Deep Thermohaline Circulation Across the Closure of the Central American Seaway

The closure of the Central American Seaway (CAS) resulted in changes of ocean‐climate dynamics since the late Miocene following the uplift of northern Andes. Reconstructing the bottom‐water temperatures (BWTs) of the Caribbean Sea illustrates feedbacks of the closure on the ocean‐climate system including deep‐water dynamics of the Caribbean Sea. Here, Mg/Ca‐derived BWTs of the Plio‐Pleistocene Caribbean Sea from the benthic foraminifer Cibicidoides wuellerstorfi are presented for the first time and interpreted along with Na/Ca and Sr/Ca as proxies of salinity and continental input, respectively. Our results highlight several warm (93, Gi15‐19, and N1) and cool (92, M2, Gi20, and CN4) marine isotope stages (MISs). Accordingly, changes in the circulation of deep‐water masses during the CAS closure developed in four main time intervals: (I) between 5.2 and 4.1 Ma (million years ago) BWT was ~1.1°C, (II) 4.1–3.2 Ma ~2.1°C, (III) 3.2–2.7 Ma ~2.7°C, and (IV) 2.7–2.2 Ma ~2.1°C. Relatively higher, gradually increased temperatures between 3.2 and 2.7 Ma correspond to late Pliocene warmth and restricted inflow of Pacific waters into the Caribbean due to shoaling of the CAS. In addition, Sr/Ca values reveal gradually escalating terrigenous input until 2.7 Ma most likely related to the increased river discharge in response to the Andean uplift. The gradual decrease of the BWTs from 2.7 Ma may have resulted from the onset of Northern Hemisphere Glaciation. Overall, BWTs match with previous sea surface temperatures from the planktic foraminifer Neogloboquadrina dutertrei. The BWTs presented here confirm intensified thermohaline circulation during the overall Pliocene warmth with increased bottom‐water Na/Ca values indicating enhanced salinity.


Introduction
The Isthmus of Panama results from the subduction of Pacific Farallon Plate beneath the Caribbean and South American plates, driving the development of the volcanic Panama Arc at the western edge of the Caribbean Plate since the upper Cretaceous from around 73 Ma. Consequent collision between the Central American Peninsula and South America resulted in the uplift of the northern Andes and continuous uplift of the Panama Isthmus (León et al., 2018;O'Dea et al., 2016), inferred, for example, from benthic foraminifer biostratigraphy (Coates et al., 1992;Collins et al., 1996;Duque-Caro, 1990;McDougall, 1996). Tectonic activity and gradual closure of the isthmus lead to the connection between the Americas and caused land mammal exchange called "The Great American Biotic Interchange" by circa 2.7-2.6 Ma (Iturralde-Vinent, 2006;Kirby et al., 2008;Schmidt, 2007;Webb, 2006;Woodburne, 2010).
The closure of the Panama Isthmus starting at circa 13 Ma is believed to have trigged the onset of Northern Hemisphere Glaciation (NHG) at about 2.7 Ma. This global climatic event has been caused by the termination of water exchange between the Pacific Ocean and the Atlantic Ocean through the CAS driving the northward heat transport via the Gulf Stream and North Atlantic Current (De Schepper et al., 2013;Haug & Keigwin, 2004;Haug & Tiedemann, 1998;Keigwin, 1982;Schmidt, 2007). When the CAS was closed and fresh water input from the Pacific Ocean had ceased, surface waters of the Caribbean Sea and Gulf of Mexico became warmer and more saline, driving intensification of the thermohaline circulation by density and temperature differences between surface and underlying water bodies as they cooled in northern latitudes (De Schepper et al., 2013;Groeneveld et al., 2008;Haug et al., 2001;Haug & Tiedemann, 1998;. Closure of the CAS also caused changes in the deep-water chemistry of the Caribbean Basin, which was filled with eastern North Atlantic Deep Water (NADW), Antarctic Intermediate Water (AAIW), and western North Pacific Intermediate Water (NPIW) (Kirillova et al., 2019;Newkirk & Martin, 2009). By~9.2 Ma, the intra-oceanic straits shoaled, which were more than 1,200 m deep, and deep-water exchange between the Pacific and Caribbean Sea had ceased (Newkirk & Martin, 2009;O'Dea et al., 2016;. As shoaling continued until the overall Pliocene warmth, more oxygenated, carbonate-preserving NADW preponderated over relatively low oxygen, corrosive southern sourced waters in the Caribbean Basin (Burton et al., 1997;Collins et al., 1996;Haug et al., 2001;Karas et al., 2017;Schmidt, 2007).
