Plio-Pleistocene Ocean Circulation Changes in the Gulf of Alaska and Its Impacts on the Carbon and Nitrogen Cycles and the Cordilleran Ice Sheet Development

Integrated Abstract The modern Gulf of Alaska (GOA) is a Cordilleran Ice Sheet (CIS) region, estimated to be important for nutrient cycling and CO 2 exchange. Little is known of the GOA evolution over the Pliocene and Pleistocene as well as its impact on the CIS development, when other evidence for changing North Pacific circulation has emerged. We analyzed Integrated Ocean Drilling Program Expedition 341 Site U1417 sediments, which extend through the Plio-Pleistocene transition (4–1.7 Ma), focusing on productivity-related biomarkers (alkenones, brassicasterol), siliceous microfossils and bulk carbon and nitrogen stable isotopes. Our results show two dominant water column regimes: one characterized by high silica and low organic matter (OM) preservation, containing microorganism remains from a mix of habitats (4–3.7 Ma) and a second characterized by low biogenic silica and increased OM preservation of microorganisms from dominantly open ocean habitats (3.33–3.32 Ma and 2.8–1.66 Ma). An increase of phytoplankton diversity (3.7–3.35 Ma, 3.19–2.82 Ma) characterizes the two transitions of water column conditions, from oxygenated to reductive, that we attribute to a change from ocean mixing to strong stratified conditions with some occasional mixing. The biogeochemical changes in the GOA follow 400 and 100 kyr eccentricity cycles which are also reflected in changes in the CIS. We conclude that the CIS expansion created high nutrient low chlorophyll conditions in the GOA during the Mid Piacenzian Warm Period and the early Pleistocene. In turn, positive feedbacks increased marine productivity export, atmospheric CO 2 drawdown and further CIS expansion.

. Modern ocean currents and onshore to offshore water properties in the Gulf of Alaska (GOA). Map of the GOA showing data collection sites (blue dots) and selected transect (red rectangle, upper panels) that give rise to the vertical plots; temperature (ºC) salinity (pps), nitrate (μmol kg −1 ), chlorophyll (μgl l −1 ) and oxygen (μmol kg −1 ) during winter (F-M, left) and summer (J-A-S, right). ACC = Alaska Coastal Current, AC = Alaska Current, AS = Alaska Stream, NPC = North Pacific Current, AL = Aleutian Low. The location of site U1417 is indicated with yellow stars and the River Yukon is represented in blue in the upper panels. Data downloaded from World Ocean Database , bathymetry data downloaded from GEBCO (GEBCO, 2020) and plotted with Ocean Data View (Schlitzer, 2016). The Cordilleran Ice Sheet (CIS) would occupy the highest Alaskan topography (darker brown colors) of Wrangell-St. Elias and McKenzie Mountains, partially the Yukon and Tanana upland and Olgivine Mountains during the Pliocene-early Pleistocene (2.9-2.6 Ma, Duk-Rodkin et al., 2004). 4 of 22 sediment recovery in this section is close to 100% (Jaeger et al., 2014). This section of the age model remains unchanged in the modified age model used by Sánchez-Montes et al. (2020) and the age errors adopted here from Jaeger et al. (2014) account to ±0.02 Ma, with a propagated error across neighboring samples of ±0.028 Ma. For the sediment record older than 2.2 Ma (288.45 m) sediment recovery reduces significantly (70%-12%; Jaeger et al., 2014). The revised age model redistributes this material across the core sections assuming no loss in material (Sánchez-Montes et al., 2020). The age model adjustments resulted in a similar age/depth of the C2r.2r (B) Gauss/Matuyama top (found in U1417 B, D and E), however, C2An.3n (B) Gilbert/Gauss transition found only in U1417D shifted from 408.26 to 410.75 m CCSF-A and from 3.75 to 3.88 Ma (Jaeger et al., 2014;Sánchez-Montes et al., 2020). While the depth error remains the same, the age error over this section of the core has been increased to ±0.12 Ma to take the Gilbert/Gauss transition age shift into account, with a propagated error across neighboring samples between ±0.23 to ±0.69 Ma. The are no big shifts in between the sedimentation rates calculated from the age model of Jaeger et al. (2014) and our stretched age model, which give confidence that our proposed age model only refines the shipboard age model (Sánchez-Montes et al., 2020). Beyond 3.88 Ma U1417's age model is poorly constrained due to poor core recovery (see Jaeger et al., 2014;Sánchez-Montes et al., 2020) and, therefore, it determines the lower age range of this study to 4 Ma.
Uncertainty of the age model and sedimentation rates presented here follow the polarity chronozone interpretations of the shipboard age model (Gulick et al., 2015;Jaeger et al., 2014) and accounts for shifts in the Gilbert-Gauss magnetic reversal in the adjusted age-depth model (Sánchez-Montes et al., 2020, Table S1 in Supporting Information S1). The sedimentation rates are calculated across our analyzed samples, almost exclusively in Hole U1417D. Sedimentation rates increase during the mid and late Pliocene (from 40 to 85 m/Myr across 3.5-2.8 Ma), and during the iNHG increase in glaciogenic inputs (from 85 to 156 m/Myr at 2.4 Ma). Our manually calculated sedimentation rates on the CCSF-A scale across neighboring samples follow the depth-age model uncertainties of Jaeger et al. (2014) but differ slightly from the shipboard age model, where average sedi mentation rates are calculated over 0.5 Myrs following the CCSF-B scale, and uncertainties (+1σ) are calculated using a Monte-Carlo sedimentation model over the whole Site U1417 (Jaeger et al., 2014). After distributing the sediment evenly across the core (Sánchez-Montes et al., 2020), our calculated sedimentation rate uncertainties become smaller when compared to the statistical approach of the shipboard age model (Jaeger et al., 2014). The sediments recovered include diatom ooze interbedded with debris flow deposits containing mud clasts and plant fragments (lithostratigraphic unit VA, 4-3.2 Ma), marine mud (unit IV and II, 3.2-2.8 Ma and 2.4-1.66 Ma) and ice-rafted diamict interbedded with mud (unit III, 2.8-2.4 Ma; Jaeger et al., 2014). Samples for organic matter analyses were selected avoiding sand, silt, ash or gravel (e.g., Ausín et al., 2019Ausín et al., , 2021.

