4) At Site U1338, the uncertainty in estimating changes in sea surface temperatures during the MMCT with planktonic foraminiferal Mg/Ca is about ±∼1°C (σ1), largely due to the variability among replicated samples (estimated here to be ∼0.9°C) and the error associated with the temperature sensitivity of the calibration (1.13°C; Anand et al., 2003). We estimate the combined error for absolute SSTs for Site U1338 to be ±2°C. Holbourn et al. (2010) report the uncertainty of paleotemperature estimates at Site 1146 to be ±1.2°C. This combines the standard deviation deduced from the calibration study (±1.13°C; Anand et al., 2003) and the analytical error of the Mg/Ca measurement. No correction was applied for the reductive step applied in our cleaning protocol, which generally results in a small decrease in Mg/Ca, leading to temperature underestimation of up to ∼0.6°C (Rosenthal et al., 2004). Changes in salinity may also have altered Mg/Ca reconstructed temperatures, however, Tierney et al. (2019) suggest that sensitivity is low between 33 and 38 PSU, we therefore did not make any adjustments for salinity changes.

Site 1146 Sensitivity of Reconstructed SSTs
We applied several equations to examine the sensitivity of reconstructed SSTs at Site 1146 ( Figure 3) (Figure 3). Adjusting for dissolution, following the equation of Dekens et al. (2002), average SSTs increase by 3°C, with a mean of 31°C, fluctuating between 33°C and 28°C. By adjusting for pH (Figure 3), the variability in SSTs becomes more extreme, between 36.5°C and 31°C with mean SSTs of 33°C. Equation 4 without the pH correction, is more consistent with the current understanding of mid Miocene temperatures and seasonal variation in the subtropics (Herold et al., 2012).

Surface Seawater δ 18 O
To determine changes in surface seawater δ 18 O (δ 18 O sw ) and the impact of evaporation and precipitation changes, we used our paired planktonic foraminifera Mg/Ca temperatures and δ 18 O to reconstruct δ 18 O sw at Site U1338. We corrected δ 18 O sw for ice volume by applying an ice volume value of 0.58‰ from the base of the record to 13.9 Ma, and 1.18‰ from 13.7 Ma and younger following Modestou et al. (2020). For the 200 kyr between 13.9 to 13.7 Ma we applied a linear correction of 0.003‰ per 1 kyr. FOX ET AL.

Results
At the east Pacific Site U1338, planktonic foraminiferal δ 18 O fluctuates between ∼−1.8‰ and −0.4‰ (Figure 4b). The δ 18 O record of Site 1146 shows higher amplitude variability than Site U1338 with δ 18 O values fluctuating between ∼−3.6‰ and −2.0‰. Comparison of the planktonic foraminiferal δ 18 O records between Site U1338 and Site 1146 reveals an average offset of ∼1.5‰, with the east Pacific recording the more positive values (Figures 4a-4b). Across Mi-3, while there is a 1‰ increase in benthic foraminifera δ 18 O, we do not see accompanying changes in the planktonic foraminifera δ 18 O of corresponding magnitude. Given that ice volume changes would impact δ 18 O in both planktonic and benthic foraminifera at the same time and to the same extent, the minor δ 18 O variations of in the planktonic foraminifera data would suggest the δ 18 O record is counterbalanced by surface water salinity variations.
Separation of the various components of the δ 18 O signal (temperature, salinity, ice volume) are required to improve understanding of the processes and feedbacks at work during this interval of climatic transition. The measured Mg/Ca ratios for the mixed layer dwelling species T. quadrilobatus at Site U1338 range from approximately 2.5 to 3.8 mmol/mol, giving a mean value of ∼3.2 mmol/mol and are consistently ∼1 mmol/ mol lower than those at Site 1146 ( Figure 5e).
Mg/Ca-derived SSTs following corrections (with adjusted paleo water depth) average at 26.4°C ± 2°C for Site U1338 (Figure 5d), these values are 2.8°C warmer than SSTs calculated following Equation 2. Similarly, Mg/Ca-derived SSTs calculated for Site 1146 following corrections are 2.6°C warmer than those originally presented in Holbourn et al. (2010). At Site 1146, SSTs calculated using Equation 4 vary between ∼28.3°C and 33.2°C with a mean temperature of 30.6°C, whereas at Site U1338 the SST record ranges from 23.6°C to 28.9°C with an average temperature of 26.4°C (Figure 5d). The temperature offset between Site 1146 and Site U1338 remains present in both the paleo water depth corrected and the uncorrected SST records.
Our corrected Mg/Ca SSTs in conjunction with planktonic foraminifera δ 18 O, allows for the calculation of δ 18 O sw . At Site U1338, δ 18 O sw averages 0.75 ± 0.25‰ from 15.5 to 13.7 Ma (Figure 5a), and decreases to 0.0‰ at 13.45 Ma. At Site 1146, δ 18 O sw is consistently lighter than at Site U1338 with mean values of 0.00 ± 0.31‰ from 15.5 to 13.9 Ma, values become more positive and reach 0.75‰ at 13.8 Ma, before decreasing to −0.55‰ at 13.45 Ma. The δ 18 O sw gradient between the two sites is approximately 0.75‰ from 15.5 to 14.2 Ma, decreasing to 0.25‰ in the younger part of our record.

