Middle Miocene Temperature and Productivity Evolution at a Northeast Atlantic Shelf Site (IODP U1318, Porcupine Basin): Global and Regional Changes

We present a high‐resolution multiproxy middle Miocene sea surface temperature (SST) and productivity (SSP) reconstruction of Integrated Ocean Drilling Program Site U1318, from the upper slope edge (∼400 m water depth) of the Porcupine Basin continental margin, eastern North Atlantic Ocean. Biomarker and dinoflagellate cyst proxies reveal warm and mostly stratified waters during the Miocene Climatic Optimum (MCO) that cooled ∼3°C across the Miocene Climate Transition (MCT). The organic biomarker (TEX86 and UK′37 ) paleothermometers document a series of 11 transient cooling events (CEs), superimposed on the long‐term climate evolution. These CEs are associated with increases in cold‐water dinocysts and correlate to global benthic δ18O shifts, including the Mi‐2, Mi‐3, and Mi‐4 events. Most CEs are also associated with increases in primary productivity. A prolonged interval of high SSP between ∼13.8 and 13.6 Ma supports the idea that carbon production (and burial) in shallow areas represents a feedback mechanism contributing to long‐term atmospheric CO2 decline and cooling during the MCT. SST comparison in three North Atlantic sites (Azores Site 608, Porcupine Basin Site U1318, and Rockall Plateau Site 982) reveals that MCO SSTs are much warmer at Site 608 than at the other two sites. The low‐resolution SST record of Site 982 shows no decrease in temperature around the MCT. This may be linked to contemporaneous tectonic changes in the Tethys, Central American, and Arctic Seaways impacting local ocean circulation, superimposed on global drivers of climate change.

different from that of today (e.g., Hamon et al., 2013;Jakobsson et al., 2007). High-latitude surface water temperatures during the MCO in both hemispheres were exceptionally warm, >∼10°C warmer than today Sangiorgi et al., 2008Sangiorgi et al., , 2018Schreck et al., 2013;Super et al., 2018;Warny et al., 2009). Models still fail to represent data-derived temperature gradients solely with an increase in CO 2 , as complex feedback mechanisms may have been in place (e.g., Gasson et al., 2016;Goldner et al., 2014). Local factors and changes in oceanography (e.g., sea-ice coverage and albedo, poleward ocean heat transport due to shifts in atmospheric and ocean circulation) can offer additional explanation for the sustained warmth at high latitudes (e.g., Bijl et al., 2018;Sangiorgi et al., 2018;Stuecker et al., 2018). High-resolution surface water temperature data for the MCO are needed to improve data-model comparison. The global warming during the MCO has been associated with the rapid eruption of the Columbia River Basalt Group, the strongest phase (95%) of which occurred between 16.7 and 15.9 Ma (Kasbohm & Schoene, 2018). The release of CO 2 into the atmosphere is suggested to have triggered changes in ocean carbon chemistry (Sosdian et al., 2018(Sosdian et al., , 2020. Partly overlapping with the MCO, a ∼3 Myr (∼16.8-13.5 Ma) positive excursion (∼1‰) in the benthic foraminifera carbon isotope ratios (δ 13 C), known as the Monterey Carbon Isotopic Excursion (Vincent & Berger, 1985), indicates a perturbation in the carbon cycle and marine organic carbon burial. The Monterey excursion is composed of six carbon isotope maxima (CM) thought to represent carbon cycle feedback processes associated with marine productivity (Holbourn et al., 2007;Sosdian et al., 2020;Woodruff & Savin, 1991).
Given the lack of high-resolution surface temperature data for the middle Miocene and the importance of shelf regions in regulating the carbon cycle via marine productivity, we reconstruct sea surface temperature (SST), sea surface productivity (SSP), and oceanographic changes across the MCO and the MCT at an upper slope site (water depth ∼400 m) of the eastern North Atlantic Ocean (IODP Site U1318, Porcupine Basin, offshore Ireland, Expedition 307 Scientists, 2006) with independent multiproxy records. The aim is to investigate regional surface water changes and detect whether they correspond to proposed mechanisms and causes for global climate and carbon cycle variations. Furthermore, we present a North Atlantic perspective by comparing our SST data with the few other available data obtained from North Atlantic deep-ocean sites (DSDP 608, Azores, andODP 982, Super et al., 2018, 2020). The analyses of Miocene climate records from different locations in the North Atlantic may help us understand the role of regional variability, superimposed on the global trends and drivers of climate change.

Site Location and Oceanography
Site U1318 (3 Holes, 51°26.16′N, 11°33.0′W; 409 m water depth; Expedition 307 Scientists, 2006) was drilled on the upper slope edge of the continental margin in the Porcupine Seabight (Figure 1), a failed rift system that originated during the opening of the North Atlantic Ocean during the Middle to Late Jurassic. During the middle Miocene the British Isles were still connected to continental Europe and there was no connection to the North Sea through the English Channel (Gibbard & Lewin, 2003;deeptimemaps.com). The Porcupine Basin is filled with 12 km of Late Palaeozoic to Quaternary sediments (Ryan et al., 2009), mainly supplied from the Irish and Celtic shelves (Rice et al., 1991). Modern surface water temperatures (SST) at Site U1318 range between ∼10°C in winter and ∼16°C in summer. At a depth of 400 m, the water temperature is ∼11°C all year round (Boyer et al., 2018, Supplemental Figure S1).
At the Porcupine bank and at Site U1318 location, the water (between 0 and 500 m depth) is composed of Continental Slope Current (CSC), which transports Eastern North Atlantic Water via the North Atlantic Current (NAC) to the Norwegian Sea (Raddatz et al., 2011, Figure 1 Raddatz et al., 2011). Background bathymetry created using the Global Multi-Resolution Topography (GMRT) Map tool: http://www.marine-geo.org/tools/GMRTMapTool/; (c) Middle Miocene (14.9 Ma) map indicating the Fram Strait (FS), the Central American Seaway (CAS), and the Tethyan (Western Mediterranean) Gateway (WMG) (modified from Scotese [2014]). contribution from the Mediterranean Outflow Water, occupy the water column below (500-1,500 m), offshore the continental margin where Site U1318 is located (García-Ibáñez et al., 2015;Pollard et al., 1996). At greater depths, Labrador Sea water and Norwegian Sea Deep Water are observed (Raddatz et al., 2011). Waters are moderately productive and productivity peaks occur in spring when seasonal stratification begins and the mixed layer depth becomes shallower (Gutknecht et al., 2019;UNEP LME Report, 2008).
During the middle Miocene (∼15 Ma), Site U1318 was located at a paleolatitude of ∼47°N (van Hinsbergen et al., 2015) at a similar to present-day water depth (Ryan et al., 2009). Gateway configuration was different. During the early Miocene, the Fram Strait started opening and the freshwater dominated, stratified Arctic Ocean became more ventilated at ∼17.5 Ma. At ∼13.7 Ma, the strait deepened to present-day depth (Jakobsson et al., 2007), with implications for the Atlantic Meridional Overturning Circulation and deepwater formation (e.g., Lozier et al., 2019). Changes in lower latitude seaways (Mediterranean gateway, eastern Tethys and Central American Seaways) also occurred in the middle Miocene (∼14 Ma) with important consequences for the ocean circulation reorganization and global climate. However, the exact timing of these tectonic changes is still debated (Bialik et al., 2019;Hamon et al., 2013;Montes et al., 2015;Zhang et al., 2014).