Previously established paleoclimate intervals of the Caribbean Sea relate to the emergence of the Panama Isthmus. For example, three distinct climate intervals were identified between 4.7 and 4.0 Ma based on SST estimations obtained through Mg/Ca of Trilobatus sacculifer (Gussone et al., 2004). Covering the latest Miocene and early Pliocene, two climate intervals were identified before and after 4.5 Ma based on SSTs obtained from T. sacculifer . Steph et al. (2010) provided a Mg/Ca temperature record measured on the thermocline dweller Neogloboquadrina dutertrei in order to reconstruct Caribbean subsurface temperature changes over the Pliocene between 5.2 and 2.7 Ma. De Schepper et al. (2013) have distinguished climate fluctuations between~3.4 and 3.2 Ma combining data from T. sacculifer and Globorotalia inflata Mg/Ca-derived SSTs, alkenone-derived SSTs, and dynoflagellate cysts. However, a continuous bottom-water paleotemperature record has not yet been produced.
BWT estimates are important not only to identify paleoclimate intervals but also to investigate the vertical mixing of the mesobathyal to deep-water column by thermohaline circulation. Vertical mixing is thought to have intensified in the Caribbean Sea in response to CAS closure, contributing to the NHG onset. With this scope, the role of the Caribbean Sea in thermohaline circulation corresponding to the shoaling of the CAS has been interpreted using various proxies to investigate, for example, warming of sea surface waters and heat-moisture transport. Planktic and benthic δ 18 O were analyzed to address the warming of sea surface waters and to decipher its impact on deep-water mass component (Haug et al., 2001;Keigwin, 1982;Schmidt et al., 2004). Neodymium isotopes were presented to discuss the deep-water sources through their radiogenic characteristics (Kirillova et al., 2019;Osborne, Haley, et al., 2014;Osborne, Newkirk, et al., 2014). Numerical models were applied to assess the oceanic and atmospheric parameters responsible for extreme climate events resulting from the closure of the CAS (Sarnthein et al., 2009;Schneider & Schmittner, 2006).
In this study, we have analyzed Mg/Ca data of C. wuellerstorfi for BWT estimates from the Caribbean Sea over the time interval between 5.2 and 2.2 Ma. We have applied femtosecond laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) for trace element analysis. BWTs are correlated to SSTs reported in the previous studies. Sr/Ca data are used to estimate river runoff driven by the uplift of the northern Andes during the same time interval. Bottom-water salinity (BWS) is obtained from Na/Ca of C. wuellerstorfi to reconstruct deep thermohaline circulation of the Caribbean Sea for the first time.

Samples and Hydrography of the Study Area
Samples from Ocean Drilling Program (ODP) Site 999A, located in the Caribbean Sea, Colombian Basin (12°44.639′N, 78°44.360′W) have been analyzed (Shipboard Scientific Party, 1997). Site 999 was drilled during Leg 165, at 2,827.9-m water depth (Figure 1a). The modern Caribbean Basin is filled with saline   Fiedler and Talley (2006), Gröger et al. (2003), Restrepo et al. (2006), and Schmuker and Schiebel (2002). Dashed black lines represent the passages where deep and intermediate waters enter the Caribbean Sea. Dashed rectangle refers to the paleogeography map (b). White line refers to mean annual salinity and temperature east-west oriented depth transects in Figures 2a and 2b, respectively. Map and sections produced using Ocean Data View (Schlitzer, 2018). (b) Simplified paleogeography map modified after Kirby et al. (2008), showing paleohydrography of the Caribbean Sea at~15 Ma according to Kirillova et al. (2019) and Newkirk and Martin (2009). Light gray shaded area represents shoaled areas following CAS closure; dark blue represents deep-water areas.

Trace Element Analyses Using fs-LA-ICP-MS
Tests of the benthic foraminifer C. wuellerstorfi were picked from the >250-μm size fraction and cleaned in the Clean Laboratory at the Max Planck Institute for Chemistry (MPIC) following the Mg-cleaning (oxidative) method (Barker et al., 2003).
