Introduction

In the equatorial Pacific, dust provenance studies based on detrital Nd and Pb isotopic signatures of marine sediments represent a well-established method for reconstructing the average latitudinal position of the paleo-ITCZ1,2,3,4, due to the ITCZ’s role as a barrier to inter-hemispheric dust transport5 and the distinct Pb and Nd isotopic signatures (206Pb/204Pb, 207Pb/204Pb, 208Pb/204Pb and εNd) of global dust sources. While these studies have established important constraints and hypotheses on ITCZ variability over the central equatorial Pacific (CEP) and westernmost EEP during the last and penultimate deglaciations, recent model simulations have shown that Pacific ITCZ responses to different climatic forcings are not zonally homogenous and that some of the largest meridional shifts of the ITCZ are observed in the easternmost EEP during abrupt climate events6. There is therefore a need for millennial-scale resolution reconstructions of paleo-ITCZ variability further east in the basin. Here others have shown there to be changing patterns of export production that may be related to changes in the fluxes of CO2 into and out of the ocean on climate timescales (e.g., Schimmenti et al.7, and references therein).

To date, South American volcanic systems from the Andean margin and loess from the deserts of China and Mongolia have been recognized as the prevailing northern and southern endmembers, respectively, of atmospheric dusts reaching the equatorial Pacific at 110°W and further west in the CEP1,2,3. Modern dust trajectory models indicate that East Asian dust does not reach east of ~110°W in the equatorial Pacific and that instead, North Africa is the only possible northern hemisphere dust endmember that can physically reach the easternmost EEP (EEEP; east of 110°W). Due to the position of the ITCZ over the region, these models (e.g., ref. 8) estimate ~90–100% of the EEEP atmospheric dust load at the equator is South American in origin and ~10% or less is North African in origin today. However, the latter proportion increases to as much as 20–30% towards the north of the equator and particularly, the ITCZ at 7°N8. Beyond these model projections, there have yet to be any studies to 1) corroborate the presence of North African dust in the EEEP and 2) evaluate the proportions of Northern Hemisphere (North African) vs. Southern Hemisphere (South American)-sourced dust delivered to the region during the last 30 kyr. This is best achieved through direct measurements of the isotope geochemistry of the fine-grained dust fraction of marine sediments, which is the aim of this paper.

Here we present new high-resolution records of detrital Nd and Pb isotopes from site MV1014-02-17JC (00° 10.8297’S, 85° 52.0042’W; see Fig. 1) in the Panama Basin. High-resolution dust fluxes to this core have already been reconstructed9. Here, working on the same samples analyzed for dust fluxes in Loveley et al.8, we reconstruct the provenance of the dust by compiling a database of the most likely potential dust source areas (PSAs) from around the globe based on available global projections, prevailing wind trajectories, and existing geochemical data (see “compilation of dust source data” section in Methods). We then execute two isotope (206Pb/204Pb-εNd and 208Pb/204Pb-εNd) Bayesian mixing envelope analyses and mixing models (MixSIAR)10,11,12 with the Pb and Nd isotopic data from 17JC and our PSA database to identify the most geochemically probable PSAs and quantify their contributions to our site’s dust load over time. We utilize these model results to constrain meridional shifts in the ITCZ for the EEEP over the past 30,000 yr. The sub-millennial resolution of our records allows us to evaluate dust provenance and ITCZ change over abrupt climate events since the Last Glacial Maximum (LGM; 19–21 ka), at Heinrich Stadial 1 (HS1; 14.6–18 ka13,14), the Bølling-Allerød (B/A; 14.6–13.6 ka15), the Younger Dryas (YD; 13.6–11.7 ka15,16), and the Early Holocene/African Humid Period (EH/AHP; 11–5 ka17) in the equatorial Pacific. Importantly, because the timing of the shifts has been suggested to be related to changing patterns of export production in the EEEP7,9,18, clues to the relationship between the sources of dust delivery and atmospheric CO2 levels may be discerned.