Sediments, Age Model, and Stable Isotope Analysis
Samples were obtained from the Integrated Ocean Drilling Program (IODP) Expedition 307, Site U1318 Hole B (cores 10-14H and 17X-27X) and Hole C (cores 7H and 8X-10X, Expedition 307 Scientists, 2006; Supplemental Figure S2 and Data sets 1 and 2) between 92.4-and 247.5-m composite depth (mcd) to target the middle Miocene. The composite record was established based on physical properties (Expedition 307 Scientists, 2006). Middle Miocene sediments belong to the lithostratigraphic Unit 3 (Subunits A-C) which mainly consists of greenish-gray clay with Total Organic Carbon content (TOC%) for selected samples ranging from 0.27% to 0.70% (uppermost sample analyzed representing near modern conditions at 3.6 mbsf, TOC% = 0.25, Mangelsdorf et al., 2011) and are divided into subunits based on their calcium carbonate content (Expedition 307 Scientists, 2006;Quaijtaal et al., 2017). When material was sufficient for all analyses, 245 samples were used for oxygen and carbon stable isotopes on benthic foraminifers, palynology (dinoflagellate cysts), and organic geochemistry (U K 37 and TEX 86 ). However, some sediment samples were not sufficient for all analyses and/or were not productive. One hundred forty-five samples were analyzed for δ 18 O and δ 13 C on the benthic foraminifers Uvigerina sp. and Cibicidoides pachyderma (methodology and complete data set presented in Quaijtaal et al., 2017). Two hundred forty-five samples were analyzed for organic geochemistry and 99 samples were analyzed for palynology additional to those reported in Quaijtaal et al. (2014) (total 222).
The age model of the portion of Site U1318 record presented in this study is based on integrated bio-, isotope-, and magnetostratigraphy and is already published in Quaijtaal et al. (2017). The tie points used to build up the age model are listed in their Table 3. Samples between 92.4-and 247.5-m composite depth (mcd) cover the interval 12.75-16.60 Ma. Average time resolution for stable isotope analyses, organic geochemistry, and palynology is 25 thousands of years (kyr), 17 and 16 kyr, respectively.

10.1029/2020PA004059
Agilent 1100 series LC/MSD SL and separation and a Prevail Cyano column (2.1 × 150 mm, 3 µm; Alltech), maintained at 30°C. The glycerol dialkyl glycerol tetraethers (GDGTs) were eluted isocratically using a changing mixture of hexane and propanol. The first 5 min proportions were 99% hexane:1% propanol, then increasing linearly toward 1.8% in 45 min. Flow rate was 0.2 mL/min. Single ion monitoring was set to scan the [M + H] + ions of the isoprenoid and branched GDGTs (dwell time 237 ms for each ion). TEX 86 values were calculated after Schouten et al. (2007) and SST values were calculated following the equation: SST H = 68.4 × TEX 86H + 38.6, where TEX 86H = log(TEX 86 ) (Kim et al., 2010). Analytical precision is 0.3°C based on repeated analysis of an in-house standard. Standard error on TEX 86H reconstructed temperatures is 2.5°C. The Bayesian calibration (BAYSPAR) (Tierney & Tingley, 2014 has also been used to calculate SST. To evaluate the reliability of the TEX 86 and SST values, we also calculated the branched and isoprenoid tetraether (BIT) index (Hopmans et al., 2004), an indicator for continental organic matter input (Supplemental Data set 1). Following Weijers et al. (2006), we assumed that at BIT values below 0.3 TEX 86 values were not biased by continental organic matter input. TEX 86 can be further biased by input of GDGTs derived from methane utilizing Archaea (Y. G. Zhang et al., 2011). For this, a "Methane Index" (MI) can be calculated (Y. G. Zhang et al., 2011), which evaluates the contribution of methanotrophic archaea to the total GDGT pool (Supplemental Data set 1). MI values of 0.3-0.5 mark the boundary between "normal marine sediments" and methane-impacted sediments. Furthermore, we calculated the Ring Index (Supplemental Data set 1), which reflects archaeal communities distinct from those present at modern core top calibration sites that might indicate nontemperature effects on TEX 86 and can be used to determine if TEX 86 temperature estimates are influenced by nonthermal factors and/or deviate from modern analogs (Y. G. Zhang et al., 2016). Finally, we used the GDGT-2/GDGT-3 ratio as indicator of the depth of GDGT production in the water column (Taylor et al., 2013) (Supplemental Data set 1) although the water column depth is relatively shallow.
For  K 37 U analysis, the ketone fraction, containing alkenones, was dissolved in hexane and injected for gas chromatography using a Hewlett Packard 6890N Network GC system equipped with a 50-m long 0.32-mm diameter silica column coated with CP Sil-5 (thickness 0.12 µm). The carrier gas used was helium. Oven temperature was programmed at 70°C for injection and increased by 20°C min −1 up to 200°C, then increased by 3°C min −1 until 320°C. This final temperature was maintained for 30 min. A constant pressure of 100 kPa was maintained and samples were injected on-column. Analytical precision is 0.2°C based on repeated analysis of selected fractions. Selected fractions were analyzed by GC-MS to confirm the identification of the long-chain alkenones. GC/MS was done on an Agilent 7890A GC coupled to an Agilent 5975C MS. The column, carrier gas, flow rate, and oven program were identical to the GC-FID but the end temperature was kept for 25 min. The MS operated at 70 eV, with an ion source temperature of 250°C and an interface temperature of 320°C. The injection was done on-column, with an injection volume of 1 µL. The long-chain alkenones were identified in full scan, scanning between m/z 50 and 600 and comparison with literature (de Leeuw et al., 1980;Volkman et al., 1980).
U was calculated after Prahl and Wakeham (1987), that is, U reconstructed temperatures is 1.5°C (Supplemental Data set 1). The symbol ± after a value represents the standard deviation of several measurements. Instead, a ±2.5°C and a ±1.5°C indicate the standard error of calibration for TEX 86 and  K 37 U , respectively.

