Response of the Central Pacific Intertropical Convergence Zone to Northern Hemisphere Cooling During the Last Glacial Maximum and Heinrich Stadial 1

The latitudinal position of the Intertropical Convergence Zone (ITCZ) reflects the energy imbalance between the hemispheres. Southward displacements of the ITCZ during the Last Glacial Maximum (LGM; 19–26.5 ka) and Heinrich Stadial 1 (HS1; 14.6–17.5 ka), are widely accepted, but their magnitude is controversial. Geochemistry of detrital fractions in down‐core sediments collected from 6°N to equator along the 131.5°W transect reveal a distinct shift in εNd, La/Yb, and La–Sc–Th composition from predominantly northern hemisphere‐sourced to mixed northern and southern hemisphere‐sourced signal at 3°N–4°N during the LGM and 3°N–6°N during HS1. These contrasting provenance signals point to the past ITCZ functioning as a dust barrier. Given that a comparable geochemical demarcation currently occurs at 6°N–7°N, our data suggest that the ITCZ migrated southward by ∼3° during the LGM and ∼1°–3° during HS1 relative to its modern position in the central Pacific.


Introduction
The Intertropical Convergence Zone (ITCZ) is a zonal belt of intense precipitation that has major influence on the hydroclimate of the tropics (Flohn, 1981;Nicholson, 2018).Formed by the convergence of northeast and southeast trade winds, it migrates seasonally and on longer timescales toward the warmer hemisphere with a weaker temperature gradient (Schneider et al., 2014).As the magnitude of latitudinal displacement increases proportional to the degree of temperature contrast between hemispheres (Donohoe et al., 2013;McGee et al., 2014), the mean latitudinal position of the ITCZ reflects the global climate state, especially interhemispheric energy imbalance.Reconstruction of the latitudinal shift of the ITCZ in the geologic past can therefore provide insights to the Earth's climate dynamics.Previous studies investigated the ITCZ position for different geologic periods where the Earth's climate at the time is of particular interest: Eocene-Oligocene Transition (Hyeong et al., 2016;Pettke et al., 2002); Mi-1 glaciation (Hyeong et al., 2014); Mid-Pleistocene Transition (Seo et al., 2015); Heinrich Stadial 11 (Jacobel et al., 2016;Reimi et al., 2019); the Last Glacial Maximum and Heinrich Stadial 1 (e.g., Arbuszewski et al., 2013;Bahr et al., 2018;Koutavas & Lynch-Stieglitz, 2003).
Eolian dust in pelagic sediments, transported by wind and settled by gravity and rainfall, is a recorder of the past atmospheric circulation.Due to the heavy precipitation and the resulting removal of aerosols from the atmosphere, the ITCZ is known to act as an effective barrier to interhemispheric dust transport (Merrill et al., 1989;Rea, 1994).This results in a distinct change in the composition of eolian dust across the ITCZ-Northern Hemisphere (NH)-sourced dust and Southern Hemisphere (SH)-sourced dust predominantly accumulate on the seafloor to the north and south of the ITCZ, respectively (Nakai et al., 1993;Stancin et al., 2006;Ziegler et al., 2008).Based on this demarcation of hemispheric dust provenance, Seo et al. (2021) recently proposed the concept of "geochemically defined mean ITCZ position" as a measure to quantitatively assess the degree of ITCZ displacement in the geologic past.Eolian particles in the surface sediments collected along a 131.5°W transect in the study revealed well-defined Asian dust signals north of 7°N and mixed Asian-South American source signals between 2°N and 6°N.Given the large hemispheric asymmetry in dust flux, the southern limit of an intact NH dust signal at around 7°N in the central Pacific was proposed as the present-day mean ITCZ position in terms of dust provenance.It was suggested that this "geochemically defined mean ITCZ position" can be used as a reference back to ca. 14 Ma during which hemispheric dust flux contrast were maintained (Seo et al., 2021).
During the Last Glacial Maximum (LGM), the cryosphere expanded markedly in the NH (e.g., Laurentide Ice Sheet) and moderately in the SH (Donohoe et al., 2013;Peltier, 2004), which led to significant NH cooling (Blunier et al., 1998).The subsequent deglaciation is interrupted by Heinrich Stadial 1 (HS1), another pronounced NH cooling related to a massive iceberg discharge in the North Atlantic (Blunier et al., 1998;Bond et al., 1993;Heinrich, 1988;Wolff et al., 2010).These two periods have been extensively studied to understand the behavior of the ITCZ on asymmetric cooling in high latitudes.A consistent southward shift of the ITCZ during the LGM and HS1 has been supported by both proxy records and model simulations (Arbuszewski et al., 2013;Bahr et al., 2018;Broccoli et al., 2006;Carolin et al., 2013;Denniston et al., 2013;Donohoe et al., 2013;Gibbons et al., 2014;Koutavas & Lynch-Stieglitz, 2003;Li & Liu, 2022;McGee et al., 2014;Partin et al., 2007;Peterson et al., 2000;Wang et al., 2004), but the estimated magnitude of the ITCZ shift varies across studies.While climate models and theoretical calculations have estimated a global mean southward displacement of <1°for the LGM and HS1 (Atwood et al., 2020;Donohoe et al., 2013;McGee et al., 2014), some marine proxy records indicate much greater displacements of up to 7°for these intervals (e.g., Arbuszewski et al., 2013;Reimi & Marcantonio, 2016;Seo et al., 2016).Yet there are also proxy data that support much smaller ITCZ shifts even in the ocean (e.g., McGee et al., 2007;Rowland et al., 2021;Xie & Marcantonio, 2012).These large discrepancies point to a paucity of research compared to the substantial regional variability in the latitudinal position of the ITCZ.
The purpose of this study is to quantitatively assess the magnitude of the ITCZ displacement in the central Pacific (131.5°W) during the LGM and HS1.We investigate the dust component in sediment cores ranging from 0°to 6°N , whose surface layers have been analyzed by Seo et al. (2021) (Figure 1).Based on the geochemical composition of the inorganic silicate fraction (composed mainly of eolian materials with minor unconstrained amount of volcanogenic and authigenic particles; Rea, 1994) isolated from sediments deposited during the LGM and HS1, the southernmost latitude of an intact NH signal for each time period is identified.The displacement of  2021); white circles are the cores from other previous studies discussed in the text (from left to right: Seo et al., 2016;Reimi & Marcantonio, 2016;Ziegler et al., 2008;Xie & Marcantonio, 2012).
this "geochemically defined mean ITCZ position" from its modern location at 7°N allow us to quantitatively estimate the magnitude of the ITCZ shift during the two periods of NH high latitude cooling.

