Marine carbon cycle evolution in the eastern equatorial Pacific over the last deglaciation

UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA Facultad de Ciencias del Mar ANEXO I Dª MARÍA ISABEL PADILLA LEÓN, SECRETARIA DE LA FACULTAD DE CIENCIAS DEL MAR, CENTRO RESPONSABLE DEL PROGRAMA DE DOCTORADO EN OCEANOGRAFÍA, DE LA UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA. CERTIFICA Que el Consejo de Doctores del Programa de Doctorado en Oceanografía, en su sesión de fecha 21 de julio de 2016, tomó el acuerdo de dar el consentimiento para su tramitación, a la tesis doctoral titulada: “Marine Carbon Cycle Evolution in The Eastern Equatorial Pacific Over The Last Deglaciation”, presentada por la doctoranda: Dª María de la Fuente García; dirigida por las Doctoras: Dª Eva Calvo Costa y Dª Isabel Cacho Lascorz. Y para que así conste, a efectos de lo previsto en el Artº 6 del Reglamento para la elaboración, tribunal defensa y evaluación de tesis doctorales de la Universidad de Las Palmas de Gran Canaria, firmo el presente en Las Palmas de Gran Canaria, a veintiuno de julio de dos mil dieciséis. Marine Carbon Cycle Evolution in the Eastern Equatorial Pacific over the Last Deglaciation Tesis doctoral presentada por María de la Fuente García para obtener el grado de Doctora por la Universidad de Las Palmas de Gran Canaria Programa de Doctorado en Oceanografía Directoras: Eva Calvo Costa e Isabel Cacho Lascorz Universidad de las Palmas de Gran Canaria Institut de Ciències del Mar (ICM-CSIC) Universitat de Barcelona Barcelona, julio de 2016 La doctoranda La directora La codirectora Portada: CLIMAP Project Members (1981). Seasonal Reconstruction of the Earth’s Surface at the Last Glacial Maximum. Geological Society of America, Map and Chart Series, 36, 18pp. Contraportada: Geographic Map of the Circum-Pacific Region, The American Association of Petroleum Geologists, 1978 Composition by Luke Skinner A todo lo que me hace sonreir.. CONTENTS List of figures 13 List of tables 17 Abstract 19 Chapter 1. General Introduction 23 1.1. Late Pleistocene Glacial/Interglacial cycles 23 1.1.1. Insolation pacing of global temperature change 23 1.1.2. CO2 as a relevant internal feedback for G/IG cycles 27 1.2. Marine carbon cycle 29 1.2.1. Seawater carbonate chemistry 29 1.2.2. Processes that set up sea surface pCO2: “Carbon pumps” 32 1.3. Marine mechanisms proposed for the glacial atmospheric CO2 decrease 37 1.4. Current understanding of G/IG marine carbon cycle change 40 1.5. Thesis objectives and Outline 43 Chapter 2. Materials and methods: the palaeoceanographic toolbox 47 2.1. Study location: ODP Site 1240 47 2.1.1. Sediment core and site description 47 2.1.2. Local modern hydrography and climatology at the Eastern Equatorial Pacific 48 2.1.3. Fossil foraminifera analysed in ODP1240 50 2.2. Proxies and Methodologies 51 2.2.1. Radiocarbon (14C) 51 2.2.1.1. The tracer 51 2.2.1.2. The technique: From the foraminiferal calcite shell to the Accelerator Mass Spectrometer 54 2.2.2. B/Ca ratio 56 2.2.2.1. The tracer 56 2.2.2.2. The technique: Trace elements performed by LA-ICPMS 59 2.2.3. Stable carbon isotopes (d13C) 61 2.2.3.1. The tracer 61 2.2.3.2. The technique: d13C by isotope-ratio (IR) mass spectrometry 61 Chapter 3. New age models for ODP1240 through reservoir age estimates 65 3.1. Introduction 65 3.2. Material and methods 67 3.3. Results and discussion 68 3.3.1. Reservoir age assessment 68 3.3.2. Age models 71 3.4. Conclusions 74 Chapter 4. Shallow sub-surface and deep radiocarbon ventilation in the EEP and its context in the whole south Pacific and Southern oceans 77 4.1. Introduction 77 4.2. Material and methods 82 4.2.1. Radiocarbon dates 82 4.2.2.Radiocarbon calculations in modern water column 82 4.3. Results 82 4.3.1. B-P offsets and reservoir ages 82 4.3.2. Ventilation ages 85 4.4. Discussion 86 4.5. Conclusions 96 Chapter 5. B/Ca foraminifer analyses by LA-ICPMS 99 Subchapter 5.1. LA-ICPMS set up for B/Ca analysis 100 5.1.1. Introduction 100 5.1.2. Material and methods 101 5.1.2.1. LA-ICPMS set up through standard material 101 5.1.2.2. Test-analyses in foraminifer calcite 103 5.1.2.2.1. Analysis by LA-ICPMS 103 5.1.2.2.2. Analysis by Solution-ICPMS 104 5.1.3. Results and discussion 105 5.1.3.1. Standard accuracy through LA-ICPMS 105 5.1.3.2. The importance of boron instrumental background conditions 107 5.1.4. Conclusions 109 17 List of tables Table 1.1. Processes affecting DIC, ALK and pCO2 36 Table 4.1. Radiocarbon ventilation records from the Pacific Ocean 79 Table 5.1. Instrumental operation conditions used in this study 102 Table 5.2. Laser spot analysis data accuracy 107 Table 5.3. Core-top sample information 112 Table 5.4. Seawater chemistry data for the B/Ca-D[CO3 2-] core-top calibration 125 Table A1. Compiled radiocarbon data from core ODP1240 163 Table A2. Reservoir age estimates for ODP1240 164 Table A3. Summary of the radiocarbon dates selected for the estimations of B-P, reservoir ages (R.age) and deep ventilation ages (B-Atm) 165 19 Abstract ABSTRACT Although the glacial-interglacial climate cycles of the late Pleistocene were very likely paced by changes in solar insolation, the full amplitude of these climate cycles was only achieved with the help of strong feedbacks within the Earth system, including variations in the atmospheric CO2 concentration. It is thought that these CO2 changes resulted primarily from perturbations of the marine carbon cycle. Thus, for example, an increase in the efficiency of the marine carbon pumps (more specifically the solubility pump and biological pumps) may have permitted the accumulation of larger amounts of CO2 in the ocean interior, away from the atmosphere. The goal of this thesis is to investigate the existence of such a theoretical stock of CO2 in the glacial ocean and to assess the mechanisms through which it may have arisen. To this end, I have analysed a combination of proxies in sediment core ODP1240 from the Eastern Equatorial Pacific (EEP) over the last deglaciation, all of which related to past changes in ocean interior carbon content. These proxies include: radiocarbon “ventilation ages” as a proxy for deep ocean turn-over time, benthic foraminiferal B/Ca ratios as a proxy for deep water Δ[CO3 2-]in situ-sat (and therefore [CO3 2-]), δ13C as a proxy for respired carbon content, and Δδ13C(epifaunal-infaunal) as an additional estimate of water oxygenation. The results demonstrate that the deepand shallow-subsurface EEP were much more poorly ventilated during the Last Glacial Maximum (LGM) relative to the Holocene epoch. This would suggest an increased residence time for carbon in the deep Pacific and an increase in the efficiency of the marine biological carbon pump via a decrease in its “leakiness”, which could have been further enhanced by an increase in local export productivity during the last glacial period. The increase in deep ocean respired carbon levels that would be expected from these changes, is confirmed by deep-water [CO3 2-] reconstructions (derived here using the novel LA-ICPMS technique), as well as benthic δ13C and oxygenation estimates. All together indicate a more respired CO2 enriched and more oxygendepleted deep EEP during the last glacial period. This greater accumulation of respired CO2 in the glacial ocean would have been achieved at the expense of the 20 Abstract surface ocean and atmosphere. However, the ∆[CO3 2-]LGM-Holocene reconstructed in this study location is relatively small compared to the changes estimated in radiocarbon ventilation for the same period, suggesting that a “counteracting” mechanism, such as carbonate dissolution at the seafloor, should have also played a role. This mechanism would have increased average ocean alkalinity, allowing even more atmospheric CO2 to be “sequestered” by the ocean during the glaciation. An initial comparison with results from the South Pacific and Southern Oceans suggests that a promising avenue for future work might be to extend the approach adopted in this study to a larger number of sites from around the global ocean, with the aim of moving closer to a quantification of the glacial marine respired carbon stock and the proposed increase in average ocean alkalinity. CLIMAP Project Members (1981). Seasonal Reconstruction of the Earth’s Surface at the Last Glacial Maximum. Geological Society of America, Map and Chart Series, 36, 18pp. 23 Chapter 1 CHAPTER 1 1. General introduction 1.1. Late Pleistocene Glacial/Interglacial cycles 1.1.1. Insolation pacing of global temperature change The last 800 kyr are characterised by cold (glacial)-warm (interglacial) cycles of ~100 kyr duration, that ultimately seem to have been paced by insolation variations, the dominant external forcing of Earth’s climate on timescales of 104-105 years (Hays et al., 1976). These insolation variations are produced by changes in the eccentricity, obliquity and precession, which modulate the amount of solar radiation that arrives at the top of the atmosphere across latitude and through the annual cycle (Milankovitch, 1941) (Fig. 1.1). Insolation varies with characteristic periodicities of ~100 kyr for the eccentricity of the Earth’s orbit, 41 kyr for the obliquity or tilt of the Earth’s spin axis and ~23 kyr for the precession of the equinoxes, which are clearly recorded in a host of proxy climate records of past climate change, including the marine stable oxygen-isotope record in particular (Hays et al., 1976; Shackleton, 2000). The striking similarity of the frequency structure of insolation and climate records demonstrates a clear link between the two, and is the basis for the hypothesis that insolation changes have paced past climate change. The original formulation of the “orbital theory of the ice ages” (Milankovitch, 1941) proposes that the intensity of peak summer insolation at 65oN (i.e. the primary latitude of ice sheet growth and decay during the Pleistocene) is crucial for pacing global climate change, via its impact on summer ice sheet melt. Therefore, the local-seasonal redistribution of solar radiation, rather than global average changes in solar energy input that is only affected by eccentricity and amounts to very little (e.g. <0.5 Wm-2), has always been conceived of as the primary means of global climate forcing in this context. 24 General Introduction Despite the very strong correspondence between astronomical/insolation periodicities and climatic variability, it has been observed that the magnitude of insolation changes (the presumed forcing) is not always related to the amplitude of glacial/interglacial transitions (G/IG, the presumed response), e.g. the so-called “Stage 7 and Stage 11 problems”, where a large/small insolation change produces a “counterintuitively” small/large climatic response, respectively (Paillard, 2001) (Fig. 1.1). Furthermore, while summer insolation variability at 65oN in particular is dominated by precession (~23 kyr), the global climate response over the last ~1 Ma has been dominated by a ~100 kyr period (resulting in the so-called “100kyr problem”), while earlier Pleistocene glacial cycles have been dominated by a 41 kyr cycle. Thus, a shift in a dominant periodicity in climate from 41 kyr to 100 kyr at ~1 Ma (referred to as the Mid-Pleistocene Transition, MPT) without any such Figure 1.1. Insolation variations in July at 65oN (orange line; Berger and Loutre, 1991; Berger, 1999) and the LR04 global stack of benthic δ 18O records indicating changes in global climate over the last 1500 kyr (blue line; Lisiecki and Raymo, 2005). Odd numbers at the top of the panel indicate Marine Isotope Stage (MIS) for interglacial states, while vertical grey bands highlight glacial states. 0 100 200 300 400 500 600 700 800 900 1000 1100 1200 1300 1400 1500 380 400 420 440 460 480 500 Insolation July 65oN (W/m2) Time (kyr) 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 5.0 4.5 4.0 3.5 3.0 Benthic δ18O (permil) 25 Chapter 1 change in insolation presents a further enigma, and points to a “non-linear” process linking insolation and climate change (Imbrie and Imbrie, 1980). Indeed, climatic archives show that G/IG cycles, apart from being clearly linked with changes in solar insolation, are also marked by changes in ice-sheet volume/sea level, landcover, atmospheric dust loading and atmospheric CO2 (Fig. 1.2), all of which affect global energy balance via changes in planetary albedo and emissivity. In light of these shortcomings of the Milankovitch theory, the contemporary conception of glacial-interglacial climate change typically emphasizes that the effects of G/IG climate change are also largely the drivers of G/IG climate change. Another way of saying this is that while insolation may act as the pacemaker of G/ Figure 1.2. Individual radiative forcings from vegetation, atmospheric dust loading, sea ice, GHGs (including CO2, CH4 and N2O) and land cryosphere (including ice-sheets and associated snow cover and sea level change) over the last 800 kyr. Horizontal dashed grey line notes 0 W/m2 (adapted from Köhler et al., (2010)). 0 100 200 300 400 500 600 700 800 -5.0 -4.5 -4.0 -3.5 -3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 Radiative forcing (W/m 2 ) Time (kyr BP) Insolation Vegetation Dust Sea ice GHGs Land ice 32 General Introduction (5) The equilibrium constants, as mentioned previously, are controlled by the temperature and salinity of the seawater, which are strongly influenced by local atmospheric temperature, evaporation-precipitation balance and global ice-volume change. The temperature/salinity dependence of the dissociation constants causes CO2 solubility to increase for colder and fresher water, such that lower temperature and salinity will lead to increased seawater pCO2 for a given DIC and ALK, and vice versa. However, what process(es) control DIC and ALK, and therefore ultimately the seawater pCO2 regardless of temperature and salinity? Understanding these other processes, and their link to climate-driven changes in CO2 solubility and air-sea gas exchange is central to the study of the climate of the past and its relation with the carbon cycle, including in particular the G/IG atmospheric CO2 changes observed in the late Pleistocene. 1.2.2. Processes that set up sea surface pCO2: “Carbon pumps” Modern vertical profiles of DIC and ALK across all oceans show a gradient between the surface and the ocean interior, as well as latitude/longitude differences in surface ocean DIC and ALK. These patterns suggest the existence of active processes in the carbonate system that maintain chemical gradients against the homogenizing influence of the ocean circulation (Sarmiento and Gruber, 2004). The processes that determine the vertical gradients in the oceans are often referred to as “marine carbon pumps” (based on the analogy of a pump that maintains a pressure gradient between two locations), which are commonly subdivided into the “solubility pump” and the “biological pump”, where the latter is further subdivided into the “soft tissue-” and “carbonate-” pumps (Fig. 1.5) (Volk and Hoffert, 1985). The so-called “solubility pump” has been partly addressed in the previous subsection, as it refers to the capacity of the ocean to hold CO2 away from the atmosphere pCO2≈K2 K 0 ⋅K 1 2DIC −ALK ( ) 2 ALK −DIC 33 Chapter 1 as a function of temperature and salinity of seawater and in the absence of any processes that might otherwise create DIC/ALK gradients in the ocean. The solubility pump therefore depends on air-sea CO2 exchange efficiency ((Wanninkhof, 1992) and CO2 solubility. It could be argued that the term “solubility pump” is not entirely appropriate in this case, as these processes do not act as pumps that maintain chemical gradients, but rather the opposite; they just equilibrate the pCO2 of seawater Figure 1.5. Schematic representation of the marine carbon pumps, including the solubility and biological pumps, where the latter is further subdivided into the soft-tissue and carbonate pumps. The arbitrary division between the surface and interior ocean is indicated by a horizontal dashed line. Yellow arrows and text indicate processes/mechanisms that tend to contribute to atmospheric CO2 uptake, while purple arrows and text indicate processes/mechanisms that tend to contribute to CO2 release. Grey boxes summarize the effect of such processes/mechanisms on DIC and ALK. Blue circles represent both organic (Corg) and calcium carbonate (CaCO3) particles, whose size represents the rapidly diminishing degree of oxidation/dissolution with depth, as the particles sink (modified from Heinze et al., 1991). Surface ocean Ocean interior Sediment Solubilitypump “Softtissue”pump “Carbonate”pump Biologicalpump Carbonpumps CaCO 3  bioformation U P W E L L I N G Coldfresh CO 2 Warmhaline CaCO 3 burial C org formation (Photosynthesis) U P W E L L I N G Corgflux C org oxidation (Respiration) é DICê DIC  NoeffectinALK ê DIC  é ALK ê DIC êê ALK CaCO3flux CaCO 3  dissolution é DIC  éé ALK é DIC êALK C org burial .. .. ... .. .. .. ... .. CO 2 CO 2 Atmosphere 34 General Introduction with that of the atmosphere as a function of seawater properties. Indeed, they are only responsible for the vertical gradients insofar as a certain amount of DIC can be directly advected into the deep ocean by the large scale overturning circulation (which occurs only in specific areas of the planet), with impacts on DIC but not on ALK. Because of this, the effects of the solubility pump are directly dependent on the ocean circulation. This notion has led some to use the term “physical pump”, in order to more clearly emphasize the important role played by ocean physics, which would combine the effects of solubility with those of ocean transport. Thus, although solubility and air-sea exchange processes are very important for oceanatmosphere carbon cycling, it is the ocean circulation and its interaction with the “soft-tissue” and the “carbonate” pumps that are responsible for most of the DIC and ALK gradients that are observed in the water column (Fig. 1.6). Figure 1.6. Graphical relationship between dissolved inorganic carbon (DIC) and total alkalinity (ALK) in the ocean at constant temperature (T = 15oC), salinity (S = 35) and pressure (P = 1 atm). The processes that modify carbonate chemistry parameters are displayed by black arrows. CO2 concentration contours (in µmol/kg) are indicated by diagonal white lines and rainbow colours (Figure from Zeebe, 2012). 