The late Cenozoic evolution of the Humboldt Current System in coastal Peru: Insights from neodymium isotopes

The Humboldt Current System along the Paciﬁc coast of South America creates one of the most productive ecosystems on Earth. To trace the origin of the water masses in this area, we measured neodymium isotope compositions ( ԑ Nd ) in tooth enameloid of two genera of coastal sharks from latest Oligocene to early Pleistocene strata in the Pisco and Sacaco basins in southern Peru. Most ԑ Nd values range from (cid:1) 4 to (cid:1) 1, with a strong negative excursion in the late Miocene ( (cid:3) 8–7 million years ago [Ma]) with values as low as (cid:1) 9.2. The overall trend of the ԑ Nd values resembles that of equatorial Paciﬁc deep waters, though with an offset of about +2 ԑ Nd units until about 6 Ma. With a major input of hinterland weathering considered unlikely, we interpret this pattern as reﬂecting a modern-type upwelling regime, though with a lower contribution of Antarctic waters than today. Starting about 6 Ma, the contribution of Antarctic waters to the upwelling waters increased approximately to present-day levels, coincident with, and possibly driven by, increased Antarctic glaciation and the Andes reaching their present-day elevation, both of which likely enhanced the counter-clockwise circulation in the South Paciﬁc Ocean. The negative excursion of ԑ Nd values in the Pisco/Sacaco basins (cid:3) 8–7 Ma coincides with a late Miocene biogenic bloom in the Paciﬁc Ocean and elsewhere, and with a strongly increased northward bottom current observed on the Nazca Drift System just offshore our sampling area. Thus, the negative excursion of ԑ Nd values in the Pisco/Sacaco basins likely resulted from a southern sourced input of nutrient-rich, unradiogenic water, which could have been an important contributor to the biogenic bloom. (cid:1) 2023 The Author(s). Published by Elsevier B.V. on behalf of International Association for Gondwana Research. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/ 4.0/).


Introduction
The Humboldt Current System comprises the surface and subsurface flows in the eastern boundary current system along the Pacific coast of South America (Fig. 1).The upwelling of cold, nutrient-rich waters along the coast of Peru and Chile creates one of the most productive ecosystems on Earth, producing more fish per unit area than any other region, and about 10 % of the world's fish catch (Thiel et al., 2007;Chavez et al., 2008;Montecino and Lange, 2009).Through the decay of the abundant organic matter and the oxygen-poor nature of the upwelling water, it also results in one of the world's largest and shallowest Oxygen Minimum Zones (Helly and Levin, 2004).Upwelling is driven by equatorward surface winds along the Peruvian coast, leading to Ekman transport offshore and thus upwelling along the coast.The upwelling water masses are composed of the underlying waters of the poleward Peru-Chile Undercurrent (PCUC), Sub-Antarctic Mode Water (SAMW) and local recirculation (Fig. 1; Karstensen and Ulloa, 2009;Montes et al., 2010;Grasse et al., 2012;Toggweiler et al., 2019).
Whereas the short-term variability of this upwelling system, for example the El Niño-southern Oscillation phenomenon or Ice Agerelated changes, are well studied (Salvatteci et al., 2016;Timmermann et al., 2018), much is still to be learned about its longer-term evolution.Its origin was linked to the onset of the Antarctic Circumpolar Current (ACC) in the Eocene (Marty et al., 1988), and upwelling analogous to the present-day condition were suggested to have persisted since then (von Huene et al., 1987;Lagabrielle et al., 2009;Armijo et al., 2015;DeVries et al., 2017).However, paleontological studies in the Pisco and Sacaco basins suggest a more complex history (Muizon and Devries, 1985;Collareta et al., 2021;Ochoa et al., 2021).They indicate the persistence of warm-water faunas well into the Neogene (Bosio et al., 2020), marked changes in diversity in the latest Miocene-early Pliocene, and faunal communities resembling those of the present-day Humboldt Current System only from the late Pliocene onward (Ochoa et al., 2021).