Materials and Methods
The studied cores were collected at ∼1°interval along the 131.5°W transect between the equator and 6°N using multiple and piston corers on the R/V Onnuri in 2005 (Figure 1; Table S1 in Supporting Information S1).The collected cores were subsampled at 1 cm intervals on board, and the subsamples were freeze-dried and stored in plastic bottles in a cool room until analysis.Because of the increasing water depth below the carbonate compensation depth, sediments at 7°N and north consist primarily of pelagic red clay and are not appropriate for millennial scale paleoclimate studies.For this reason, we used the same cores studied by Seo et al. (2021), but only those from 6°N and south.Ages were resolvable to millennial scale for these cores consisting of calcareous ooze.

Chronology
We used 14 C measurements of mixed species of planktic foraminifera to establish the core chronologies (Text S1 in Supporting Information S1).The 14 C ages were obtained by accelerator mass spectrometry (AMS) at the Beta Analytic Radiocarbon Dating Laboratory, Miami, USA (Table S2 in Supporting Information S1).The radiocarbon ages were corrected for isotopic fractionation and calibrated to calendar years before present (cal yr BP) using Calib 8.2 software (Stuiver & Reimer, 1993).This correction was based on the Marine20 data set (Heaton et al., 2020) and a local reservoir effect (ΔR) of 150 ± 36 years estimated from the average models of 10 nearby sites (Broecker & Olson, 1961;Burr et al., 2009;Petchey et al., 2008;Zaunbrecher et al., 2010).Oxygen isotope ratios (δ 18 O) of planktic foraminifer, P. obliquiloculata (355-425 μm, ∼10 specimens) were measured for MC5111 core to supplement radiocarbon ages using a Finnigan MAT 251 mass spectrometer at the University of Michigan, Ann Arbor, USA.The 1σ standard deviation of repeat analyses of the reference material (NBS-19; National Bureau of Standards) was 0.05 (n = 21).