Carbon and Nitrogen Bulk and Isotope Analyses
Approximately 70 mg of freeze-dried and homogenized sediment was weighed into silver capsules and acidified in-situ with 5%-6% sulfurous acid (H 2 SO 3 ) to remove carbonate phases (Verardo et al., 1990). Additional aliquots of acid were repeated until no reaction was observed using a binocular microscope, which ensured removal of inorganic carbon. No method was implemented for inorganic nitrogen removal. Samples were oven-dried between acidifications at 40°C. 6 mg of Tungsten VI oxide (WO 3 ) was added to each sample to facilitate combustion. Samples were then measured using a Varian elemental analyzer coupled to a Europa Scientific continuous-flow isotope-ratio mass spectrometer. The average standard deviation of replicate samples is 0.58 ‰ for δ 15 N, 0.006% for total nitrogen (TN), 0.26 ‰ for δ 13 C and 0.031% for total organic carbon (TOC) (n = 8 pairs). The TOC and TN were normalized to the accumulation rates of the sediment analyzed (Equations 1-3), where "material" refers to TOC or TN: SÁNCHEZ MONTES ET AL.

Siliceous Microfossil Assemblages and Biogenic Silica
Microplaeontological counting standard methods and techniques (Schrader & Gersonde, 1978) were followed to prepare sediments after freeze-drying. Siliceous microfossil species identification and counts were performed on pre-acid cleaned permanent slides (Mountex® mounting medium). Several traverses across each slide were examined on a Zeiss®Axioscop with interference illumination at x1000 magnifications (MaruM, University of Bremen). Depending on valve abundances, between ca. 400 and ca. 700 valves per slide were counted. Duplicate slide counting quantified the concentration estimate analytical error ≤10.0%. Counting outputs were converted in sedimentary abundance of individual diatom taxa, total diatom (in valves per g −2 ) and total silicoflagellate (in skeletons per g −2 , Equation 4) which were then converted to mass accumulation rates (MAR, Equation 3). The relative abundance (%) of each species was calculated as the fraction of the diatom species versus the TC in a particular sample (Equation 4).  (Sancetta & Calvert, 1988).
Diatom taxa were grouped in 6 palaeo-habitats: benthic, coastal high productivity, coastal moderate productivity, pelagic high productivity, pelagic warm water and freshwater (Table S2 in Supporting Information S1). The coastal high and moderate productivity palaeo-habitats describe coastal diatoms that occur at intervals of high and moderate productivity due to high and moderate nutrient availability in surface coastal waters, respectively. The pelagic high productivity palaeo-habitat is composed by pelagic diatoms that occur at intervals of high productivity and high nutrient availability in surface pelagic waters. The pelagic warm waters is a palaeo-habitat that contains subtropical pelagic diatoms species that thrive in warm waters (Ren et al., 2014), representing the possible northward transport of warm to temperate waters into the GOA. The Shannon Weaver Index (SWI) was calculated to quantify diatom diversity (Shannon & Weaver, 1949).
Diatom preservation is reconstructed to assess whether a water column silica is rich or poor, with longer or lower exposure of silica to degradation due to slower or more rapid sediment burial. Following observations with light microscopy, three main states of valve preservation were defined as: (a) good or no significant enlargement of the areole or dissolution of the valve margin; (b) moderate where valves show areole enlargement, dissolution of the valve margin and valve fragmentation; and (c) poor or strong dissolution of the valve margin and areole enlargement (Crosta et al., 2012;Romero et al., 2005Romero et al., , 2009Romero et al., , 2012Romero et al., , 2015. In addition, two intermediate states of dissolution characterize valves whose state of preservation does not fully fit the three above-mentioned categories: good/moderate (good preservation predominates over moderate preservation) and moderate/good (moderate preservation predominates over good preservation).
Biogenic silica was determined with a sequential leaching technique with 1M NaOH at 85°C (Müller & Schneider, 1993) and normalized to the weight of the sample (wt%) and accumulation rates of the sediment (Equation 3). The precision of the overall method based on replicate analyses varies between ±0.2 and ±0.4%, depending on the material analyzed.

Results
During the Pliocene-early Pleistocene, biogenic silica export to Site U1417 exceeds TOC by 15 times (Figures 2a  and 2b). TOC is overall 10 times more abundant than TN ( Figure 2b). As part of the TOC, the sum of long-chain (C 27 , C 28 , C 31 ), or terrigenous, n-alkanes is overall 3 times higher than that of short chain (C 15 , C 17 , C 21 ), or aquatic, n-alkanes ( Figure 2d). Aquatic n-alkane MAR are 2 times higher than alkenone MAR (from haptophyte algae), which, in turn, are 3 times more abundant than brassicasterol (from diatoms and haptophyte) (   Table S1 in Supporting Information S1).
Diatoms and silicoflagellate MAR are highest during the early part of the record (4.0-3.8 Ma, Figure 3a), with coastal high productivity diatoms as the most abundant group at Site U1417 followed by benthic, coastal moderate productivity, pelagic warm and freshwater diatoms ( Figure 3, Table S2 in Supporting Information S1). Across 3.7 to 2.8 Ma, diatoms and silicoflagellate MAR decrease progressively. Silicoflagellates disappear by 3.35 Ma, whereas diatoms almost disappear by 2.8 Ma (Figure 3a). The diatom assemblages recorded at Site U1417 switch from coastal high productivity to dominantly pelagic high productivity groups across 3.7 to 2.8 Ma (Figure 3, Table S2 in Supporting Information S1), particularly between 3.3 and 3.2 Ma. Coscinodiscus marginatus, the main species contributing to the pelagic high productivity assemblage at Site U1417, is a minor component of the diatom communities in the North Pacific across the Cenozoic, but it has been found abundant in the subarctic Pacific and Bering Sea during the Late Pliocene (Shimada et al., 2009).
The preservation of diatoms follows the same stepwise decline as the total diatom abundances, where preservation is high during 4-3.7 Ma, decreases gradually across 3.7-3.2 Ma and increases slightly between 3.19 and 2.8 Ma before diatoms almost completely disappear from the record between 2.8 and 1.7 Ma (Figure 3b). Biogenic silica MAR decreases in response to the decline in diatom MAR at 2.8 Ma (from an average of 0.39 ± 0.05 7-0.31 ± 0.016 g cm −2 kyr −1 ; Figure 3a). The decrease in siliceous microfossils occurs during an increase of organic matter concentrations (TOC and TN) across the PPT ( Figure 4). The early Pleistocene increase in organic matter content ( Figure 4d) is also evident in the increasing biomarker concentrations (Figures 4a-4c). Signs of different production or selective organic matter degradation are suggested by different patterns in the biomarkers that is, peaks in alkenone often coincide with the absence of sterols (Figures 4a and 4b).