35 Chapter 1 The “soft-tissue pump” maintains a chemical gradient in the water column by fixing CO2 from the upper ocean through photosynthesis and releasing it back into the ocean interior through organic matter respiration (also called oxidation). More specifically, the photosynthetic organisms that live in the upper photic zone of the ocean use light, water, nutrients and CO2 to grow, thus turning dissolved inorganic carbon (DIC) into particulate organic carbon (POC or organic matter), that is transferred through the marine food web in the upper layers. When the photosynthetic and non-photosynthetic organisms die, they sink so that POC is exported into the deep ocean where it is progressively degraded back into CO2 and nutrients by microorganisms. A small percentage of this organic carbon remains captured in marine sediments, and is eventually returned to the ocean, for instance, via volcanism and more recently anthropogenic emissions. Photosynthesis therefore fixes CO2 making the DIC decrease in the sea surface, while respiration produces a commensurate increase in DIC in the deep ocean by releasing this CO2 into the ocean interior (Fig. 1.6). These processes also produce a slight change in ALK (in the opposite direction to DIC) as they consume/release protons in association with nutrients fixation/remineralization (charged species such as NO3 -) (Fig. 1.6). Thus, photosynthesis (respiration) is analogous to CO2 release to (invasion from) the atmosphere in the way that it decreases (increases) DIC, but photosynthesis (respiration) differs as it also affects ALK slightly. The processes mentioned so far have primarily produced changes in DIC, with no change or relatively little in ALK. What therefore creates the significant vertical and lateral ALK gradients observed in the ocean? The main process responsible for these gradients is the other biological pump called the “carbonate pump”, which refers to the formation/dissolution of biogenic calcium carbonate (CaCO3) in the surface/deep ocean, respectively. Calcifying organisms such as coccolithophores and foraminifera build their shells by removing CO3 2and Ca2+ from the surface (i.e. CO3 2 + Ca2+ ↔︎ CaCO3), which are returned to the deep ocean when they die and sink. Thus, carbonate formation in the sea surface reduces both DIC and ALK, and vice versa in the deep ocean, affecting each in a ratio of 1:2 (DIC:ALK) (Fig. 1.6). 36 General Introduction Having surveyed the processes that influence DIC and ALK distributions in the ocean, we can go back to what concerned us in the beginning, i.e. what can drive significant changes in seawater pCO2? Looking at equation 5 and Figure 1.6, seawater pCO2 would decrease with decreased DIC (from CO2 release to the atmosphere and/or enhanced photosynthesis), as well as with increased ALK (from increased CaCO3 dissolution), and vice versa (see Table 1.1 for a summary).   Solubilitypump Biologicalpump SofttissuepumpCarbonatepump Invasion CO2 Release CO2 Production OrganicC (Photosynt.) Oxidation OrganicC (Respirat.) Formation CaCO3 Dissolution CaCO3 DICéêêéêé ALK  éêêêéé pCO2 (seawater)éêêééê   Table 1.1. Processes affecting DIC, ALK and pCO2. Decreasing arrows are indicated in grey and increasing arrows in black. The size and number of arrows for DIC and ALK indicate the magnitude of change, as also shown in Figure 1.6. This shows a counter-balance effect of the carbonate pump that exists within the biological pump, such that the impact on surface ocean pCO2 will depend on the abundance of calcifying vs non-calcifying organisms in the upper layers of the ocean. Nevertheless, the biological pump will always tend to decrease the pCO2 of the upper ocean, as the bulk of biological export productivity consists of organic carbon rather than carbonate (the latter being about a quarter of that of organic carbon flux; Li et al., 1969), with DIC:ALK being exported from the mixed layer on average at a ratio of ~0.2-0.3 (Kwon and Primeau, 2008). In the ocean interior, however, the balance of DIC and ALK supply at a given depth will depend more on the different remineralization profiles that apply to organic carbon and carbonate, which will cause the ratio of DIC:ALK delivery to the ocean interior to vary with 37 Chapter 1 water depth. A further influence will come from changes in the position of the carbonate compensation depth (CCD), which can be caused by changes in the ocean circulation and the soft-tissue pump and can result in the addition of carbonate ions to the deep ocean via the dissolution of previously deposited carbonate sediments. Having reviewed the processes that control the current partitioning of CO2 between the oceanic and atmospheric carbon reservoirs, the proposed mechanisms for explaining the atmospheric CO2 drawdown over the last glacial period are in turn addressed below. 1.3. Marine mechanisms proposed for the glacial atmospheric CO2 decrease An atmospheric temperature drop during the glacial inception due to solar insolation variations and attendant radiative feedbacks would have cooled the sea surface, thus enhancing CO2 solubility and causing an estimated drop of ~30 ppm in the atmospheric CO2 content (Sigman and Boyle, 2000). However, reduced glacial temperatures would have also been accompanied by a saltier ocean on average, due to the accumulation of fresh water in the ice-sheets, which would have partially counter-balanced the temperature-driven enhanced solubility of CO2, adding ~18 ppm to the atmosphere (Sarmiento and Gruber, 2004; Sigman and Boyle, 2000). Thus, glacial estimates taking into account lower temperature and higher salinity in the sea surface, along with the lower content of carbon in terrestrial vegetation mentioned previously, would end up producing a net CO2 change close to 0 ppm (Sigman and Boyle, 2000). However, it is worth noting that these calculations depend on assumptions regarding air-sea exchange efficiency and ocean circulation (as discussed above), whereas: 1) air-sea gas exchange may well have been modified in glacial periods by other physical processes such as sea-ice cover and ocean stratification; and 2) the precise impact of altered solubility of CO2 in the glacial ocean will have depended on changes in the ocean circulation (Ito and Follows, 2013; Skinner, 2009), which 38 General Introduction remain uncertain. Nevertheless, it is clear that lower glacial atmospheric CO2 cannot be accounted for easily by temperature, salinity and terrestrial biosphere changes alone. In light of the above issues and given that the ocean seems to have been the key place to store a significant amount of CO2 over glacial times, several marine carbon cycle mechanisms have been invoked to explain most of the atmospheric CO2 drawdown observed over the last glacial period (e.g. Kohfeld and Ridgwell, 2009; Sigman and Boyle, 2000). They can be categorised broadly as deriving primarily from: 1) changes in CO2 solubility and the rate or efficiency of air-sea gas exchange (i.e. the “solubility pump”; e.g. Volk and Hoffert, 1985); 2) changes in the efficiency of the “biological carbon pumps” (i.e. “soft-tissue” vs “carbonate” pumps; e.g. Volk and Hoffert, 1985); 3) changes in whole ocean carbonate chemistry (e.g. Archer, 1991; Broecker and Peng, 1989; Keir, 1995). The specific mechanisms included in these categories are many, but the most significant ones are listed below: Category 1. a) Change in sea-ice cover at high southern latitudes (Stephens and Keeling, 2000); b) Change in the rate of ocean overturning due to an altered windfield (Toggweiler, 1999); c) Change in ocean density-stratification due to brine formation in the Southern Ocean (Adkins et al., 2002; Roberts et al., 2016); d) buoyancy forcing in the Southern Ocean (Ferrari et al., 2014; Watson and Naveira Garabato, 2006); e) solubility effects (Sigman and Boyle, 2000); f) volumetric effects (Skinner, 2009). Category 2. a) Increase in biological export productivity due to increased iron supply from dust (Calvo et al., 2004; Martínez-Garcia et al., 2014); b) increased Si-leakage from the Southern Ocean to lower latitudes (Brzezinski et al., 2002; Matsumoto et al., 2002); c) remineralization depth changes (Kwon et al., 2009; Matsumoto, 2007); d) whole ocean phosphate changes (Broecker, 1982b; Wallmann, 2010). Category 3. a) Increase in the ocean’s average alkalinity due to carbonate compensation/carbonate dissolution at the seafloor (Keir, 1995) and/or loss of biogenic carbonate and coral deposition in marine shelf areas (Berger, 1982; Opdyke 39 Chapter 1 and Walker, 1992); b) Increase primarily in the surface Southern Ocean’s alkalinity due to changes in the AMOC (Broecker and Peng, 1989); c) rain ratio (Archer and Maier-Reimer, 1994). How does the ocean produce the above changes to exert a control on atmospheric CO2? Are physical mechanisms primarily responsible for controlling atmospheric CO2, or is it marine biology that plays the dominant role, or perhaps changes in the whole ocean chemistry do? While the categories described above might be useful for organising our thinking about past marine carbon cycle change, it is important to emphasize that all these mechanisms need not have operated in isolation from each other. Furthermore, despite substantial differences between all of them, most would directly or indirectly involve changes in the efficiency of the biological pump (Volk and Hoffert, 1985). Defining biological pump efficiency in terms of its ability to store CO2 in the deep ocean away from the surface ocean and thus from the atmosphere, we find that such changes could be achieved by variations in the “leakiness” of the biological carbon pump (the soft tissue pump primarily, as opposed to the carbonate pump) or by changes in its “strength”. Here, leakiness can be defined as the return flux of remineralised nutrients and carbon from the deep interior to the sea surface achieved as a result of the ocean’s large scale overturning circulation and upwelling, and strength refers to export production rates. Thus, for example, the hypothetical mechanisms listed above in category 1) would refer primarily to changes in the leakiness of the soft tissue pump, while those in category 2) would imply changes in its strength. The mechanisms listed under category 3) could in principle arise as a positive feedback on soft-tissue pump efficiency (i.e. via respired carbon accumulation in the deep ocean driving carbonate ion decrease and carbonate dissolution), although they could also arise independently as a result of a reduction in carbonate deposition. A more efficient biological carbon pump can occur as a result of either an increase in its strength or a reduction in its leakiness, with a direct impact on surface ocean 40 General Introduction DIC and ALK and therefore atmospheric CO2. However, if increased biological pump efficiency is achieved by a change in the ocean’s large-scale overturning circulation, it would very likely co-occur with a change in the conditions of air-sea gas exchange and the solubility pump. Furthermore, regardless of how biological carbon pump efficiency is increased (and as long as the soft-tissue pump remains dominant), this would lead to more respired carbon storage in the deep ocean (thus sequestering CO2 from the atmosphere), which could eventually reduce deep ocean carbonate ion concentrations enough to cause seafloor carbonate dissolution. The latter would eventually increase the whole ocean alkalinity budget and lead to further atmospheric CO2 drawdown. This illustrates the point made above, that the various processes (physical, biological and sedimentary/geochemical) need not, or perhaps even could not, have operated in isolation from each other. Indeed, an increase in the biological carbon pump efficiency, due to a decrease in its leakiness (possibly in association with a reduction in air-sea CO2 exchange rates) and/or an increase in its strength, leading to an increase in the ocean’s average alkalinity as a consequence, represents a particularly effective way to cause an increase in deep-ocean carbon storage over the last glacial period, which would involve a combination of physical, biological and geochemical processes. 1.4. Current understanding of G/IG marine carbon cycle change Theory points to CO2 variability as the most important mechanism to amplify insolation-albedo driven G/IG climate changes, and the ocean as the key reservoir for storing and releasing atmospheric CO2. Most of the mechanisms proposed have a base in modeling studies (both conceptual and complex numerical models), which are extremely useful for hypothesis development, but which need to be assessed against observations. We currently lack a complete explanation of the causes of G/IG atmospheric CO2 variations, let alone a clear confirmation of the hypothesis that a significant amount of additional CO2 was indeed stored in the ocean interior during the last glacial. Thus, in a practical way, evidence from real data must be 41 Chapter 1 sought. Are we able to find such evidence in the deep glacial ocean? What areas might be important to look at? What proxies might be essential to infer the state of the carbon cycle and the marine carbon inventory in the past? Are there novel potential successful techniques to explore that might help us to understand proxies better? The Southern Ocean is thought to be a key place to look for such evidence, as it is currently the area where “old” water masses highly charged in respired CO2 make their first contact with the atmosphere and where the bulk of the ocean interior undergoes air-sea exchange (Gebbie and Huybers, 2011; Primeau, 2005). However, other locations are equally important as monitoring locations, including the deep Pacific for example, which is relatively homogeneous and accounts for the bulk of the ocean volume, or the Eastern Equatorial Pacific which is directly connected to both the Southern Ocean and the North Pacific via the transport and upwelling of waters derived from these high latitude regions. There are some specific tracers for carbon cycle reconstructions that are extremely useful when combined together. Radiocarbon activities from the bottom of the ocean compared to the coeval radiocarbon in the atmosphere over time is one of these, as it gives us an idea of the ventilation state of the ocean, whereby higher radiocarbon differences between the two reservoirs indicate longer carbon residence times in the ocean interior and therefore a greater potential for respired CO2 accumulation via the biological carbon pump (see next chapters for further explanation). If some carbonate system parameters such as [CO3 2-] or pH can also be inferred, as well as approximations of the amount of carbon respiration in the deep ocean through variations in O2 content or stable carbon isotopes (d13C), then relative, or even quatitative changes in ocean interior respired carbon storage could in principle be constrained. Using such techniques, evidence for a larger deep ocean CO2 inventory during the last glaciation has been advanced for the Atlantic sector of the Southern Ocean, for example based on radiocarbon evidence of reduced ventilation across the 48 Material and Methods 2.1.2. Local modern hydrography and climatology at the Eastern Equatorial Pacific The Panama Basin presents two main inflow passages for the deepest waters; one along the Ecuador Trench and another one across the broad central saddle of the Carnegie Ridge (sill depths ~2,900 m and ~2,300 m, respectively) (Laird, 1971; Lonsdale and Malfait, 1974). Thus, sediment core ODP1240 is currently bathed by a mixture of Upper Circumpolar Deep Water (UCDW) and Pacific Deep Water (PDW; also called Common Water (Montgomery, 1958)) as both occupy approximately the same density and depth range in the Pacific (Talley et al., 2011) (Fig. 2.1). Both UCDW and PDW are characterized by low oxygen (especially the UCDW), high nutrients and high 14C age although their origins are different. PDW is formed internally in the North Pacific through very slow vertical mixing of bottom and intermediate waters, while UCDW represents a mixture of PDW and Indian Deep Water (IDW) that upwells just below the surface in the Antarctic Circumpolar Current (ACC) (Talley et al., 2011). The shallow sub-surface circulation over the core location is governed by the Equatorial Undercurrent (EUC) between 50-300 m water depth flowing eastward along the tropics (Lukas, 1986; Toggweiler et al., 1991; Tsuchiya et al., 1989). The EUC is partially fed by intermediate water masses (Antarctic Intermediate Water (AAIW) and Subantarctic Mode Water (SAMW)) originated from upwelled Circumpolar Deep Water (CDW) in the Southern Ocean (Kessler, 2006), that are subsequently subducted by winter convection across the Antarctic Polar Front (APF) and Subantartic Front (SAF), respectively, spreading its signature into the Pacific interior. The EUC is also partially fed by Subantarctic Water (SAAW) which is a central water mass originating in the Subantartic zone that subducts at the Subtropical Front (STF; around 30°S) with its core flowing at 150-200 m water depth within the Subtropical gyre (Llanillo et al., 2013; Tsuchiya and Talley, 1998). The surface circulation in this area is a complex system of jets flowing eastand 49 Chapter 2 westward. ODP1240 is situated south of the equatorial front in the equatorial divergence zone (Mix et al., 2003), where warm and low-salinity water masses confront with the cold saline waters from the so-called “Cold Tongue”. The latter is the result of convergence of the cold upwelled waters in this area and the South Equatorial Current (SEC) flowing westward between 4°N and 20°S. The ITCZ is also located in this region a bit north of the equator mainly due to the American geometry continent, but fluctuates over time depending on ocean-atmosphere interactions (Philander et al., 1996), like for instance ENSO (El Niño-Southern Oscillation system), and thus making the latitude surface salinity to change as the ITCZ position controls the amount of precipitation. Latitude Depth Depth Salinity 14C NEC NECC EUC ACC SPC PC SEC AAIW UCDW PDW SAAW P19 UCDW diffusion ODP1240 South America AAIW UCDW PDW SAAW CHC P19 WOCE section Figure 2.1. EEP hydrography. In the left panel, a map of the area displaying the bathymetry (blue scale), the location of the sediment core ODP1240 (red dot), the WOCE P19 transect (yellow meridional line), and a scheme of the upper (black arrows) and deep (white arrows) ocean circulation of the area: North Equatorial Current (NEC), North Equatorial Countercurrent (NECC), Equatorial Undercurrent (EUC), South Equatorial Current (SEC), Peru Current (PC), South Pacific Current (SPC), Cape Horn Current (CHC), Antarctic Circumpolar Current (ACC) and the northward diffusion of the Upper Circumpolar Deep Water mass (UCDW). In the right panel, two vertical sections of the WOCE P19 transect showing salinity and ∆14C. Water masses are indicated: Antarctic Intermediate Water (AAIW), UCDW, SubAntarctic Water (SAAW) and Pacific Deep Water (PDW). The red vertical dashed line indicates ODP1240 location. 50 Material and Methods Figure 2.2. Images of the foraminifer species used in this thesis. The left panel shows the benthic foraminifer Cibicides wuellerstorfi (Schwager, 1866), Source: www.foraminifera.eu, Author: Robert W. Jones. The right panel shows the planktonic foraminifera Neogloboquadrina dutertrei (d’Orbigny, 1839), Source: www.marinespecies.org, Author: Bruce Hayward. The white horizontal line at the bottom left of each panel represents ~100 µm length. 2.1.3. Fossil foraminifera analysed in ODP1240 The elemental composition of the foraminifer’s calcite shell has been widely used as a proxy for past environmental conditions of seawater. These organisms build their shell depending on the hydrographic, geochemical and climatic conditions, which can be inferred by the analysis of different elements relative to calcium, or by the isotopic composition of their tests for instance. The two foraminifer species used to develop this thesis are the benthic species Cibicides wuellerstorfi and the planktonic species Neogloboquadrina dutertrei. C. wuellerstorfi (Schwager, 1866) is an epibenthic foraminifer species that usually lives attached to structures on the seafloor (Lutze and Thiel, 1989) and has a robust calcareous shell with conspicuous hooked sutures (Fig. 2.2a). It has a trochospiral form being different between sides; the dorsal side is rather flat while the ventral side is conical, with perforations and a bilamellar radial wall (Galloway, 1933). It has 51 Chapter 2 been widely used in paleoceanographic studies as its habitat at the sediment-water interface makes it an ideal recorder of bottom water properties. N. dutertrei (d’Orbigny, 1839) is a planktonic foraminifer species with a variable number of globular chambers that grow in a spiral form (Parker, 1962) (Fig. 2.2b). Its habitat is mainly restricted to tropical areas, with temperatures above 21-22°C, although it may also be found in some subtropical areas, and with no apparent specific range of salinity (Hilbrecht, 1997). It is common at the deep chlorophyll maximum (Fairbanks et al., 1982), as well as in upwelling zones in warm surface waters, and its diet is based on phytoand zooplankton prey (Hemleben et al., 1989). 