For trace element analyses, single specimens of C. wuellerstorfi have been analyzed using femtosecond LA-ICP-MS (fs-LA-ICP-MS) at the Geochemistry Laboratory at the MPIC. Na, Mg, and Sr of 142 specimens of C. wuellerstorfi have been measured at high-resolution on (up to) each of the final five chambers to eliminate possible inaccuracies that are likely to be caused by varying sample size (cf. Elderfield et al., 2002). Average values of all chambers analyzed in the same specimen were used for further analyses (Ogretmen et al., 2020). In the case of multi-specimen analyses (see supporting information), samples with the smallest standard deviation were considered for further analyses. Measurements were repeated on multiple specimens from nine selected samples ranging from two to four specimens chosen from cold and warm time intervals to test the reproducibility of single-specimen data ( Figure S2 and Table S1). After applying oxidative cleaning, specimens were placed on a sand bath (>250-μm-sized pure quartz grains) to avoid contamination of the tests for further analyses. The fs-LA-ICP-MS analyses were performed on 25-μm-diameter spots on each chamber with a pulse repetition rate of 15 Hz at low fluence (0.1-0.3 J/cm 2 ). Calibration was performed with the microanalytical synthetic reference material MACS-3 for carbonate (Jochum et al., 2019) and NIST612 for silica (Jochum et al., 2011). Measurements with fs-LA-ICP-MS yield reproducible results with a relative standard error~3-4%.  (Zweng et al., 2013) and temperature (Locarnini et al., 2013). Depth distribution of the AAIW and NADW are inferred from the mean annual salinity of the less saline AAIW (~34.8 PSU) entering the Caribbean Basin at shallower water depths (~1,000 m) and more saline NADW (~35 PSU) in the deeper Caribbean Basin (>1,000 m). White dashed lines represent the temperature distribution in the eastern and western part of the Isthmus of Panama. Modern-day mean annual temperature in the deep Caribbean Basin below 1,000 m is~5°C. Transect produced using Ocean Data View (Schlitzer, 2018).

Preservation of Samples and Terrestrial Input
Diagenesis and recrystallization of the calcareous foraminifer tests are indicated by Sr/Ca. In addition, Sr/Ca is affected by, and indicator of, terrestrial input due to continental weathering processes. For example, Sr/Ca values increase and decrease in relation to high and low weathering rates, respectively, rather than temperature (Lear et al., 2003;Martin et al., 2000;Rosenthal et al., 1997). Benthic foraminifer Sr/Ca has been discussed as a measure of seawater carbonate chemistry (Dissard et al., 2010;Dueñas-Bohórquez et al., 2011;Raitzsch et al., 2010). Yu et al. (2014) suggested that Sr/Ca of deep-water benthic foraminifers such as C. wuellerstorfi may be used as an auxiliary proxy for deep-water carbonate ion saturation if seawater Sr/Ca is stable. Recently, in an experimental study, Sr/Ca of the shallow benthic Ammonia sp. has been considered as a potential indicator for carbonate system showing positive correlation with bicarbonate ion concentration or dissolved inorganic carbon, DIC (Keul et al., 2017). An increase in atmospheric CO 2 and seawater pCO 2 would result in an increase in DIC and consequently an increase in foraminifer Sr/Ca. Major changes in pCO 2 occurred between 3.2 and 2.5 Ma corresponding to the intensification of NHG at 2.7 Ma (de la Vega et al., 2020;Seki et al., 2010). However, long-term increase in Sr/Ca values ( Figure 4) do not correlate with reconstructed pCO 2 as they do not reveal any significant  change (Seki et al., 2010). Thus, our long-term foraminifer Sr/Ca values are interpreted as an indicator of terrestrial input due to continental weathering. The largest river transporting a significant amount of terrestrial matter from the Colombian Andes into the eastern Caribbean Sea is the Magdalena River (Mora & Martínez, 2005;Restrepo et al., 2006). Therefore, the long-term increase in Sr/Ca may be related to the enhanced terrigenous material input from the Magdalena River due to elevated topography caused by the uplift of the northern Andes and the shoaling of the Isthmus of Panama resulting in reinforced continental weathering (Peters et al., 2000;Shipboard Scientific Party, 1997).