Fig. 1: Location map for area of study.
figure 1

ΔpCO2 [seawater-air] flux from Takahashi et al.48 with location of study site MV1014-02-17JC (red star) and the maximum (September) and minimum (March) mean seasonal positions of the Intertropical Convergenze Zone (ITCZ, black dashed lines)25. Figure modified from Schimmenti et al.7. Source data are provided as a Source Data file in the supplemental section.

Full size image

Results and discussion

Dust sources to the EEEP over the past 30,000 years

In their reconstruction of dust fluxes over the last 30,000 years at this equatorial site, Loveley et al.9 found that, on average, during the last glacial, dust fluxes were higher than those in the Holocene by a factor of about two. Punctuated throughout this longer-term climatic relationship were pulses of high dust fluxes during HS1, and at the LGM (~20 kyr). Dust fluxes decreased during the B/A but then stalled during the YD, until continuing their decrease again during the EH/AHP (lowest dust fluxes measured in the record). Using Pb and Nd isotope ratios, our goal here is to determine the sources of the dust delivered to the EEEP at our site during each of these climatic intervals.

There is considerable variability in detrital 206Pb/204Pb, 208Pb/204Pb, and εNd over the past 30 kyr at 17JC (Fig. 2). We note 206Pb/204Pb and 208Pb/204Pb are most radiogenic (highest) during the LGM and

Fig. 2: Detrital Pb and Nd isotope signatures at 17JC versus time over the past 30,000 yrs.
figure 2

a (top panel) 206Pb/204Pb, b (middle panel) 208Pb/204Pb, c (bottom panel) εNd. Cold periods (Last Glacial Maximum-LGM, Heinrich Stadial-HS1, Younger Dryas-YD) are highlighted with gray bars and warm periods are highlighted with yellow bars (Bolling/Allerod-B/A and Early Holocen/African Humid Period-EH/AHP). Lines through data points represent error bars (see Methods). Black data points in panel (c) represent replicate measurements of εNd (Supplementary Data Table 1). Source data are provided as a Source Data file in the supplemental section.

Full size image

HS1 and least radiogenic at the YD and pre-HS1. For εNd, values are least radiogenic (more negative) during the LGM, HS1, and YD and more radiogenic during the B/A and Early Holocene. Dust trajectory models indicate that the two major dust sources to this corner of the EEP today are South America and North Africa (e.g., refs. 8,19). The range of 206Pb/204Pb and 208Pb/204Pb at 17JC (Supplementary Data Table 1) is consistent with a South American dust source endmember to the equatorial Pacific as has been established in the CEP and further west of our site at 110°W2,4,20. These young volcanic sources cannot, however, explain the significantly unradiogenic εNd values in the 17JC detrital record (Supplementary Data Table 1), which are indicative of mixing with older continental material. The two candidates for this older continental material are North Africa and Asia. Models consistently rule out the ability for Asian dust sources (e.g., Chinese loess) to reach east of 110°W8. Instead, we suggest here that the older continental material represented by detrital unradiogenic Nd isotope ratios (and radiogenic Pb isotope ratios) are derived from North African dust sources most distinctly during HS1 and the early YD.