Geochemistry
The inorganic silicate fractions deposited during the LGM and HS1 were isolated from bulk sediments by sequential treatment with a 25% acetic acid, a hot sodium citrate-dithionite buffered with a sodium bicarbonate, and a hot 1.5 M sodium hydroxide solution to dissolve out carbonate, oxides and hydroxides, and biogenic silica, respectively (Hovan, 1995).The isolated samples were digested by a standard HF-HNO 3 -HClO 4 method and were analyzed for elemental chemistry using a Perkin-Elmer ELAN 6100 quadrupole inductively coupled plasma-mass spectrometer at the Korea Basic Science Institute (KBSI), Daejeon, Korea.Analytical accuracy was within 10% of the certified values (NIST 1646a), and the relative standard deviation of triplicate samples was 3%, 5%, 4%, and 6% for La, Yb, Sc and Th, respectively. 143Nd/ 144 Nd ratios were determined via multiple collector thermal ionization mass spectrometry (VG54-30, Isoprobe-T) at KBSI using standard cation exchange techniques.The 143 Nd/ 144 Nd data are reported as ε Nd relative to the chondrite uniform reservoir (CHUR) where ε Nd = ([ 143 Nd/ 144 Nd 0.512638]/0.512638)× 10 4 (Jacobsen & Wasserburg, 1980).The total procedural blank was 0.05 ng and replicate analyses of JNdi-1 standard yielded 143 Nd/ 144 Nd = 0.512111 ± 0.000006 (2σ, n = 10).Analytical reproducibility associated with column chemistry and instrumental measurements was 143 Nd/ 144 Nd of ± 0.000011 (1σ, n = 9) based on routine duplicate analyses.

Results
The measured 14 C dates and δ 18 O of P. obliquiloculata picked from MC5111 are presented in Tables S2 and S3 in Supporting Information S1.While precise dating is difficult due to low sedimentation rates (∼0.6-3.4 cm/kyr, Figure S1 in Supporting Information S1), we focused on selecting sediment samples deposited during the LGM and HS1 (19-26.5 ka and 14.6-17.5 ka, respectively; Palacios et al., 2020) for each core.Ages of the analyzed sediment samples were estimated by linear interpolation between 14 C age-control points that bracket the sampled interval (Figure S1 in Supporting Information S1).The exceptions were MC5107 (2°N) 32 cm and MC5111 (6°N) 31 cm, whose ages were estimated by linear extrapolation from two age-control points at 18.5 and 30.5 cm and 32.5 and 45.5 cm of each core, respectively (Figure S1; Table S4 in Supporting Information S1).Both of these samples are only 1.5 cm away from the nearest age-control points, and thus can be considered to correspond to HS1, even given the uncertainties in the age estimates.
It was originally intended to analyze both the LGM and HS1 intervals for each core collected at 1°latitude intervals, but a few data points are missing.MC5107 (2°N) lacks sediments deposited during the LGM due to the short length of the core.We failed to obtain a reliable 14 C age that postdate HS1 from MC5110 (5°N) due to poor preservation of planktic foraminifera in the upper part of the core, and therefore HS1 interval could not be specified with confidence.MC5118 (1°N) 31.5 cm and MC5109 (4°N) 17.5 cm samples were initially assumed to correspond to the LGM and HS1, respectively, and were subjected to inorganic silicate fraction extraction and geochemical analysis.However, their estimated ages have been adjusted to 18.8 and 13.8 ka, respectively, following further dating and correction.Geochemical data of these two samples are presented in the supporting information, but will not be discussed further due to their ages outside of the LGM and HS1.
The geochemical compositions of inorganic silicate fractions are summarized in Table S4.We report Post-Archean Australian average shale (PAAS; Taylor & McLennan, 1985)-normalized La to Yb ratio ((La/ Yb) SAMPLE /(La/Yb) PAAS , hereafter expressed as La N /Yb N ) as an indicator of the enrichment of light rare earth elements relative to heavy rare earth elements.The inorganic silicate fractions analyzed show distinct compositional groupings according to sampling latitudes.For both the LGM and HS1 periods, samples from 4°N to 6°N exhibit distinctly higher La N /Yb N (0.94-1.15) and La/Sc ratios (0.97-1.41) than samples from 0°to 3°N (La N / Yb N : 0.53-0.63;La/Sc: 0.45-0.75)(Figures 2b and 3).In terms of Nd isotope ratios, 4°N-6°N samples show a narrow range of ε Nd values ( 9.31 to 8.91), while 0°N-3°N samples show higher and more dispersed ε Nd values ( 8.34 to 4.30) that tend to increase toward the south (Figure 2a).There are no marked compositional differences in ε Nd , La N /Yb N , and La-Sc-Th compositions between the LGM and HS1 intervals at sites where both periods are analyzed (Figures 3 and 4b).