Deep-Water Mg/Ca Paleothermometry
Mg/Ca values of foraminifer carbonate tests are a proxy of temperature changes of ambient seawater (Billups & Schrag, 2003;Nürnberg et al., 1996;Rosenthal et al., 1997). Given the long residence time of Mg in seawater of~13 Myr, Mg/Ca is considered a reliable tool for paleoclimate studies due to salinity-independence of Mg incorporation in the calcite shell (Lear et al., 2000(Lear et al., , 2010; (Billups & Schrag, 2002;Broecker & Peng, 1982;Evans & Müller, 2012). So far, effects of changing seawater Mg/Ca on longer-term BWT estimates (~17 Myr) have only been investigated on the infaunal benthic foraminifer Oridorsalis umbonatus, proving low sensitivity to changing seawater Mg/Ca and suggesting strong biological control on Mg uptake during the calcification process (Lear et al., 2015). Here, the epifaunal species C. wuellerstorfi is used to estimate the temperature variability of the bottom waters on longer time scales, due to its direct contact with bottom water (Jorissen et al., 2007). Since foraminifers from different regions show particular offsets in the Mg/Ca Chamber-to-chamber variations of the trace element ratios related to ontogenetic effects during calcification revealed by LA-ICP-MS display a significant offset from r 2 = 1, that is, r 2 > 0.85 ( Figure 6). Average standard deviation for all the samples is σ = 0.13°C, ranging between σ = 0.6°C and σ = 0.01°C (see supporting information Figure S1). Standard error for the Mg/Ca-BWTs of the all samples amounts to σ = 0.9°C, that is, σ = 1.01°C in Interval I, σ = 0.77°C in Interval II, σ = 0.63°C in Interval III, and σ = 0.87°C in Interval IV. Varying standard deviations comprise analytical uncertainties in the technique and environmental/biological factors such as preservation state of the shell and ontogenetic effect during calcification, respectively (Jochum et al., 2019, and references therein).
As application of the Mg/Ca proxy for paleothermometry has increased, concerns about its reliability and potential biases are discussed. Increasing carbonate ion saturation (Δ½CO 2 − 3 Þ may result in underestimation of the Mg/Ca paleotemperature (Elderfield et al., 2006;Healey et al., 2008;Yu & Elderfield, 2008). For example, low Δ½CO 2 − 3 associated with colder glacial temperatures causes lower Mg/Ca in C. wuellerstorfi calcite, whereas BWTs show weak correlation with Mg/Ca (Yu & Elderfield, 2008). For the Caribbean Sea, Broecker and Clark (2002) reconstructed high ½CO 2 − 3 in the glacial deep waters, whereas the planktic foraminifers have heavier tests as they are enriched in calcium ( 40 Ca) rather than magnesium ( 24 Mg) in cold temperatures. In addition to factors that drive the uptake of Mg during the production of foraminifer shell calcite, dissolution is assumed affect Mg/Ca in Caribbean seafloor sediments below 3,000-m water depth (Regenberg et al., 2006). However, it has been demonstrated that planktic foraminifer tests experience different degrees of alteration during their vertical transport to the seafloor due to the wide range of water chemistries (Schiebel et al., 2007). Therefore, any trace element alteration of planktic foraminifer tests cannot be directly linked to the deep-water masses (Davis & Benitez-Nelson, 2020), and benthic foraminifers cannot be directly compared to planktic foraminifers of differential preservation states.
According to the Mg/Ca-BWT, Sr/Ca, and planktic foraminifer Mg/Ca (N. dutertrei), the development of deep-water masses in the eastern Caribbean during the closure of the CAS between 5.2 and 2.2 Ma may be subdivided in four time intervals (Figures 4 and 7a).
Interval I is characterized by oscillating and rather low BWTs with average BWT around 1.1°C slightly increasing toward 4.1 Ma from 0.5°C up to 2-2.5°C (Figure 7a). This increasing trend in BWT confirms previous studies relating the changes in dominant water masses in the Caribbean Basin from relatively warm and oxygen depleted AAIW, to cooler, more saline, and more oxygenated NADW (Haug & Tiedemann, 1998). A gradual SST increase (Figure 7b) is assumed to result from restricted inflow of relatively cold Pacific waters and increasing SSTs in the Caribbean Sea (Steph, Tiedemann, Prange, et al., 2006). The BWT record presented here occurred at the same time as the formation of the Caribbean Warm Pool around 4.8 Ma from SSTs of 16-17°C to 23-24°C (Steph et al., 2010). The decreasing influence of the AAIW and the increasing contribution of the NADW has also been demonstrated by increasing benthic foraminifer δ 13 C values from 5.2 until 4-3.5 Ma (Figure 7f) associated to enhanced ventilation. Finally, all of these developments, increasing BWTs and SSTs and increasing ventilation, result from an intensification of the large-scale thermohaline circulation.