2.2. Proxies and Methodologies 2.2.1. Radiocarbon (14C) 2.2.1.1. The tracer Radiocarbon (14C) is a natural radioactive isotope that is produced in the stratosphere by the interaction of cosmic rays with nitrogen (14N) (Lal and Peters, 1967). As any other carbon isotope, 14C atoms get oxidised forming molecules of CO2, which spread to all carbon reservoirs including the ocean. Radiocarbon is a very useful and widespread tool for chronometric dating and as a carbon cycle tracer, since it is a radioactive isotope and decays over time with a half-life of 5,730 yr (Godwin, 1962) (although a “conventional” half-life of 5,568 yr (Libby, 1955) is used for reporting radiocarbon ages). The 14C budget in the atmosphere is not constant over time due to variations in its production rate, which in turn are due to changes in the solar and earth’s magnetic fields (which shield the earth from cosmic radiation) and changes in solar intensity (Kovaltsov et al., 2012), as well as due to variations in the exchange of carbon between different reservoirs with different characteristic radiocarbon activities. Indeed, 52 Material and Methods using records of other cosmogenic isotope production that do not participate in the carbon cycle like 10Be and 36Cl (Hughen et al., 2006; Muscheler et al., 2005), it has been shown that the observed atmospheric radiocarbon activity changes over the last ~40 ka cannot be explained by production rate changes alone, suggesting that carbon cycle changes also contributed to past atmospheric radiocarbon variability. To assess the occurrence of such carbon cycle changes, reconstructions of the marine radiocarbon inventory are essential, as the ocean is the biggest reservoir of carbon and radiocarbon, so that small variations in its degree of carbon exchange with the atmosphere should produce significant (and opposite) changes in the atmospheric radiocarbon activity over time. Radiocarbon goes into the ocean through atmospheric CO2 dissolution in the surface of the ocean depending on the partial pressure difference of this gas between the atmosphere and ocean, which in turn depends on physical and chemical seawater properties (see Chapter 1 for further details). Once in seawater, the carbon atoms that are part of the new chemical species formed when CO2 is dissolved in seawater are transported by ocean circulation, while radiocarbon decays. Thus, in general terms, radiocarbon can be used as a tracer of the “age” of any parcel of water in the ocean. It is important to note however, that any given radiocarbon age estimate in a water parcel will be the result of the mixing of different water masses, each one contributing with its own radiocarbon signal depending on its specific transport time (the amount of time taken to arrive to that specific parcel), as well as with its correspondent atmospheric radiocarbon activity and air-sea gas exchange rate at the moment each water parcel sank into the ocean interior. Despite this complexity, radiocarbon measurements provide very useful information about the amount of time, on average, that a parcel of water (or more accurately the carbon in that water) has been out of contact with the atmosphere. In order to take account of the fact that the timescale for air-sea radiocarbon equilibration is different (i.e. ~100x longer) than that for e.g. oxygen, the radiocarbon activity of a water parcel is commonly referred to in terms of “radiocarbon ventilation”. There are several metrics to estimate the radiocarbon ventilation of a specific 53 Chapter 2 water parcel, which in addition can be applied over time allowing the evolution of such ventilation to be inferred. One of the most straightforward and available metrics to measure radiocarbon ventilation, that is applied in this thesis, is the radiocarbon offset between coeval planktonic and benthic foraminifera (B-P). When measured along a marine sediment core, this metric can provide an estimate of the changing deepversus surface ocean radiocarbon ventilation gradient over time (Fig. 2.3). In the same way, the radiocarbon offset between planktonic foraminifera and the atmosphere (often referred to as the surface ocean ”reservoir age”) offers information about the degree of equilibration between the marine and 14C(bent 14C(plk B$P P$Atm!!! (Reservoir!age) B$Atm!!!!!!!!!! (Ventilation!age) VENTILATION(AGE(via(B5P(14C(offsets 14N 14C 14CO2( (atm) Figure 2.3. Simple schematic representation of the atmosphere-ocean radiocarbon (14C) cycle and the utility of radiocarbon as a tracer for ventilation age estimates via benthic-planktonic foraminifera (B-P) and benthic-atmospheric (B-Atm) radiocarbon offsets. Radiocarbon atoms, which are produced in the atmosphere by cosmic ray impacts on 14N atoms, incorporate into CO2 molecules (green circle), which may enter into the surface ocean where they will gradually decay back to N atoms as they are advected into the ocean interior, away from exchange with the at- 54 Material and Methods atmospheric reservoirs. However, reservoir ages are not always easy to constrain, as they need independent calendar age control to be accurately estimated, which are often not available or are difficult to obtain depending on the region for instance (see Chapter 3 for further details). This is the reason why modern regional values are commonly applied over time, under the implicit assumption that local air-sea equilibration and sub-surface carbon entrainment have not changed over time. In any case, by combining these two radiocarbon offsets it is possible to infer deep ocean ventilation changes relative to the atmosphere (i.e. benthic-atmosphere radiocarbon offset, B-Atm) across the last 40-50 kyr, the upper dating limit of this technique (Fig. 2.3). All three of these metrics can help to shed light on the glacialinterglacial carbon cycle and ocean circulation changes, as well as their role in past climate dynamics. 2.2.1.2. The technique: From the foraminiferal calcite shell to the Accelerator Mass Spectrometer In order to be able to measure the radiocarbon content of foraminifer shells, the carbon from the carbonate shell has to be concentrated in graphite form, which can readily be ionized in an (ion counting) accelerator mass spectrometer (AMS). Samples from this thesis were graphitised in the radiocarbon preparation facility set up at the Godwin Laboratory for Palaeoclimate Research at the University of mosphere. The size of the red circles represents the level of radiocarbon activity (the smaller, the little activity, i.e. less radiocarbon, and vice versa). Vertical white arrows show the metrics applied in this thesis to estimate ventilation ages (B-Atm) based, in a first step, on the radiocarbon offset between benthic and planktonic foraminifera (B-P), and secondly, on the radiocarbon offset between planktonic foraminifera and the atmosphere, the so-called “Reservoir age”. Note that radiocarbon activity at the shallow sub-surface is already lower than in the atmosphere (represented by different red circle size), which is why reservoir age estimates are required for accurate ventilation ages, as well as accurate age models. 55 Chapter 2 Cambridge, UK for the analysis of small calcite samples (Fig. 2.4) (Freeman et al., 2016). The first step consists of dissolving the calcite shells in phosphoric acid and thereby releasing the carbon content as CO2. The CO2 is then introduced into a vacuum line, removing any water by contact with an ethanol bath at -60°C. Once into the vacuum line, the samples are sized and frozen down using liquid nitrogen into a reactor containing pre-conditioned iron powder and water-absorbing magnesium perchlorate. Pure hydrogen is then introduced into the reactor and heated to 550-650°C for up to 7 hours, causing the CO2 to reduce to graphite, which grows in filaments on the iron catalyst, while water that is produced as a reaction by-product is absorbed by the magnesium perchlorate (Santos et al., 2004). The graphite/iron mixture produced is then pressed into stainless steel cathodes backed by aluminium wires, which are loaded into wheels for subsequent AMS analysis. Sample unknowns are typically graphitised in parallel with a set of oxalic acid (OXII) primary standards, a set of secondary standards (e.g. IAEA standards) and a set of “radiocarbon blanks” for assessing background levels. Samples and standards of various sizes are used in order to correct for size-dependent modern and dead carbon contamination (Santos et al., 2007). Figure 2.4. a) The radiocarbon preparation facility at the Godwin Radiocarbon Laboratory (University of Cambridge), showing the reactors, the vacuum line, the heaters and the tubes that contain the Mg-perchlorate and the Fe powder where the CO2 reacts. b) Schematic diagram of the reactor setup. (Image and description from Freeman et al., 2016). 56 Material and Methods The samples were then sent for analysis by accelerator mass spectrometry (AMS) at the Chrono Centre, Queen’s University Belfast, UK. In this centre, radiocarbon ages are calculated using the on-line d13C value measured in the AMS, following Stuiver and Polach (1977), which permits to monitor any fractionation that may occur in the accelerator. Thus, the 14C/12C and 13C/12C ratios were measured on a NEC 0.5 MV AMS and afterwards, the 14C/12C sample was background corrected and normalised to the OXII standard (Reimer et al., 2004). In this thesis, this technique has been applied to coexisting planktonic (N. dutertrei) and mixed benthic foraminifera from ODP1240, both in order to derive new age models for the core by taking into account plausible reservoir ages (described in Chapter 3), and to estimate benthic-planktonic and benthic-atmosphere radiocarbon age-offsets as indicators of past ocean ventilation changes (described in Chapter 4). 2.2.2. B/Ca ratio 2.2.2.1. The tracer The boron to calcium ratio (B/Ca) in foraminiferal calcite has been theoretically and experimentally proposed as a proxy for the carbonate system. The theory of the B/ Ca is rooted in the dissolved boron species in seawater and their relative concentrations and isotopic composition dependence on pH (Hemming and Hanson, 1992) (Fig. 2.5). Thus, a theoretical relationship between B/Ca and the carbonate system might be summarised as follows: 1) Boron is mainly found in seawater in two dissolved molecular species, boric acid (B(OH)3) and borate ion (B(OH)4 -), which depend on seawater pH and whose stoichiometric equilibrium constant (KB) is close to seawater pH (i.e. pKB = 8.6) (Fig. 2.5). 2) Assuming that only the charged species B(OH)4 - is incorporated into biogenic 57 Chapter 2 calcite, following the consistency found between modern isotopic marine carbonate values and the isotopic composition of B(OH)4 - in seawater (Hemming and Hanson, 1992), the next equation can be derived: (1) 3) Thus, if the seawater pH increases, the amount of boron available for incorporation into calcite as B(OH)4 - increases, thus increasing B/Ca ratios in marine carbonate material. 4) Therefore, as seawater pH is controlled by the partial pressure of CO2 (pCO2) in seawater, which in turn is determined by alkalinity and dissolved inorganic carbon ratios (ALK/DIC), B/Ca ratios seem to be a potential tool to reconstruct the past carbonate system in the ocean. Although the theoretical assumption of B(OH)4 - as the primary species controlling Figure 2.5. Proportions of dissolved boron species B(OH)3 and B(OH)4 - changing with seawater pH. The curves are calculated with typical deep ocean conditions: T =2 °C, S = 35, water depth = 3500 m using [B]total =416 μmol/kg, δ11Bseawater = 39.5‰, and α = 1.0272 (Figure from Yu et al., 2010b). CaCO3+B OH ( ) 4 −↔CaHBO3+HCO3 −+H2O 65 Chapter 3 CHAPTER 3 3. New age models for ODP1240 through reservoir age estimates Synopsis In this chapter, variable surface reservoir ages for ODP1240 have been assessed for the last 30 kyr and used to revise the previous age model for this core, which applied constant regional reservoir age corrections. 3.1. Introduction Accurate age models for marine records are often difficult to constrain. The most common approach to obtain a chronology for the last 40-50 kyr is by radiocarbon dating in planktonic foraminifera, usually using species that live in the ocean’s mixed layer or photic zone, and therefore should record the atmospheric radiocarbon activity. However, this approach suffers from two drawbacks: 1) even the mixed layer or photic zone will be somewhat out of equilibrium with respect to the atmosphere, resulting in the so-called “reservoir age effect” that must be corrected for when calibrating radiocarbon dates to calendar ages using an atmospheric radiocarbon calibration curve; and 2) this “reservoir age effect” can vary over time, not only due to radiocarbon activity changes in the atmosphere, but also due to variable mixing with deeper sub-surface waters with a different isotopic signal (generally more depleted 14C), particularly in upwelling areas. Thus, in order to constrain accurate age models in marine sediment cores using radiocarbon, reservoir ages and their temporal variability must be estimated. However, this requires independent calendar age controls at different depths of the sediment core, and due to the difficulty of obtaining these, reservoir ages are not commonly estimated, instead 66 Age models for core ODP1240 the modern regional reservoir age is taken as a constant value. One of the most useful and accurate ways to constrain the temporal variability of the reservoir ages in sediment cores is via tephrochronology, permitting marine radiocarbon ages to be compared with correlative atmospheric radiocarbon ages that have been established for well-identified tephra (Siani et al., 2013, 2001; Sikes et al., 2000; Skinner et al., 2015). Unfortunately, the oceanic regions where tephra layers can be found are quite restricted. An alternative method for reservoir age estimation is by chronostratigraphic alignment (Skinner et al., 2010). This method consists on providing calendar ages to the sediment core in study by matching its major climatic changes (through stratigraphic tie-points selection either by hand (e.g. Shackleton et al., 2000) or by dynamic programming (Lisiecki and Lisiecki, 2002), with proxies/records from well-dated archives, like ice core or speleothem records. These derived calendar ages have an atmospheric radiocarbon activity (or “age”) associated that can then be subtracted from the planktonic radiocarbon dates at each core depth, thus providing the so-called reservoir age (Soulet et al., 2016). The limitation of this methodology relies on the scarce availability of well-resolved records that can be correlated with well-constrained archives, as well as the assumption of event synchronicity, and the subjectivity involved in the alignment process itself. It is interesting to note that the implementation of either the tephrochronology or the chronostratigraphic alignment methods already provide accurate age models that intrinsically constrain reservoir age variability. Thus, if achieving an age model is the goal and these methods are available, neither planktonic radiocarbon dates nor reservoir age estimates might be in principle necessary. However, radiocarbon age controls from planktonic foraminifera that can be corrected for reservoir age changes, can add detailed constraints to the age models. Furthermore, as addressed in the next Chapter 4, reservoir age estimates are useful in themselves as they provide information on shallow ocean circulation and air-sea carbon exchange rates, both of which are relevant to mechanisms of climate change. 67 Chapter 3 The previous age model for core ODP1240 performed by Pena et al., (2008b) was based on 17 calibrated radiocarbon dates of N. dutertrei, corrected for a constant global-regional surface reservoir (R+∆R) of ∼472 yr, as is often assumed. In this thesis, core ODP1240 chronology for the last 25 kyr has been rebuilt since, after assessing the variability of shallow sub-surface reservoir ages using independent chronostratigraphic constraints, the results indicate that shallow sub-surface reservoir ages varied significantly over the last 25 kyr. New age-models that take account of this variability have therefore been produced here, using the Bayesian depth-age modeling program Bchron (Parnell et al., 2008) and the IntCal09 radiocarbon calibration curve (Reimer et al., 2009). 3.2. Material and methods 19 radiocarbon dates on N. dutertrei (a thermocline-dwelling planktonic foraminifer species, thriving at 100-150 m depth (Fairbanks et al., 1982)) from core ODP1240 were graphitised (See Chapter 2 for further detail on this process) at the Godwin Laboratory for Palaeoclimate Research of the University of Cambridge (UK) (Table A1 in Appendix). N. dutertrei was picked from the >212 µm fraction (4-6 mg carbonate per sample) and, in order to avoid sample loss, a minimal cleaning approach was carried out: removal of particles, fibres, small carbonate clasts or other detritus with a fine hair brush followed by repeated rinses in deionized water. Cleaned foraminifera were placed into clean glass vials, rinsed with methanol, dried and evacuated prior to graphitisation. The calcite was hydrolysed in 0.5 ml of phosphoric acid and then reduced to graphite using a standard hydrogen and iron catalyst method (Vogel et al., 1984). Primary and secondary standards were graphitised in parallel with samples. NIST OXII was used as a primary reference standard (for normalisation) and IAEA-C7 and -C8 were used as secondary standards (for quality control). When samples were < 3 mg carbonate, size-matched aliquots of radiocarbon-dead Iceland Spar calcite were graphitized for small sample background (modern contamination) correction (Santos et al., 2007). The radiocarbon analyses were carried out at The 68 Age models for core ODP1240 14CHRONO Centre of the Queen’s University of Belfast (UK) using a National Electrostatic Corporation (NEC) compact model 0.5MV Accelerator Mass Spectrometer (AMS), with online δ13C measurement. 