Palynology and Palynological Indices for SST and SSP
Samples were treated with standard palynological preparation techniques. The samples were oven-dried at 60°C and ∼10-15 g of dry-weight material was used, to which one or two Lycopodium clavatum tablets (batch no. 177745, X = 18,584 ± 829 and batch no. 1031, X = 20,848 ± 2,186) were added (Wood et al., 1996). Carbonates were removed using hydrochloric acid (HCl, 30%); the samples were then allowed to settle overnight and rinsed with demineralized water until pH was neutral. Samples were then processed with 40% hydrofluoric acid (HF). Fluorosilicates were removed with HCl. The residues were sieved over a 10-µm nylon mesh screen and mounted on a microscopic slide in a drop of liquid glycerol gelatin. The slides were covered with a cover slip and sealed with nail polish. The samples are stored in the collection of the research group Paleontology and Paleoenvironment, Department of Geology, Ghent University, Belgium.
Dinocysts are resting stages of dinoflagellates, marine protists, with multiple trophic strategies (e.g., De Vernal & Marret, 2007;Head, 1996). The phototrophic, autotrophic dinoflagellates mostly live in the upper ∼25-30 m of the water column. They are most abundant in the continental shelf, coastal to neritic environments . For the purpose of this study, that is, to provide qualitative SST and SSP reconstructions based on dinocysts, we use the ecological preferences of dinocysts as known in literature (e.g., De Schepper et al., 2009Zonneveld et al., 2013) and calculate temperature and productivity indices as introduced in Quaijtaal et al. (2014). We selected three dinocyst species (Nematosphaeropsis labyrinthus, Polysphaeridium zoharyi, and Operculodinium centrocarpum sensu stricto, s.s.) for additional environmental interpretations, as their ecological niches in the modern ocean (first two species) or possible affinity to the modern species O. centrocarpum sensu Wall and Dale (1966) (for the third one) may aid paleoceanographic interpretations. Abundances of N. labyrinthus, O. centrocarpum s.s., and P. zoharyi are given with a 95% confidence interval based on the determination of a multinomial distribution for multiple taxa (following Heslop et al., 2011).

Sea Surface Temperature (SST DINO ).
A Warm/Cold index (W/C) has often been used to evaluate changes in SST calculated as W/C = nW/(nW + nC), where n is the number of specimens counted, W are warm-water indicating species, and C are cold-water indicating species (Sangiorgi et al., 2003;Versteegh, 1994). The list of warm-water species (W) and cold-water species (C) is reported in Table 2 of Quaijtaal et al. (2014). SST DINO index can vary between 0 (only cold species) and 1 (warm species only). We have also calculated the concentrations (dinocysts/gram sediment) of warm and cold-water dwelling species to further support the dinocyst-based temperature reconstructions. Data are reported in Supplemental Data set 2 and Figure S3.
Sea Surface Productivity (SSP DINO ). A commonly used qualitative index for SSP is the P/G index (P/G; Versteegh, 1994), where P/G = nP/(nP + nG), where n is the number of specimens counted belonging to the peridinioid (P) or gonyaulacoid (G) (Sangiorgi & Donders, 2004;Sluijs et al., 2005). The P/G ratio (also called H/A ratio) is based on the assumption that peridinioid species have a heterotrophic (H) feeding strategy, while most of the gonyaulacoids are phototrophs, autotrophs (A). Peridinioids feed on organic matter including phytoplankton (e.g., Jacobson & Anderson, 2008) and they usually dominate areas where high annual primary productivity is high due to high nutrients availability brought for instance by upwelling, river discharge, or high remineralization (e.g., De Vernal & Marret, 2007;Lewis et al., 1990;Pospelova et al., 2008;Sangiorgi & Donders, 2004;Zonneveld et al., 2013). The species used to represent the P group (heterotrophs) belong to the genera Barssidinium, Brigantedinium, Cristadinium, Echinidinium, Lejeunecysta, Palaeocystodinium, Selenopemphix, Sumatradinium, and Trinovantedinium. G group includes all other dinocysts, except the goniodomacean species (e.g., Capisocysta spp., P. zoharyi, and Tuberculodinium vancampoae). SSP DINO index can vary between 0 (only Gonyaulacoid phototrophic species) and 1 (only Peridinioid heterotrophic species). The higher the P/G ratio, the higher the productivity. We calculated the P-cyst concentrations (P-cysts/gram) to strengthen the productivity reconstruction. P-cysts are notably more sensitive than G-cysts to bottom water oxygenation and their % in the record (hence the P/G ratio) could be biased by changes in oxygen availability through time (e.g., Versteegh & Zonneveld, 2002) although a strong relationship between P/G and primary productivity in modern settings has been found (Pospelova et al., 2008). Evaluating P-cysts concentration next to P/G ratio provides a way to distinguish between productivity and preservation (Reichart & Brinkhuis, 2003). Data are presented in Supplemental Data set 2 and Figure S3.