Interval II, between 4.1 and 3.2 Ma, is characterized by higher and less variable BWTs (around 2.1°C) than the Interval I and can be attributed to the late Pliocene warmth (Figure 7a), corresponding SSTs with high temperatures around 23-24°C (Figure 7b). Gradual SST increase is assumed to result from an increasingly limited inflow of relatively cold Pacific waters leading to the formation of Caribbean Warm Pool with successively increasing subsurface temperatures (Steph et al., 2010;Steph, Tiedemann, Prange, et al., 2006). An overall increased NADW component characterizing the deep-water masses in the Caribbean Basin results in higher sand content, that is, preservation of the sand-sized calcareous tests of planktic foraminifers (Haug & Tiedemann, 1998). Preservation of the calcareous foraminifer tests results from decreased CO 2 − 3 Â Ã and dissolution, as discussed above. Our findings are supported by lowered Nd isotope values in the Caribbean Sea after~4 Ma, indicating an enhanced contribution of NADW carrying less radiogenic waters into the Caribbean Basin (Osborne, Newkirk, et al., 2014). Accordingly, a slight increase in δ 13 C values until 3.5 Ma support our results (Figures 7e and 7f; Haug & Tiedemann, 1998).
The Interval II comprises several cool stages including KM6, M2, MG4, Gi8, and Gi20-22 (Figure 7a), which would be readily explained by an enhanced AAIW contribution and weakened NADW formation during cold stages. However, more AAIW would result in poorer not better carbonate preservation as discussed above (cf. Haug & Tiedemann, 1998). Another explanation is provided by the cooling of northern waters during a short-term re-opening of the CAS interlinked with closing of the Indonesian Seaway and the opening of the Bering Strait (De Schepper et al., 2009, 2013. However, models fail to explain the M2 glaciation if the Pliocene pCO 2 levels did not sufficiently drop from 405 to 220 ppm (De Schepper et al., 2013;Tan et al., 2017), which suggests that the impact of the CAS closure on the M2 glaciation was rather weak. Even if the opening of the Bering Strait and the closure of the Indonesian Seaway are taken into account, without a drastic pCO 2 decrease, an expansion of the Northern Hemisphere glaciers is unlikely (Tan et al., 2017). Also, our findings fail to provide a coherent explanation for such short-term glaciation/cooling intervals.
Interval III represents the final stage of the CAS closure, with increased BWTs (2.7°C) as ensuing response to the ceased fresh-water inflow from the Pacific Ocean. As a result, Atlantic water masses NADW and AAIW filled the Caribbean and heated up in the semi-enclosed marginal basin. Surface waters cooled and got saltier on their way to the North Atlantic (Figure 1), where they sank to depth triggering the thermohaline circulation (De Schepper et al., 2013;Haug et al., 2001). Continuous heating of surface waters caused intensification of thermohaline circulation during the late Pliocene warmth and resulting in increased SSTs and BWTs (Figures 7a and 7b). However, if temperature increase was caused by the contribution of the warmer AAIW, increased carbonate dissolution should be seen instead of increased sand-sized carbonate (i.e., foraminifer tests) preservation as reported by Haug and Tiedemann (1998).

Paleoceanography and Paleoclimatology
ÖĞRETMEN ET AL.

Bottom-Water Paleosalinity: A New Tool for Thermohaline Circulation?
Application of Na/Ca to paleoceanographic interpretations from benthic foraminifer calcite is a novel approach (Bertlich et al., 2018;Mezger et al., 2016;Mojtahid et al., 2019) to reconstruct past thermohaline circulation of the Caribbean Sea. Na/Ca has been measured using fs-LA-ICP-MS. So far, paleosalinity has been calculated from stable isotope data of planktic foraminifers that live in surface waters (Malaizé & Caleye, 2009;Rohling, 2000). Since surface waters are directly affected by precipitation/evaporation processes, salinity differences are often much larger and easier to identify than in deep waters (i.e., Bertlich et al., 2018;Mezger et al., 2016;Mojtahid et al., 2019). Relatively large differences in salinity have also been analyzed from Na/Ca in cultured larger foraminifera and shallow water benthic foraminifers (Geerken et al., 2018;Hauzer et al., 2018;Wit et al., 2013).