Discussion
Geochemistry of detrital material deposited during the LGM and HS1 shows contrasting provenance signals depending on sampling latitude.This is in parallel with the geochemical patterns shown by surface sediments, only differing in the latitude at which the provenance is divided.Seo et al. (2021) previously identified the same compositional grouping in core-top sediments at our sampling sites and at sites further north, and interpreted this as a difference between NH-sourced materials and mixed NH-and SH-sourced materials (Figures 3 and 4a).In their study, samples from 7°N to 16°N exhibited ε Nd , La N /Yb N , and La-Sc-Th compositions consistent with those of central North Pacific surface sediments, explained by predominant Asian dust with minor addition of volcanic materials incorporated over the Japan arc.In contrast, samples from 0°to 6°N showed compositions similar to southern and equatorial Pacific sediments, best explained by the mixing of Argentine loess and south Andean ash as well as Asian dust (refer to Seo et al. (2021) for detailed discussion).This latitudinal change in hemispheric provenance was later supported by the distinct compositional grouping of smectite in the surface samples (Jung Figure 3. La-Sc-Th ternary diagrams for (a) the surface samples from the equator to 16°N along the 131.5°W transect (Seo et al., 2021) and (b) HS1 and (c) LGM samples of this study.Stars mark the composition of the upper continental crust (UCC) and oceanic crust (OC) (Taylor & McLennan, 1985).Brown and yellow shaded areas represent compositions of predominantly NH-sourced and mixed NH and SH-sourced dust, respectively.Numbers indicate approximate latitudes of each core site.S4 in Supporting Information S1. Brown and yellow shaded areas represent compositions of predominantly NH-sourced and mixed NH and SH-sourced dust, respectively.Numbers indicate approximate latitudes of each core site.et al., 2022).Following the above interpretation of ε Nd , La N /Yb N , and La-Sc-Th compositional data, our results for the LGM samples indicate a predominance of NH-sourced dust at 4°N-6°N sites, and mixed contributions of NH-and SH-sourced dust at 0°and 3°N sites.Results for HS1 samples are not available at 4°N and 5°N sites, but a clear NH signal is found at 6°N site and a mixed NH-SH signal at 0°-3°N sites (Figures 3 and 4).
Taken together, we find that a compositional demarcation occurs further south during the LGM and HS1 compared to the present.The dominant influence of Asian dust appears to have reached ∼3°farther south during the LGM and ∼1°-3°farther south during HS1 than present.The greater penetration of NH dust to the south could result from one of the following four mechanisms: (a) symmetrical contraction of the ITCZ range (Collins et al., 2010); (b) weakened role of the ITCZ as a barrier to interhemispheric dust transport (McGee et al., 2007;Xie & Marcantonio, 2012); (c) enhanced hemispheric asymmetry in dust flux; or (d) southward shift of the mean ITCZ position (Reimi & Marcantonio, 2016).The first mechanism is improbable because the present ∼6°seasonal range (4°N-10°N) of the ITCZ in the study area is too narrow to explain the ∼3°latitudinal shift of the provenance boundary.The second case would likely result in increased mixing of dust from both hemispheres and can be evaluated in a comparative sense from the sharpness of the spatial transition of NH to SH source signals (e.g., Xie & Marcantonio, 2012).However, our data exhibit similar gradients in La N /Yb N and clear groupings in ε Nd -La N /Yb N and La-Sc-Th plots in all recent, HS1, and LGM samples, which suggests insignificant changes in the dust scavenging efficiency of the ITCZ since the LGM.The third mechanism requires an increase in the proportion of NH-sourced dust to a level that would erase the SH signal currently observed at 4°N-6°N.However, the available records of dust fluxes near the study sites do not support this possibility.Dust flux increases at the 110°W and 158°W transects during the LGM and HS1 were assessed to be similar at northern sites where NH dust dominates (7°N and 7.2°N) and southern sites where SH dust prevails (3°S, 0.2°N, and 0.5°N) (Jacobel et al., 2017;McGee et al., 2007).A greater increase in NH dust therefore cannot account for the more southerly position of compositional demarcation during these periods.Based on the above discussion, we propose the last case, a southward shift of the mean ITCZ position, as the main cause of the more southerly propagation of NH dust signal during the LGM and HS1.Seo et al. (2021) proposed the southernmost latitude with an intact NH source signal as "geochemically defined mean ITCZ position", which reflects hemispheric energy imbalance and dust flux contrast at the