Biogenic phosphates are among the most commonly used archives of the Nd isotope composition of seawater for past oceans (Staudigel et al., 1985;Martin and Haley, 2000;Dopieralska et al., 2016).While vertebrate bones may be subjected to substantial diagenetic exchange that compromises their utility as a seawater e Nd archive (Kocsis et al., 2010;Tütken et al., 2011), fluorapatite forming fish (including shark) teeth, and most notably the enamel and enameloid, show high resistance to late-diagenetic alteration, providing a reliable recorder of seawater Nd isotope signatures for Cenozoic timescales (e.g., Scher and Martin, 2006;Kocsis et al., 2010;Scher and Delaney, 2010;Moiroud et al., 2013;Kim et al., 2020).The MREE-enriched patterns commonly observed in fish teeth reflect acquiring their REEs largely from shallow-level, early-diagenetic pore waters, rather than directly from seawater; in most cases, fish-tooth fluorapatite can, however, be expected to reflect the seawater e Nd signal (Martin and Haley, 2000;Kocsis et al., 2010;Huck et al., 2016).This is because, in marine settings, the early-diagenetic pore water REE inventories, and thus Nd isotope signals, are typically dominated by reduction of marinederived, authigenic Fe-Mn oxides and sedimentary organic matter, and thus resemble the local seawater signature (Martin and Haley, 2000;Gutjahr et al., 2007;Molina-Kescher et al., 2014).As a result, as long as no exotic Nd is locally introduced by seepage of deepseated fluids (cf., Jakubowicz et al., 2019), the offset between deep seawater and shallow-burial pore water values is typically insignificant and does not exceed 1-1.5 e Nd units even for profiles exceptionally rich in volcanic ash dispersed within the sediment (Abbott et al., 2015).Consequently, fish teeth provide a valuable archive of secular changes in the Nd isotope signature of seawater in marine basins.
Here, we present a late Oligocene to early Pleistocene record of oceanographic conditions in the south-eastern Pacific Ocean based on the Nd isotope composition of shark teeth enameloid from the Pisco and Sacaco basins in south-central Peru.

Material and methods
The Pisco and Sacaco basins (Fig. 2) are located at 13-16°S, contain Eocene to Pleistocene shallow marine sedimentary sequences deposited at depths not deeper than $150 m, and are renowned for their diverse marine mammal fauna (Bosio et al., 2021;Collareta et al., 2021;Ochoa et al., 2021).For our analysis, teeth of a genus of coastal shark (Carcharhinus) without inter-oceanic migratory behavior (Schultz et al., 2008;Benavides et al., 2011) were used to ensure that the isotope signal reflects local water masses.Most teeth belong to Carcharhinus brachyurus, with additional specimens of C. leucas and C. aff.macloti, and a few that could not be assigned with confidence to any of these species, but belong to Carcharhinus.The material was partly collected during a field trip in 2018 and supplemented by specimens housed in the Departamento de Paleontología de Vertebrados del Museo de Historia Natural, Universidad Nacional Mayor de San Marcos de Lima (MUSM) in Peru (Table 1).
The shark-tooth fluorapatite samples (5-18 mg) were spiked with 150 Nd-149 Sm enriched tracer and dissolved in concentrated HNO 3 .Separation of Nd and Sm was performed on 2 ml columns packed with EICHROM Ln resin.Both elements were eluted with ultrapure HCl, 0.25 N and 0.75 N, respectively.Nd and Sm (loaded as phosphate) were measured on Re in a double filament configuration.The Nd-Sm isotope analyses were performed at the Isotope Laboratory of the Adam Mickiewicz University, Poznan ´(Poland), on a Finnigan MAT 261 multi-collector thermal ionization mass spectrometer running in a static (Sm) and dynamic (Nd) mode.The AMES standard yielded 143 Nd/ 144 Nd = 0.512128 ± 10 (2r; n = 15).The 143 Nd/ 144 Nd ratios were normalized to 146 Nd/ 144 Nd = 0.7219 and Sm isotope ratios to 147 Sm/ 152 Sm = 0.56081.Total procedure blanks were < 40 pg for both Nd and Sm.Nd isotope data are reported in the standard e notation calculated using 143 Nd/ 144 Nd = 0.512638 and 147 Sm/ 144 Nd = 0.1967 for present-day CHUR (Jacobsen and Wasserburg, 1980) (see Table 2).