To determine the predominant mixtures of dust sources to 17JC, we first consider and compile all possible North African and South American PSAs to the EEP21,22,23,24,25,26 (see Methods for complete description of PSAs). We find that 17JC detrital data are best captured in 206Pb/204Pb - εNd and 208Pb/204Pb - εNd space by 11 PSAs from across North Africa26 and the Andean margin of South America21,22,23,24,25 that include: the West African Coast (PSA-WAC), Mali Center (PSA-MC), the Bodele depression (PSA-BD), Libya Algeria Mali (PSA-LAM), Libya Egypt (PSA-LE), North, Middle and Southern Central Western Argentina (CWA), the Puna, the Altiplano, and the Paracas High (data for each PSA are reproduced in the Supplementary Data Table 7). After choosing the PSAs we use the Bayesian isotope mixing model, MixSIAR10,11, to justify our choices (see Methods). Briefly, we first statistically validate that our 17JC Pb and Nd isotope data are enveloped by the mean Pb and Nd isotope signatures of the PSAs (Supplementary Fig. 1). We run Bayesian mixing polygon simulations of the 206Pb/204Pb - εNd and 208Pb/204Pb - εNd systems with data from the 11 PSAs and the 17JC detrital data11,12 (see Methods and Supplementary Figs. 1 and 2). On average, we find that there is a 95% chance that the 17JC detrital data fall within the two-dimensional mixing polygon formed by the PSA dataset in 206Pb/204Pb – εNd space and about an 80% chance within 208Pb/204Pb – εNd space (Supplementary Fig. 2). We proceed with the construction of our isotope mixing models based on our analysis that, in these two isotope systems, we have captured the majority of the endmembers for the full 30,000-year detrital record at 17JC.

To quantify the contributions of each of the 11 PSAs from South America and North Africa to the dust load at 17JC through time, we construct two two-isotope (206Pb/204Pb – εNd and 208Pb/204Pb – εNd) Bayesian mixing models with MixSIAR (see Methods). The results of the 206Pb/204Pb – εNd model are shown in Figs. 3 and 4. The results of the 208Pb/204Pb – εNd model, which are virtually the same, are shown in Supplementary Figs. 5 and 6. We find that the majority of the dust load at 17JC over the past 30,000 years can be attributed to contributions from South American PSAs (~75% on average; Fig. 3), with smaller contributions from North African PSAs (~25% on average; Fig. 3). The major South American sources over our record are the Paracas High and broadly, CWA, with lesser contributions from the Puna and Altiplano (Fig. 3a), whereas the relative contributions of the five North African PSAs are quite similar on average with larger variations occurring during discrete climate events such as HS1 and the YD (Fig. 4; a posteriori climate event grouping described further below).

Fig. 3: Provenance of 17JC dust in each sample over the past 30,000 years.
figure 3

Results from the 206Pb/204Pb- εNd MixSIAR model. a % South American dust vs. time, color-coded by PSA. b % North African dust vs. time, color-coded by PSA. Sample ages are plotted on a discrete scale on the x-axis. Source data are provided as a Source Data file in the supplemental section.

Full size image
Fig. 4: Provenance of 17JC dust during each climate event of the past 30,000 years.
figure 4

Results from the 206Pb/204Pb- εNd MixSIAR model. a % South American dust vs. time, color-coded by PSA. b % North African dust vs. time, color-coded by PSA. MIS 2 Marine Isotope Stage 2, pre-LGM, LGM Last Glacial Maximum, Dgl Deglacial, pre-HS1, HS1 Heinrich Stadial 1, B/A Bølling-Allerød, YD Younger Dryas, EH Early Holocene. Source data are provided as a Source Data file in the supplemental section.