Sea Surface Temperature at IODP Site U1318
TEX 86 -based sea surface temperature (SST TEX-H ) varies between ∼18°C and ∼27°C ( Figure 2) and between ∼15°C and ∼28°C when using the BAYSPAR calibration (SST TEX-BAY , Figure 3). Minimum and maximum SSTs are reached at 12.8 and 16.4 Ma, respectively, regardless of the calibration used. Indeed, the two calibrations show very similar values, although the BAYSPAR calibration generally depicts higher variability and lower SSTs after 13.9 Ma. The overall record corresponds in remarkable detail to the Site U1318 benthic oxygen isotope data published earlier (Quaijtaal et al., 2017, Figure 2).
Superimposed on a long-term ∼9°C cooling, SST TEX-H records several transient 1.5°C-6°C CEs, numbered here CE1-11 ( Figure 2). A drop in SST TEX-H in multiple (≥3) samples is indicated with blue bars, while cooling episodes based on few (2) datapoints are marked in light blue. In the interval between 16.6 and ∼14.7 Ma, average SST TEX-H (24.5°C ± 1.3°C) values are generally higher than in the younger part of the record (average 21.6°C ± 1.7°C).
The most pronounced event, CE7, is part of a 6°C decline in SST TEX-H from 14.76 to 14.61 Ma. Another remarkable event is CE8, which is superimposed on a long-term cooling trend, occurs between 14.20 and 13.66 Ma and corresponds to a temperature drop of ∼3°C. A third pronounced cooling step (CE11) of ∼3°C is recorded between 12.83 and 12.78 Ma. SANGIORGI ET AL.  U -based SST reconstruction (SST UK ) based on the P. J. Müller et al. (1998) calibration (error bars are calibration errors). Qualitative SST (SST DINO ) and SSP (SSP DINO ) based on dinoflagellate cyst assemblages; percentages of three representative dinoflagellate cyst species (Nematosphaeropsis labyrinthus, Polysphaeridium zoharyi, and Operculodinium centrocarpum s.s.) plotted with a 95% confidence interval based on a determination of a multinomial distribution for multiple taxa (following Heslop et al., 2011). SST, sea surface temperature; SSP, sea surface productivity.
To check for any bias in the TEX 86 record, we screened the data for several criteria (Supplemental Data set 1 and Figure S4). No TEX 86 values were discarded because of terrestrial input, as BIT was <0.1 throughout the record, which is well below the accepted threshold of 0.3 (Weijers et al., 2006). The values for the MI indicate that there is no substantial input of methanotrophic or methanogenic archaea (Y. G. Zhang et al., 2011). TEX 86 is primarily derived from marine Thaumarchaeota as shown by the small (<0.3) deviation in the Ring Index (Y. G. Zhang et al., 2016). The GDGT-2/GDGT-3 ratio (Taylor et al., 2013) is quite constant (1.9-3.0) implying there is no remarkable difference in the depth of GDGTs production as expected in a relatively shallow water location. In the part of the record between 15.5 and 12.7 Ma, GDGT-2/GDGT-3 ratios are generally higher (2.7 ± 0.2) than in the interval 16.6-15.6 Ma (2.2 ± 0.2), but this shift does not correspond to any remarkable changes in the SST TEX-H (Figure 2). There is a weak significant correlation between GDGT-2/GDGT-3 ratios and TEX 86 (R 2 = 0.31; p < 0.0001, supplemental Figure S4) suggesting no major bias due to depth changes.   (Super et al., 2020), and IODP Site U1318 plotted using BAYSPAR calibration (Tierney & Tingley, 2015) for the interval 16.6-12.7 Ma. CO 2 (μatm) concentrations based on stomata (Kürschner et al., 2008), alkenones (Y. G. Zhang et al., 2013), paleosols (Retallack, 2009), and Boron isotopes (from Sosdian et al., 2018Sosdian et al., , 2020. SST, sea surface temperature. show a major cooling step at ∼13.64 Ma (min SST UK37 = 26.6°C), which separates an interval with low (∼1°C) SST UK37 variability (16.60-13.64 Ma) from one with higher SST UK variability (up to 3°C at 13.2 Ma, between 13.64 and 12.7 Ma).
SST relative changes, as calculated with the index based on dinocyst (SST DINO ) assemblages and warm dinocyst concentrations (Supplemental Figure S3), show prevailing warm waters between 16.60 and 14.04 Ma, followed by a gradual cooling, a major cooling step at 13.88 Ma, and others at 13.53 and 12.85 Ma. The cold-water dinocyst concentrations are always higher than those of warm dinocysts in younger part of the record (13.88-12.75 Ma). Considering the species responsible for the CEs, the CEs 1-7 are caused by a decrease in warm-water species and a sudden appearance of cold-water indicators Bitectatodinium tepikiense and Impagidinium pallidum, although in low abundances (Supplemental Data set 2 and Figure S3). In the modern ocean, these species are associated with (sub)polar cold, well-ventilated nutrient-rich waters. However, I. pallidum has also been found in high abundances at SSTs as high as 10°C-15°C in the Pliocene and Miocene of the North Atlantic and in low abundances even at higher temperatures Schreck et al., 2017). The younger CEs (CEs 8-11) are instead mostly caused by peaks (up to 35% of the assemblage) of Habibacysta tectata and lower abundances of Filisphera filifera, both extinct species. Habibacysta tectata abundances similar to those found in our record have been associated to SST of ∼15°C-20°C in other Pliocene and Miocene records Schreck et al., 2017). The temperature information obtained from dinocysts is comparable to that derived from SST TEX-H ; during CEs mean annual temperature rarely dropped below 15°C.

Sea Surface Productivity and Oceanographic Conditions at Site U1318
The Peridinioid/Gonyaulacoid (P/G) ratio based on dinocysts varies between 0 and 0.3 (Figure 2, Supplemental Data set 2 and Figure S3). Compared to the average value of 0.07, the highest values (>0.2) in the P/G ratio occur at 16.58, 16.10, 15.71, 14.83 Ma, and between 13.84 and 13.57 Ma and at 13.26 Ma. Increases in P/G ratio are supported by peaks in P-cyst concentrations (Supplemental Figure S3), particularly high at 15.70 Ma (∼3,600 cysts/g), 14.8 Ma (∼6,600 cysts/g), and between 13.84 and 13.57 Ma (4,000-8,300 cysts/g).
To further constrain interpretation of paleoenvironments, the ecological preferences of three dinocysts species are considered ( Figure 2). It is worth mentioning that the dinocyst assemblages in the Miocene record of Site U1318 are dominated by Spiniferites/Achomosphaera complex, indicating the proximity to the shelf (e.g., Zonneveld et al., 2013;Supplemental Figure S3). Nematosphaeropsis labyrinthus, a species abundant in the modern ocean frontal systems, is present in percentages generally lower than 2% between 16.6 and 13.61 Ma and significantly increases to values up to 6% after 13.61 Ma. Polysphaeridium zoharyi prefers warm and stratified waters. Here, it is almost constantly present between 16.6 and 14.6 Ma and virtually absent after 13.9 Ma. Operculodinium centrocarpum s.s. is never very abundant, but its percentages tend to increase from about 13.8 Ma and reaches its highest value (11%) at the top of the record (12.7 Ma). If we assume that O. centrocarpum s.s. shares the same ecological niche with the extant species O. centrocarpum sensu Wall and Dale (1966), it represents a cosmopolitan species, well adapted to unstable conditions, and found where oceanic and neritic waters mix at continental margins (e.g., Dale, 1996). In the North Atlantic, high abundances of O. centrocarpum sensu Wall and Dale (1966)