3.3. Results and discussion 3.3.1. Reservoir age assessment An assessment of shallow sub-surface reservoir age variability at the location of ODP1240 has been performed by chronostratigraphic alignment between the δ18Oice and the [Ca2+] records from three Greenland ice cores (NGRIP, GRIP and GISP2 (Rasmussen et al., 2008)) placed on the recent GICC05 time scale (Svensson et al., 2006) and two independent records from ODP1240: a) Sea Surface Temperature (SST) based on the index (Calvo et al., unpublished) and b) δ18Osw obtained from the planktonic foraminifer Globigerinoides ruber (Pena et al., 2008b) (Fig. 3.1). The proposed alignments are based on a host of previous studies, including previous palaeoclimate reconstructions/interpretations and numerical model studies all of which support the underlying premise of the stratigraphic alignments presented here (Dubois et al., 2014; Kienast et al., 2006; Leduc et al., 2007). The outcome of this previous work is that anomalies in sea surface temperatures recorded by alkenones and surface salinities reflected in surface water δ18O reconstructions at ODP1240 location should coincide in time with North Atlantic/Greenland temperature anomalies. Thus, on one hand, the basis to select the index calls on the millennial-scale link between cold events in the North Atlantic and the Tropical Pacific SST (Kienast et al., 2006). On the other hand, the basis to select δ18Osw for this correlation calls on the apparent relationship between changes in Greenland temperatures and surface water salinity in the EEP through shifts in the mean latitude of the Intertropical Convergence Zone (ITCZ) (Leduc et al., 2007). There is therefore a set of well tested physical mechanisms underlying this correspondence that relate to changes in cross-equatorial temperature gradients and their impact on easterly 69 Chapter 3 Figure 3.1. Surface reservoir age assessment for sediment core ODP1240 through the alignment between two independent ODP1240 records and three Greenland ice cores over the last 25 kyr. a) [Ca2+] and b) δ18O records from Greenland ice cores (Rasmussen et al., 2008): NGRIP (grey), GRIP (yellow) and GISP2 (green). Small vertical arrows above Greenland δ18O records indicate the apparent HS1 onset for each ice core. (c) SSTs from the index and (d) δ18Osw (G.ruber) (Pena et al., 2008b) records from ODP1240 core. In blue, the alignment based on the SST tie-points and in orange, based on the δ18Osw (G.ruber) tie points. (e) Estimated reservoir age scenarios for ODP1240 core using the two alignments: Thick central lines represent the best reservoir age estimate and stippled indicate their upper and lower reservoir age limits (by taking an alignment uncertainty of ±200 yr) (in orange, based on the SST correlation and in blue, based on the δ18Osw (G.ruber) correlation). Selected tie-points are indicated at the bottom by black triangles. An additional core-top date (from Pena et al., 2008b), corrected and calibrated is also shown (grey triangle). Light grey bands highlight the main cold stadials of the last 25 kyr as defined in Greenland records. 0 5000 10000 15000 20000 25000 LGM HOLOCENE Calendar age (yr) YD B-A HS1 HS2 LAST GLACIAL PERIOD DEGLACIATION PB -46 -44 -42 -40 -38 -36 -34 b) Ca 2+ δ 18 O δ 18 O (permil) 0 5000 10000 15000 20000 25000 21.5 22.0 22.5 23.0 23.5 24.0 24.5 25.0 c) SST (U k 37 ) ODP1240 SST (ºC) Calendar age (yr) 2.0 1.5 1.0 0.5 0.0 d) δ 18 O sw G.ruber ODP1240 Tie points from SST (U k 37 ) Tie points from δ 18 O sw G.ruber δ 18 O sw (permil) 2500 2000 1500 1000 500 0 R.age (from SST (U k 37 ) alignment) R.age (from δ 18 O sw G.ruber alignment) e) Reservoir age (yr) 1200 1000 800 600 400 200 0 -200 a) NGRIP GRIP GISP2 [Ca 2+ ] (ppb) 70 Age models for core ODP1240 winds, equatorial upwelling and moisture transport across the Isthmus of Panama. For this alignment, four stratigraphic tie-points from the ice-core records were selected coinciding with major climatic changes at the Heinrich Stadial 2 (HS2) end, the Heinrich Stadial 1 (HS1) onset, the HS1/Bølling-Allerød (BA) transition and the Younger Dryas (YD) end (Fig. 3.1; Table A2 in the Appendix). The HS1 onset was designated only through the change in the [Ca2+] record at 17.6 kyr since precipitation δ18O records from the various Greenland ice-cores present differences between 19-16 kyr due to different site locations (Fig. 3.1a,b). An additional age constraint for the ODP1240 core-top was derived based on a single calibrated radiocarbon date (from Pena et al., 2008b), corrected for the modern shallow sub-surface reservoir age in the area (this is ~700 yr; see Material and Methods from Chapter 4 for further details). On the basis of the above chronologies, two sets of reservoir age estimates were generated (i.e. one from each ODP1240 record used for stratigraphic alignment) (Fig. 3.1e; Table A2 in the Appendix). These differ due to slight mismatches between the patterns of SST and δ18Osw variability before 17 kyr (Fig. 3.1c,d). Reservoir ages were calculated as the difference between the interpolated ODP1240 planktonic 14C for each tie-point (Skinner et al., 2010) and the atmospheric 14C-age (IntCal09 curve (Reimer et al., 2009)) at the calendar age of each tie-point (note that the more recent IntCal13 radiocarbon calibration curve (Reimer et al., 2013) has not been used in order to avoid possible problems of inconsistency when comparing with previously published radiocarbon studies that are referenced to IntCal09; only minor differences would arise if IntCal13 curve was used instead). This approach provides reservoir age estimates only at the four proposed calendar age tie-points. An alternative approach of interpolating calendar ages (and on this basis 14Catm) for all the planktonic radiocarbon dates in ODP1240 would yield spurious reservoir ages where no direct calendar age constraints exist (as would occur in the Holocene for example, where no tie-points were designated). As observed in Figure 3.1e, despite small quantitative differences between the 71 Chapter 3 two sets of reservoir ages estimates generated from SST and δ18Osw alignments, they both show a broadly similar pattern of variability across the last deglaciation indicating large changes in the shallow sub-surface ocean ventilation over this time interval. Furthermore, these reservoir ages begin to decline coincident with the beginning of the deglaciation (HS1 onset), continuing to drop across HS1 until minimum reservoir ages are reached around the HS1/B-A transition. Quantitatively, they show glacial shallow sub-surface reservoir ages of ~970 and ~1,105 yr respectively, that is ~405 and ~540 yr older than during the Holocene (when reservoir ages are ~565 yr on average, including the modern value of ~735 yr). These results allow the pre-existing radiocarbon based age model for ODP1240 to be refined for the last 25 kyr. 3.3.2. Age models Four age models were thus derived using the Bayesian age-depth modelling program Bchron (Parnell et al., 2008) and the IntCal09 radiocarbon calibration curve (Reimer et al., 2009) (Fig. 3.2; Table A3 in the Appendix): 1) two of them by calibrating 17 N. dutertrei radiocarbon dates corrected for the two sets of reservoir age estimates (Fig. 3.2a); and 2) two more using only the stratigraphic tie-points (Fig. 3.2b). The purpose of this was to check the consistency of the age models when ignoring the radiocarbon age constraints between stratigraphic tie points, as well as the radiocarbon age reversals that are automatically rejected by Bchron as outliers in the first two age models. Notably, all of these four chronologies are in good agreement, indicating a good degree of consistency between the various chronostratigraphic constraints that have been adopted, within their respective uncertainty ranges. Only across the Holocene, where no stratigraphic tie-points were selected, do large differences between the radiocarbonand stratigraphic age models emerge, reflecting significant changes in sedimentation rate that cannot be captured by the stratigraphic alignment performed in this thesis over this interval. 72 Age models for core ODP1240 Finally, circumstantial support for the age-models is also provided by the fact that deglacial minima in the N. dutertrei δ13C record from ODP1240 (Pena et al., 2008b) correspond with similar G. bulloides δ13C minima from the sub-Antarctic Atlantic (Skinner et al., 2014) and peak upwelling intensities indicated by the opal flux record Figure 3.2. Age models for core ODP1240. Four possible Bchron age models and their uncertainties (95% HDR range) for core ODP1240 derived from; a) two from the δ18O (Greenland ice cores) – SST (ODP1240) records alignment (orange curves) and b) two from the δ18O (Greenland ice cores) – δ18Osw (G.ruber) (ODP1240) records alignment (blue curve). Thick solid lines and stars represent the age models derived only from the tie-points (the error is 200 yr). Thin solid lines and squares denote age models using the planktonic radiocarbon measurements corrected for the correspondent interpolated reservoir age obtained from the alignments. c) Sedimentation rates obtained from the four age models; Shaded areas correspond to sedimentation rates using only the tie-points and solid lines using the radiocarbon dates. As for a) and b), orange colors refer to the SST alignment, and blue colours refer to the δ18Osw (G.ruber) alignment. 0 50 100 150 200 250 300 350 0 5000 10000 15000 20000 25000 30000 0 50 100 150 200 250 300 350 0 5000 10000 15000 20000 25000 30000 a) BCHRON AGE MODELS FROM SST (Uk 37) ALIGNMENT Tie points Calibrated dates age model TP 95% HDR range age model dates 95% HDR range Calendar age (yr) Depth (cm) BCHRON AGE MODELS FROM δ18Osw G.ruber ALIGNMENT Tie-points Calibrated dates age model TP 95% HDR range age model dates 95% HDR range Calendar age (yr) Depth (cm) b) HOLOCENE DEGLACIATION LAST GLACIAL PERIOD 0 20 40 60 80 c) Sed.rates (cm/kyr) 73 Chapter 3 observed in the Southern Ocean (Anderson et al., 2009) (Fig. 3.3). The two minima also broadly coincide with periods of CO2 rise and with minima in the atmospheric δ13CO2 record (Schmitt et al., 2012) (Fig. 3.3). Figure 3.3. Comparison of marine and atmospheric δ13C records. a) Atmospheric δ13CO2 record compilation (Schmitt et al., 2012). b) N. dutertrei δ13C record from ODP1240 (Pena et al., 2008b) placed on the revised age models of this study (orange/blue dashed lines with solid/hollow dots from SST/δ18Osw alignments, respectively) and G. bulloides δ13C record from MD07-3076 (Skinner et al., 2014) from the sub-Antarctic Atlantic (green dashed line). Solid thick lines indicate the 5-point running averages. c) TN057-13 opal flux record in a reverse scale (Anderson et al., 2009). d) Atmospheric CO2 concentrations from EPICA Dome C (EDC) ice core (for the deglacial period (Monnin et al., 2001); for the Holocene period (Flückiger et al., 2002)), placed on the age scale of Lemieux-Dudon et al. (2010). 0 5000 10000 15000 20000 25000 180 200 220 240 260 280 CO 2 atm (EDC) CO 2 atm (ppmv) Calendar age (yr) 0 5000 10000 15000 20000 25000 LAST GLACIAL PERIODHOLOCENE Calendar age (yr) YD B-A HS1 DEGLACIATION PB a) b) c) d) 7 6 5 4 3 2 1 0 -1 Opal flux (TNO57-13) Opal flux (g/cm 2 kyr) 0.8 1.0 1.2 1.4 1.6 EEP δ 13 C (N.dut ODP 1240) EEP Planktonic δ 13 C (permil) -7.0 -6.8 -6.6 -6.4 -6.2 δ 13 C (EDC) MCMC ave. δ 13 CO 2 (permil) -0.4 -0.2 0.0 0.2 0.4 0.6 S. Atl δ 13 C(G. bull MD07-3076) S. Atlantic Planktonic δ 13 C (permil) 80 Shallow sub-surface and deep radiocarbon ventilation in the EEP radiocarbon ventilation records from the Pacific are based on the commonly used method of calculating benthic-planktonic radiocarbon age offsets (B-P) (Table 4.1), which ignores the possible impact of variable surface reservoir ages on the inferred radiocarbon disequilibrium of deep waters relative to the atmosphere (i.e. the benthic-atmosphere radiocarbon age offset, B-Atm). Furthermore, it is also notable that studies adopting this method or the 14C projection-age method (Adkins and Boyle, 1997), and assuming invariant surface reservoir ages, tend not to find evidence of poorly ventilated glacial mid/deep Pacific ocean waters (Broecker et al., 2008, 2007, 2004a, 2004b, 1990, 1988; De Pol-Holz et al., 2010; Okazaki et al., 2012) (Table 4.1). However, numerical model experiments have suggested that surface reservoir age should in principle have varied since the last glacial period (Butzin et al., 2012), and this has been confirmed via a range of techniques applied, not only in the Pacific Ocean (Lund et al., 2011; Sarnthein et al., 2013; Siani et al., 2013; Skinner et al., 2015), but also in the Southern Atlantic Ocean (Skinner et al., 2010) and other basins (Bard et al., 1994; Siani et al., 2001). Notably, Siani et al. (2013) suggested that data from a previous study at a nearby location (core SO161-SL22; De Pol-Holz et al., 2010) would indeed show significant changes in ventilation ages over the glacial period if their surface reservoir ages were taken into account. These observations serve to underline the importance of assessing the variability of the surface reservoir ages, especially when using the B-P offset approach to calculate radiocarbon ventilation ages. Today, the Southern Ocean is the main region where water from the ocean interior makes its first contact with the surface, and where surface water first enters the ocean interior (Primeau, 2005). It is also the main source of deepand abyssal water exported to the rest of the global ocean (Orsi et al., 1999), and it therefore plays a key role in the water mass budget and the ventilation (i.e. the mixing of surface water properties into the ocean interior) of the Pacific Ocean (Talley et al., 2011). It is for this reason that the Southern Ocean has been proposed as the main area for the release of “old” carbon-rich deep water from the ocean interior 81 Chapter 4 at the close of the last glacial, possibly due to increased upwelling and/or air-sea gas exchange as a result of altered winds (Anderson et al., 2009) and/or sea-ice/ shallow sub-surface density stratification (Skinner et al., 2014, 2010). A connection between the highand low latitude Pacific is provided through upwelled shallow sub-surface/intermediate waters from the Southern Ocean (See Material and methods for further details) in the Eastern Equatorial Pacific (EEP) via the Equatorial Undercurrent (EUC), identified for both the modern- (Lukas, 1986; Toggweiler et al., 1991; Tsuchiya et al., 1989) and the glacial circulation (Anderson et al., 2009; Calvo et al., 2011; Pena et al., 2013, 2008b; Spero and Lea, 2002). This far-field dynamical connection would imply that changes occurring in the Southern Ocean should exert an influence on the hydrography of the thermocline in low latitude areas like the EEP. The EEP thus provides an excellent study location for assessing past changes in Pacific radiocarbon ventilation, in particular where these are believed to originate in the Southern Ocean. Here it is presented deepand shallow sub-surface (i.e. deep-thermocline) radiocarbon ventilation age reconstructions for the EEP, from sediment core ODP Site 1240 located in the Panama Basin (0°01.31’N, 86°27.76’W; 2,921 m water depth). This core site is currently bathed by Upper Circumpolar Deep Water (UCDW) and/or Pacific Deep Water (PDW), with southern sourced shallow sub-surface/intermediate waters contributing to the EEP thermocline via the EUC. Deep water radiocarbon ventilation ages are estimated from B-P age offsets (B: mixed benthic foraminifera; P: N. dutertrei, a deep thermocline-dwelling planktonic foraminifer, 100-150 m water depth (Fairbanks et al., 1982)), plus a variable shallow sub-surface reservoir age (yielding B-Atm age offsets), which are derived on the basis of chronostratigraphic alignment to Greenland ice core records (see Chapter 3). Notably, the revised chronology that it is derived in this thesis for this core implies a significant increase in both deep water and shallow sub-surface radiocarbon reservoir/ventilation ages during the last glacial period in the EEP, consistent with a significant change in the radiocarbon concentration of waters exported from the Southern Ocean to the rest of the global ocean during the last glacial period. 82 Shallow sub-surface and deep radiocarbon ventilation in the EEP 4.2. Material and methods 4.2.1. Radiocarbon dates 19 radiocarbon dates on N. dutertrei (which have been described in the previous Chapter 3 when assessing the variability of the surface reservoir age) and 17 samples of mixed benthic foraminifera were performed from core ODP1240 (i.e. 17 paired B-P radiocarbon dates). Mixed benthic foraminifera were picked from the >150 µm fraction (1-6 mg carbonate per sample; adjacent sediment samples were combined when the amount of carbonate was less than 1 mg). The samples were cleaned, analysed and measured according to the description made for N. dutertrei in Chapter 3. 4.2.2. Radiocarbon calculations in modern water column The modern radiocarbon values for ODP1240 core site were obtained from section P19 of the World Ocean Circulation Experiment (WOCE) (Tsuchiya and Talley, 1998), very close to this core site. Using the “Δ14C natural” and applying the equation of Stuiver and Polach (1977) (t=-8033ln(1+(Δ14C/1000)), the current shallow sub-surface and deep ventilation ages are calculated in this site. Thus, the ventilation ages would summarise as: modern shallow sub-surface ventilation age (Reservoir age) at ~150 m water depth is ~735 radiocarbon years; modern deep ventilation age (B-Atm) at ~3,000 m water depth is ~1,985 radiocarbon years; modern B-P radiocarbon offset between 150 m and 3,000 m water depth is ~1,250 radiocarbon years. 4.3. Results 4.3.2. B-P offsets and reservoir ages 17 paired planktonic and benthic radiocarbon measurements and the offset 83 Chapter 4 between them (B-P) are shown in Figure 4.1a and b respectively, and summarised in Table A1 in the Appendix. Of the data shown in Figure 4.1a and 1b, three B-P pairs (depth intervals 26, 223 and 289; see Fig. 4.1b in grey) were flagged up as potential outliers in the present study, on the basis that: these three pairs yield B-P offsets Figure 4.1. Summary of planktonic and benthic radiocarbon dates and their offsets from core ODP1240. a) Radiocarbon dates from N. dutertrei (red triangles) and mixed benthic foraminifera (black triangles) (all dates were δ13C-normalised, without reservoir age correction and uncalibrated to calendar years), with associated 1σ-uncertainties. The discarded B-P offsets are indicated by a grey vertical arrow. Note that planktonic dates from intervals 77/78 and 117/118 were averaged for B-P calculation in order to be able to compare with the coeval benthic radiocarbon dates which were combined due the low carbonate abundance of each interval (see Material and methods and Table 1 in the Appendix). b) Benthic-Planktonic radiocarbon offset (B-P) (thick light blue line and stars) plus the discarded B-P offsets (stapled light grey line) (error bars represent combined 1σ error in 14C dates). Modern B-P offset is shown by a dashed horizontal line and relevant climatic periods are separated by dashed vertical lines. 