Discussion
Across 4.0-1.7 Ma, the marine productivity export to Site U1417 is characterized by changes in the relative contribution of siliceous microfossils and organic matter ( Figure 4). These changes culminate with an increase in organic matter export and almost complete disappearance of siliceous remains across the PPT and early Pleistocene ( Figure 4). A range of factors could explain the siliceous/organic matter export changes, including (a) changes in nutrient supply and phytoplankton production, (b) water column stratification and carbon/silica preservation, and/or (c) the impact of sedimentation rate on phytoplankton burial efficiency and preservation. Here we examine chronologically the forcings and responses behind the siliceous and organic matter productivity export changes. According to these changes, the record can be divided into three intervals: the early Pliocene (4.0-3.8 Ma), the late Pliocene (3.7-2.82 Ma), and the PPT and early Pleistocene (2.8-1.66 Ma).

The Early Pliocene (4.0-3.8 Ma): High Biogenic Silica Export of Coastal High Productivity Habitats
The maximum MAR of diatoms (up to 100 × 10 6 valves cm −2 kyr −1 , Figure 3a) and silicoflagellates (up to 0.3 × 10 6 valves cm −2 kyr −1 , Figure 3a) suggests increased marine productivity at Site U1417 during the early Pliocene. The high SWI indicates a highly diverse diatom community ( Figure 3b) at Site U1417, where diatoms originate from a range of habitats: coastal high productivity, benthic, pelagic high productivity, coastal moderate, pelagic warm and freshwater (from most to least abundant, Figure 3, Table S2 in Supporting Information S1). The biogenic silica concentrations are also high, which accounts for the productivity, preservation and valve sizes of a number of siliceous microorganisms such as diatoms, silicoflagellates and radiolarians, the last two being less abundant at Site U1417 than diatoms.
The most prolific diatoms are Chaetoceros resting spores (up to 68 × 10 6 valves cm −2 kyr −1 , Figure 3c, Table  S2 in Supporting Information S1), which are abundant in highly productive coastal waters in the GOA (e.g., Ren et al., 2014). Chaetoceros dominate the diatom community of the GOA when iron is added to the North Pacific (Tsuda et al., 2003) and they develop resting spores after nutrient consumption (Margalef, 1978). Therefore, the prolific coastal diatom productivity suggests iron supply from the nearby continent and shelves to the neritic zone of the GOA (e.g., Lam & Bishop, 2008;Ren et al., 2014). The marine productivity species at Site U1417 suggest a productivity gradient during this time, less steep towards the coast compared to present (Figure 1), where the maximum productivity export of diatoms occurs.
The low contribution of pelagic diatoms and marine biomarkers (brassicasterol, alkenones and aquatic n-alkane MAR) suggests a limited open-ocean productivity in the GOA (Figures 4a-4c). This causes a contrast between a highly productive coastal zone which supported diatom productivity in coastal habitats to the detriment of pelagic diatom, coccolithophores and other marine phytoplankton productivity (Herbert, 2001). The regime of abundant micro-nutrients and low macro-nutrients known as LNHC characterizes the modern coastal Alaska (Weingartner, 2007). The source of micro-nutrients during the Pliocene could be from two terrestrial sources: glacial flour and/or aeolian dust (Müller et al., 2018;Romero et al., 2022). Iron-rich glacial flour is abundant in modern coastal Alaska estuaries for example, the Copper River (Crusius et al., 2011) and the continental margin (Lam & Bishop, 2008). Wind transport is an important mechanism today for transporting dust from Alaska hundreds of kilometres beyond the shelf break to the GOA (Crusius et al., 2011. However, high pollen . Detailed Pliocene (4.0-2.8 Ma) marine productivity export at Site U1417. (a) Silicoflagellate (million skeletons cm −2 kyr −1 , green) and total diatom (million valves cm −2 kyr −1 , gray) mass accumulation rates (MAR); (b) the Shannon-Weaver index of diversity of species (turquoise) and the preservation value (brown); (c) coastal high productivity diatom species relative abundance (%; orange) and MAR (million valves cm −2 kyr −1 , gray); (d) pelagic high productivity diatom species relative abundance (%; green) and MAR (million valves cm −2 kyr −1 , gray); (e) benthic diatom species relative abundance (%; in black) and MAR (million valves cm −2 kyr −1 , gray); (f) coastal moderate productivity diatom species relative abundance (%; in blue) and MAR (million valves cm −2 kyr −1 , gray); (g) pelagic warm diatom species relative abundance (%; in red) and MAR (million valves cm −2 kyr −1 , gray) and (h) freshwater diatoms relative abundance (%, blue) and MAR (million valves cm −2 kyr −1 , gray). Vertical lines correspond to glacial stages (gray) and key events on the Cordilleran Ice Sheet (CIS) runoff history (blue, see text). Blue vertical shadings indicate higher (deeper blue) and high (light blue) biogenic silica preservation, the non-shaded interval indicates poor silica preservation. Shadings indicate uncertainties in MAR associated with the age and depth uncertainties of the tie-points in Jaeger et al. (2014) and age-depth adjustments in Sánchez-Montes et al. (2020) (see Table S1 in Supporting Information S1).
abundance at Site U1417 during the early Pliocene suggests higher riverine terrestrial input from coastal Alaska to the GOA (Pisias et al., 2001;Sanchez-Montes et al., 2020). The presence of freshwater diatoms contributing around 1% to the total diatom concentration in the GOA is most likely attributed to Alaskan aeolian dust transport ( Figure 4b) (e.g., Crusius et al., 2011;Crusius et al., 2017). Coastal diatoms have been previously interpreted as downslope transport inputs to Site U1417 from the coastal area (Jaeger et al., 2014). Another explanation is the ocean fertilisation by dust nutrient delivery and increase in coastal diatom productivity.