Foraminiferal Preservation and Calcification Depth
Dissolution has been shown to impact foraminiferal Mg/Ca values (Dekens et al., 2002;Fehrenbacher et al., 2006;Regenberg et al., 2006); indeed, a recent study on bottom water temperatures by Kochhann et al. (2016) found evidence for minor carbonate dissolution at the event of peak warmth (15.6 Ma) at Site U1338. However, high resolution SEM analysis of planktonic foraminiferal specimens from Site U1338 reveals exceptional preservation of the original test wall texture and no visible etching of the wall surfaces ( Figure 2) at 15.56 Ma, where the Site U1338 planktonic foraminiferal record begins. A detailed study of Site U1338 planktonic foraminiferal assemblages also reveals numerous dissolution susceptible species are present throughout the studied interval (Fox & Wade, 2013). Furthermore, a reconstruction of the paleo-water depth and the carbonate compensation depth (CCD) in the equatorial Pacific suggests that Site U1338 has been above the CCD from 20 Ma to the present day (Pälike et al., 2012). Preservation at Site 1146 is recorded as "good" to "very good" over the studied interval, this means <20% fragmentation of foraminifera specimens and no visible evidence of etching. Whilst detailed SEM analyses conducted on multiple specimens reveal minimal recrystallization on foraminifer tests at both Sites U1338 and 1146 (Figure 2), the SST data presented in Figure 5 have been calculated using a dissolution correction (Equation 3) to ensure the potential effects of contrasting water depths on our specimens has been considered. However, we argue that the impact was minimal at Site U1338, and that preservation of foraminiferal specimens at Site U1338 is comparable and often superior to those found at the shallower Site 1146 (as illustrated in Figure 2). Mg/Ca estimates for Site U1338 are in good agreement, though slightly lower (∼1°C) than SST reconstructions for the same site using alkenones, which record temperatures of ∼28°C during this interval (Rousselle et al., 2013) (Figure 5d). Whilst these data are well within the uncertainty of each other, the calcification depth of T. quadrilobatus may also have played a role in the difference in SST estimates calculated from alkenones (Rousselle et al., 2013) and foraminiferal Mg/Ca (this study) at Site U1338.   of stepwise cooling during the MMCT. A major expansion of the EAIS occurred at ∼13.9-13.8 Ma following an initial cooling step at ∼14.7-14.6 Ma (e.g., Holbourn et al., 2007;Shevenell et al., 2004;Westerhold et al., 2005). The major glaciation at ∼13.9-13.8 Ma or Mi-3 event, highlighted on Figure 4, is recorded as a ∼1‰ increase in the benthic foraminiferal δ 18 O at both sites.