Comparison Among Different Temperature Proxies
The best and most striking correlation between temperature proxies is found between SST TEX-H and benthic δ 18 O records (R 2 = 0.88; p < 0.0001) (supplemental Figure S5). The correlations between SST TEX-H and SST UK37 (R 2 = 0.66; p < 0.0001), SST TEX-H and SST DINO (R 2 = 0.54; p < 0.0001), and SST UK37 and SST DINO (R 2 = 0.42, p < 0.0001) are lower but still significant. These correlations suggest that there is likely one dominant controlling factor on these individual proxies, that is, temperature, although other factors might contribute to the lower degree of correlation between some proxies (e.g., seasonality, water column depth, biases related to ammonium oxidation rates on TEX 86 , maximum temperature limit of , ice volume component on benthic oxygen isotope records).
At present, CSC, ENAW waters occupy the entire water column at Site U1318 (51.5 N; 11.5 W, Figure 1). Measured mean annual SST is 12.6°C, 10.5°C in winter and 15.6°C in summer when a thermocline develops in the upper (∼30 m) layer as consequence of atmospheric warming (Boyer et al., 2018, Supplemental Figure S1).
We tested the paleothermometry proxies on modern surface sediment samples. Unfortunately, the only nearby surface sediments where In our Miocene record, SST TEX-H is several degrees (average 10.0°C ± 2.1°C) higher than modern mean annual SST at Site U1318, while SST UK temperatures are on average 15.3°C ± 0.6°C higher than modern mean annual SST. One explanation for such difference between the two lipid-based paleothermometers could be the different growth seasons of the organisms on which the temperature proxies are based. Coccolithophorids, the organisms producing alkenones (SST UK ), usually bloom in spring and summer (April-July) in the North Atlantic (e.g., Shutler et al., 2013), while the Thaumarchaeota (SST TEX-H ) can be abundant all year round, even in winter (Bale et al., 2013;O. Müller et al., 2018;Pitcher et al., 2011). Alternatively, SST TEX-H may represent a deeper than surface water signal (e.g., Besseling et al., 2019;Ho & Laepple, 2016). Site U1318 is located at shallow water depth (400 m) and measured mean annual SSTs at surface and at maximum depth differ by 2.5°C (Boyer et al., 2018). Locations where high contribution of deep Thaumarchaeota may affect SST TEX-H as surface signal are usually upwelling systems and/or areas where the water column is >1,000 m depth (Taylor et al., 2013), unlike our study site. An additional cause for the offset between SST TEX-H and SST UK may be due to the import of allochthonous material from warmer regions and/or alkenones preferential selective degradation, which has shown to produce temperature increases of up to 1.8°C (Kim et al., 2009). Several studies have shown that the TEX 86 is less impacted by lateral transport than U K 37 as GDGTs degrade quicker than alkenones, leading to relatively higher allochthonous input for alkenones (Kim et al., 2009;Mollenhauer et al., 2008). Another potential bias could be due to archaeal ammonium oxidation rates on TEX 86 (Hurley et al., 2016) which may have varied over the Miocene and introduce amplifications of temperature trends in TEX 86 records and relates to productivity (Lawrence et al., 2020). Cross correlation between TEX 86 and the dinocyst-based productivity (P/G ratio) proxy shows poor (R 2 = 0.24; p < 0.0001) correlation. It should also be noted that  K 37 U index is well above the value 0.9 in all samples and equal to or higher than 0.95 in 201 of the 245 samples analyzed (Supplemental Data set 1). This could imply that saturation is reached and temperatures experienced by haptophytes could have been even higher, which would explain the lower variability of the SST UK compared to SST TEX-H . If SST UK represents surface water summer temperature, summers temperatures during the middle Miocene would have been 27.9°C ± 0.6°C (average SST UK in the record), ∼12°C warmer than at present. Such high summer temperatures throughout the Miocene record would question the presence of cold-temperate water dinocysts, which become very abundant (up to ∼8,000 dinocysts/g sediment) in the interval between 13.84 and 13.57 Ma (Quaijtaal et al., 2014(Quaijtaal et al., , 2017, SST DINO , Figure 2; Supplemental Figure S3). Dinoflagellates live in the upper ∼30 m and are most abundant during summer in the North Atlantic (e.g., Barton et al., 2015). The production of different dinocyst species in different seasons in coastal sites has been investigated with sediment traps studies (e.g., Fujii & Matsuoka, 2005;Pospelova et al., 2018;Ribeiro & Amorim, 2008). Records usually indicate that an increase in (heterotrophic) P-cysts is mostly dependent on prey (e.g., diatoms) availability and usually occurs in autumn (or winter). Autotrophic (phototrophic) G-cyst species depend directly on local physical and chemical environment factors, and cyst production occurs throughout the year.
Rather than a clear seasonal pattern, interannual variability is more important (Ribeiro & Amorim, 2008). We included 20 different species in the SST DINO. We are hence confident that a strong seasonal bias due to seasonal production of temperature related dinocyst species can be excluded . Although qualitative, SST DINO (supported by the concentration of cold-water loving species, Supplemental Figure S3) indicates shifts toward colder water conditions in several parts of the record.
Given the above, we speculate that SST TEX-H values may be more indicative than SST UK of mean annual temperature changes though we cannot exclude complications due to changes in archaeal ammonium oxidation rates.
The benthic foraminiferal δ 18 O (Figure 2) at Site U1318 shows good similarity with the benthic foraminiferal δ 18 O stack (De Vleeschouwer et al., 2017) but has also clear differences. The different resolution between the two records makes a direct comparison difficult. The major trends are similar, the δ 18 O values are lower (as expected at a shallower site) but variability is similar (average 0.46‰ ± 0.72‰ compared to 1.58‰ ± 0.71‰, in De Vleeschouwer et al., 2017). The strong correlation between SST TEX-H and benthic foraminiferal δ 18 O may suggest that SST TEX-H includes a deeper component and reflects subsurface water signal. At present, the difference between mean annual temperature at the surface (0 m, 12.6°C) and at the seafloor (400 m, 10.7°C) is low (∼2°C), and the water below 100 m has virtually the same temperature all year round. The calibration uncertainty of the TEX 86 -based SST proxy is ±2.5°C, which precludes distinguishing between surface and subsurface signal. Furthermore, we note that most studies indicate that TEX 86 in surface sediments is generated predominantly in the upper waters (100-200 m) and not in the sediment itself (e.g., Lengger et al., 2013). Given that the SST TEX-H from the modern sediment sample best reflects surface and not bottom water temperatures, SST TEX-H at this site is likely to represent temperature of the upper waters.
Most of the CEs recorded at Site U1318 in both the SST TEX-H and benthic foraminiferal δ 18 O data are accompanied by SST DINO shifts toward lower values, suggesting cooling of the entire water column, with SST DINO representing surface conditions, SST TEX-H (sub)surface conditions and benthic δ 18 O bottom water (400 m depth) conditions. SST UK may be mostly indicative of summer conditions.