The 30% contribution of benthic diatoms to the total diatom MAR at Site U1417, despite its location at 4,218 m water depth (Jaeger et al., 2014), points to the transport of shallow water habitat assemblages toward the Surveyor Fan (McGee et al., 2008). The early Pliocene lithology at Site U1417 contains gravity flow deposits, while the tectonic history in the GOA and coastal Alaska during the Pliocene (Enkelmann et al., 2015) may result in slope instability and distal transport of shallow species to Site U1417, which lies ∼700 km from the coastline (comparable to distances observed in turbidity current transports offshore West Africa, Talling et al., 2007). Relatively high terrestrial n-alkane MAR (Figure 4c) suggest enhanced flux of terrestrial OM input and rapid burial during gravity flows, which may also account for the peaks in TOC (e.g., Hage et al., 2020). Plant fragments were also observed at Site U1417, and attributed to both mature (Rea, Basov, Krissek, & the Leg 145 Scientific Party, 1995;Rea, Basov, Scholl, & Allan, 1995;Gulick et al., 2015) and fresh (Jaeger et al., 2014)  The early Pliocene contains the highest concentration of benthic and freshwater diatoms (average 15%) known to synthesize brassicasterol (e.g., Piepho et al., 2012;Rampen et al., 2009) found at U1417 (Table S2 in Supporting Information S1), yet brassicasterol concentrations are low during the early Pliocene. Low brassicasterol concentrations could reflect organic matter degradation or less favorable (nutrient) conditions for brassicasterol producers (Goad & Withers, 1982;Kanazawa et al., 1971;Lei et al., 2012;Volkman, 1986Volkman, , 2006. The high diatom productivity (high diatom MAR) may have been supported by an abundant ocean carbon pool with widely available 12 C, as shown by low bulk δ 13 C org at this time (−24.7‰, Figure 2c). A large carbon pool could suggest a very well mixed and oxygenated ocean. An oxygenated water column would also favor silica preservation rather than organic matter preservation, as observed in Site U1417 (Figures 4a and 4b, Figures 5b  and 5c). The oxygenated water column, perhaps aided by oxygen transport via turbidity currents, downwelling movement of water masses and ocean mixing, could explain the low concentration of marine biomarkers in the GOA during the early Pliocene ( Figure 6). The low sedimentation rates during the early Pliocene might also have favored organic matter degradation at the sediment-water interface due to longer exposure to oxidising conditions outside of the gravity flow events (Figure 4a).

The Late Pliocene (3.7-2.82 Ma): Decrease in Terrestrial and Coastal Biogenic Silica and Increase in Marine Productivity
The decrease in diatom and silicoflagellate MAR (Figure 3a) during the late Pliocene seems to be caused by a decrease in flux and silica preservation (e.g., Thunell et al., 1994; preservation value, Figure 3b) and is accompanied by an increase in the number of pelagic diatom species (Figure 3d). The main contributor to the total diatom MAR thus shifts from coastal high productivity species during the early Pliocene (average of 54%) to pelagic high productivity species during most of the late Pliocene (average of 56%; Figures 3c and 3d). These ecological changes are accompanied by a higher diatom diversity during the late Pliocene than during the early Pliocene (Figure 3b), which could partially explain a small increase in biogenic silica preservation from the previous period ( Figure 2a).
The disappearance of silicoflagellates from the record at 3.35 Ma ( Figure 3a) occurs with an increase in organic matter inputs to Site U1417. The progressive increase in productivity indicators from coccolithophores (e.g., alkenones; Figure 4b) and other phytoplankton (aquatic n-alkanes, Figure 4c) alongside increasing TOC and TN (Figure 4d) suggests better organic matter preservation and/or increasing export production from marine sources. Terrigenous n-alkanes decrease during most of the late Pliocene ( Figure 4c) suggesting a reduction in land vegetation and/or a reduced riverine/aerial terrestrial plant transport to Site U1417, also suggested by an overall decrease in freshwater diatoms (Figure 4b). Progressively increasing δ 13 C but decreasing δ 15 N across the late Pliocene suggests a limited carbon but abundant bioavailable nitrogen pool for phytoplankton consumption (Figure 2c). A reduced ocean mixing/more stratified water column, caused by slow ocean circulation or reduced gravity flow movement, would limit the atmosphere-ocean 12 C exchange and increase bacterial nitrogen fixation (Galbraith et al., 2004).
The continuous presence of benthic diatoms suggests that downslope transport towards Site U1417 still occurred during the late Pliocene but was reduced. The decrease in river transport could reflect an increasingly glaciated landscape, as suggested in Sánchez-Montes et al. (2020), which is supported by higher sedimentation rates (Figure 5d; Gulick et al., 2015). Reduced downwelling is suggested by decreased silica preservation and increased organic matter preservation. Further, the LNHC region in the GOA reduced in extent due to a higher macronutrient availability and allowed the expansion of marine pelagic communities across a range of producers (including diatoms and coccolithophores, Figures 3d and 4b) and increased the overall ocean diversity (Figure 3b). In addition, the less mixed water column would have favored bacterial nitrogen fixation (Galbraith et al., 2004), phytoplankton nitrogen consumption and phytoplankton nitrogen burial ( Figure 4d).
As the glaciation progressed during the late Pliocene, there is evidence of tectonic uplift and glaciations causing changes in the Alaskan landscape such as a shift in the Yukon River flow, from originally southward direction to the GOA to flow westward to the Bering Sea (Duk-Rodkin et al., 2004), which would have reduced riverine terrestrial and freshwater input to the GOA. Similarly, reversal of the Bering Strait throughflow from southward to northward at 3.6 Ma (Horikawa et al., 2015) away from the North Pacific to the Arctic Ocean, may also have impacted the ocean circulation in the GOA. The resulting enhanced organic matter MAR from reduced river runoff and decrease in strength of ocean currents in the GOA could reflect improved preservation resulting from a more stratified water column.
The Mid Piacenzian Warm Period (MPWP) (3.33-3.19 Ma) is noted separately here as it has several specific characteristics which are different from the rest of the Pliocene at Site U1417. Although there is a sharp reduction in diatom MAR (3.25-3.19 Ma; Figure 4a) elevated marine export productivity is indicated by high abundances of pelagic diatom species (with an average of 88%, Figure 3d) accompanied by peaks in brassicasterol (Figure 4a), alkenone ( Figure 4b) and aquatic n-alkane (Figure 4c) MAR, alongside peaks in Ba/Al (330-350 ppm % −1 ) and CaCO 3 (<2%) at Site U1417 (Jaeger et al., 2014;Zindorf et al., 2019). The higher susceptibility of brassicasterol than alkenones to degradation (Gaskell & Eglinton, 1974.;Wakeham et al., 2002) suggest a combined favouring of water chemistry and nutrient regime for brassicasterol producers (diatoms) to the detriment of alkenone producers (haptophyte). The reduced diatom diversity and shift to increasingly open ocean productivity conditions, dominated by pelagic high productivity environments, is consistent with the reduction of diatom biodiversity in other marine habitats (Figure 5b) (Nakov et al., 2019).