Full size image

To more fully consider the relationship between millennial climate events and changes in dust sources to the EEEP, we performed a posteriori grouping of the model results by climate event (see Methods). We note significant and opposite variations in the contributions of South American vs. North African PSAs during abrupt cooling (HS1 and the YD) and warming events (the B/A and EH/AHP) of the deglaciation and early Holocene (Fig. 4). Furthermore, the contributions of the South American PSAs decrease and those of the North African PSAs increase from the onset of the glaciation during Marine Isotope Stage (MIS) 2 through the LGM and early deglacial, reaching a trough/peak (~30:70, North Africa:South America), respectively, at HS1 (Fig. 4). At HS1, the majority of South American dust is sourced from the Paracas High and the MCWA, with lesser contributions from the Altiplano, NCWA, and SCWA (Fig. 4a), whereas the majority of the North African dust is sourced from the Bodelé depression, Libya-Egypt, and Mali center with lesser contributions from Libya-Algeria-Mali, and the West African Coast (Fig. 4b). At the B/A, the proportion of South American vs. North African dust increases again (~80:20). There is a marked increase in the contribution of NCWA, smaller increases in that of the Paracas High and SCWA, slight decreases in the contributions of the Altiplano, and no significant change in the contribution of the Puna or MCWA PSAs (Fig. 4a). Contributions of all the North African PSAs decrease at the B/A, such that the largest contributions come from the Bodelé depression and Libya-Egypt with lesser contributions from PSA-MC followed by PSA-LAM and WAC (Fig. 4b). The proportion of North African dust to South American dust increases yet again at the YD (~30:70). Relative to the B/A, the contributions of PSA-WAC and PSA-LAM double, PSA-MC increases by 1.5-fold, and PSA-BD and LE remain largely the same (Fig. 4b), whereas the contributions of the Puna and NCWA decrease and that of the Altiplano and SCWA increase slightly (Fig. 4a). Finally, the onset of the EH and AHP are marked by elevated proportions of South American vs. North African dust (~85:15). This time, all North African PSA contributions, except for Libya-Egypt, are reduced to relatively equal proportions (Fig. 4b). The contributions of SCWA, MCWA, and the Paracas High increase at the onset of the EH whereas that of the Puna, Altiplano, and NCWA decrease slightly or remain constant (Fig. 4a).

Millennial migration of the mean position of the ITCZ in the EEEP

Today, models suggest that ~20–25% and ~75–80% of the eolian dust deposited in the EEEP near 17JC comes from North Africa and South America, respectively8. We have detrital Pb and Nd isotope ratio observational data to support these models. South American dust dominates the dust load to the equator because the ITCZ sits north of it and acts as an effective barrier to African dust (Figs. 3 and 4). Using our Nd-Pb isotope ratio mixing model results, from 30 kyr to the present, the proportion of dust from Africa increases and culminates with a significantly higher proportion during HS1. At the B/A, the proportion of African dust supplied to our site decreases before rapidly increasing again during the YD. After the YD, the proportion of African dust retreats to its lowest values in the early Holocene/AHP (Figs. 3 and 4). Each of these intervals of climate change associated with changes in the proportion of African to South American dust delivery likely signifies a considerable migration of the ITCZ since the onset of the last glaciation 30,000 years ago.

We see evidence for a potential southward shift of the mean ITCZ relative to today at the LGM. During the LGM, 17JC dust is a mix of dust sourced from African and South American sources at an approximate proportion of 22:78 (Fig. 4). This contrasts with the lower proportions of African dust (relative to South American dust) delivered to our site during the Holocene (~15:85) (Fig. 4). This is in agreement with modeling studies6,27, dust provenance records1, and SST reconstructions28,29 over both the Atlantic and Pacific Oceans. This is also consistent with evidence from 17JC for reduced upwelling and therefore silica delivery/diatom productivity during the LGM due to the ITCZ shifting south, closer to the equator thus dampening cross-equatorial flow of the trade winds7. Notably, the proportion of African to South American dust deposited at 17JC does not appear to change after the LGM during the early deglaciation, suggesting the ITCZ may not have moved significantly from its LGM position. Hypotheses concerning changes in the mean position of the ITCZ in the Pacific Ocean, as shown here and by others1,2,3,4,7,9,18,20, is in stark contrast to the Atlantic Ocean where there is little evidence for a shift in the mean position of the ITCZ on glacial-Holocene timecales30.