Impact of Ice Sheet
At Site U1338 Mg/Ca derived SSTs average 27°C between 15.5 and 13.7 Ma. Surprisingly, we do not see any decrease in SSTs in the eastern or western Pacific associated with Mi-3 (Figure 5d). Minimal sea surface cooling occurs between 13.9 and 13.6 Ma, whereas the enrichment in benthic foraminiferal δ 18 O marks the onset of a long-term climate shift at high latitudes, primarily due to deep ocean temperature cooling and polar ice growth. The lack of SST excursion at Mi-3 at low latitude sites, as recorded in both Mg/Ca and alkenone-based temperature proxies across the interval of ice sheet expansion (Figure 5d), contrasts with high latitude records from the Southern Ocean, which indicate a ∼6°C-7°C cooling (Shevenell et al., 2004), thus signifying a large increase in the meridional temperature gradient, as the Earth transitioned toward icehouse conditions. Furthermore, the divergence between the degree of positive δ 18 O change at ∼13.9-13.8 Ma in the surface and deep water records suggests that variations in local salinity are also imprinted on the surface δ 18 O data, as indicated by the δ 18 O sw records (Figure 4a). Following Mi-3, SSTs in the eastern Pacific (Site U1338) decreased to ∼26°C at 13.5 Ma and to ∼25°C in the youngest part of our record (13.35 Ma), although there is some variability with two of the warmest SST values (28°C) occurring at 13.5 Ma. The SST offset of ∼4°C between Sites 1146 and U1338 remained stable for over 1 million years (Figure 5c), however, the overall reduction in SSTs at Site U1338 after 13.7 Ma results in an increase in SST gradient to 7°C in the youngest part of our record. Significantly, the decrease in Mg/Ca SSTs occurs after Mi-3, and not in tandem with the benthic foraminiferal δ 18 O changes. The sea surface cooling recorded by planktonic foraminifera Mg/Ca is not synchronous with ice growth and deep sea cooling inferred from the benthic foraminiferal δ 18 O record, but occurs ∼150 kyr later (Figure 4c). The delayed SST cooling at Site U1338 is inconsistent with low-latitude records from ODP Sites 806 and 761, where surface cooling occurs synchronously with deep water temperature decrease and ice growth . The delayed SST cooling at Site U1338, indicates a heterogeneous temperature response across Mi-3 and suggests that changes in ocean circulation played a role.
The delayed 2°C cooling between ∼13.7 and 13.4 Ma displayed in the Mg/Ca derived SST record at Site U1338 (Figure 5d) indicates that deep sea cooling and Antarctic ice growth preceded the SST decrease. This contrasts with the dynamics of the Eocene-Oligocene Transition (34-33.5 Ma), when SSTs decreased prior to the interval of major ice sheet expansion (Lear et al., 2008;Wade et al., 2012). Unlike the EOT, Mi-3 is not associated with a turnover in plankton assemblages.