The driver(s) for the enhanced MPWP marine production at Site U1417 could reflect enhanced nutrient supply to the center of the Alaskan gyre as productivity gradients relax during warmer intervals ( Figure 1) and/or improved preservation of the productivity proxies. There is a sharp reduction in benthic and coastal high productivity diatoms during the MPWP since the marine isotope stage M2, suggesting that gravity flows decreased, which may have been aided by a decrease in riverine inputs due to the expansion of mountain glaciation in Alaska (Horikawa et al., 2015;Sánchez-Montes et al., 2020). Peaks in terrestrial n-alkane MAR show that erosion and transport of terrestrial material to the site was still occurring, potentially by wind and/or glaciers if riverine inputs were reduced (e.g., Müller et al., 2018;Sánchez-Montes et al., 2020). These terrestrial inputs would likely still have provided a source of terrestrial iron. The general increase in ocean productivity (Figure 5b) during colder periods of the MPWP suggests increased atmosphere and ocean circulation and increased deep nutrient availability to the photic zone of the GOA under a HNLC region. Stratification during warmer periods of the MPWP would have been achieved by ice melting and increased ocean heat absorbance (e.g., Behera et al., 2021) while the MPWP poleward shift of the westerly winds might have been able to occasionally break the stratification in the GOA (Abell et al., 2021).
The overall sluggish circulation during the late Pliocene may suggest an ocean circulation reorganization in the GOA after the northward shift of the Yukon River outlet and Bering Strait throughflow. An interesting increase in pelagic warm water productivity (up to 45% of diatoms) might shed some light on the origin of the first C 37:4 peak above 5% in the GOA at 3.0 Ma, interpreted as the first evidence of glacier tidewater freshwater runoff in the GOA since 4.0 Ma (Sánchez-Montes et al., 2020). This could indicate that ice (i.e., tidewater glaciers) might have developed in the coastal GOA during the MPWP, the terminus of which might then have been melted by the northward advection of warm Pacific waters, such as the NPC (Figure 1). The injection of warm waters from lower latitude North Pacific to the GOA (shown by pelagic warm diatoms) and an increase in coastal moderate productivity diatoms seem to anticipate blooms in the gyre diatom communities (peaks in pelagic high productivity diatoms) characterising the transitional diatom communities during the late Pliocene. The new oceanography and reduction of freshwater input to the GOA allowed the Alaskan gyre to bring warm NPC waters to Site U1417. It might be that the new ocean configuration in the GOA with higher influence from lower latitudes during the late Pliocene allowed the rapid expansion of the CIS (Sánchez-Montes et al., 2020).

The Late Pliocene -Early Pleistocene (2.8-1.66 Ma): Increase in Organic Matter Export
The change from higher siliceous to higher organic matter productivity export, first observed in our record during the MPWP, becomes a permanent characteristic during the PPT and early Pleistocene (2.8-1.66 Ma; Figures 4a and 4b). The dominant pelagic contribution from diatoms to the productivity export since 2.8 Ma suggests that the alkenone MAR indicates total productivity export (Raja & Rosell-Melé, 2021) where the productivity was higher in central GOA and the productivity gradient shifted toward the coast (Figure 1). From 2.8 Ma, diatoms almost completely disappear from the record and TOC, TN and biomarkers increase more rapidly (Figure 4). Concentrations in terrestrial n-alkanes also increase but also become more variable, suggesting a mixture of transport mechanisms at play, for example, wind and glacial runoff (Figure 4d). The synchronous response of increasing terrigenous and aquatic n-alkanes seems to reflect ocean fertilisation through wind and/or glacial runoff-derived nutrients (Figures 4c and Figure 5c and 5d).
We propose that the increase in pelagic productivity from 2.8 Ma occurred under a strong Alaska gyre fueled by an Aleutian Low (AL) centered in the GOA (Sancetta & Silvestri, 1986) and highly functioning ocean-ice-climate linkages (Sánchez-Montes et al., 2020) which favored ocean fertilisation ( Figure 6). Frequent glacial meltwater influence in the GOA is supported by peaks in C 37:4 above 5% after 2.8 Ma (Sánchez-Montes et al., 2020, Figure 5c), where sea-ice (Wang et al., 2021) at Site U1417's climatic setting likely played a minor role. A stratified and warmer water column (Figures 5a and 5c) could suggest a decrease in oxygen availability in the surface ocean and a shallower remineralization depth that favored the carbon pump . Warm intervals of the GOA have previously been linked to a decrease in dissolved oxygen (Barron et al., 2009;Galbraith et al., 2004;Zindorf et al., 2020). Since the late Pliocene until 2.4 Ma, the δ 13 C and δ 15 N display a variable but an overall decreasing trend, suggesting a progressively larger pool of carbon and bioavailable nitrogen feeding the marine phytoplankton. The increase in organic nitrogen during 2.8-2.4 Ma suggests a slow ocean circulation and an overall stratified water column (Galbraith et al., 2004), storing nutrients in the deeper ocean, where nitrogen could become bioavailable by increased nitrogen fixation and would become available to the surface with episodes of ocean mixing. During the early Pleistocene (2.4-1.66 Ma) the δ 13 C remains low, however the δ 15 N increases suggesting an increase in inorganic nitrogen supply (Figure 3c).
The higher alkenone and brassicasterol (haptophyte and diatoms) export suggests that the GOA iron abundance decreased during the late Pliocene-early Pleistocene similarly to during the MPWP and at modern times (Hinckley et al., 2009;Martin et al., 1991;Martin & Fitzwater, 1988). The marine productivity, controlled by the availability of nutrients accumulated in the deep GOA, is made available when increased wind speeds increased ocean mixing, interrupting an otherwise highly stratified water column configuration (peaks in alkenones and brassicasterol, Figures 4a and 4b; Abell et al., 2021;Sánchez-Montes et al., 2020). The highest TOC and TN of the record are found between 2.4 and 1.66 Ma suggesting the highest carbon and nitrogen deep ocean storage. However, TOC and TN show a slowly decreasing trend which, together with similar δ 13 C and an increase in δ 15 N from the previous interval (2.8-2.4 Ma), suggests a shift towards reducing organic carbon export to the deep sea, perhaps reflecting enhanced respiration in the water column and return of carbon to the atmosphere under an intensified ocean circulation (Galbraith et al., 2004). A better ventilated ocean is consistent with the frequent peaks in terrestrial n-alkanes attributed previously to strong winds, which could have promoted oxygenated conditions in the GOA, consistent with our observed decrease in brassicasterol preservation and increased diatom concentrations (although still very low, 0.04-0.4 Million valves cm −2 kyr −1 ) (Figure 4). The increase of diatoms (2.0-1.66 Ma) supports the aeolian supply of terrestrial nutrients, possibly aided by glacial meltwater input. Upwelling centered in the GOA and ocean fertilization is a characteristic of the northeast Pacific at present under the influence of the AL atmospheric circulation and HNLC configuration.