Productivity and Oceanographic Changes at Site U1318
The majority of the CEs are accompanied by an increase in productivity (SSP DINO and P-cysts concentrations, Figure 2, Supplemental Figure S3). This correspondence is particularly clear for events CE1, CE2, CE3, CE6, and CE8, suggesting that cooling and an increase in nutrients in the upper water column were related. Nutrient-rich surface waters can be the result of more vigorous mixing of the water column, upwelling, changes in ocean currents, proximity to land during global cooling, and sea level drop and consequent displacement of shelf material, or river input. At present, dinocyst assemblage from a site close to Site U1318 (A125, Zonneveld et al., 2013, ∼50.3°N, 11.6°W, Figure 1) is composed of 1% Peridinioid cysts only (SSP DINO = 0.01, Supplemental Figure S1). SSP DINO values as high as 0.2-0.3 (compared to average 0.07 of the entire Miocene record) and increases in P-cysts concentrations (Supplemental Figure S3 and Data set 2) demonstrate that primary productivity was on average higher than at present and increased significantly during some of the cold events (Figure 2), in particular between 13.84 and 13.57 Ma. We correlated SSP DINO and BIT to test whether higher terrestrial organic matter input (e.g., via river input) could have caused the increase in productivity. The explained variation is very low (R 2 = 0.09; p < 0.0001, supplemental Figure S5), dismissing this cause. Moreover, the few pollen present in the record (Supplemental Figure S3) are also indicative of moderate to low terrestrial input and certainly negligible after 15.5 Ma. The lithology of the core does not show major changes (clay to fine sand), and the observed increases in fine grain sand particles (>63 μm) do not correspond to peaks in productivity (Supplemental Figure S2). This is particularly true for the highest productivity peak between 13.84 and 13.57 Ma. At present, SSP DINO (P/G ratio) values of ∼0.2-0.3 in shelf areas are found in the North Adriatic Sea (Sangiorgi & Donders, 2004), Northern Gulf of Mexico (Limoges et al., 2014) both offshore a river plume, and in the Northern Jorge Gulf (Patagonia) where waters from offshore well up on the shelf (Faye et al., 2018). Excluding the influence of nutrient-rich riverine waters as cause for increased productivity, and in absence of a relationship between productivity and sediment grain size or pollen/spores (Supplemental Figures S2 and S3), which could point to a coastline closer to the site and higher displacement of material from the shore, we submit that increased mixing of the water column during cold intervals provided the nutrients to sustain productivity in the surface waters. Rearrangements in the Tethys Seaways at ∼ 14 Ma initiated the Mediterranean Outflow Water, a water mass flowing through the western Mediterranean Gateway into the Atlantic Ocean (MOW, Figure 1, Bialik et al., 2019). At present, MOW is a high salinity, nutrient-rich water mass (Coste et al., 1988;Huertas et al., 2012) that occupies the water column below 500 m depth to the west of drill site. Whether this water mass could have influenced the nutrient content of the surface waters at Site U1318 around ∼13.8 Ma is unknown.
Modern dinocyst assemblage at A125 is dominated by O. centrocarpum (89%), followed by Spiniferites spp. (4%), N. labyrinthus (3%), and P. zoharyi is absent. Polysphaeridium zoharyi is the cyst of a red tide forming, potentially toxic dinoflagellate, Pyrodinium bahamense (e.g., Usup et al., 2011). It thrives in warm (summer temperatures up to 29°C), saline (seasonally) stratified water environments, and is never observed north of 30°N (Zonneveld et al., 2013). Polysphaeridium zoharyi's near-continuous presence at Site U1318 between 16.6 and 14.6 Ma confirms that waters must have been much warmer and more stratified than today's during the MCO. Interestingly, when P. zoharyi is present, it generally decreases in abundance when productivity increases (Figure 2). This could confirm that more vigorous water mixing during cooling episodes sustained productivity. Its disappearance after ∼13.8 Ma is in line with the average ∼3°C SST cooling recorded by SST TEX-H during the MCT, while SST UK is still high (∼28°C). In our Miocene record, the highest (3.5%-10%) percentages of O. centrocarpum s.s. occur in the record between ∼13.75 and 12.75 Ma. If its ecological affinity is similar to that of the extant O. centrocarpum, significantly higher percentages (exceeding the confidence interval) may indicate environmental conditions became generally more unstable. Interestingly, the age of ∼13.7 Ma coincides with that of the MCT and that of a fully ventilated Arctic Ocean following complete opening and deepening of the Fram Strait (Jakobsson et al., 2007;Sangiorgi et al., 2008). Other major tectonic rearrangements (Tethys and Central American Seaways) seem to have occurred around 14 Ma (Bialik et al., 2019;Montes et al., 2015;Zhang et al., 2014). Simultaneous changes in the oceanography at Site U1318 are also recorded by a small, but significant increase in N. labyrinthus, which best thrives in proximity of frontal systems, when waters with different temperature come into contact (Boessenkool et al., 2001;Prebble et al., 2013).
In general, the dinocyst assemblages of the entire middle Miocene record remain different from the modern ones but show clear changes at ∼13.8 Ma.