Results and interpretation
Overview.Most measured late Oligocene to Pleistocene shark teeth e Nd values from the Pisco and Sacaco basins range from À4.3 to À0.7, with a remarkable excursion at around 8-7 million years ago (Ma), when ԑ Nd values dropped to À9.2 (Fig. 3A; Table 2).
The Pisco/Sacaco e Nd curve shows the same overall trend towards more radiogenic values as the deep equatorial Pacific e Nd curve, though with an offset of +2 e Nd units from 25 to 9 Ma.This offset disappears after the late Miocene negative excursion.From about 2.5 Ma onward, the Pisco/Sacaco e Nd curve shows the same trend toward less radiogenic values as the deep equatorial Pacific e Nd curve (Fig. 3A).
Weathering inputs.Interpreting these data requires careful consideration of potential Nd sources, which include both the coastal water masses and local input.Ehlert et al. (2013) proposed that e Nd values found in Mn-Fe coatings of particles and benthic foraminifers along the Peruvian coast follow the local water and detritus values, and concluded that the coastal e Nd signatures Fig. 1.Schematic oceanography off northwestern South America and ԑ Nd values of key ocean currents and water masses (after: Karstensen and Ulloa, 2009;Montecino and Lange, 2009;Grasse et al., 2012;Bostock et al., 2013;Basak et al., 2015;Grasse et al., 2017).PaC: Panama Current; pSSCC and sSSCC: primary and secondary Southern Subsurface Countercurrent.
broadly reflect local weathering inputs and hinterland geology.The main rocks exposed in the drainage area of the Pisco and Sacaco basins include Precambrian basement (with e Nd values of $À5 to À4), Paleozoic crust (e Nd $À1.5 to -1), Mesozoic volcanics (e Nd $1 to 3), the Cretaceous Coastal Batholith (e Nd $0 to 3) and Cenozoic volcanics (e Nd $ À5 to +2) (Soler and Rotach-Toulhoat, 1990;Vatin-Pérignon et al., 1992;Martínez Ardila et al., 2019).A recent provenance analysis showed that material of the most Miocene to Pleistocene clastics in the Pisco/Sacaco basin is derived from Mesozoic and Cenozoic sources, though until 6 Ma there was also input from a somewhat enigmatic late Neoproterozoic source (Ochoa et al., 2022).
Weathering inputs from the whole range of hinterland rocks could potentially have produced most of the values seen in the Pisco/Sacaco shark teeth, but not the Mesozoic and Cenozoic sources alone.None of the known hinterland rocks could have produced values as low as those of the late Miocene negative excursion, at least assuming that the unknown Neoproterozoic source has e Nd values similar to those of the known Precambrian basement.The history of Andean uplift is still under debate, but most studies agree that the Western Cordillera reached a near-modern elevation well before or during the mid-Miocene, with relatively little change afterward, and the more easterly located Eastern Cordillera and Altiplano reached their present-day elevation during  1.Modified from www.geomapapp.org(Ryan et al., 2009).Thouret et al., 2007;Prudhomme et al., 2019;Sundell et al., 2019;Boschman, 2021).Furthermore, the south-central Peruvian margin has, on average, been very dry since the Miocene (Hartley et al., 2005;Garreaud et al., 2010;Amidon et al., 2017).Therefore, large changes in the input of weathering products into the Pisco and Sacaco basins may not be expected.A significant part of the upwelling waters along the Peruvian coast originate from the southward-flowing Peru-Chile Undercurrent, which in turn is fed by the Equatorial Undercurrent (Fig. 1; Montecino and Lange, 2009).The e Nd signature of this current is around À3 to À1 (Grasse et al., 2012;Haley et al., 2021) and thus not unlike that of the Pisco/Sacaco shark teeth.Interestingly, the e Nd values of extant foraminiferans and Mn-Fe coatings published by Ehlert et al. (2013) from the area with the strongest permanent upwelling along the Peruvian coast (5-15°S, cf.Montecino and Lange, 2009) are remarkably uniform with a mean around À1, consistent with the e Nd signature of the PCUC and EUC (Grasse et al., 2012;Haley et al., 2021), which nearly reaches the surface in this area (Chaigneau et al., 2013).Thus, with (i) major changes in the input of weathering products and associated changes in the e Nd signature of the Pisco and Sacaco basin sediments since the late Oligocene being unlikely, and (ii) proxies for the present-day e Nd signature of the waters along the central Peruvian coast (foraminiferans and Mn-Fe coatings) showing values of the upwelling, underlying water masses, we assume that the longer-term trends seen in our shark teeth ԑ Nd data should provide insights into the general oceanographic conditions along the south-central Peruvian coast.