At HS1 there is about a 30–40% increase in the proportion of African to South American dust relative to both the LGM and deglacial (ratio increases from about 22:78 to 30:70), indicative of a further southward shift of the ITCZ. Evidence for enhanced upwelling at 17JC7 and strengthened NE/ weakened SE trade winds27 support the possibility of the ITCZ being further south during HS1 than during the LGM. Further, model simulations of the ITCZ, produce zonal mean and regional EEP ITCZ shifts that are greater under Heinrich Stadial-like forcings than LGM-like forcings6. The EEP ITCZ, specifically, exhibits southward shifts as large as 10° of latitude that are reproduced across all models for HS events; and most of the zonal mean shift of the ITCZ during HS events is due to shifts of the ITCZ over the CEP and EEP. We, therefore, favor a southward shift of the average position of the EEP ITCZ of at least 7° to a position south of the equator to produce the significant increase in African dust to the dust load that we observe at the equator at 17JC during HS1.

During the B/A, the proportion of African to South American dust delivered decreases to LGM levels (~20:80), suggesting a northward migration of the ITCZ. Given the similar proportions for LGM dust, we expect that the average latitudinal positions of the B/A and LGM ITCZs were similar. This represents an improved constraint on the relative magnitude of ITCZ shift between HS1 and the B/A over the EEP, which was previously unclear from existing high-resolution records of Pacific ITCZ migrations over this timeframe7,31.

At the YD, the proportion of African to South American dust is once again elevated (~30:70), suggesting a significant southward shift in the ITCZ like that interpreted for HS1. Similar to HS1, a strengthening of the NE trade winds/SE trade winds occurred with cooling and ITCZ displacement at the YD32. Upon closer examination, one might conclude that the smaller ratio of African to South American dust relative to HS1 at the YD (Fig. 4) indicates a smaller southerly displacement of the YD ITCZ relative to the HS1 ITCZ. However, there is evidence from the West African margin that this difference can be explained by latitudinal differences in peak northeasterly wind anomalies over North Africa between these two events16. In particular, the wind anomalies peak south of 20°N during HS1 and north of 20°N during the YD16.The dust provenance record at 17JC may be capturing these differences. At HS1, the increase in African dust is primarily driven by contributions from PSA-MC and PSA-BD, both of which lie south of 20°N, whereas at the YD peak African dust proportions are driven by elevated contributions from PSA-WAC and PSA-LAM (these increase while the contributions of PSA-BD, MC, and LE all decrease relative to that at HS1), both of which extend north of 20°N. We cannot unequivocally decipher relative differences in ITCZ position during the YD versus HS1 at this time, but our results are consistent with changes in the latitudinal placement of North African wind anomalies16.

Finally, during the Early Holocene, the lowest proportion of African to South American dust (~15:85) suggests the northernmost displacement of the ITCZ in our record. This is in agreement with evidence from 17JC for enhanced upwelling and silica delivery/diatom productivity (Pa/Th ratios, Fig. 4a in ref. 7) with strengthening of the SE trade winds and northward displacement of the ITCZ over the EEEP driven by peak boreal summer insolation at ~10 ka7. The reduction in African dust is also consistent with evidence for intensification of the African monsoon and therefore enhanced precipitation and decreased aridity of the Sahara associated with the AHP.33

To summarize, our observation-based Bayesian isotope mixing model suggests that dust delivered over the past 30,000 yr to the EEEP oscillates between delivery of varying proportions of dust sources from North Africa and South America. The shifts in the proportion of dust supplied from each continent can be explained by meridional shifts of the ITCZ in the eastern Pacific Ocean across distinct climate events. Indeed, our work suggests southward shifts of the ITCZ during cold periods (LGM, HS1, and YD), with the most extreme shift of at least 7° during HS1 when the most significant proportion of African dust is apparent. Conversely, northward shifts of the ITCZ during warm climate events of the EH/AHP and B/A can be discerned based on the highest proportion of South American sources of dust delivered during these times. Overall, our radiogenic Pb and Nd isotope data on dust supply source changes provide the strongest evidence to date for significant migrations of the ITCZ over the EEEP during abrupt climate change events.