Cooling Events at Site U1318: Regional Versus Global
The majority of the surface CEs (CE, Figure 2) at Site U1318 are mirrored by increases in global benthic δ 18 O record (De Vleeschouwer et al., 2017), including Mi-2 (∼16 Ma, CE2), Mi-2a (∼14.8 Ma, CE6/CE7), Mi-3 (∼13.8 Ma, CE8), and Mi-4 (∼12.8 Ma, CE11). We will now focus the discussion on relating these global events to regional climate and oceanography changes. All the other CEs are considered regional changes, with minor expression in the global δ 18 O record.
The first, transient cooling (CE2, 16.03 Ma, Figures 2 and 3) can be correlated to Mi-2. A first cooling step in SST DINO and an increase in productivity (SSP DINO, P-cyst concentration, Supplemental Figure S3) occur at the beginning of a 0.93‰ benthic δ 18 O excursion at Site U1318, accompanied by a ∼3°C SST TEX-H (∼0.3°C SST UK ) cooling. During the maximum temperature drop, reduced stratification (decrease in P. zoharyi) and waters masses with contrasting characteristics (N. labyrinthus peak) are present. The benthic δ 18 O global record change in the same time interval is 0.43‰, about half of the signal recorded at Site U1318. Using backstripping solutions and benthic δ 18 O and Mg/Ca records, eustatic sea level is estimated to have dropped ∼30-40 m during Mi-2 (John et al., 2011;Miller et al., 2020), under atmospheric CO 2 concentrations (∼400-600 ppmV, Sosdian et al., 2020, Figure 3) clearly lower in absolute values and range compared to the intervals before and after. Productivity is relatively high during the entire interval.
The second marked cooling phase (CE5,CE6,and CE7,15.05,14.84,and 14.61 Ma,respectively Figure 2), is a ∼400 kyr interval centered on the age of Mi-2a (14.8 Ma, Miller et al., 2020). CE5 is mainly detectable in SST TEX-H and corresponds to a small rise in productivity, while changes in SST UK , benthic foraminifer δ 18 O and SST DINO are less evident. Instead, CE6 represents a significant ∼2-3°C SST TEX-H (∼1°C SST UK ) decrease coeval with a positive δ 18 O excursion of 0.5‰ (Figure 2). Similarly to CE2, CE6 is identified by a sharp initial SANGIORGI ET AL.
10.1029/2020PA004059 decrease in SST DINO and a prominent but temporary increase in productivity, while P. zoharyi indicates a general decrease in warmth and water column stratification. Surface cooling and productivity correspond to a maximum global deepwater cooling (∼1‰ increase in global benthic δ 18 O record change, De Vleeschouwer et al., 2017) associated with Mi-2a. Mi-2a has been linked to an eustatic sea level fall of ∼30 m and a cooling of ∼0.7°C in deep waters (John et al., 2011;Miller et al., 2020). CO 2 concentration values are extremely sparse and seem to point to values generally lower than 300 ppmV (Figure 3).
During CE7, a sharp decrease in SST DINO marks the beginning of a long gradual SST TEX-H cooling of ∼5°C down to 21°C (SST UK decrease of ∼1°C down to 28°C), accompanied by an increase in benthic δ 18 O of 0.8‰ at Site U1318. A comparable δ 18 O excursion of 0.75‰ is observed in the deep ocean (De Vleeschouwer et al., 2017). Productivity is constantly, but minimally, above average values, and water stratification seems to be in place (P. zoharyi is present). The few CO 2 reconstructions available for this interval depict a drop from ∼400 ppm precooling (at ∼14.7 Ma, Y. G. Zhang et al., 2013) to 116 ppm (at ∼14.6 Ma, based on 1 measurement on paleosols, Retallack, 2009; Figure 3). Regardless of the CO 2 concentration values, the remarkable drop in SST observed at Site U1318 does not seem to be accompanied by any major changes in the oceanography, including productivity, which is only slightly above average at Site U1318.
After CE7, proxies indicate an environment with generally colder and less stratified surface waters. SST DI-NO shows gradual cooling starting at 14.12 Ma, followed by a sharp decline while SST TEX-H decreases from ∼13.89 Ma toward the CE8 cooling at 13.8 Ma (Figure 2). The overall surface water SST TEX-H temperature decrease is ∼5°C-6°C Ma (SST UK ∼ 2°C). The sharp SST DINO decline at ∼13.89 Ma is simultaneous with the start of a gradual positive excursion of ∼1‰ in the global benthic δ 18 O (δ 18 O shift at Site U1318 is ∼0.9‰), which corresponds to the Mi-3 event. Mi-3 is associated with cooling in the deep ocean of 1.2°C, ∼50 m eustatic sea level fall (De Vleeschouwer et al., 2017;Miller et al., 2020), and declining atmospheric CO 2 concentrations (from ∼600 to ∼400-500 ppm, Sosdian et al., 2020, Figure 3). Dinocyst proxies (Figure 2, Supplemental Figure S3 and Data set 2) show remarkably high productivity, which lasted for a period of ∼ 200 kyr. High productivity on continental margins at global scale, if accompanied by carbon burial, can certainly contribute to CO 2 withdrawal from the atmosphere and to the global cooling at the MCT. Indeed, this interval corresponds to the so-called CM6 event, one of the most pronounced CM detected in the deep ocean, which indicates a drastic change in the global carbon cycle (Holbourn et al., 2004;Lear et al., 2010;Sosdian et al., 2020). CE11 at ∼12.78 Ma is part of a cooling trend that started at ∼13.12 Ma. This cooling, especially pronounced in the surface water proxies, is of a smaller magnitude than CE8 (Mi-3), both in benthic δ 18 O (0.71‰) and in SST TEX-H and SST UK (∼3°C; Figure 2). This cooling corresponds to the Mi-4 event (∼12.9 Ma), which has been associated with a shift in global benthic δ 18 O of ∼0.5‰, a drop in deepwater temperature of ∼1°C and sea level fall of ∼20-30 m (De Vleeschouwer et al., 2017;Miller et al., 2020) and mostly preindustrial atmospheric CO 2 concentrations ( Figure 3). The presence of a small hiatus prior to the Mi-4 (Quaijtaal et al., 2014) has possibly prevented to capture this event fully at the Porcupine Basin. During this event, dinocysts show low productivity and O. centrocarpum s.s. percentages reach ∼10%.
CE8 represents the peak of a SST cooling trend (6°C in total) that started ∼14.2 Ma at Site U1318 ( Figure 2) and affected the entire water column (∼400 m deep). At a global scale, CE8 is associated with a ∼50 m sea level drop (Miller et al., 2020). Reconstructed CO 2 concentration data for much of the Miocene are quite sparse if we exclude the MCO and the MCT intervals ( Figure 3). Interestingly, when considering boron-based CO 2 reconstructions (e.g., Sosdian et al., 2018), both the CE2 cooling (∼16 Ma) and the CE8 event (∼13.8 Ma) occurred at CO 2 values of ∼400-500 ppm (Figure 3). Instead, other proxies, reconstruct lower CO 2 concentrations for all CEs after the MCO (Figure 3). It appears that CO 2 values of ∼400 ppm (or lower) may have been able to shift the relatively warm Miocene climate to a temporarily colder mode.