0 50 100 150 200 250 300 350 400 25000 20000 15000 10000 5000 0a) N.dutertrei Mixed benthic 14C age (yr) Core depth (cm) 0 50 100 150 200 250 300 350 400 2500 2000 1500 1000 500 0 b) Modern B-P LAST GLACIAL PERIOD HOLOCENE B-P B-P outliers B-P (yr) Core depth (cm) DEGLACIATION 84 Shallow sub-surface and deep radiocarbon ventilation in the EEP that fall anomalously far away from the three-point running mean of the dataset and; the benthic ages converge on their planktonic counterparts far more rapidly than could be expected due to real changes in deep ocean circulation/ventilation. However, these tentative flyers are shown in Figure 4.1a and 1b in order to note that it is possible (if not presently demonstrable) that other studies in the region might in future provide support for their veracity. The 14 retained radiocarbon B-P pairs (Fig. 4.1b) indicate a change of ~870 yr between the last part of the last glacial period (average B-P ~1,805 yr) and the Holocene (average B-P ~935 yr, including the modern B-P offset of ~1,250 yr; see Material and methods). Across the deglaciation, decreasing B-P offsets indicate enhanced deep ocean ventilation, that peaks in the early Holocene and then drops again to modern values across this period. These results thus provide clear evidence for a drop in deep equatorial Pacific radiocarbon ventilation ages since the Last Glacial Maximum. As described in Chapter 3, two reservoir ages were estimated and were used to both perform the age models and to estimate deep ocean ventilation ages (Fig. 4.2b, figures in Chapter 3 and Table A2 and A3 in the Appendix). The two sets of reservoir age estimates show a broadly similar pattern of variability across the last deglaciation, indicating large changes in the shallow sub-surface ocean ventilation over this time interval (Fig. 4.2b). Quantitatively, small differences are found between both SST and δ18Osw alignments, yielding glacial shallow sub-surface reservoir ages of ~970 and ~1,105 yr respectively, that is ~405 and ~540 yr older than during the Holocene (when reservoir ages are ~565 yr on average, including the modern value of ~735 yr). As shown in Fig. 4.2b, shallow sub-surface reservoir ages begin to decline coincident with the beginning of the deglaciation (HS1 onset), continuing to drop across HS1 until minimum reservoir ages (i.e. maximum ventilation) are reached around the HS1/B-A transition. 85 Chapter 4 4.3.2. Ventilation ages Using the observed radiocarbon offsets between benthic-planktonic (B-P; Fig. 4.1b and 4.2a) and between planktonic-atmosphere (i.e. reservoir ages; Fig. 4.2b), the offset between benthic-atmosphere (i.e. deep-ocean ventilation; calculated Figure 4.2. Ventilation scenarios for core ODP1240 over the past 25 kyr using the four estimated age models from Chapter 3 (b-spline smoothed). a) B-P offset; b) Shallow sub-surface reservoir age; c) Ventilation age (B-Atm offset). Thick solid lines with stars denote age models using the planktonic radiocarbon measurements corrected for the correspondent interpolated reservoir age obtained from the alignments, and thin stapled lines with squares represent the age model derived only from the tie-points, (hollow/solid in all cases corresponds to UK37’ /δ18Osw (G.ruber) alignments, respectively). Modern offsets are shown by dashed horizontal lines. Light grey bands highlight the coldest periods of the last 25 kyr and delimit relevant climatic periods. Note the different Y-axis scale in a-c. 0 5000 10000 15000 20000 25000 30000 Calendar age (yr) 0 5000 10000 15000 20000 25000 30000 4000 3500 3000 2500 2000 1500 1000 500 0 LGM LAST GLACIAL PERIODHOLOCENE c) b) B-Atm Modern R.age Modern B-Atm R.age Shallow sub-surface and deep ventilation age (yr) Calendar age (yr) YD B-A HS1 HS2 DEGLACIATION PB 2500 2000 1500 1000 500 0 a) Modern B-P B-P B-P (yr) 86 Shallow sub-surface and deep radiocarbon ventilation in the EEP as B-Atm = B-P + R.age; Fig. 4.2c) can be estimated. Thus, Fig. 4.2 displays the ventilation history at the location of ODP1240, taking into account all four possible age models and their associated reservoir age implications. As observed in the shallow sub-surface reservoir age reconstructions, the deep-ocean ventilation (Fig. 4.2c) also exhibits different amplitude and slightly different trends since the LGM depending on the various alignments (SST or δ18Osw), but still broadly similar patterns of variability. Quantitatively, the averages from both SST and δ18Osw alignments indicate B-Atm values of ~2,775 and ~2,910 yr, respectively, during the last part of the glacial period, which would correspond to a ~1,275 and ~1,410 yr older deep glacial ocean than during the Holocene (on average ~1,500 yr including the modern B-Atm value of ~1,985 yr at 3,000 m water depth; see Material and methods). Enhanced ventilation of the deep eastern Pacific is thus observed from the start of HS1, with peak ventilation (minimum B-Atm) being reached at the end of the deglaciation, in the early Holocene (Fig. 4.2c). 4.4. Discussion The ventilation reconstructions from ODP1240 indicate a glacial-interglacial B-P change of ~870 yr that, taking into account the shallow sub-surface (deep-thermocline) reservoir age change of ~405/540 yr (from both SST and δ18Osw alignments, respectively), imply a glacial-interglacial B-Atm change of ~1,275/1,410 yr (Fig. 4.2; Table A3 in the Appendix). Taken together, these results indicate that both deepand shallow sub-surface reservoir ages (Fig. 4.2b,c) were significantly increased at the LGM and that both evolved in a similar manner over the last 25 kyr, albeit with a larger change in the deep ventilation ages (as indicated by the change in B-P across the deglaciation; Fig. 4.2a). A poorly ventilated glacial shallow sub-surface and deep EEP is thus demonstrated by the reservoir age reconstructions and by both the B-P offsets and the B-Atm offsets, with implications for the glacial ocean’s total radiocarbon inventory and its role in glacial-interglacial CO2 change (Broecker and Barker, 2007). 87 Chapter 4 The changes in B-P offsets between the LGM and modern observed in ODP1240 clearly reflect a discernable trend in deep equatorial Pacific ventilation across the last deglaciation that is consistent with other B-P records from the northernand southern high latitude Pacific Ocean (SWP (Sikes et al., 2000; Skinner et al., 2015); SEP (Siani et al., 2013); NEP (Galbraith et al., 2007)) (Fig. 4.3). All of these studies indicate slightly higher B-P age offsets during the last glacial period (~200-600 yr) and, although they are not extremely large, they should not be downplayed. Indeed, it has been argued that the full 190 ‰ drop in atmospheric Δ14C since the LGM (ignoring changes in radiocarbon production) could in principle be explained Figure 4.3. B-P radiocarbon offsets from several locations in the Pacific and Southern Oceans over the last 25 kyr (B-spline smoothed): ODP1240 (2,921 m; blue hollow stars from SST alignment; this study); Bay of Plenty (1,675-2,065 m; green stars; Sikes et al., 2000); MD97-2121 (2,314 m; yellow dots; Skinner et al., 2015); MD07-3088 (1,536 m; orange diamonds; Siani et al., 2013); MD07-3076 (3,770 m; grey dots; Skinner et al., 2010); ODP887 (3,648 m; pink squares; Galbraith et al., 2007). Light grey bands highlight the coldest periods of the last 25 kyr and delimit relevant climatic periods. 0 5000 10000 15000 20000 25000 30000 0 5000 10000 15000 20000 25000 30000 2500 2000 1500 1000 500 0 Calendar age (yr) LGM LAST GLACIAL PERIOD HOLOCENE EEP ODP1240 (2,921 m) SO MD07-3076 (3,770 m) NEP ODP887 (3,648 m) SWP Cores Bay of Plenty (1,675-2,065 m) MD97-2121 (2,314 m) SEP MD07-3088 (1,536 m) B-P (14C yr) YD B-A HS1 HS2 DEGLACIATION PB 88 Shallow sub-surface and deep radiocarbon ventilation in the EEP Figure 4.4. Comparison of stable isotopes (from N. dutertrei) between cores ODP1240 (grey lines) and TR163-31B (blue lines; Shackleton et al., 1988), each one placed on its own radiocarbon age scale, showing broad coherence of signals and therefore of reservoir age histories. Upper panel displays δ18O records and lower panel δ13C records. by rather small changes in the mean ventilation age of the whole upper ocean (e.g. ~100 yr over 60 % of the ocean) if these were combined with a slightly larger change in the mean ventilation of the deepest ocean (e.g. ~3,000 yr on average, over 35 % of the ocean) (Skinner et al., 2010). Although B-P offsets in ODP1240 are clearly observed to decrease across the last deglaciation (Fig. 4.1b and 4.2a), a larger drop in B-Atm offsets is inferred on the basis of the shallow sub-surface reservoir age reconstructions estimated in this thesis. In a first instance, the regional coherence of these shallow sub-surface reservoir age reconstructions is explored via a comparison of planktonic stable isotope records from the studied core and core TR163-31B (Shackleton et al., 1988) (located in the same area and also with N. dutertrei radiocarbon dates), performed by placing each core on its own conventional 14C age scale (Fig. 4.4). For this purpose the 89 Chapter 4 radiocarbon data from Shackleton et al. (1988) has been used. The fact that the isotope stratigraphies and radiocarbon data from the two cores are consistent (Fig. 4.4) indicates that both locations have experienced broadly similar reservoir age changes. On this basis, a “regional radiocarbon calibration” can be used to transfer calendar ages (and therefore reservoir ages) from ODP1240 to TR163-31B. This in turn allows B-Atm offsets to be derived for core TR163-31B, using the few available benthic radiocarbon dates from Shackleton et al. (1988), as described below. While the comparison described above demonstrates that similar reservoir age changes were experienced at the locations of ODP1240 and TR163-31B, it does not provide an independent confirmation of the reservoir ages derived for ODP1240. However, direct support for the observation made in this thesis of significantly enhanced glacial shallow sub-surface reservoir ages is provided by the few studies in the South Pacific Ocean that have also sought to constrain reservoir age variability using paleoceanographic observations (Fig. 4.5c), as well as direct support for a deglacial change in deep equatorial Pacific ventilation (Figure 4.6d). Both comparisons are discussed below. As shown in Figure 4.5, the radiocarbon ventilation rate of both the thermocline and the deep EEP are observed to increase broadly in time with the onset of deglaciation, the rise in atmospheric CO2 (Monnin et al., 2001) and the drop in atmospheric 14C (Hughen et al., 2004; Reimer et al., 2013) (Fig. 4.5a,b). These results would thus support the hypothesised role of reduced ocean-atmosphere CO2 exchange rates in sequestering carbon in the glacial ocean interior, and the release of this carbon to the atmosphere across the last termination. The demonstration of reduced radiocarbon ventilation rates in both the deep and shallow sub-surface Pacific Ocean is crucial in this respect, as this basin accounts for ~50 % of the ocean’s total volume. Thus, small ventilation or carbon sequestration effects in the Pacific can have a particularly large impact on global ocean chemistry, as well as that of the atmosphere. Going back to the reservoir age support from other Pacific records, in the southwest 96 Shallow sub-surface and deep radiocarbon ventilation in the EEP 4.5. Conclusions In this chapter it is argued that even relatively small changes in B-P radiocarbon ventilation ages are relevant to the question of deglacial radiocarbon and carbon cycling, especially where these are observed over a large ocean volume like the Pacific. Furthermore, due to the possibility (indeed the likelihood) of significant changes in shallow sub-surface reservoir ages since the last glacial period, relatively small deglacial changes in B-P offsets may not necessarily rule out the occurrence of much larger changes in ocean interior radiocarbon ventilation, even in the low latitudes, as it is demonstrated here for the EEP. Indeed, the results presented in this thesis underline the crucial importance of constraining surface reservoir ages (or radiocarbon-independent calendar ages). Furthermore, these findings are found to be consistent with other surfaceand deep water ventilation age reconstructions from the southern Pacific and Atlantic. This may reflect glacial circulation patterns that were not very different from modern, albeit with altered surface boundary conditions (air-sea exchange efficiency) in the Southern Ocean in particular, which it is identified here as the main source region for poorly ventilated deep glacial waters in the shallow sub-surface and deep EEP. These results have implications for the ocean’s role in glacial-interglacial carbon cycling, demonstrating an increase in the Pacific Ocean’s mean carbon residence time, which very likely would have contributed to lower atmospheric CO2 during the last glacial period. Furthermore, it would appear that this change in the ventilation state of the glacial Pacific may have stemmed in large part from less effective ocean-atmosphere CO2 exchange in the Southern Ocean in particular. 99 Chapter 5 CHAPTER 5 5. B/Ca foraminifer analyses by LA-ICPMS - Subchapter 5.1: LA-ICPMS set up for B/Ca analysis - Subchapter 5.2: Application of the LA-ICPMS method to down-core and core-top B/Ca analyses Synopsis As mentioned in the previous Chapter 4, carbonate system proxies would offer valuable information to complement the ventilation inferences obtained from radiocarbon in the EEP over the last 30 kyr. In order to achieve this, B/Ca ratios from benthic foraminifer calcite shells, as a proxy for ∆[CO3 2-] (and therefore [CO3 2-]), are investigated here. The primary focus of this chapter is on the novel application of the LA-ICPMS technique in this field, and its comparison to the more commonly encountered solution-ICPMS method. This chapter is subdivided into two subchapters: Subchapter 5.1, which investigates the accuracy of the laser ablation analytical technique and assesses the best LA-ICPMS set up for the analysis of B/Ca ratios in foraminifera; and Subchapter 5.2, which discusses laserablation profiles obtained from the benthic foraminifer species C. wuellerstorfi and compare them with the values obtained on the same samples analysed by the solution-ICPMS technique. Subchapter 5.2 also shows core-top ”proxy calibration” analyses in C. wuellerstorfi and preliminary information on B/Ca distribution in the planktonic foraminifera N. dutertrei. 100 B/Ca foraminifer analyses by LA-ICPMS 5.1. Subchapter 5.1: LA-ICPMS set up for B/Ca analysis 5.1.1. Introduction Laser-ablation inductively-coupled-plasma mass spectrometry (LA-ICPMS) is a micro-analytical technique applied to many disciplines that permits the analysis of samples in a solid state at very high spatial resolution. Over the last decade, this technique has become more and more prevalent in the study of the trace element composition of foraminifer shells for paleoenvironmental reconstructions (Eggins et al., 2003; Hathorne et al., 2003; Pena et al., 2005; Raitzsch et al., 2011; Sadekov et al., 2009). The LA-ICPMS technique provides high-resolution depth profiling, offering accurate and precise measurements of a wide range of elements at very high spatial resolution, avoiding laborious cleaning procedures, and using only a few foraminifer specimens per sample (which depending on the species can be very valuable). The most noticeable advantage, however, is probably that this technique allows element distributions to be scrutinized on a single specimen through the calcite test ablation, thus providing information about biomineralization processes via their influence on trace element variability within a foraminiferal test (Anand and Elderfield, 2005; Eggins et al., 2004, 2003; Hathorne et al., 2003; Sadekov et al., 2008, 2005). It is important to constrain such intra-test composition variability in order to properly understand the relationship between the bulk chemistry of the test and the supposed environmental controls, which might be masked by significant “vital effects”, with implications for the accuracy of paleoenvironmental reconstructions. However, a “handicap” that is usually found when using LA-ICPMS to analyse carbonate material stems from the need to match the sample material with the material used for analytical calibration, i. e. the standards (e.g. Hathorne et al., 2008). Ideally, calcite standards should be used in the context of foraminifer calcite analyses, but the chemical homogeneity of the available standard material remains poorly constrained (Hathorne et al., 2013). For this reason, glass standards are commonly used instead and, although in some cases they have been described to 101 Chapter 5 present matrix effect issues as well as isotopic fractionation, this problem seems to have been fixed by using laser systems with wavelengths of 193 nm (Hathorne et al., 2008; Jochum et al., 2012, 2007). In any event, the accuracy of the glass standards should be assessed for any analytical set-up by calibrating well-known material with these standards. Among the elemental ratios in benthic foraminiferal calcite, the boron to calcium ratio (B/Ca) has been proposed to relate to the carbonate system, and more specifically the degree of carbonate ion unsaturation (D[CO3 2-]) (Yu and Elderfield, 2007). Foraminifer B/Ca ratios are therefore a potential tool for reconstructing past ocean carbonate unsaturation changes. However, B is found in very low concentrations in foraminifer shells and is a ubiquitous element, presenting challenges with respect to maintaining a sufficiently low instrumental B background and therefore maintaining good reproducibility of foraminifer analyses. In this subchapter, the optimal conditions for foraminifer B/Ca measurements are set up by testing how different parameters (such as the type of tubing used to transfer the ablated material from the LA system to the ICPMS) may cause the instrumental B background to vary. Then, the accuracy of LA-ICPMS B/Ca measurements is assessed using glass standard materials. Finally, how such instrumental B background variations might have repercussions for paleoceanographic interpretations is assessed by analysing N. dutertrei specimens from ODP1240 (as this species is highly abundant in this core), by both LA-ICPMS and solution-ICPMS techniques in the same specimens. 5.1.2. Material and methods 5.1.2.1. LA-ICPMS set up through standard material The RESOlution M-50 prototype 193 nm ArF laser ablation system coupled to an Agilent 7500ce ICPMS available at Royal Holloway University of London, UK (Müller et al., 2009) (see Figures in Chapter 2) has been used for B/Ca analysis in foramin- 102 B/Ca foraminifer analyses by LA-ICPMS ifera. This analytical set-up (193 nm of wave length) minimises the possibility of element fractionation in trace elements like Boron (Hathorne et al., 2008; Jochum et al., 2007). All the ablations were done at 2 Hz pulse rate, at a fluence of 4.4 J/cm2 and were analysed for trace element abundances relative to Ca (11B, 24Mg, 25Mg, 27Al, 43Ca, 55Mn, 88Sr, 89Y, 138Ba and 238U), although with a primary focus on B/Ca ratios. The analyses were carried out under a Helium atmosphere and Argon was used as a carrier gas. An additional di-atomic gas was also applied to enhance the sensitivity (nitrogen or hydrogen) and, with the purpose of checking which of these two gases produced better sensitivity for the elements of interest, some tests were made. The system at Royal Holloway has a manifold coupled called “squid” that smooths laser pulses and is especially important for repetition rates lower than 10 Hz, as it is the case for foraminifer analysis (Eggins et al., 1998; Evans and Müller, 2013; Müller et al., 2009). All the instrumental operating conditions and the acquisition time for each isotope are summarised in Table 5.1. National Institute of Standards and Technology (NIST) SRM 610 and 612 glasses were used as external standards for calibration using the concentration data from Jochum et al. (2011). In order to assess the standard accuracy and be able to select  Laserparameters N2/H2gasflow6ml/min Fluence4.4J/cm 2  Repetitionrate 2Hz Laserspotsize7496μm ICPMSparameters Analysedisotopes(m/z)and acquisitiontime(s) 11 B(0.05), 24 Mg(0.03), 25 Mg(0.03), 27 Al(0.01), 43 Ca (0.02),55Mn(0.01),88Sr(0.005),89Y(0.01),138Ba (0.005)and238U(0.01) RFpower~1,270W Carriergasflow 0.45l/min   Table 5.1. Instrumental operation conditions used in this study. 