The general resemblance of the ԑ Nd curves of the Pisco/Sacaco basins and deep equatorial Pacific Ocean (except for the late Miocene negative excursion and the +2 offset before 9 Ma) is striking.Although the PCUC and EUC are not fed by equatorial Pacific bottom waters (most EUC waters originate in the Solomon basin just east of Papua New Guinea, cf.Tsuchiya et al., 1989;Grenier et al., 2011), we interpret this resemblance as further indication that the Pisco/Sacaco shark ԑ Nd values reflect an open marine signal, rather than hinterland weathering.Furthermore, we consider it unlikely that changes in weathering in the hinterland should coincidentally have reproduced the ԑ Nd trend of the deep equatorial Pacific Ocean.Rather, we interpret the Pisco/Sacaco shark ԑ Nd data as evidence that upwelling occurred along the Peruvian coast at least since the late Oligocene, consistent with earlier studies (von Huene et al., 1987;Marty et al., 1988;Lagabrielle et al., 2009;Armijo et al., 2015;DeVries et al., 2017).
The +2 offset before 9 Ma.The offset of +2 e Nd units before 9 Ma in the Pisco/Sacaco e Nd record compared to the equatorial Pacific record and its subsequent disappearance (Fig. 3A) suggests changes in the origin, proportions, and/or intensity of the upwelling waters.The source area of the EUC, the Solomon Basin, is surrounded largely by young volcanic rocks, and the present-day NE directed South Solomon Arc volcanism started about 6 Ma (Smith, 1982;Petterson et al., 1999).Thus increased input from this source, if any, should have resulted in higher rather than the observed lower e Nd values after 6 Ma.
Strengthened or even permanent El Niño conditions have been discussed for the Pliocene and Miocene (Fedorov et al., 2006;Von Der Heydt and Dijkstra, 2011).Such conditions could have subdued upwelling along the Peruvian coast and could potentially account for the higher e Nd values in the Miocene Pisco record.However, the evidence for permanent El Niño conditions is controversial (Steinthorsdottir et al., 2021) and other studies instead indicate present-day El Niño variability for the Mio-Pliocene (Batenburg et al., 2011;Watanabe et al., 2011;Okamura et al., 2013;Pérez-Rivarés et al., 2019).Thus, the 'permanent El Niño' is here not considered further to explain the +2 e Nd units offset in the Pisco/Sacaco data, though it may not be completely ruled out.
The disappearance of the +2 offset at 8-6 Ma broadly coincides with the time when the Altiplano reached its present-day elevation and with increased glaciation in Antarctica (Fig. 3).Andean uplift has been associated with increased upwelling along the Peruvian coast, because it enhances northward along-shore wind strength (Sepulchre et al., 2009).Likewise, an expanding West Antarctic ice sheet would steepen the latitudinal temperature gradient and increase counter-clockwise circulation in the southern Pacific Ocean (Holbourn et al., 2018;Steinthorsdottir et al., 2021).This in turn could have enhanced the transport of Antarctic waters towards the Pacific margin of South America, including Sub-Antarctic Mode Water that feeds into the upwelling waters along the Peruvian coast.Enhanced input of SAMW with its negative e Nd signature into the upwelling water mass could thus have eliminated the +2 offset at 6 Ma in the Pisco/Sacaco record.Another potential explanation for the disappearance of the +2 offset is the exhaustion of an enigmatic Neoproterozoic source rock in the hinterland at this time (cf., Ochoa et al., 2022) that could have contributed unradiogenic Nd to the Pisco/Sacaco signal.But because the disappearance of the +2 offset coincides with a gradual change in the Pisco/Sacaco marine vertebrate record toward more coldadapted faunas (Collareta et al., 2021;Ochoa et al., 2021), we consider increased upwelling of colder water the more likely scenario.