As a new approach, Na/Ca as a paleosalinity proxy of deep-sea benthic foraminifers requires confirmation from a well-established method, that is, benthic foraminifera δ 18 O, to assess its value for seawater salinity reconstruction (see supporting information Figure S3. The foraminifer δ 18 O depends on the calcification temperature and the composition, including salinity, of ambient seawater from which foraminifer shell CaCO 3 is precipitated (Lear et al., 2000;Rohling, 2000). The temperature relationship is represented by (paleo) temperature equations (i.e., Shackleton, 1974). Together with Mg/Ca-derived BWT, δ 18 O can be used to deconvolve the temperature component of the equation and compute the residual δ 18 O for the potential seawater paleosalinity.
Due to the much longer residence time of Na in ocean waters in comparison to Ca, the Na record may have less temporal resolution than the Ca record (Lécuyer, 2016). For deep waters, Na/Ca may serve as an indicator of thermohaline circulation depending on the ocean currents and interactions between water masses. Overall changes in the Na/Ca curve of long-term record of deep waters, therefore, may be a promising paleosalinity proxy (see supporting information Figure S3).
The changes observed in the Na/Ca data are in agreement with the above identified four time intervals deduced from Mg/Ca and Sr/Ca (Figure 7g). Interval I between 5.2 and 4.1 Ma is represented by 4.9 mmol/mol Na/Ca; Interval II between 4.1 and 3.2 Ma is characterized by increased Na/Ca values of 5.4 mmol/mol with rather unchanged trend throughout the interval. Interval III between 3.2 and 2.7 Ma shows a slight but significant decrease of Na/Ca to 5.02 mmol/mol. Interval IV between 2.7 and 2.2 Ma, which is considered the onset of the NHG, is represented by the highest here recorded Na/Ca values of 5.5 mmol/mol (Figure 7g). After 5.6 Ma, C. wuellerstorfi, which inhabits oxygenated, moderately saline waters like NADW, was reported dominant species within the benthic foraminiferal fauna in the Caribbean Sea (DSDP Site 502 shown in Figure 1a; McDougall, 1996). The increased paleosalinity values shown here are in agreement with this finding (Figure 7g). The same study reports a decrease in C. wuellerstorfi and increased contribution of stress tolerant genera such as Uvigerina after 4.1 Ma until 3.2 Ma, corresponding to our Interval II, and suggesting increased dominance of Antarctic water masses. Since varying temperatures and salinities caused by replacement/mixing of the NADW and AAIW may not be sufficient to change the benthic foraminifer assemblages, concomitant changes in oxygenation could play an important role for the presence or absence of oxygen-sensitive Uvigerina species (Loubere, 1996). If this was the case, episodically increased AAIW contribution within Interval II (4.1-3.2 Ma) and weakening of the NADW component in the deep Caribbean water bodies might have sufficiently altered the ecological conditions (i.e., increasing T, decreasing oxygen content, and decreasing S) in the benthic habitat affecting the benthic foraminifer assemblage and Na/Ca signature (Figures 7a, 7d, and 7g).
Even though the AAIW may have been the dominant water mass during the cold episodes when the NADW formation was reduced, carbonate preservation was not decreased during this time interval (Haug & Tiedemann, 1998). Therefore, the AAIW does not correlate with increased BWTs and the Na/Ca record. Changing benthic foraminifer assemblages with decreasing abundances of C. wuellerstorfi have been discussed to be related to the vertical flux of organic (phytodetrital) matter to the seafloor and velocity of the bottom-water currents during the time interval between 3.9 and 1.8 Ma (DSDP 502A; Bornmalm et al., 1999). Being characteristic of well-oxygenated conditions, C. wuellerstorfi may indicate low-nutrient and well-oxygenated NADW waters and rather weak stratification of the deep-water bodies during this time interval (Bornmalm et al., 1999;McDougall, 1996).