time of interest.Adopting the proposed concept, the geochemically defined mean ITCZ position would be 3°N-4°N for the LGM as intact NH source signals appear at 4°N and north.In the absence of 4°N and 5°N data points, it is difficult to pinpoint the geochemically defined mean ITCZ position for HS1.However, it must be between 3°N and 6°N because the 6°N sample clearly shows an intact NH source signal.Using the present-day geochemically defined mean ITCZ position of 6°N-7°N as a reference, it is concluded that the annual mean ITCZ position was about 3°s outh during the LGM and 1°-3°south during HS1 relative to the present (Figure 2).The maximum 3°latitudinal shift observed in this study is much larger than the global average of <1°southward shift during the LGM and HS1 estimated by models and paleoclimate reconstructions (Atwood et al., 2020;McGee et al., 2014;Wang et al., 2023).However, these studies pointed out potential greater shift of the precipitation centroid in the central Pacific.Our estimate of the magnitude of the ITCZ shift is consistent with proxybased estimates from other central Pacific sites as well (locations in Figure 1).For example, Seo et al. (2016) estimated a minimum 2°southward shift of the ITCZ during HS1 based on the reconstruction of water column structure using Mg/Ca-derived temperature proxy and planktic foraminiferal assemblages from MC931 (6°40'N, 177°28'W).Reimi and Marcantonio (2016) estimated a minimum 2.5°southward shift of the ITCZ during the LGM based on the radiogenic isotope composition of detrital component in cores collected along the 158°W meridional transect.Their surface ε Nd data at 4.7°N and 7.0°N indicate a South American source, whereas LGM ε Nd data at the same sites indicate an Asian source (Figure 2a).Ziegler et al. (2008) analyzed the terrigenous component in down-core sediments along the 140°W transect, with the LGM data available only for 5°S, 0°, and 4°N sites and HS1 data for 2°N site, making it difficult to locate the past ITCZ position.Nevertheless, their LGM sample from 4°N shows characteristics of nearly intact NH dust with a limited contribution from SH sources (ε Nd = 8.3), as does our LGM sample from 4°N.In the eastern Pacific, Xie and Marcantonio (2012) presented a differing opinion on the ITCZ migration based on the radiogenic composition of detrital component in cores collected from 3°S to 7°N along the 110°W transect (locations in Figure 1).In the study, latitudinal gradient in ε Nd values was sharpest between 2.77°N and 5.29°N during the LGM and HS1, the same as during the Holocene, leading to the conclusion that the ITCZ did not move significantly.However, due to the greater influence of Central America than Asia among NH sources in this region (Stancin et al., 2006), ε Nd values of NH-and SH-sourced dust do not show marked contrast as seen in the central Pacific, making it difficult to trace the ITCZ position using ε Nd .In fact, all ε Nd values observed in the study fell within a narrow range of 5.46 to 3.25, with the lowest values being obtained at the southernmost site at 3°S.Interestingly, our data support that the southward displacement of the ITCZ in the central Pacific was either similar or greater during the LGM than HS1, which is unexpected given that most previous studies have suggested the opposite.Reconstruction of rainfall intensity using reflectance data from the Cariaco Basin and Arabian Sea sediments and δ 18 O of cave stalagmites from Borneo indicated greater ITCZ displacement during HS1 than the LGM (Carolin et al., 2013;Deplazes et al., 2013).Calculations using past sea surface temperature (SST) estimates and the modeled relationship between the ITCZ position and the tropical SST gradient gave the global average of ∼0.3°and ∼0.6°southward shift of the ITCZ during the LGM and HS1, respectively (McGee et al., 2014).A transient climate simulation on changes between the LGM and HS1 indicated ∼1.2°southward shift of the ITCZ during HS1 (McGee et al., 2018).Our data from the 131.5°W transect challenge the above findings, but at the same time it could be a pattern confined to the central Pacific, a region known for its high variability in precipitation centroid for a given forcing (Atwood et al., 2020;McGee et al., 2014).Reimi and Marcantonio (2016), another study from central Pacific that traced the ITCZ position based on dust provenance, also estimated that the ITCZ moved southward by more than 2.5°during the LGM as well as HS1, although exact values could not be obtained due to limitations in spatial resolution of sediment core collection.Future quantitative studies in other tropical oceans as well as other parts of the central Pacific are needed to get a full picture of the global ITCZ movement during the LGM and HS1.