Evolution of Pacific Ocean Temperature and δ 18 O Gradients
The planktonic foraminiferal δ 18 O gradient between Sites U1338 and 1146 averages ∼1.5‰, with more positive values in the east Pacific (Site U1338). Using our Mg/Ca results, ∼1‰ of the offset in planktonic foraminiferal δ 18 O between the two sites can be explained by temperature (assuming 0.22‰ per °C, Kim & O'Neil, 1997). The remaining 0.5‰ must therefore relate to differences in regional evaporation-precipitation budgets and/or ocean circulation. On the basis of the modern δ 18 O-salinity relationship (Conroy et al., 2014), a change of 0.5‰ in the δ 18 O of seawater reflects a surface salinity shift of between 1 and 1.5 p.s.u (Stott et al., 2004). Present-day surface water salinities ( Figure 1)  Transient increases in the planktonic foraminiferal δ 18 O gradient between Sites 1146 and U1338 following the global cooling steps at ∼14.7-14.6 Ma and ∼13.9-13.8 Ma (Figure 3b) additionally support latitudinal shifts or intensification of the Intertropical Convergence Zone (ITCZ) coupled to waxing and waning of the WPWP through time. The divergence between the benthic and planktonic foraminiferal δ 18 O gradients after ∼14.6 and ∼13.8 Ma (Figure 4b) is primarily driven by decreases in planktonic foraminiferal mean δ 18 O values at Site 1146. These decreases, also clearly expressed in the δ 18 O seawater record at Site 1146, have been related to rises in regional precipitation, associated with intensification of the Southeast Asian monsoonal subsystem following global cooling (Holbourn et al., 2010). These temporal changes in local precipitation and δ 18 O seawater patterns also imply that the Southeast Asian monsoonal climate was highly sensitive to changes in the interhemispheric thermal gradient during the middle Miocene.
The average temperature offset between the records from Sites U1338 and 1146 ( Figure 5) is approximately ∼4°C, comparable with the modern temperature difference between these two sites ( Figure 1). However, the spatial extent of the WPWP may have differed on a warmer Miocene Earth, with a substantially reduced latitudinal thermal gradient (e.g., Fedorov et al., 2015). Middle Miocene temperatures at Site 1146 were ∼2°C-3°C warmer than present-day, supporting a northwards expansion of the WPWP throughout this extended interval of warmth. During the MMCT, differing rates of cooling between hemispheres following Antarctic glacial expansion probably favored a northwards displacement of the annual mean position of the ITCZ, away from the cooler hemisphere (e.g., Chiang & Bitz, 2005;Broccoli et al., 2006;Yang et al., 2020). This northwards shift would have maintained relatively warm temperatures at Site 1146 during the MMCT. In the eastern equatorial Pacific Ocean, temperatures were also ∼2°C-3°C warmer than present through the MCO and MMCT, suggesting that upwelled intermediate waters had different properties and origin compared to today.

Eastern Equatorial Pacific Paleoceanography
The paleoceanography of the eastern equatorial Pacific Ocean is complex, and sea surface conditions prior to 8 Ma are not well known or understood, due to a lack of high resolution proxy records (Drury et al., 2018). Frontal systems, upwelling and the cold tongue are all responsive to changes in atmospheric circulation, particularly the trade winds, which alter with changes in the meridional gradients (for instance high latitude cooling associated with Mi-3). Furthermore, the volume of throughflow through the Central American Seaway, which would impact salinity gradients in the EEP, is not well constrained. From 15.5 to 13.3 Ma, Site U1338 was located ∼1° south of the equator (Pälike et al., 2010), a region affected by the seasonal migration of the ITCZ. The EEP is a region of low sea surface salinity (∼33.5-34‰) due to high annual rainfall, low evaporation and advection of low salinity water (Alory et al., 2012;Martins & Stammer, 2015). Salinity changes both temporally and spatially (Melnichenko et al., 2019), and low salinity anomalies in the EEP are driven by strong and transient storms (Riser et al., 2019). Salinity interactions between advecting water and heavy rainfall are varied, and subject to entrainment and nonlinear effects (Martins & Stammer, 2015).
The "cold tongue" (Wyrtki, 1981) is an area of cool SSTs within the EEP. It is caused by divergence of surface Ekman currents resulting in water upwelling from the subsurface and advecting from the east (Zheng et al., 2012). The strength of the cold tongue has been shown to vary on glacial-interglacial timescales (Koutavas & Lynch-Stieglitz, 2004). Shoaling of the EEP thermocline and expansion of the cold tongue have been linked to cooling and the expansion of subpolar oceans on more recent timescales (Martinez-Garcia et al., 2010), though processes that influence the range and intensity of the cold tongue are not well understood. Whilst a cold tongue comparable to the modern may not have developed until the late Miocene or younger (Liu et al., 2019;Steph et al., 2010), there is widespread evidence that upwelling in the EEP initiated in the mid Eocene (Lyle et al., 2002;Moore et al., 2004;Wade et al., 2020;Zhang et al., 2017), thus, the EEP would have developed a proto-cold tongue, due to advection of cold, deep water.
The cooling and freshening of the EEP, could have been driven by enhanced advection of deep waters in response to high latitude cooling and strengthening of the trade winds. Intensification of upwelling is suggested at 14 and 13.8 Ma by Holbourn et al. (2014) based on the productivity proxies of opal mass accumulation rates from X-ray fluorescence (XRF) scanning, silica content and carbon isotopes. However, our SST cooling is not associated with higher biogenic opal accumulation rates (Figure 5b), thus the data do FOX ET AL.