During 2.4-2.0 Ma, the increase in alkenone MAR (Figures 4a and 4b) and the fragmentary brassicasterol record suggest a less stratified ocean column and increased upwelling conditions in the HNLC region. Intensified winds during 2.4-1.66 Ma (deduced from high terrestrial n-alkane concentrations) under the influence of the AL and increased deep ocean mixing due to increased gyre circulation might have been responsible for the increases in terrestrial input and deep nutrient availability in the photic zone, respectively, corresponding to a HNLC region. The increase in nutrient availability resulted in a higher marine productivity export and high carbon and nitrogen deep storage (high TN, TOC; Figure 5e) which suggests a still strongly reductive ocean interior.

Implications for Ocean Circulation and CO 2 Storage in the GOA During the CIS Development
We have outlined several scenarios of changing Pliocene and early Pleistocene productivity in the GOA, as well as changes in likely nutrient sources. The biogeochemical changes in the water column described at Site U1417 follow an eccentricity-driven ∼400 kyr cyclicity (Figure 3). Although the patterns are complex, we note that low eccentricity periods can be linked to surface ocean cooling and changes in the CIS extent ( Figure 5). We suggest here that this, in turn, resulted in impacts on nutrients and productivity offshore linked to changes in the AC and the AL (Barron, 1998;Sancetta & Silvestri, 1986). We propose that eccentricity minima and cooler periods on land triggered an increased glaciation and changed the regional oceanography of the GOA via internal feedback mechanisms. For example, the disappearance of silicoflagellates (3.48 Ma), the reduced diatom productivity export (M2 and KM2) and the disappearance of diatoms (2.8 Ma) all mark the Pliocene ∼400 kyr eccentricity cycles of silica dissolution under increased ocean reductive conditions linked to a shift toward increased gyre circulation.
The 100 kyr cycles of lower amplitude eccentricity changes can be linked to a slightly more active gyre circulation and are also reflected in relatively cool periods such as the Gi4 at 3.65 Ma and Gi2 at 3.60 Ma. Stage Gi4 impacted productivity export by decreasing the terrestrial freshwater diatom and increasing marine OM inputs to Site U1417. Peaks in alkenone MAR at Site U1417 mark the 400 kyr and some 100 kyr low eccentricity events and could be attributable to a higher biological pump during cooler climate (Boscolo-Galazzo et al., 2021;Crichton et al., 2021, Figure 4b). These low eccentricity conditions intensified overall during the CIS expansion at 2.8 Ma (Figure 5), where the Alaskan gyre became more active driving pelagic productivity at the center of the GOA. Decrease in diatom and increases in pelagic high productivity at the center of the Alaskan gyre suggests a steep productivity gradient shifted toward the coastal region (Figures 4 and 5) (Barron, 1998). Productivity peaks in the subarctic Pacific have been associated with lower dissolved oxygen under warmer climate (Barron et al., 2009;Knudson et al., 2021), which suggests a more reductive ocean during the PPT and an increase in carbon burial efficiency (Lopes et al., 2015). Lower productivity during warmer sea surface temperatures (SSTs), with a stratified ocean and increased atmospheric oxygen, has also been suggested due to a higher carbon remineralization in the water column and lower C pump (Boscolo-Galazzo et al., 2021;Crichton et al., 2021;Fakhraee et al., 2020;Komar & Zeebe, 2021). However, in the northeast Pacific, ecosystem and sea-floor controls have been identified as playing a bigger role than SST to increase carbon export despite lower productivity (Lopes et al., 2015). In addition, under a warmer climate and more stratified ocean, productivity export has been suggested to increase due to an increase in carbon remineralization in the upper ocean resulting in increase in productivity, whereas the carbon pump remains largely similar before and after warming . SSTs increase ∼1°C at Site U1417 during the early Pleistocene in comparison with the late Pliocene (Sánchez-Montes et al., 2020). An SST increase of 0.6°C is estimated to reduce the particulate organic carbon at a 1 km water column depth by 5% . Carbon remineralization at 4 km deep (Site U1417) is therefore likely to have contributed minimally to decrease C export across the Plio-Pleistocene, where other sites in the northeast Pacific register lower SSTs (Sánchez-Montes et al., 2020). As a result, we do not think that changes in organic matter remineralization in response to changing ocean temperatures can account for the shift in MAR we identify here, leaving a change in export production as the most likely driver of the changes we observe. During the CIS retreat at 1.9 Ma, offshore productivity is high as well as the transport of coastal productivity diatoms returns to Site U1417 suggesting an increase in iron delivered to the GOA (Costa et al., 2016) and a return to slightly more oxygenated conditions. Site U1417 sits in the Surveyor Fan while Ocean Drilling Program (ODP) 887 sits in the Aleutian Abyssal Plain (Rea & Snoeckx, 1995). While we note an apparent higher terrigenous input at ODP 887 than U1417 across the Pliocene and early Pleistocene (Figure 7), these records need to be considered to reflect different environmental proxies (coarse and fine mineral clasts vs. long-chain n-alkanes, respectively) and settings. Sedimentation rates are higher at Site U1417 than ODP 887 across the Pliocene and early Pleistocene (Jaeger et al., 2014;Rea & Snoeckx, 1995) suggesting higher terrigenous inputs to Site U1417 than ODP 887, which is explained by proximity to the CIS (ODP 887 is located 200 km southwest of U1417, further away from the CIS; Figure 7). Site U1417 recorded similar biogenic silica MAR to ODP 887 (3,634-m water depth, Figure 7). Compared to other sites of the North and equatorial Pacific, Site U1417 and ODP 887 contain the lowest biogenic silica and alkenone MAR during the Pliocene-early Pleistocene (Figure 7). This could suggest a similar biogenic silica preservation across the east subarctic Pacific due to more oxygenated/reduced water column (Galbraith et al., 2004), where ODP 887, closer to the centre of the AL suffered larger variations in silica preservation than Site U1417 which is located under the AC (Figure 7). Comparing the subpolar gyre, which expands across Site U1417, ODP 887 and ODP 882, there is an order of magnitude higher biogenic silica MAR at ODP 882, under the influence of the Kamchatka Current in the west subarctic Pacific (3,244 m water depth) than in the GOA (Figure 7). The highest alkenone MAR are recorded at Site 1012, under the influence of the California Current (1,772 m water depth, Figure 7), followed by ODP 846 under the Peru Current and close to the Equatorial Undercurrent (3,307 water depth, Figure 7). According to these patterns, the sites located in shallower water depths exhibited the highest productivity MAR, which suggests better preservation but could also suggest that shallower water columns were more easily mixed or had greater terrestrial nutrient supply to trigger productivity blooms.