The late Miocene negative excursion.The most remarkable deviation from the general 'equatorial Pacific' trend occurred in the late Miocene about 8-7 Ma, when ԑ Nd values dropped to -9.2.These values are lower than those of any of the rocks in the hinterland -including the Paleozoic basement (Ehlert et al., 2013;Martínez Ardila et al., 2019;Robinson et al., 2021).Also dust input from the Andes to the equatorial eastern Pacific does not show a marked increase at this time (Fig. 3B; cf.Tiedemann and Mix, 2007).Thus, changes in terrestrial input from hinterland weathering are an unlikely driver of this negative excursion.The values are also well below those of coeval Pacific deep water (Fig. 3A; see also McKinley et al., 2019), Pacific surface waters and shallow marine sediments, and of rocks in the hinterlands of the Pacific continental margins (Robinson et al., 2021), thus indicating an input of oceanic water masses from elsewhere.
Potential sources with such low ԑ Nd values are a direct influx of Antarctic water or an influx from the Atlantic Ocean via the Central American Seaway (Lacan et al., 2012).Flow through the Central American Seaway is typically considered to run from the Pacific into the Atlantic Ocean, rather than vice versa (Iturralde-Vinent and MacPhee, 1999;Heydt and Dijkstra, 2005), and with progressive shallowing of this seaway during the late Miocene, the flux should have decreased, rather than increased.Furthermore, Caribbean seawater ԑ Nd values reconstructed based on fish teeth and debris of this age are only as low as -7.3 (Newkirk and Martin, 2009), which is too high to explain the Pisco/Sacaco basin values of -9.2.However, oceanographic modeling of the Central American Seaway indicates that at a certain depth of the seaway ($50 m), wind-driven surface waters would flow westward from the Atlantic into the Pacific Ocean (Sepulchre et al., 2014), from where they could potentially reach the Peruvian coast.Interestingly, this would be a short-lived event due to the continued shallowing and ultimate closure of the seaway, thereby providing a mechanism to end the negative excursion.
The other source of water masses with ԑ Nd values negative enough to explain our data is Antarctic water (Basak et al., 2015;McKinley et al., 2019).This explanation could involve either a direct influx of surface water (i.e., a proto-Humboldt current) or an increased contribution of deeper Antarctic waters (AAIW/SAMW) to the upwelling water.Enhanced input of surface water would be possible given increased atmospheric and oceanic circulation (see discussion of mechanisms below) as it would result in increased export of northern ACC water into the Humboldt Current System (Lamya et al., 2015).
The main area of AAIW and SAMW formation today lies northwest of the Antarctic Peninsula, from where it is transported NW-ward into the South Pacific (Hartin et al., 2011;Bostock et al., 2013).Among these water masses, SAMW is known to feed into the upwelling water along the Peruvian coast (Montes et al., 2010;Toggweiler et al., 2019).Although the northern limit of the AAIW near the South American coast was considered to lie between 18 and 22°S today (Bostock et al., 2013), Grasse et al. (2012) reported potential AAIW from as far north as 9°S, albeit only at depth of c. 800 m and thus not reaching the surface mixing zone.Furthermore, modeling of Nd isotope distributions indicates that AAIW flows northward into the Pacific Ocean at depths between 500 m and 1000 m and nearly reaches the equator (Arsouze et al., 2007, fig. 5a).