103 Chapter 5 the more appropriate NIST standard for the foraminiferal analyses, MPI-DING komatiite glasses GOR128 and GOR132 (Jochum et al., 2006) were also measured. Thus, GOR reference materials were treated as unknowns and calibrated with NISTs standards by using all possible combinations between unknowns-standards-spot sizes (e.g. GOR128-NIST612-96mm, GOR128-NIST610-74 mm, and so on). Two types of tubing were used during different sessions in order to test its influence on boron instrumental backgrounds: the common Nylon tubing, which produced high B background conditions and signal/noise ratios between 4-15, and Nylon6 free of sulphur tubing (Nylon6 free-S), which achieved low B background conditions and ratios up to 100. 5.1.2.2. Test-analyses in foraminifer calcite 5.1.2.2.1. Analysis by LA-ICPMS In order to test if the different background conditions might quantitatively affect B/Ca results in foraminifer calcite, 8-10 N. dutertrei specimens from nine ODP1240 samples were picked from fraction size 315-355 mm. Three samples (intervals 33, 89 and 223) were analysed under high background conditions (i.e. using common Nylon tubing), and six more samples (intervals 9, 45, 89, 129, 203 and 283) under low background conditions (i.e. using Nylon6 free-S tubing). All specimens were measured under the instrumental conditions described above and with a spot size of 74 mm, as this was considered a good balance for the depth profiling analysis in this species. Before ablation, the individuals were rinsed in milliQ water (with no ultrasonication step as the specimens were too delicate) and secured on double-sided adhesive tape mounted onto glass slides (see Chapter 2). Samples were ablated for a maximum of 120 s with delays of 30 s between measurements, while NIST610/612 were ablated after ~ 6-8 samples for a duration of 120 s with delays of 40 s between measurements. Data reduction was carried out by using an Excel template, where 104 B/Ca foraminifer analyses by LA-ICPMS shell surface enrichments were avoided, and the remaining data were “despiked” (to remove anomalous flyers), standardised with NIST glasses, checked for machine drift, and normalised to Ca. Further details about laser data reduction performed in this thesis can be found in the next subchapter 5.2. 5.1.2.2.2. Analysis by Solution-ICPMS The specimens from most of the samples analysed by LA-ICPMS were removed from the double-sided tape after laser ablation, cleaned and analysed by solutionICPMS. A quick test to check if the double-sided tape might bring any contamination to the solution-ICPMS measurements was performed by picking new individuals from the same sample vials (the same amount of individuals per sample), cleaned and analysed in parallel with the previously laser-ablated material. The cleaning procedure for solution-ICPMS analysis, carried out at the University of Barcelona (Spain), consisted of crushing all foraminifer tests between two clean glass slides, transferred to acid leached 0.5 ml Eppendorf vials and rinsed with MilliQ water and methanol in order to remove clays and other detrital material. The samples were then cleaned applying the “Oxidative” cleaning protocol (Barker et al., 2003) with a final weak acid leaching. The reductive step was not applied, as it has been shown to produce negligible effect on benthic B/Ca measurements (Yu and Elderfield, 2007; Yu et al., 2007a). Cleaned samples were dissolved in 0.1 M HNO3 and centrifuged before analysis at the Godwin Laboratory for Palaeoclimate Research at the University of Cambridge (UK). The solution-ICPMS analyses were performed by both ICP-OES (Varian® VISTA) and ICP-MS (Thermo® Element XR). ICP-OES was used for calcium determination. In order to minimize instrumental memory effects and boron background contamination, ICP-MS was used to determine B/Ca at 10 ppm [Ca] in a matrix of nitric and hydrofluoric acid (0.1 M HNO3 and 0.3 M HF) (Misra et al., 2014). The trace elements 7Li, 25Mg, 27Al, 55Mn, 87Sr, 111Cd, 137Ba and 238U were measured at Low Resolution (LR), 23Na, 47Ti, 55Mn, 56Fe, 66Zn at Medium Resolution (MR) and 11B and 43Ca at both resolutions. 105 Chapter 5 5.1.3. Results and discussion 5.1.3.1. Standard accuracy through LA-ICPMS Accuracy was assessed over the course of 15 months (in 3 periods of time) by calibrating 18 analyses of the MPI-DING glasses GOR132 and GOR128 (taken as unknowns) to both NIST610 and NIST612 (through all possible combinations) measured under the same ablation conditions and using the reported values from Jochum et al. (2011) for NIST standard glasses. Two extra variables were also introduced in order to find the better accuracy for the elements of interest, as well as the most suitable standard for each element: 1) the use of two different additional di-atomic gases (N2 and H2), and 2) the use of two types of tubing. Thus, accuracy was estimated with Nylon tubing and N2 (period 1), Nylon6 free-S tubing and N2 (period 2) and Nylon6 free-S tubing and H2 (period 3) (Fig. 5.1). The results are given as a percentage offset between the reported value from Jochum et al. (2006) for MPI-DING glasses and the mean value measured here (such as in Evans et al. (2013), i. e. the lower percentage the better accuracy. As shown in Figure 5.1, a striking trade-off between the heaviest isotopes 138Ba and 238U versus the other elements is observed when using the Nylon6 free-S tubing (yellow and green bars). B accuracy highly improves when the new tubing Nylon6 free-S is used and under H2 conditions. For Mn and Mg the largest change seems to depend on the type of NIST standard used. The use of H2 clearly improves Mn accuracy. For Al, H2 makes the accuracy slightly worse and for Mg, the result depends on the NIST used. Sr and Y accuracy is quite high in all cases. It is therefore remarkable how the B accuracy improves when the new tubing is used, and how Mn accuracy is enhanced when the additional di-atomic gas H2 is used (B accuracy is slightly enhanced in this case too). With the exception of Ba and U that accuracy gets worse, the rest of the elements hardly experience significant changes with the use of either the different tubing or gas. The reason why Nylon6 112 B/Ca foraminifer analyses by LA-ICPMS nine core-top samples (Table 5.3) were picked from the >212 mm size fraction and drilled with a 96 mm spot size. For N. dutertrei, as described in subchapter 5.1, 8-10 specimens from six ODP1240 samples were picked from the 315-355 mm size fraction and drilled with a 74 mm spot size. Samples were ablated for a maximum of 200 s and 120 s for C. wuellerstorfi and N. dutertrei, respectively, with delays of 30 s between measurements in both cases, and applying the analytical setup described in subchapter 5.1 (using H2 as additional di-atomic gas and Nylon6 free-S tubing). Table 5.3. Core-top sample information. A location map can be seen in Figure 5.8 below.      RegionandCruiseSite/StationLatitudeLongitudeDepth(m)           Artic     HERMIONEARTIC stC 77˚14.90'N 09˚30.98'E 2,011 stE 76˚22.67'N14˚35.72'E615     NorthAtlantic    CATARINA PALEOACID KTACG0152˚06.38'N42˚51.67'W4,147 KTABC0254˚34.96'N39˚07.53'W3,002 KTABC0345˚26.39'N34˚33.14'W4,014 KTABC0546˚11.16'N18˚18.83'W3,938      IberianMargin     JC089 JC08903 SHAK033B37˚42.54'N10˚29.56'W3,731 JC08911 SHAK119B37˚51.51'N09˚20.14'W618      SouthWestAtlantic          GEOTRACESst1029˚S 35˚W 3,879        that elevated Al could indicate contamination by aluminosilicates (clays), as has been done in previous studies of Mg/Ca in planktonic foraminifera, although with a lower contamination threshold, as planktonic species are usually less susceptible to Al incorporation (Bolton et al., 2011; Boyle, 1983; Marr et al., 2011). Once the data reduction has been done for the whole data set, a last step to screen 113 Chapter 5 Several steps to reduce the data by removing any contaminant phase were applied after using an Excel template to “despike”, standardise with NIST glasses and normalise data to Ca. Thus, the first step of the data reduction process, and an advantage when using laser ablation, is to eliminate the surface enrichments of the shells. Apparently, this contamination is either adhered onto the outer part or goes into the pores once the specimens are dead (Reichart et al., 2003), or it could also be an original part of the shell as has been observed in samples from sediment traps that were never deposited or incorporated into the sediments (Hathorne et al., 2009). Regardless of their origins, these surface enrichments are a potential factor to bias measurements away from those representative of pure foraminiferal calcite, and therefore to bias paleoenvironmental inferences if they are included in the final averages. A second step of data reduction is to select the segment of the profile suitable for integration. In this thesis two approaches have been applied: 1) selecting the part of the profile with constant calcium (hereafter referred as “flat calcite” approach) and, 2) selecting also the part of the ablation that presents a progressive element intensity decrease, as long as it does not correspond to the ablation of another material rather than foraminifer shell, such as the tape or the glass slide indicated by very noisy element signals (hereafter referred as “whole ablation” approach). The decrease in element intensity along the ablation is a typical feature in laser ablation analysis since the material extraction becomes harder as ablation depth increases, and it is commonly excluded from integration (retaining only the part of the ablation that present “flat calcite”), as it might potentially produce isotopic fractionation and/or might alter the element/Ca ratios by triggering a variation in the volatility of the different elements. However, as solution-ICPMS analyses are made by using whole specimens, long ablations have been performed in this thesis so that different profile integrations can be further compared between techniques. The next step of this process consists of discarding the profile segment from each analysis with Al/Ca ratios higher than 500 µmol/mol, even when this restriction might imply disregarding a whole profile. This criterion is based on the premise 114 B/Ca foraminifer analyses by LA-ICPMS for potential contaminant phases that might exclusively bias B/Ca estimates would consist of removing all chambers with B/Ca that are higher than the interval average plus 2SD, defined as outliers (Evans and Müller, 2013). This is on the premise that outliers may occur as a result of effects additional to or independent of aluminosilicate contamination specifically. The final suitable profile segments selected for integration are then averaged by chamber, then by specimen and finally by sediment interval for subsequent palaeoreconstructions, which are discussed in Chapter 6. 5.2.2.2. Solution-ICPMS analyses The C. wuellerstorfi individuals from 12 of the 15 ODP1240 samples analysed by LA-ICPMS at RHUL were removed from the double-sided tape, cleaned by using the “oxidative” approach at the University of Barcelona, and analysed by solution at the University of Cambridge, as described in Subchapter 5.1 for N. dutertrei. The N. dutertrei samples used in subchapter 5.1 are also screened here. 5.2.3. Results and discussion 5.2.3.1. Downcore B/Ca measurements in C. wuellerstorfi C. wuellerstorfi chambers are typically narrow and contain low B concentration. This fact initially invited the use of a 96 mm spot size for ablation, which covered parts of the shell rather than single chambers. The nomenclature adopted for each ablation in every specimen is the commonly used in these type of studies, which is related to the direction of shell growth and goes from the foraminifer’s youngest chamber (f, from the “final” chamber that was built) towards the foraminifer’s oldest chamber (f-X, where the larger the X, the older the chamber). Thus, two to three depth profiles were performed per shell in this species: one on the oldest part 115 Chapter 5 (f-2), one on the middle (f-1) and one on the youngest part (f) of each individual. They will be referred to as “chambers” in this species from now on, as they are not real chambers (Fig. 5.3a). 5.2.3.1.1. Element laser profiles and intra-/inter-“chambers” variability Figure 5.3b displays representative examples of the trace element depth profiles across three “chambers” of two C. wuellerstorfi specimens from ODP1240, where different thickness between chambers is observed (the older the chamber, the thicker the wall and vice versa) due to the bilamellar type of growth of this species, where a new layer of material is secreted over all existing chambers when a new chamber is formed (Saraswati and Srinivasan, 2015). Surface enrichments are present in all samples and are especially visible for 27Al, 55Mn and 238U and less pronounced for 24Mg and 25Mg (Creech et al., 2010; Eggins et al., 2003; Pena et al., 2005; Reichart et al., 2003; Sadekov et al., 2008). However, 11B surface enrichments are hardly perceptible, as also reported by Hathorne et al. (2009) for several planktonic foraminifera. They are not observed for 88Sr either. In any event, they were always removed before data integration. The remaining element signals across the shell generally present regular profiles, with the exception of Al (and other elements like Mg or U that follow Al trends in some cases) that, depending on the specimens, may present internal variations. B signal, although generally relatively flat, and therefore suggesting that is not affected by other trace element’s incorporation, seem to consistently present a “bump” across the profile, specifically in the oldest “chamber”, which is also the thickest one, and therefore represent the whole period of life of each specimen. The progressive element intensity decrease typical from laser ablation analysis is observed in this study. The two data reduction approaches applied in this thesis, “flat calcite” and “whole ablation”, correspond to the integration of around 3-10 µm and 16-32 µm, respectively, as the ablation depth per second is ~0.16 µm 116 B/Ca foraminifer analyses by LA-ICPMS (Fig. 5.3b). In Figure 5.4, the resultant integrated segments of B related to Ca from “chambers” displayed in Figure 5.3 can be seen for both approaches. Intra-chamber 0 100 200 300 400 10 100 1000 10000 100000 1000000 2200 2300 2400 2500 2600 10 100 1000 10000 100000 1000000 "Chamber" f Surface contamination "Chamber" f-1 "Chamber" f-2 Oldest part 129W1 Intensity (cps) 11 B 24 Mg 25 Mg 27 Al 43 Ca 55 Mn 88 Sr 238 U Individual Integrated part/Integrated part Youngest part Intensity (cps) Ablation Time (s) 259W6 a) b) f f-1 f-2 Figure 5.3. Illustration of laser ablation approach in C. wuellerstorfi. a) C. wuellerstorfi specimen before and after ablation displaying the nomenclature used in this thesis. b) Trace element profiles across the foraminifer shell of two C. wuellerstorfi individuals (129W1 and 259W6), displaying three “chambers” per specimen. The two approaches applied in this thesis for selecting the segment to integrate are shown by black (“whole ablation”) and grey lines (“flat calcite”), both removing the surface enrichments. 117 Chapter 5 B/Ca variability of any “chamber” is between 12-13% for both approaches, representing much lower variability than previous reported LA-ICPMS measurements in this species (Raitzsch et al., 2011: ~40% of intra-chamber variability). The “bump” observed in B profiles in the thickest “chambers” in Figure 5.3 is translated into a cyclical B/Ca rise and drop in the case of the “whole ablation” approach and only the first rise (or part of it) in the case of the “flat calcite” approach, where a shorter part of the profile is selected (Fig. 5.4). If the consistently observed cyclical B/Ca rise and drop in the oldest-thickest “chamber” reflects composition changes associated with bilamellar growth of the Cibicides genus on either side of an equatorial plane (towards dorsal and ventral faces; Saraswati and Srinivasan, 2015), then a long ablation at this spot might cover not only one “chamber”, but also some other “primary chambers” underneath. If this is correct, this would reflect an interesting decreasing pattern in B incorporation as the chamber is built (in both Figure 5.4. LA-ICPMS B/Ca profiles for selected segments of two C. wuellerstorfi specimens’ “chambers”. a) and c) “Whole profile” approach. b) and d) “Flat calcite” approach. Colours repre-Flat calcite” approach. Colours repre-” approach. Colours represent the different “chambers” of one specimen, from the oldest (f-2) to the youngest (f). 118 B/Ca foraminifer analyses by LA-ICPMS dorsal and ventral directions), that might be due to changes in the microenvironmental conditions, but more likely due to ontogenetic processes. Elucidating these is beyond the scope of this study. However, as B/Ca has been empirically proved to relate to D[CO3 2-], what it is interesting here is how this feature might modify the final average values between approaches, as only part of the dorsal face is usually integrated when taking the “flat calcite” approach. Besides the intra-“chamber” variability, the B/Ca composition between different “chambers” (or inter-chambers) also present some differences, with a gradient from low to high B concentrations (in average) as you move from the oldest part of the shell (f-2) to the youngest part (f) (Fig. 5.5a). Note that this way of showing the “chambers” trend is not ideal, as it includes individuals from different sediment depth and therefore “chambers” variability and potential climatic/hydrographic signals might conflated (e.g. higher B/Ca from “chambers” f-2 in the Holocene period might be coincident with lower B/Ca from “chambers” f during the glacial). However, the few “chambers” analysed in each specimen does not allow for a more representative inter-chamber comparison within intervals. Despite of this, Figure 5.5 gives us an idea of the inter-“chamber” trends in this species, which is actually supported by an unpublished study focused on inter-(real) chamber comparison (Kerr et al., unpublished) that shows the same variability range and trend, from lower to higher B/Ca ratios along the chambers formation. When looking at other trace elements, a similar trend between “chambers” is also found in Al/Ca ratios, which also display higher values from f-2 towards f (Fig. 5.5b), while Mg/Ca and Mn/Ca ratios hardly differ between “chambers” (Fig. 5.5c and d, respectively). B/Ca does not correlate with Mg/Ca or Mn/Ca, and despite the fact that B/Ca presents a similar pattern to Al/Ca along the foraminifer shell, it does not exhibit a significant correlation across the shell walls nor between “chambers” (R2 = 0.1) (also pointed out by Rae et al. (2011) and Yu and Elderfield (2007) in bulk analysis). This indicates that, in principle, the reason for an apparent B/Ca gradient between 119 Chapter 5 “chambers” might be attributed to “natural” changes, either micro-environmental or ontogenetic and that B/Ca ratios are not biased by diagenetic coatings or silicate contamination in this species. However, it should be pointed out that the youngest “chamber” (f) typically contains very high Al and often B, indicating a higher susceptibility of this youngest “chamber” to authigenic coatings or metal adsorption. This is important since the rejection of many of these last “chambers” for further integration due to either a higher than 500 µmol/mol of Al/Ca content or a higher 2SD B/Ca bias will affect the B/Ca final average. Figure 5.5. Element/Ca per “chamber” for all C. wuellerstorfi measured in this thesis. a) B/Ca, b) Al/Ca, c) Mg/Ca, d) Mn/Ca. Colours represent the different “chambers”, from the oldest (f-2) to the youngest (f). Black dots represent the average element/Ca ratios for each type of “chamber”. Black line indicates the trend in element/Ca ratios through shell growth. 