Comparison With Other North Atlantic SST Records
Few biomarkers-based SST reconstructions are available for the middle Miocene in the North Atlantic. Existing records are from ODP Site 982 (Rockall Plateau, 57°30.992′N, 15°52.001 W, 1,135 m, Super et al., 2018;calculated paleolatitude ∼54°N at 15 Ma, van Hinsbergen et al., 2015), DSDP Site 608 (Azores, 42°50.205′N, 23°05.252 W, 3,526 m, Super et al., 2018, calculated paleolatitude ∼39°N at 15 Ma, van Hinsbergen et al., 2015. SSTs in these records were calculated using the Bayesian calibration (BAYSPAR) (Tierney & Tingley, 2014. For a direct comparison, we also recalibrated our data from Site U1318 using BAYSPAR (Figure 3 and Supplemental Data set 1). As reported before, values obtained with BAYSPAR do not differ much with the calibration of Kim et al. (2010) shown in Figure 2 (Supplemental Data set 1). Although significant differences in resolution exist among the records, some SST patterns can be recognized. In the MCO interval (16.6-14.5 Ma), SST at Site 608 is much higher (average ∼30°C) than that of the other two more northerly sites. Counterintuitively, however, SSTs at Site U1318 are 2°C-3°C lower between 16.3 and 15.5 Ma than those of the northern Site 982. Between 14.4 and 13.9 Ma, the three sites show comparable temperatures. More interesting is the SST record between 13.8 and 12.6 Ma. There is a remarkable similarity in SST trends between the Azores Site 608 and Site U1318, while temperatures at Site 982 do not show major changes. Future research should focus on increasing the resolution of the Rockall Plateau record to detect the influence of the (proto)NAC at this site. If higher resolution SST data confirm no major changes in SST also during global CEs, tectonic changes occurred around 14 Ma (Bialik et al., 2019;Hamon et al., 2013;Jakobsson et al., 2007;Montes et al., 2015;Zhang et al., 2014) may have been responsible for the regional variability recorded in the North Atlantic, superimposed on global drivers of climate change. The 2°C-3°C SST difference between Sites 608 and U1318 records, especially during the warmer phases, is similar to the present-day SST difference at the two sites (mean annual SST at Site 608 = 16.6°C and mean annual SST at Site U1318 = 12.6°C).

Conclusions
We presented the first high-resolution multiproxy middle Miocene SST and SSP record from a shelf site (Site U1318) in the Porcupine Basin, in the eastern North Atlantic. Warm (average SST TEX-H = ∼24°C) and mostly stratified waters are reconstructed for the MCO, ∼12°C higher than at present, which cooled by ∼3°C (average SST TEX-H = ∼21°C) across the MCT. Proxies show transient and long-term coolings, some of which correspond to global increases in benthic δ 18 O including Mi-2, Mi-3, and Mi-4 events. Most of the CEs at Site U1318 are associated with intervals of high surface water productivity. Dinocyst proxies record a ∼200 kyr-long period (∼13.8-13.6 Ma) with primary productivity significantly higher than that recorded in the modern system (SSP DINO 0.2-0.3 compared to SSP DINO = 0.01). High productivity begins ∼100 kyr after a temperature (SST DINO ) drop at ∼13.9 Ma (CE8). CE8 is synchronous with the global event known as Mi-3 (∼13.8 Ma) and coincides with the MCT. The high productivity interval is also synchronous with the global carbon isotope maximum event CM6 (Holbourn et al., 2004) recorded in the deep ocean and indicative of a major perturbation of the carbon cycle. If carbon production in the surface waters of shelf areas led to carbon burial globally, shelves indeed represent the locations where an important feedback mechanism contributing to long-term atmospheric CO 2 decline occurs, as recently suggested by Sosdian et al. (2020). This aspect merits scientific attention and more records from shelf areas can provide the answer.
We compared the surface temperature records of three North Atlantic sites and detected concurrent cooling at ∼13.8 Ma in the Azores and Porcupine Basin sites. The record of Site 982 does not seem to show any major cooling at ∼13.8 Ma, but the temporal resolution is too low to draw conclusions.
High-resolution temporal records, similar to those produced for the Late Miocene (Herbert et al., 2016), are needed to understand the link between tectonic, global climate, and oceanography changes in different locations of the North Atlantic during the MCT. Interestingly, tectonic changes in seaways, including the complete opening and deepening of the Arctic gateway (Fram Strait, Jakobsson et al., 2007), changes in the Tethys Seaways and in the Mediterranean outflow (Bialik et al., 2019;Hamon et al., 2013;Zhang et al., 2014), and possibly the complete closure of the Central Atlantic Seaway (Montes et al., 2015) occurred around 14 Ma.

Acknowledgments
The samples for this study were provided by the Integrated Ocean Discovery Program. Walter Hale and Alex Wülbers kindly supported the sampling at the Bremen Core Repository. Ivo Vandemoortel is thanked for helping with isotope analyses. Anchelique Mets, Jort Ossebaar, Monique Verweij (NIOZ), and Sabine Van Cauwenberghe (UGent) are thanked for their technical assistance. We thank James Super and Matt Huber for sharing the biomarker data of ODP Site 982 and Sindia Sosdian for sharing CO 2 data. We thank Stijn de Schepper and one anonymous reviewer for their comments, which greatly improved the manuscript. This work was supported by the Research Foundation-Flanders (FWO) under project number G.0179.11N. SS thanks the Netherlands Earth System Science Center, funded by the Ministry of Education, Culture and Science (OCW).