There are two potential mechanisms and triggers that might have increased upwelling and allowed AAIW to reach surface waters: a coeval pulse of uplift of the Altiplano of the Andes (Kar et al., 2016;Schildgen and Hoke, 2018;Boschman, 2021) resulting in increased wind stress along the Peruvian coast (cf.Montecino and Lange, 2009;Sepulchre et al., 2009), and growth of the Antarctic ice sheet, resulting in a steeper equator to pole temperature gradient (Holbourn et al., 2018;Steinthorsdottir et al., 2021).Climate modeling suggests that a climate warmer than today (as expected for the late Miocene) would increase the thermal stratification of the water column and decrease the depth of the upwelling source waters (Oerder et al., 2015;Wang et al., 2015), making it less likely for AAIW to reach surface waters.However, an analysis of sediment distribution on the Nazca Drift System offshore Peru showed a period of intense northward bottom current c. 7.7 to 9.4 Ma (Calvès et al., 2022), indicating an enhanced Antarctic Circumpolar Current and associated northward export of Antarctic waters coincident with the late Miocene negative ԑ Nd excursion in the Pisco and Sacaco basins.Also consistent with an increased input of deeper, nutrient-rich, Antarctic waters to the upwelling water is the increased deposition of diatomite mud in the Pisco/Sacaco basin from around 8 Ma (Di Celma et al., 2017).
The late Miocene negative excursion of the Pisco/Sacaco shark ԑ Nd record shows a remarkable coincidence with a global, late Miocene to early Pliocene biogenic bloom and associated high opal and CaCO 3 deposition (Peterson et al., 1992;Tiedemann and Mix, 2007).The timing and duration of this biogenic bloom varied somewhat between ocean basins (Karatsolis et al., 2022); in the Pacific Ocean, it was strongest between about 8 and 6.4 Ma (Suto et al., 2012;Karatsolis et al., 2022), consistent with the negative ԑ Nd excursion in the Pisco/Sacaco basin.The biogenic bloom has been attributed to various causes, including ocean fertilization by dust input due to increased aridity in central Asia and South America (Suto et al., 2012), the closure of the Central American Seaway (Lyle and Baldauf, 2015), and the growth of the Antarctic ice sheet, resulting in the strengthening of trade winds and associated increased upwelling (Holbourn et al., 2018).Our data and interpretation of the late Miocene negative ԑ Nd excursion in the Pisco/ Sacaco basin would support the hypothesis of increased upwelling during this time (Holbourn et al., 2018;Steinthorsdottir et al., 2021).Furthermore, increased input of AAIW/SAMW to the upwelling waters along the Pacific coast of South America could have been an important contributor to the biogenic bloom.
The HCS ecosystem.In the Pisco and Sacaco basins, the late Miocene negative ԑ Nd excursion was associated with a decrease in the diversity of cetaceans (baleen and toothed whales) that is mirrored in a global decrease in cetacean diversity (Marx and Uhen, 2010;Villafaña and Rivadeneira, 2014).The negative excursion and the disappearance of the +2 offset between the Pisco/ Sacaco and the equatorial Pacific ԑ Nd record coincide with a gradual change in the Pisco/Sacaco marine mammal record toward more cold-adapted faunas (Fig. 3; Collareta et al., 2021;Ochoa et al., 2021).After a 14-million-year record of two marine crocodylian species in the Pisco/Sacaco basins, their disappearance from coastal environments at about the earliest Pliocene (Salas-Gismondi et al., 2022) provides further evidence for our hypothesis of an increased contribution of Antarctic waters to the upwelling waters along the Peruvian coast.
The present-day Humboldt Current upwelling ecosystem has a characteristic structure in which a few species of extremely abundant, plankton-feeding 'forage fishes', such as sardines and anchovy, link the planktonic primary production to the generalized predators (Alheit and Niquen, 2004).The suggested increased contribution of Antarctic waters into the Pisco/Sacaco basins since 9 Ma -carrying along abundant silica -might have facilitated the onset of this present-day food web structure.This is suggested by the stomach content of a diversity of large predatory vertebrates (baleen whales, sharks, ziphiid odontocetes) from the upper part of the Pisco Formation (8-7 Ma), which consisted mainly of the same sardine fish, despite their different feeding anatomies (Collareta et al., 2021).Thus, increased primary productivity induced by the suggested input of silica-rich Antarctic water could have catalyzed the onset of the forage fish-based food web seen in the Humboldt Current upwelling ecosystem today.