0.14 0.16 0.18 0.20 0.22 0.0 0.2 0.4 0.6 0.6 1.2 1.8 2.4 0.0 0.8 1.6 2.4 3.2 d) c) b) B/Ca (mmol/mol) f-2 f-1 f a) Al/Ca (mmol/mol) Mg/Ca (mmol/mol) Mn/Ca (mmol/mol) 120 B/Ca foraminifer analyses by LA-ICPMS A slight significant correlation is consistently found between Mg/Ca and Mn/Ca (R2 = 0.3). This has been pointed out before in this EEP area, as a result of the existence of Mn-Mg-rich phases in the foraminifera shell, that prevents the reconstruction of accurate temperature records using the Mg/Ca-palaeothermometer, unless the reductive cleaning method is applied (Pena et al., 2008a, 2005). 5.2.3.1.2. Comparison between laser and solution analyses A comparison between laserand solution-ICPMS B/Ca analyses is shown in Figure 5.6. The results between approaches and techniques are quantitatively different, with higher B/Ca values obtained by solution and lower values by the LA-ICPMS “flat calcite” approach. Between the two LA-ICPMS approaches applied in this thesis, the “whole ablation” approach better represents the common solution analysis, both in shape and magnitude. In terms of magnitude this might be expected, as a greater part of the shell is included in the total average. Importantly, this would suggest that isotope fractionation and/or differences in element volatilities in deep laser ablations is not an issue when using LA-ICPMS for B/Ca studies. In terms of shape, with the exception of a few intervals, an intriguing constant offset is observed between solutionand laser “whole ablation” approach measurements (Fig. 5.6). What are the reasons for such a consistent offset between techniques? Firstly, it seems it cannot be attributed to the rejection of outliers, as the resulting curve is, in both approaches, nearly identical to that given by the curve including them, as only some “chambers” lie beyond the 2SD range of its respective interval in both approaches. An obvious and simple reason for the offset observed might be the type of standard used as, apparently, the use of NIST standards instead of calcite in LA-ICPMS analysis usually gives results less consistent with solutionICPMS measurements (Hathorne et al., 2003). However, another possible reason for this consistent offset (as this study points out) is the removal of many of the final “chambers” during the data reduction of LA-ICPMS analyses, or at least part of them, due to the high Al content (which is usually coincident with higher B/Ca 121 Chapter 5 ratios). These final “chambers” are thinner than the rest, but also represent a large area of the shell, and solution-ICPMS analysis includes them all in the final average. This fact, therefore, might increase the final values obtained by solution creating the offset observed between both techniques. In other words, solution-ICPMS measurements would consistently include high-B/Ca final “chamber” calcite (which would have to be offset by a constant amount relative to the rest of the tests), while LA-ICPMS measurements would omit these. In support of this idea, interestingly, the largest discrepancies between solution and LA-ICPMS “whole ablation” Figure 5.6. B/Ca comparison between the two data reduction LA-ICPMS approaches and solution analysis of the same C. wuellerstorfi samples downcore. a) Solution-ICPMS. b) “Whole ablation” approach. c) “Flat calcite” approach. Dashed and thick solid lines in b) and c) indicate data with and without outliers, respectively. Note that not all the intervals measured by LA-ICPMS were also measured by solution. 0 50 100 150 200 250 300 350 400 450 0.13 0.14 0.15 0.16 0.17 0.18 0.19 0.20 0.21 0.22 0.23 0.24 0.25 solution-ICPMS "whole profile" "flat calcite" c) b) a) B/Ca (mmol/mol) Core depth (cm) 128 B/Ca foraminifer analyses by LA-ICPMS 2900 3000 3100 3200 3300 3400 1 10 100 1000 10000 100000 1000000 2700 2800 2900 3000 3100 3200 3300 3400 1 10 100 1000 10000 100000 1000000 Instrumental background Crossing the shell Surface contamination Chamber f-1 Chamber f-2 Oldest part 9D6 Intensity (cps) 11 B 24 Mg 25 Mg 27 Al 43 Ca 55 Mn 88 Sr 238 U Chamber f Chamber f-3 Chamb f-4 Individual Integrated parts Youngest part Intensity (cps) Ablation Time (s) 283D9 Chamber f-4 Chamber f-3 Chamber f-2 Chamber f-1 f f-1 f-2 f-3 f-4 a) b) 129 Chapter 5 Two representative examples of the trace element depth profiles across each chamber of two N. dutertrei specimens from ODP1240 core are shown in Figure 5.10b. Wall thickness varies consistently and significantly between chambers (also reported by Eggins et al., 2003), with the central chambers f-2 and f-3 presenting the thickest walls and therefore the longest profiles. Chamber f-4 (the oldest one) is usually hidden and very small (although when visible it is also thick), and chamber f (the youngest one) is often very thin and fragile. For these reasons, f-4 and f chambers were not always ablated. As for C. wuellerstorfi, surface enrichments are visible for all elements (with the exception of 11B and 88Sr, indicating again that these are not major constituents of surface contaminant phases), and were always removed before data integration. The B signal is very uniform in all chambers indicating that, apparently, B incorporation does not change over chamber growth and that it is not affected by other trace elements’ incorporation. Al presents an irregular profile with internal variability, as it was also observed for C. wuellerstorfi. The B/Ca shape for all chambers from the resulting integrated parts of the two N. dutertrei specimens seems to be either pretty flat or with a discrete slope from the inner to the outer part (Fig. 5.10c). B/Ca variability within a single profile (in a given chamber) is between 21-26 %, being not very different between chambers. As for C. wuellerstorfi, B/Ca ratios are different between chambers, but following a different pattern. Thus, chamber-specific patterns show that, on average, the central chambers f-2 and f-3 chambers tend to present higher B/Ca as compared to the youngest chamber f and the oldest f-4, which tend to have the lowest (Fig. 5.11). Lower B/Ca in the oldest chamber has also been reported in Globorotalia sacculifer (Allen et al., 2012). Figure 5.10. Illustration of laser ablation approach in N. dutertrei. a) N. dutertrei specimen before and after ablation displaying the nomenclature used in this thesis. b) Trace element profiles across the foraminifer shell of two N. dutertrei individuals (9D6 and 283D9), with 4-5 chambers per specimen. c) N. dutertrei B/Ca selected segments of the two specimens’ chambers. Colours represent the different chambers of one specimen, from the oldest (f-4) to the youngest (f). 130 B/Ca foraminifer analyses by LA-ICPMS Other chamber-specific elemental ratios, such as Mg/Ca, seem to behave in the opposite manner, i.e. the chambers with higher Mg/Ca contain the lowest B/Ca and vice versa. This might suggest that either the variable/variables controlling B incorporation is/are related to ontogenetic processes rather than microenvironmental changes, as temperature and carbonate system parameters are usually positively related, or that B/Ca in N. dutertrei is not a consistent proxy for the carbonate system. Al/Ca and Mn/Ca from different chambers do not follow any consistent pattern, though the youngest chamber f typically presents very high values in some intervals, indicating the higher susceptibility of these chambers to authigenic coatings or metal adsorption. Figure 5.11. Element/Ca per chamber for all N. dutertrei measured in this thesis. a) B/Ca. b) Al/ Ca. c) Mg/Ca. d) Mn/Ca. Colours represent the different chambers, from the oldest (f-4) to the youngest (f). Black dots represent the average element/Ca ratios for each chamber. Black line indicates the apparent trend in element composition through shell growth. 0.05 0.06 0.07 0.08 0.09 0.10 0.11 0.00 0.05 0.10 0.15 0.20 0.0 0.6 1.2 1.8 2.4 0.0 0.3 0.6 0.9 f-4 f-3 d) c) b) B/Ca (mmol/mol) f-2 f-1 f a) Al/Ca (mmol/mol) Mg/Ca (mmol/mol) Mn/Ca (mmol/mol) 131 Chapter 5 A poorer relationship between laser and solution measurements on the same N. dutertrei specimens compared to C. wuellerstorfi is observed for all elements (Fig. 5.12). The reason of this disagreement is not well constrained, but it might be related to the wall structure of this species, as the fact of being less dense and flat, as well as more porous and ornamented, usually yield worse laser ablation profiles than in benthic foraminifera. Other possible reasons for the observed laser versus solutiondiscrepancy in N. dutertrei may be possible, but are as yet unknown. 0 50 100 150 200 250 300 55 60 65 70 75 80 B/Ca (µmol/mol) Core depth (cm) 0 20 40 60 80 U/Ca (nmol/mol) 0.6 0.9 1.2 1.5 Mg/Ca (mmol/mol) 50 100 150 200 250 300 350 Mn/Ca (µmol/mol) 0 15 30 45 60 75 Al/Ca (µmol/mol) 0 50 100 150 200 250 300 Core depth (cm) Figure 5.12. Comparison of element/Ca ratios between LA-ICPMS (solid lines) and solution-ICPMS (dashed lines) in N. dutertrei. 132 B/Ca foraminifer analyses by LA-ICPMS 5.2.4. Conclusions The B signal across the “chambers” of C. wuellerstorfi is generally quite uniform, with no surface enrichments, and no correlation with Al. Mg correlates with Mn, indicating the importance of the reductive cleaning application for further Mg/ Ca analysis in this area (Pena et al., 2008a, 2005). Wall thickness and B/Ca ratios in C. wuellerstorfi seem to be related with the bilamellar growth of this species, going from lower B/Ca in the thicker-oldest part (f-2) to lower values in the thinneryoungest part (f). When comparing B/Ca final averages between the two LA-ICPMS approaches and the solution analyses performed on the same samples, longer ablations yield results that are more consistent with solution-ICPMS analyses, presenting similar variations but with a constant offset. It is suggested that, apart from the potential influence of using a non-calcite standard, this consistent lower offset in “whole profile” approach versus solution analyses might be due to the different rejection thresholds applied in both techniques to avoid Al contamination, which are typically higher in solution-ICPMS. This study has shown that final “chambers” f usually contain higher Al/Ca than the rest of the shell, as well as higher B/Ca, which are rejected during the data reduction of LA-ICPMS analyses. This calls for a revision of the Al/Ca rejection threshold in solution analysis, which the results from this study suggest should be lower. Although longer ablations seem to be more consistent with solution measurements, further assessment of which ablation selection method is optimal, and indeed how calibrations are affected by these choices, is probably warranted, as the “flat calcite” approach decreases the analysis time by 3 times, thus increasing analysis efficiency and significantly reducing analysis costs. In order to achieve a low and constant error and to be statistically representative of the time interval being analysed for B/Ca ratios, especially in the context of benthic foraminifera, a minimum of 5-6 individuals should be analysed, as often 1 or 2 individuals per sample cause significant biases from the average in many cases. Core-top C. wuellerstorfi B/Ca ratios by LA-ICPMS are in good agreement with corresponding modern seawater D[CO3 2-] values. Furthermore, although offsets between 133 Chapter 5 LA-ICPMS and solution analyses may exist, calibrations made using core-top solutionand LA-ICPMS analyses are indistinguishable, suggesting that data using both approaches might be combined. However, a new calibration for LA-ICPMS B/ Ca analysis may be needed in the future in order to quantitatively compare results between techniques accurately. Regarding N. dutertrei analysis, no surface enrichments or correlation with Al are found and B signal across the shell is typically quite flat. Mn is not as high as in C. wuellerstorfi. Wall thickness also varies depending on the chamber, but going from thicker in the central chambers (f-3 and f-2) to thinner in the youngest (f) and oldest (f-4) chambers. B/Ca also seems to relate to this distribution, but curiously in the opposite way than C. wuellerstorfi, i. e. N. dutertrei shows lower B/Ca values in the thinnest chambers. When the samples are analysed by solution-ICPMS, the similarities presented between techniques are not as striking as in the case of C. wuellerstorfi, which might be related to the wall structure of this planktonic species for example, which is not as “homogeneous” as in benthic foraminifera. This preliminary study at the B/Ca distribution in N. dutertrei, while yielding typical values obtained for this species by solution (Dai et al., 2016; Foster, 2008), might suggest that it is not a viable carbonate system proxy (or at least not for D[CO3 2-]; Dai et al., 2016). However, culture experiments should be performed in this species in order to understand what oceanographic/climatic variable, if any, B/Ca might be related to. As a take home message, this study shows the suitability and advantages of LA-ICPMS technique for B/Ca analysis in C. wuellerstorfi. Importantly, it permits the rejection of chambers or even entire specimens with anomalous values that might quantitatively bias the final averages for palaeoceanographic interpretation. This may be especially important for analyses of very small samples. 135 Chapter 6 CHAPTER 6 6. The evolution of deep ocean chemistry in the Eastern Equatorial Pacific over the last deglaciation Synopsis In this chapter, the B/Ca results from C. wuellerstorfi analysed by LA-ICPMS and described in Chapter 5 are interpreted. They are supported by new d13C measurements and discussed in the context of the radiocarbon ventilation records addressed in Chapter 4 (using the age model described in Chapter 3). The new results are also complemented with other existing records from core ODP1240 and with B/Ca records from other locations in the Pacific Ocean. 6.1. Introduction Several marine carbon cycle mechanisms have been invoked to explain, at least, a major part of the atmospheric CO2 drawdown observed over the last glacial period (e.g. Kohfeld and Ridgwell, 2009; Sigman et al., 2010). These mechanisms can be categorised broadly as deriving primarily from: 1) changes in the rate or efficiency of air-sea gas exchange (i.e. the “solubility pump”; e.g. Volk and Hoffert, 1985); 2) changes in whole ocean carbonate chemistry (e.g. due to carbonate compensation/dissolution; e.g. Archer and Maier-Reimer, 1994; Broecker and Peng, 1989; Keir, 1995); or 3) changes in the efficiency of the “biological carbon pumps” (i.e. “soft-tissue” vs “carbonate” pumps; e.g. Volk and Hoffert, 1985; see Chapter 1 for further details). While these categories might be useful for organising our thinking about past marine carbon cycle changes, it is important to emphasize that they need 136 Evolution of deep ocean chemistry in the EEP not have operated in isolation from each other. Indeed, a more efficient biological carbon pump can occur as a result of either an increase in its strength (i.e. average export production rates; e.g. Martínez-Garcia et al., 2014), or a reduction in its leakiness (i.e. the return flux of remineralised nutrients and carbon from the deep interior to the surface ocean; e.g. (Stephens and Keeling, 2000; Toggweiler, 1999). If the latter is achieved by a change in the ocean’s large-scale overturning circulation, it would very likely co-occur with a change in the conditions of air-sea gas exchange and the solubility pump. Furthermore, regardless of how biological carbon pump efficiency is increased, and as long as the soft-tissue vs carbonate pump remains dominant, this would lead to more respired carbon storage in the deep ocean (thus sequestering CO2 from the atmosphere), which could eventually reduce deep ocean carbonate ion concentrations enough to cause seafloor carbonate dissolution (a mechanism also referred to as “respiratory calcite dissolution”, (Archer and MaierReimer, 1994; Archer, 1996, 1991). The latter would eventually increase the whole ocean alkalinity budget and lead to further atmospheric CO2 drawdown. Therefore, changes in the efficiency of the biological carbon pump due to variations in its leakiness and/or its strength, through their association with ocean circulation and air-sea exchange effects and their potential impact on whole ocean carbonate chemistry, represents a particularly effective way to cause changes in deep-ocean carbon storage over the last glacial period. As discussed in Chapter 4, the EEP has been shown to be less ventilated over the last part of the last glacial period, getting better ventilated at the onset of the Heinrich Stadial 1 (de la Fuente et al., 2015). This would indirectly support the idea of a more efficient biological pump during the last glacial, due to a longer residence time for carbon in the deep ocean, thus reducing the leakiness of the biological pump. As a consequence, this would have stored more carbon in the form of dissolved inorganic carbon (DIC) in the deep ocean, on the condition that the average rate of carbon export to the deep ocean did not decrease significantly. If this had been the case, an increase in respired DIC should be indirectly inferred from changes in other parameters from the deep EEP carbonate system during the last glaciation, such as 137 Chapter 6 the DIC d13C and [CO3 2-], both of which are dependent on respired CO2 variations in the deep ocean (see discussion below and Chapter 1 for more details). Seawater d13C estimates can be derived from foraminiferal d13C measurements, which are easy to perform as long as the right species is available in sufficient abundance. However, [CO3 2-] estimates have been shown to be more challenging to obtain. Over the past few years, B/Ca ratios in C. wuellerstorfi calcite shells have been experimentally proved to correlate with deep water carbonate saturation state (D[CO3 2-]in situ-sat) in the modern ocean, based on a simple linear correlation from global core-top measurements (Yu and Elderfield, 2007), thus providing a proxy for past deep-water [CO3 2-] changes where saturation [CO3 2-] can be assumed to have remained relatively constant. As explained in Chapters 2 and 5, the theoretical basis for such a relationship relies on the seawater pH dependency in the proportion of the two boron molecular species, boric acid, B(OH)3, and borate ion, B(OH)4 -, and the suggestion that only B(OH)4 - is incorporated into biogenic calcite material (Hemming and Hanson, 1992), although this aspect is still not clear (Klochko et al., 2009; Uchikawa et al., 2015). Thus, changes in pH, which is a function of alkalinity (ALK) and DIC, and are therefore closely related to seawater pCO2 and [CO3 2-] (note that [CO3 2-] ~ ALK minus DIC), should produce changes in the proportion of B(OH)4 - and B(OH)3, and thus affect the B/Ca ratios in foraminifer calcite shells. This novel proxy has been applied to several cores from the North Atlantic and the Pacific oceans showing, e.g. small [CO3 2-] changes in the Pacific and larger decreases in the deep Atlantic at the LGM vs the Holocene, in accordance with estimates of sedimentary calcium carbonate preservation (e.g. Allen et al., 2015; Yu et al., 2010a). In this thesis chapter, estimates of [CO3 2-]B/Ca-based and d13C have been analysed in the benthic species C. wuellerstorfi from core ODP1240, in order to investigate if the carbon content of deep waters was indeed greater during the LGM, as suggested by the radiocarbon ventilation ages from this area. B/Ca measurements were carried out by the more challenging LA-ICPMS technique, as described in Chapter 5. 