In general, stratification of the water column increased during warm time periods and decreased during cold periods. Accordingly, the increased SSTs result in enhanced stratification of the upper waters, decreased overall salinities due to freshwater input from the melting ice caps, enhanced riverine water inflow, and decreased ventilation of the deeper parts of the water column starting from at 4.8 Ma (Figure 7b; Capotondi et al., 2012;Steph et al., 2010). Consequently, increased Na/Ca values in comparison to the Interval I may result from the stratification (Figure 7g). However, δ 13 C values do not show any significant decrease, which would be indicative of decreased ventilation (Figure 7d), and require further investigation.

Caribbean Circulation and Water Masses During the CAS Closure
Intensification of the thermohaline circulation triggering the Pliocene warm period between 4.7 and 3.1 Ma is assumed of decisive importance in setting present-day climate conditions (Haug et al., 2001) and characterizing the modern-day Caribbean Sea deep waters by mixing warm and relatively fresh AAIW (2-6°C, 33.8-34.8 PSU) and cold and more saline NADW (~2-4°C, 34.8-35 PSU) (Figures 2a and 2b; Gallegos, 2013). Accordingly, the Na/Ca increase from 4.9 to 5.4 mmol/mol (Figure 7g) might correspond the intensification North Atlantic thermohaline circulation throughout the studied time interval. Caribbean surface waters heated up and got saltier due to the gradually restricted Pacific fresh water inflow and were transported northward via the Gulf Stream and North Atlantic Current transporting heat to northern latitudes and resulting in Pliocene warmth (Figure 7b;De Schepper et al., 2013). Consequently, thermohaline circulation intensified as waters were cooled and became denser given their increased salt content (Haug et al., 2001) resulting in increased bottom-water salinities (Figure 7g). Consequently, the slight decrease in Na/Ca content within the Interval III corresponding to mid-Piacenzian warmth when the global sea level was relatively high (De Schepper et al., 2013;Dwyer & Chandler, 2009) might have caused relatively low Na/Ca as the freshwater input from the melting ice and rivers increased. A major sea surface salinity difference was also observed through planktic δ 18 O circa 3.3 Ma corresponding to the failed M2 glaciation (Haug et al., 2001). Around the same time in Interval II, increasing benthic δ 18 O is in agreement with higher paleosalinity (~5.3 mmol/mol at M2). Increasing benthic δ 18 O values continued through the NHG onset pointing to a cool time interval with enhanced bottom-water salinities and a lower sea level (Figures 7f and 7g; Bintanja & van de Wal, 2008;De Schepper et al., 2013;Haug et al., 2001).

Conclusion
The deep thermohaline circulation in the Caribbean Sea during closure of the Central American Seaway (CAS), from 5.2-2.2 Ma, has been reconstructed using Mg/Ca derived BWTs and Na/Ca through BWS from single tests of C. wuellerstorfi. Trace element analyses have been conducted using fs-LA-ICP-MS.
Our data indicate several cold and warm episodes corresponding to the LR04 benthic δ 18 O marine isotope stages (MISs) and confirm local records from the Caribbean Sea. We subdivide four time intervals of changing BWTs, bottom-water salinities, and river runoff from the South American continent into the Caribbean Sea before and after the closure of the CAS from changes in Sr/Ca and Mg/Ca values, and resulting Mg/Ca-derived BWTs.
1. Interval I during CAS shoaling includes low but gradually increasing BWT and salinity gradients in response to increased North Atlantic Deep Water contribution through the overall Pliocene warming from 5.2 to 4.1 Ma. 2. Interval II is characterized by ceasing inflow of fresh Pacific waters into the Caribbean Basin. Increased Magdalena River runoff indicates uplift of the northern Andes in the time interval from 4.1-3.2 Ma during the late Pliocene warmth. Reinforced thermohaline circulation of more saline deep-water bodies is punctuated by episodic contributions of cooler waters during cold MISs including the M2. However, absolute BWTs may not have changed much, but surface marine conditions might have been altered during long-term changes in SSTs and surface water stratification. 3. Interval III frames the final stage of CAS closure and mid-Piacenzian warmth with increased Magdalena River runoff and BWTs but relatively low paleosalinities and a high sea level.

4.
Interval IV displays intensification of the NHG with declining river runoff and BWTs, and enhancing paleosalinity gradients related to the lowered sea level.