However, as noted for Site U1417, the PPT development of the CIS and the NHG more generally affected the preservation of siliceous and organic matter remains in the whole North Pacific, where biogenic silica decreases and organic matter increases (Figure 7). In addition, terrestrial inputs during the PPT increase, suggesting ocean fertilisation and an increase in marine productivity in the northeast and possibly the equatorial Pacific (Figure 7). In particular across 2.4-2.0 Ma, ocean fertilisation and productivity at the North Pacific Site U1417 and ODP 1012 are maximum during the increase in gyre circulation (e.g., Barron et al., 2002) and increase in AL and North Pacific High systems, which suggest an increase in the westerly wind strength (Abell et al., 2021). The east equatorial Pacific, however, shows a decrease in coccolithophore productivity probably explained by weaker trade winds at ODP 846 and disruptions in the Pacific Cold Tongue (Liu et al., 2019). From all regions represented in Figure 7, ODP 846 is the only region that cooled across the PPT (Sánchez-Montes et al., 2020), supporting weaker trade winds and a rapid switch from La Niña to El Niño-like conditions. The same northeast Pacific productivity and terrigenous input pattern observed for 2.4-2.0 Ma also appears during the MPWP, when a weaker east equatorial Pacific upwelling system suggests an expansion of the equatorial warm pool (Liu et al., 2019). The expansion of the equatorial warm pool also suggests a nutrient leakage of deep ocean nutrients (macronutrients) to northern latitudes and a drier Asian continent to increase in dust (micronutrients) from the Loess Plateau to the North Pacific (Abell et al., 2021), driving the highest sustained coccolithophore productivity between 2.4 and 2.0 Ma and the shift to open ocean productivity during the MPWP and HNLC conditions. The nutrient leakage from the equatorial to the North Pacific seems to have also played a role in the biogeochemistry changes observed at Site U1417 discussed above. Furthermore, these mechanisms aid a detachment from the aeolian and the riverine/glacial terrigenous n-alkane signal. The highest peaks in terrigenous n-alkanes at Site U1417 seem to be driven by an increase of terrigenous inputs from the Loess Plateau under a stronger atmospheric and ocean circulation (Abell et al., 2021), whereas the background (smaller) peaks seem to correlate with increase in riverine/glacial terrigenous inputs from the GOA (Figure 7d). Source studies of terrestrial input to Site U1417 during the PPT and early Pleistocene suggest similar to modern coastal detrital provenance of Alaskan coast and Asia (Rea, Basov, Krissek, & the Leg 145 Scientific Party, 1995;Rea, Basov, Scholl, & Allan, 1995;Horikawa et al., 2015) of the inorganic nitrogen and iron input to Site U1417 ( Figure S1 in Supporting Information S1). The erosion of lithologies at lower altitudes as the glaciation progresses ( Figure S1 in Supporting Information S1, Perry et al., 2009;Chapman et al., 2012) is consistent with more recent source interpretations in the GOA (Huber & Bahlburg, 2021).
In addition, the change in heat supply from the equator to the North Pacific due to oceanographic changes might have resulted in the CIS glaciation attempt during the M2 (De Schepper et al., 2013) and CIS build up across the 2.4-2.0 Ma (Site U1417 highest sedimentation rates, Figure 5d). The Kuroshio Extension characteristic microfossil tropical species (Lam & Leckie, 2020) and its increase at Site U1417 during stronger North Pacific gyre circulation (Gallagher et al., 2015) suggest the Kuroshio Extension influence in the GOA since the late Pliocene (Gallagher et al., 2015). The expanded storm track which characterizes modern El Niño (Joh et al., 2021), in addition to a the permanent negative Pacific Decadal Oscillation-like climate in the North Pacific across the Pliocene-early Pleistocene (Sánchez-Montes et al., 2020) and mountain building (Enkelmann et al., 2015) would increase ice accumulation on land under lower atmospheric CO 2 concentrations (Figure 5a).
The eccentricity and biogeochemical changes observed in the GOA are observable in the North Pacific, at least at the 400 kyr cycles with some other cycles that are marked by eccentricity minima at 3.8, 3.3, 3.2, 2.8, 2.4 and 2.0 Ma, where the subpolar gyre has been suggested to be the driver of changes in the subtropical Pacific gyre (Sancetta & Silvestri, 1986). Despite higher biogenic silica concentrations than Site U1417 before the PPT, ODP 882 and ODP 887 biogenic silica MAR decrease to comparable concentrations than Site U1417 after the PPT resulting in an average homogeneous biogenic silica MAR of 0.4 g cm −2 kyr −1 across east and west subarctic Pacific (Figure 7). The southward component of the Bering Sea (Horikawa et al., 2015) and the Yukon River flow during the early Pliocene (Duk-Rodkin et al., 2004), together with the productivity characteristics of the subarctic Pacific suggests that the early Pliocene circulation in the North Pacific was different compared to modern, probably caused by a different topography due to ongoing tectonic changes in coastal Alaska (Enkelmann et al., 2015). However, during the PPT, the west subarctic Pacific and the Alaska gyre unified under a HNLC region during the ocean reorganization across the late Pliocene and water column stratification during the intensification of the CIS (iCIS). This suggests a closer cycling of ocean currents between east and west Pacific, with the AS traveling westward in a subpolar gyre similar to present. We further suggest that the new ocean configuration of strong Alaska Gyre and increased ACC transport from the GOA through the Bering Sea toward the Arctic (Horikawa et al., 2015) may have contributed to freshening in the Arctic Ocean and sea-ice formation (Matthiessen et al., 2009).