Conclusions
We quantified the latitudinal movement of the mean ITCZ in the central Pacific in response to northern highlatitude cooling during the LGM and HS1.Our data reveal that the geochemically defined mean ITCZ position, which reflects hemispheric energy imbalance and dust flux asymmetry, was located at 3°N-4°N during the LGM and 3°N-6°N during HS1.This is consistent with previous proxy data and model simulations that indicated southward displacement of the ITCZ on northern high-latitude cooling.However, it was unexpected that the mean ITCZ position during the LGM was similar or more southerly than during HS1 because previous studies have generally suggested the opposite.This is the first proxy study to quantitatively address ITCZ migration during the LGM and HS1 together at the same sites, which necessitates additional studies in other parts of tropics to assess whether the comparable magnitude of the ITCZ migration had occurred in other settings.

Figure 1 .
Figure 1.Bathymetric map of the central-eastern equatorial Pacific with the location of sediment cores (modified from GEBCO, 2023 Grid).Red circles mark the cores analyzed in this study.Yellow rectangles are the 16 cores whose surface layers were analyzed by Seo et al. (2021); white circles are the cores from other previous studies discussed in the text (from left to right:Seo et al., 2016;Reimi & Marcantonio, 2016;Ziegler et al., 2008;Xie & Marcantonio, 2012).

Figure 2 .
Figure 2. (a) ε Nd and (b) La N /Yb N of the silicate fraction of samples from the equatorial central-eastern Pacific.Data for 158°W, 140°W, 131.5°W (surface), and 110°W transects are from Reimi and Marcantonio (2016), Ziegler et al. (2008), Seo et al. (2021), and Xie and Marcantonio (2012), respectively.Refer to Figure 4 and Table S4 for errors.Red, blue, and gray arrows and horizontal lines indicate the inferred range of geochemically defined mean ITCZ position (Seo et al., 2021) at 131.5°W during the LGM, HS1, and present, respectively.

Figure 4 .
Figure 4. La N /Yb N versus ε Nd diagrams for (a) surface samples (data from Seo et al., 2021) and (b) LGM and HS1 samples (this study).Error bars of ε Nd are omitted since they are smaller than the symbols; seeTable S4 in Supporting Information S1. Brown and yellow shaded areas represent compositions of predominantly NH-sourced and mixed NH and SH-sourced dust, respectively.Numbers indicate approximate latitudes of each core site.