10.1029/2020PA003924
not support enhanced upwelling at 13.75 Ma and younger as the mechanism for the ∼2°C SST cooling. The decrease in Mg/Ca values and δ 18 O sw is not synchronous with the major benthic foraminifera δ 18 O increase (Mi-3) (Figure 4), or with high latitude SST changes (Shevenell et al., 2008), suggesting that the EEP surface water cooling and freshening are in response to the initiation of the new climate regime of the icehouse mode. We suggest that alterations in wind patterns changed the boundaries between frontal systems and caused SST cooling in the EEP.

Changes in the Hydrological Cycle During the Middle Miocene
We find a freshening of EEP surface waters after Mi-3, suggesting a change in the precipitation-evaporation balance. It is surprising that the cooling and change in δ 18 O sw occurred after 13.7 Ma, postdating the deep water temperature and Antarctic ice volume changes. A freshening of ocean surface waters has now been recorded at multiple sites associated with/around Mi-3. These include western Pacific Sites 1146 (Holbourn et al., 2010) and 806 , eastern Pacific Site U1338 (this study) and Indian Ocean Site 761 . These global changes in low-latitude surface water δ 18 O sw suggest that major disruptions of the hydrological cycle and global precipitation patterns accompanied ice sheet expansion in the mid Miocene.
The meridional gradient and heat transfer between the hemispheres have a direct influence on the position of the ITCZ (McGee et al., 2014). The drop in δ 18 O sw of this magnitude (0.75‰) in the EEP surface waters (13.8-13.35 Ma), is unlikely to have been driven solely by increased precipitation, and hydrological changes would not explain the reduction in SSTs. High latitude ice cover has been shown to impact EEP through wind-evaporation-SST feedbacks (Chiang & Bitz, 2005). Southern hemisphere cooling and Antarctic ice expansion at ∼13.9-13.8 Ma was associated with a general improvement in deep water ventilation and strengthening of the meridional overturning circulation (Holbourn et al., 2013). These changes likely caused a re-organization of atmospheric circulation that led to major changes in the intensity and distribution of regional precipitation, ITCZ migration, and expansion of the cold tongue following Mi-3. The surface water cooling and freshening is more probably due to changes in surface ocean currents and frontal systems in response to changes in the meridional gradient. We suggest that increased meridional gradients and enhanced atmospheric transport led to dynamic reorganization in the hydrological cycle, with increased precipitation in the tropics, coupled with alterations in surface currents in the EEP. Further research is currently underway to extend the planktonic foraminifera isotope and Mg/Ca record at Site U1338 to determine the asymmetry of the western to eastern Pacific changes through the middle and late Miocene, and whether the reduction in δ 18 O sw represents a short or longer term modification in the EEP.

Conclusions
Here, we present a new middle Miocene record of planktonic foraminiferal δ 18 O and Mg/Ca for the eastern equatorial Pacific Ocean (IODP Site U1338). The Mg/Ca derived SST records generated for Sites U1338 and 1146 ( Figure 5) reveal a clear temperature asymmetry of ∼4°C across the Pacific Ocean throughout the middle Miocene (15.6-13.3 Ma). This implies that the oceanographic features, such as a relatively shallow thermocline and upwelling, were present and active in the eastern Pacific Ocean during the MCO, the warmest interval of the Neogene. Furthermore, the E-W gradient was unaffected by major Southern Hemisphere ice growth and the transition to a substantially colder climate mode after ∼13.9 Ma. Hydrological changes and variations in frontal systems caused a freshening of the sea surface following Mi-3. Significantly, sea surface cooling in the equatorial Pacific is not synchronous with the major benthic foraminifera d18O shifts, but is delayed by ∼150 kyr. Throughout our records (15.6-13.3 Ma), the E-W gradient never approaches zero, as implied for a "permanent El Niño" like state.

Data Availability Statement
The data that support the findings of this study are openly available via the National Geoscience Data Center (NDCD), Data ID: 164526. https://webapps.bgs.ac.uk/services/ngdc/accessions/index.html#item164526. FOX ET AL.