144 Evolution of deep ocean chemistry in the EEP Figure 6.2. Comparison of deep ocean records from ODP1240 with records of atmospheric CO2 and radiocarbon activity over the last 30 kyr. a) Atmospheric CO2 concentrations from EPICA Dome C (EDC) ice core (for the deglacial period (Monnin et al., 2001); for the Holocene period (Flückiger et al., 2002)), placed on the age scale of Lemieux-Dudon et al. (2010). b) Atmospheric radiocarbon activity (Δ14C) changes (IntCal09 calibration curve; (Reimer et al., 2009). c) Deepwater ventilation reconstruction (B-Atm, from SST alignment; de la Fuente et al., 2015). d) δ13C from C. wuellerstorfi. e) B/Ca and [CO3 2-](B/Ca-based). f) Oxygenation proxies: 238U from ME0005-24JC (same location as Site 1240; Bradtmiller et al., 2010; Kienast et al., 2007; black line/triangles), and ∆δ13C between epibenthic and infaunal foraminifera species (C. wuellerstorfi (this study) and Uvigerina sp. (Povea et al., unpublished); grey line/dots). All lines are B-spline smoothed. Vertical dashed lines delimit relevant climatic periods, and light grey vertical bands highlight the coldest periods of the last 30 kyr. 0 5000 10000 15000 20000 25000 30000 160 170 180 190 f) B/Ca C.w (by LA-ICPMS) B/Ca (µmol/mol) Calendar age (yr) 0 5000 10000 15000 20000 25000 30000 -0.6 -0.4 -0.2 0.0 0.2 0.4 e) d) c) δ13CC.w δ13CC.w(permil) Calendar age (yr) 3500 2800 2100 1400 700 B-Atm B-Atm (14C yr) 180 210 240 270 300 b) a) CO2 atm (EDC) CO2 atm(ppmv) HOLOCENE DEGLACIATION LAST GLACIAL PERIOD 0 250 500 750 ∆14Catm (Intcal09) ∆14C (permil) 60 70 80 90 [CO3 2- ]B/Ca [CO3 2- ] (µmol/kg) 6.0 4.5 3.0 1.5 0.0 238U 238U(ME0005-24JC) (permil) -0.1 0.0 0.1 0.2 0.3 0.4 0.5 ∆δ13C (C.w-Uvi) ∆δ13C(C.w-Uvi) (permil) YD HS1 HS2 145 Chapter 6 waters to the insoluble species U(IV) as the medium depletes in oxygen (Cochran et al., 1986; Langmuir, 1978; Morford and Emerson, 1999). In support of the aU measurements, ∆d13C estimates between epibenthic and infaunal foraminifera species (Hoogakker et al., 2014; McCorkle and Emerson, 1988), C. wuellerstorfi and Uvigerina from core ODP1240, show a similar trend (Fig. 6.2f, grey line). The ∆d13C between the sediment/water interface and the sedimentary anoxic boundary should vary with the oxygen concentration of bottom waters. Thus, a smaller ∆d13C at the LGM would indicate less oxygen penetration into the sediments (where Uvigerina lives), thus driving less organic carbon respiration below the seawater-sediment interface and therefore less negative infaunal (Uvigerina) d13C as compared to the seawatersediment interface (where C. wuellerstorfi lives). It should be noted that quantitative calculations of changes in [O2] based on the ∆d13C results presented here (e.g. using the calibration of Hoogakker et al., 2014) might actually yield slightly biased results, as the infaunal species used in this thesis is Uvigerina, which is not an obligate zero-oxygen species such as Globobulimina affinis. The reason for this choice of species is the scarcity of G. affinis found in ODP1240. Nevertheless, the qualitative relative oxygenation changes indicated by ∆d13C seem robust, and are supported by the good agreement with the aU results (Fig. 6.2f). Thus, all the records presented here, i.e. [CO3 2-], d13C and oxygenation proxies, are in good agreement with the deep radiocarbon ventilation in the EEP, and lend support to the hypothesis of a higher efficiency of the biological pump due to changes in its leakiness. Despite such remarkable agreement, it is interesting to note that both oxygenation proxies, aU and ∆d13C, do not change synchronously with the rest of the proxies at the beginning of the deglaciation, but later at the termination. A plausible explanation for this apparent mismatch might rely on a decoupling of pore-water and deep water chemistry, for example due to a productivity pulse between ~17-14 kyr (as seen by Calvo et al., 2011) that would have depleted O2 in the pore-fluids of the sediment (which the oxygenation proxies would record), even after the radiocarbon ventilation and carbonate saturation of the ambient deep-water started to increase. Nevertheless, taken together, all the records 146 Evolution of deep ocean chemistry in the EEP presented here point to the deep EEP as a plausible contributor to the decreased atmospheric CO2 during the LGM, by storing more respired carbon at the expense of the surface ocean and atmospheric carbon inventories. Other possible contributions to the apparent increase in the respired carbon content of the deep EEP are further investigated below. 6.3.2. Export productivity in the EEP The higher residence time of the water masses above core ODP1240 over the LGM indicated by the decrease in radiocarbon ventilation (considering that radiocarbon is not especially sensitive to changes in the soft-tissue or carbonate pumps as mentioned above), along with the quantitative and qualitative consistency of this record with ventilation records from the Southern Ocean and the South Pacific (de la Fuente et al., 2015), suggest that changes in the efficiency of the biological pump due to changes in ocean circulation affected a large volume of the ocean and thus played an important role in the glacial/interglacial atmospheric CO2 variations. However, several studies in core ODP1240 have pointed out an increase in the strength of the biological pump (i.e. export productivity rates) in this area as a mechanism contributing to the glacial atmospheric CO2 drawdown (Calvo et al., 2011; Pichevin et al., 2009; Robinson et al., 2009). The glacial changes observed in deep [CO3 2-], d13C and oxygenation records from ODP1240 described in the previous subsection of this chapter might indeed be compatible with an increase in surface productivity over the LGM, with the consequent higher export to the bottom of the ocean and therefore an increase in the efficiency of the biological pump. Thus, a larger flux of organic matter to the deep ocean in this area might have contributed to the observed drop in [CO3 2-] and d13C described in this chapter as a consequence of higher remineralisation rates that would have also increased the release of respired CO2 to deep water and the consumption of dissolved O2. These processes would be independent of, or in addition to, contributions from changes in ocean ventilation constrained using radiocarbon measurements. 147 Chapter 6 Focusing on the LGM changes, these studies advocate for an increase in surface productivity, with a consequent higher export to the deep ocean, caused by the documented increase in dust-borne iron delivery to the iron-limited low-latitude Pacific Ocean at the last glacial period (McGee et al., 2007). A first inspection of the opal % in ODP1240 as a proxy for silica-fixing organisms abundance (such as diatoms) was however contradictory, as it presented lower values over the LGM, opposite to the proposed increase in surface productivity in the EEP (Fig. 6.3d). However, the also lower opal d30Si during the glacial (as a proxy for the relative utilization of the silica) was interpreted as an indication of a decline in the Si:C uptake ratio due to an iron-replete glacial ocean. This interpretation would thus be compatible with the assumption of higher productivity over this period (Pichevin et al., 2009) (Fig. 6.3d). In the same way, nitrogen isotopes (d15N) from bulk sediments in ODP1240 core show lower values over the LGM compared to the Holocene. Lower sedimentary d15N values could be interpreted as indicative of lower nutrient consumption, but this would be inconsistent with both the absence of iron-silicon limitation (Pichevin et al., 2009) and the observed higher Corg fluxes (Robinson et al., 2009) (Fig. 6.3b,c). However, the correction of the nitrogen isotope signal for denitrification effects occurring closer to the Central American margin (using the sedimentary d15N signal from core ODP1242, situated in this denitrification region), reverses the glacial/ interglacial trend and indicates the highest d15N in ODP1240 during the LGM. This would suggest highest local nutrient consumption at that period, which would be consistent with higher surface productivity over the LGM (Robinson et al., 2009) (Fig. 6.3c). An increase in organic carbon (Corg) accumulation rates in the EEP over the LGM has been documented in core ODP1240 (Pichevin et al., 2009) (Fig 6.3b), as well as in the nearby core P6 (Pedersen, 1983). At the same time, higher abundances of organic biomarkers (alkenones and brassicasterol) are shown in this core during the LGM, as compared to the present (Calvo et al., 2011) (Fig. 6.3e). Both the higher Corg accumulation and phytoplankton abundances might have been the result of an increase in surface productivity, consistent with the isotopic evidence described 148 Evolution of deep ocean chemistry in the EEP 0 5000 10000 15000 20000 25000 30000 g) Calendar age (yr) HOLOCENE DEGLACIATION LAST GLACIAL PERIOD 0 500 1000 1500 2000 2500 f) Brassicasterol Brassicasterol (ng/g) higher diatome productivity 0 5000 10000 15000 20000 25000 30000 Calendar age (yr) 60 70 80 90 [CO 3 2- ] [CO 3 2- ] (µmol/kg) 4 5 6 7 c) bulk δ 15 N Bulk δ 15 N (permil) 0.8 1.0 1.2 1.4 e) d) higher N consumption δ 30 Si δ 30 Si (permil) less relative utilization 0.5 1.0 1.5 2.0 2.5 3.0 3.5 C org Organic C (wt %) 4000 3500 3000 2500 2000 1500 1000 500 -0.4 -0.2 0.0 0.2 0.4 δ 13 C δ 13 C (permil) 5.0 4.5 4.0 3.5 3.0 40 50 60 70 b) a) CaCO 3 % % CaCO 3 5 10 15 20 25 30 Opal % % Opal 0 5000 10000 15000 20000 Alkenones Alkenones (ng/g) 7 6 5 4 3 2 1 0 aU 238 U (ppm) 0.8 1.0 1.2 1.4 1.6 2500 2000 1500 1000 500 0 2.0 1.6 1.2 0.8 0.4 -2 -1 0 1 Residual δ 15 N Res. δ 15 N (permil v. air) 149 Chapter 6 above. Thus, these proxies suggest that an increase in surface productivity over the LGM might have taken place in the EEP area, which would have contributed to increase the efficiency of the biological pump and the atmospheric CO2 drawdown. However, the exact magnitude of this contribution remains unclear. Indeed, the question arises: was the observed increase in the biological pump efficiency in the glacial EEP mainly due to changes in ocean circulation, or changes in the carbon export productivity in this area; how much did the proposed increase in organic carbon export productivity affect carbon storage in the deep ocean? Arguably this question cannot be fully resolved without accurate quantitative estimates of export productivityand deep ocean apparent oxygen utilization changes. At present it is clear at least that ocean ventilation changes made some contribution, though only qualitative conclusions, such as those drawn above, are possible. 6.3.3. Carbonate dissolution as a complementary mechanism for [CO3 2-] EEP changes A decrease in deep ocean ventilation along with an increase in carbon export productivity in the EEP at the LGM are therefore proposed as the mechanisms responsible for the decrease in both deep [CO3 2-] and d13C observed in this area. The action of any of these mechanisms (alone or together) might be expected to achieve a Figure 6.3. ODP1240 records reflecting both surface and deep ocean changes over the last 30 kyr. a) % CaCO3 (orange line) (Pichevin et al., 2009); b) % Corg (blue line) (Pichevin et al., 2009); c) bulk (green line) and residual δ15N (orange line) (Robinson et al., 2009); d) δ30Si (light blue line), % opal (dark blue line) (Pichevin et al., 2009); e) Concentration of brassicasterol (red line) and alkenones (green line) (Calvo et al., 2011); f) authigenic uranium (238U) (brown line) (Bradtmiller et al., 2010; Kienast et al., 2007); g) [CO3 2-](B/Ca-based) by LA-ICPMS (blue line) and δ13C (orange line), from C. wuellerstorfi (this thesis). All lines are B-spline smoothed. Vertical dashed lines delimit relevant climatic periods. Light grey vertical bands highlight the coldest periods of the last 30 kyr. 150 Evolution of deep ocean chemistry in the EEP relatively large [CO3 2-] glacial-Holocene difference. However, the results from this thesis indicate only a small [CO3 2-] change between these two periods (~12 µmol/ kg), suggesting that a “counteracting” mechanism, such as seafloor respiratory calcite dissolution, might have also played a role. When carbonate dissolves in seawater, both [CO3 2-] and ALK increase (Broecker and Peng, 1989; Zeebe and WolfGladrow, 2001). If this was the case, a decrease in the CaCO3 concentration might be initially expected. However, CaCO3 % indicates no changes between 25-15 kyr. This might suggest that any CaCO3 dissolution that occurred in the deep EEP at the LGM would have been masked, for instance, by higher carbonate export productivity rates, which in turn would be consistent with higher abundance of alkenones, Fig. 6.3e. Furthermore, the contribution of any such CaCO3 dissolution to increased average ocean [CO3 2-], would have also been masked to some extent by opposing changes in deep ocean carbonate ion concentrations driven by organic carbon respiration. Thus, CaCO3 seafloor dissolution in the deep Pacific, as recorded in core ODP1240, not only might have ”buffered” glacial [CO3 2-] changes without affecting d13C in the EEP, but also might have played a role in the atmospheric CO2 drawdown during the LGM by contributing to increased average ocean alkalinity via [CO3 2-] supply to the ocean (Keir, 1995; Sigman and Boyle, 2000). This would have further reinforced the effects of higher biological pump efficiency achieved by the circulation and surface productivity changes proposed above. 6.3.4. ∆[CO3 2-](Hol-LGM) derived from B/Ca across the Pacific Ocean Although B/Ca ratios in benthic foraminifera is a relatively novel proxy for deep water [CO3 2-], it has been used already a number of times in marine sediment samples from the Pacific Ocean over the last years (Allen et al., 2015; Doss and Marchitto, 2013; Elmore et al., 2015; Yu et al., 2010a). In Figure 6.4, a compilation of averaged [CO3 2-] from these studies is presented where, LGM averages (open 151 Chapter 6 squares) across the equatorial-south Pacific Ocean are generally consistent with each other. These records indicate a LGM variation between sites of only ~5 µmol/ kg, with the exception of the deepest core in the central Pacific (PC61; 4,300 m), which displays a larger LGM value (Fig. 6.4). In fact, this deep core also presents an almost identical [CO3 2-] for the Holocene, which might point to a constant higher degree of CaCO3 dissolution in this central deep area. In contrast, Holocene-modern Figure 6.4. Deep [CO3 2-] records from several Pacific Ocean sediment cores, showing relatively small changes in carbonate saturation between the Holocene and LGM in the Pacific: DSDP593Z (Elmore et al., 2015), RR0503-83 (Allen et al., 2015), GGC15 and GGC48 (Yu et al., 2010a), PC61 (Yu et al., 2013) and ODP1240 (this thesis, highlighted in red). LGM data points are marked by a dashed line/open squares and Holocene by a solid line/filled squares. Modified and updated from Allen et al. (2015) and Yu et al. (2013). Note that the 120 m sea level difference between the Holocene and the LGM was taken into account. Map location at the top right of the figure. 50 60 70 80 90 100 110 4000 3000 2000 1000 RR0503-83 (WP) Holocene LGM Water depth (m) [CO3 2- ] (μmol/kg) ODP1240 (EEP) TTNO13 PC61 (CP) MW91-9 GGC48 (WP) MW91-9 GGC15 (WP) DSDP593Z (WP) 152 Evolution of deep ocean chemistry in the EEP [CO3 2-] averages (filled squares in Fig. 6.4) from all the other cores indicate higher variation between sites, which might suggest a more “homogeneous” glacial deep equatorial-south Pacific Ocean. Regarding glacial-Holocene/modern [CO3 2-] differences, these studies overall suggest little [CO3 2-] change in the deep Pacific between the two periods (Fig. 6.4). This “unexpected” small variation has been attributed to the large buffering capacity of carbonate sediments in the Pacific Ocean (Allen et al., 2015; Yu et al., 2013, 2010a). [CO3 2-] changes between the LGM and the Holocene at ODP1240 site agree well with this hypothesis. Core GGC15 at the South West Pacific SWP (2,300 m) is the only one displaying, in average, lower [CO3 2-] at the Holocene compared to the LGM. The authors have speculated that this core site might have been influenced by glacial NPDW/southern-sourced differences in the mixing ratio (Yu et al., 2010a) as well as by carbonate compensation effects. The ∆[CO3 2-](Hol-LGM) changes observed in Figure 6.4 are smaller than the changes observed in the Atlantic and Southern Oceans (Gottschalk et al., 2015; Yu et al., 2014, 2013). This is expected, precisely because of the current shallower calcite compensation depth (CCD) in the Pacific due to the greater carbonate dissolution in the modern Pacific Ocean as compared to the Atlantic. 6.4. Conclusions In this chapter, a possible increase in the biological pump efficiency over the LGM has been investigated as a mechanism for the atmospheric CO2 drawdown over that period. Radiocarbon ventilation is not affected by any biochemical change and indirectly supports higher biological pump efficiency through a decrease in its leakiness over the LGM. In the same way, productivity proxies from this area also suggest an increased efficiency by an increase in the strength of the biological pump. Both [CO3 2-] and d13C changes observed in this thesis, along with published oxygenation proxies, support these observations by presenting lower values in the 153 Chapter 6 EEP at the LGM compared to the Holocene. However, the decrease in [CO3 2-] in this area is relatively small, suggesting that these mechanisms were not the only ones operating or indeed contributing to the drop in the glacial atmospheric CO2. This has also been observed in other Pacific [CO3 2-] records and has been suggested to be a consequence of larger carbonate dissolution over the LGM in this basin. Thus, respiratory calcite dissolution not only might explain the relatively small change in deep ocean [CO3 2-] despite a larger d13C difference between the glacial and Holocene periods, but also would have likely and directly contributed to the glacial atmospheric CO2 decrease by increasing the ocean average ALK and therefore decreasing the surface ocean pCO2, thus allowing the surface ocean to take up more CO2. In conclusion, the deep EEP, and probably much of the wider deep Pacific, very likely sequestered a significant amount of atmospheric CO2 during the LGM, specifically due to a more efficient biological carbon pump caused by both lower ventilation rates and enhanced export productivity, but also due to an increase in average ocean alkalinity by greater carbonate dissolution.