Unlike the rest of the Pacific sites in Figure 7, the development of the subarctic Pacific HNLC is a key region because of the increases in phytoplankton productivity and preservation associated with the development of the CIS and water column stratification and the impacts on the C budget via atmospheric CO 2 drawdown. This is especially important considering that before the iCIS and ocean stratification, the subarctic Pacific's effective respiration of organic matter would have contributed to maintaining the high Pliocene CO 2 concentrations via ocean carbon degassing ( Figure 6). The subarctic Pacific HNLC region is subject to availability of micronutrients to the photic zone  and therefore, the GOA closer to the CIS (in particular Site U1417, Figure 7. Productivity and ocean circulation in the north Pacific. Upper panel: (a) ODP 882 biogenic silica mass accumulation rates (MAR) (g cm −2 kyr −1 ) located in the subarctic west (Haug et al., 1999), (b) Site U1417 biogenic silica MAR (g cm −2 kyr −1 ), (c) ODP 887 biogenic silica MAR (g cm −2 kyr −1 ), located 200 km southwest of U1417 (Rea & Snoeckx, 1995), (d) Site U1417 terrigenous n-alkane MAR (μg cm −2 kyr −1 ), (e) ODP 887 total terrigenous MAR (g cm −2 kyr −1 ) (Rea & Snoeckx, 1995), (f) Site U1417 alkenone MAR (μg cm −2 kyr −1 ), (g) Site U1012 alkenone MAR (μg cm −2 kyr −1 ), located in the east Cortez Basin, 100 km southwest of San Diego (Liu et al., 2008) and (h) ODP 846 alkenone MAR (μg cm −2 kyr −1 ), in the equatorial east Pacific (Liu & Herbert, 2004) where references to La Niña and El Niño refer to La Niña and El Niño-like conditions. Blue vertical shadings indicate the transitional water columns across more oxygenated (4.0-3.8 Ma) to more reductive (3.3-3.2 Ma, Mid Piacenzian Warm Period (MPWP)) and from the more reductive MPWP through more oxygenated (3.2-2.82 Ma) to more reductive ( which registers sedimentation rates four times higher than ODP 887, Jaeger et al., 2014;Rea & Snoeckx, 1993) plays a key role in C fixation and burial. Considering the GOA defined as the area contained within a line across the Kodiak Island and the Dixon Entrance (USGS, 1981), the Surveyor Fan extends across two thirds of the GOA, the other one third is occupied by the Baranoff Fan and ODP887, in the Aleutian Abyssal Plain, is excluded (Rea & Snoeckx, 1993). The Surveyor and Baranoff Fans share similar climatic and tectonic history, where they transitioned from riverine toward higher glacial inputs through the Pleistocene, however the sediment provenance of the Baranoff Fan in the Coast Mountains (Walton et al., 2014) is different to the Surveyor Fan from the St. Elias Mountains (Enkelmann et al., 2015) and there is a lack of understanding of the area to date. Thus, based on the Surveyor Fan alone and assuming similar sedimentation rates and marine productivity responses to the climate changes across the fan, increases in TOC MAR at Site U1417 across the PPT results to an estimated 404 ± 23 Pg carbon export increase in the Surveyor Fan during the iCIS and early Pleistocene in comparison to the rest of the Pliocene. The increase in organic carbon burial of the order of 186 ± 20% in the Surveyor Fan occurred during a period of 13 ± 2% decrease in the global atmospheric CO 2 across the iCIS and early Pleistocene (de la Vega et al., 2020). This might suggest that despite the GOA is an small area, the increase in the sediment rates and organic matter burial across the early Pleistocene could have contributed to the global atmospheric CO 2 reduction. However, these numbers are indicative only, and more research is needed to quantify the maximum and minimum extent of carbon burial more accurately during the Plio-Pleistocene transition. Both elevated productivity and better organic matter preservation could have been important for increasing deep ocean and sediment storage of organic matter, and potentially drawing down atmospheric CO 2 and cooling the climate (e.g., Burdige, 2007) under short (orbital) time scales (e.g., during low eccentricity intervals in the MPWP; de la Vega et al., 2020, Figure 5b), and longer time scales (e.g., across the PPT). In addition, high marine diversity during the late Pliocene might have slowly contributed to atmospheric CO 2 drawdown as suggested for the present day (Palevsky et al., 2013). Globally, the atmospheric CO 2 decreased about 100 ppm during the PPT (Figure 5b), where 40% of the modern decadal atmospheric CO 2 variability has been attributed to ocean forcing (DeVries et al., 2019). Similarly as what we find in the GOA during the PPT, carbon sinking at modern is accelerating in the Pacific (Carter et al., 2019). Further work is needed to estimate the GOA's contribution to global atmospheric CO 2 decrease during the Pliocene and Pleistocene.

Conclusions
The Pliocene and early Pleistocene productivity at Site U1417 is characterized by a decrease in siliceous microfossil export from coastal habitats and an increase in pelagic organic matter productivity. We attribute this change to a biogeochemical shift from (i) an oxygenated, high micro-nutrient availability but low nitrogen bioavailability (low nutrient high chlorophyll, LNHC region), to a (ii) reductive deep GOA, with restricted macro and micro-nutrients but increased nitrogen bioavailability (high-nutrient-low-chlorophyll, HNLC region), during the Plio-Pleistocene transition (PPT). We conclude that both tectonic uplift in the St. Elias mountains since the Pliocene and short-lived low eccentricity cycles increased the CIS glaciation and altered atmospheric and ocean circulation patterns, ocean biogeochemistry, and marine productivity. The stronger eccentricity cycles (∼400 kyr) marked glacial events, with associated peaks in marine productivity export which could have drawn down CO 2 from the atmosphere during, for example, the KM2 at 3.2 Ma, at 2.8 Ma, 2.4 Ma and 2.0 Ma. In contrast, during the shorter (100 kyr) higher eccentricity period glacials, climate feedback mechanisms of increased glaciation on land, decrease riverine and terrestrial nutrient inputs to the GOA derived in enhanced HNLC conditions and increased reductive conditions in the GOA. Over longer time scales (PPT), the GOA is potentially an important region due to the variability and, crucially, the increase in ocean fertilisation taking place. Changes in bottom water conditions, in particular trends to more reductive conditions can potentially help to account for increasing glaciation in Alaska, potentially having an impact on decreasing atmospheric CO 2 concentrations and contributing to cooling the climate.

Data Availability Statement
The new datasets in this article are available at Pangaea (Sánchez-Montes et al., 2021).