Also remarkable is that the abrupt shift in the Sacaco basin shark ԑ Nd values toward the less radiogenic present-day values at 2.5 Ma (Fig. 3) coincides with the first appearance of taxa related to present-day top predators (sea lions, bottlenose dolphins) of the Humboldt Current upwelling ecosystem (Ochoa et al., 2021) and with a massive increase in bioproductivity offshore northwestern South America, as indicated by increased biogenic opal deposition (Fig. 3B, cf.Tiedemann and Mix, 2007).The coeval shift toward less radiogenic ԑ Nd values in the deep tropical Pacific Ocean was interpreted as an increased NADW component via the ACC (Ling et al., 1997).Thus, increased Northern Hemisphere glaciation (Fig. 3C, cf.Zachos et al., 2001;Steinthorsdottir et al., 2021) resulting in enhanced input of cold, nutrient-rich NADW into the ACC and from there toward the South America Pacific coast, might ultimately have shaped the modern Humboldt Current ecosystem.

Conclusions
The late Oligocene through Pleistocene ԑ Nd signatures of shark tooth enameloid from the Pisco and Sacaco basins in southern Peru provide new insights into the origin of the water masses bathing these basins.Overall our data are consistent with upwelling since the late Oligocene.The upwelling waters today are a mix of equatorial subsurface waters and deeper Antarctic waters (SAMW).The Pisco/Sacaco ԑ Nd data indicate that the contribution of southernsourced waters was lower than today until the late Miocene, and the change to approximately present-day mixing proportions was coincident with increased Antarctic glaciation and with the Andes reaching their present-day elevation.Both of these factors likely enhanced the counter-clockwise circulation in the South Pacific Ocean, thereby increasing the contribution of deeper Antarctic waters to the upwelling waters along the Peruvian coast.
A remarkable negative excursion of ԑ Nd values occurred about 8-7 Ma, indicating an even stronger, though short-lived, influx of Antarctic waters.This event coincides with a late Miocene biogenic bloom recorded across the Pacific Ocean.We propose that the northward spread of Antarctic Intermediate Waters likely was an important contributor to this biogenic bloom.The general trend in the Pisco/Sacaco ԑ Nd data toward more radiogenic values was reversed at about 2.5 Ma, coincident with extensive glaciation in the Northern Hemisphere, resulting in increased transport of cold North Atlantic Deep Water with a very unradiogenic ԑ Nd signature, into the Southern Ocean.These Northern Hemisphere water masses may ultimately have established the present-day forage fish-based food web of the Humboldt Current upwelling ecosystem.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 2 .
Fig. 2. Sample locations in the Pisco and Sacaco basins in south-central Peru.Numbers refer to locality numbers in Table1.Modified from www.geomapapp.org(Ryan et al., 2009).

Fig. 3 .
Fig. 3. A. Late Oligocene to Pleistocene shark teeth ԑ Nd values from the Pisco and Sacaco basins, Peru (filled circles); horizontal bars indicate ranges of values pooled by locality; numbers indicate sampling sites given in Fig. 2 and Table 1; present-day values for the Pisco/Sacaco basins from stations 30 and 78 of Grasse et al. (2012), at 3 and 2 m water depth, respectively.For comparison, we include ԑ Nd values from the deep equatorial Pacific [data for sites VA13/2, D11-1 and CD29-1 from Ling et al. (1997); for ODP 807 from Le Houedec et al. (2016)], the Antarctic Circumpolar Current [data from ODP site 689 for 25-17 Ma, from Scher and Martin (2004), and for site U1370 from McKinley et al. (2019)] and North Atlantic Deep Water (NADW) from ODP hole 1264 on the Walvis Ridge in the South Atlantic Ocean, the nearest water mass feeding into the Pacific part of the ACC for which Neogene ԑ Nd data are available [data from Thomas and Via (2007)].Locations are shown in inset D. B. Sediment mass accumulation rates (MAR) offshore Peru; terrigenous MAR as a proxy for dust flux, biogenic opal MAR as a proxy for paleoproductivity; note different scales for each (data from Tiedemann and Mix, 2007, fig.F9). C. Related local and global trends.Events and developments from Steinthorsdottir et al. (2021); Pisco/Sacaco marine vertebrate fauna from Collareta et al. (2021) and Ochoa et al. (2021); Andean elevations from Boschman (2021).