Methods

Site and sampling

Deep-sea sediment core MV1014-02-17JC (17JC) was retrieved from south of the Carnegie Ridge in the Panama Basin (00° 10.8297’S, 85° 52.0042’W; Fig. 1) at a water depth of 2846 m; just above the lysocline in this region34. The average sedimentation rate at 17JC is 18 cm/kyr9 corresponding to an age resolution of ~60–500 yr for an average sampling resolution of 3 cm downcore.

Age model

Using the Bayesian statistical package for age depth modeling ‘rbacon’35 in R we extended the age model for the upper core (355 cm) at 17JC from Schimmenti et al.7 to the full core record (930 cm) with a combination of 21 radiocarbon ages (N. duetertrei)7,9, 5 tie points between Heinrich Stadial (HS) events in the 17JC 230Th MAR record9 and Hulu Cave36, and ash from the Los Chocoyos eruption (75 ± 2 kyr)37 identified at 853 cm in core 17JC (Supplementary Data Table 6). We ran bacon with a section thickness of 10 cm and an average accumulation rate of 90 yr cm−1. Radiocarbon ages were calibrated to calendar year using the Marine20 calibration curve and its built-in 500-year reservoir age correction in bacon. Four thousand Markov Chain Monte Carlo (MCMC) iterations yielded a Gelman and Rubin Reduction Factor of 1.025, below the 1.05 safety threshold, indicating robust MCMC mixing for our model. Calibrated age errors were calculated by the model based on a 95% confidence interval, which yielded an average uncertainty for the calibrated radiocarbon ages of +469 years/−485 years and an average age model uncertainty of +683 years/−700 years (Supplementary Data Table 6) for the period of 5–30 ka.

Isolation of the detrital Nd and Pb

We measured the Nd and Pb isotopic composition of the operationally defined detrital fraction (the non-biogenic and non-authigenic component) of 17JC sediment, to best capture the signature of fine-grained eolian dust deposited over time at our site. To isolate the operationally defined detrital fraction, bulk sediment samples were wet-sieved to collect the <63 µm fraction, then sequentially leached to remove carbonates, Fe-Mn oxyhydroxide coatings, and biogenic silica per the following procedure. ~500 mg of sediments (<63 µm) were dried down and homogenized, then sequentially leached with 10% buffered acetic acid, 0.02 M hydroxylamine hydrochloride with 25% acetic acid, and at 80 °C in a water bath with 2 M KOH. The detrital fraction was digested using a series of trace metal grade or distilled 8 N HNO3, concentrated HF, concentrated H2O2, and aqua regia additions. Pb was eluted on anion exchange (AGIX8, 100–200 mesh) columns. Nd was eluted, after being washed off the Pb columns, through additional RE Spec (Eichrom) and Ln Spec (Eichrom) columns.

Nd and Pb isotope analysis

Nd and Pb isotope ratios were measured by Multi-Collector Inductively Coupled Mass Spectrometry (MC-ICP-MS) on a ThermoFisher Neptune Plus MC-ICP-MS in the Williams Radiogenic Laboratory at Texas A&M University. Approximately 200 ng Pb and Nd were analyzed per sample. Pb isotopes, expressed as the ratio of the radiogenic isotope (206, 207, or 208Pb) to the stable isotope (204Pb), were first corrected for instrumental mass bias using the exponential law through Tl addition. Samples and the NIST SRM 981 standard were spiked with 10 ng of Tl (NIST SRM 997). Data were internally mass-fractionation corrected using a 205Tl/203Tl ratio of 2.3889 based on the certified ratio of NIST SRM 997. An additional correction for sample Pb isotope ratios was then performed by bracketing every fourth sample with the NIST SRM 981 standard (and correcting to 206Pb/204Pb = 16.9406, 207Pb/204Pb = 15.497, 208Pb/204Pb = 36.7184; Yuan et al.38).

Instrumental mass bias for Nd was corrected by first normalizing internally to a 146Nd/144Nd of 0.7219 after correcting for any Sm interferences. Sample 143Nd/144Nd ratios were then corrected by bracketing every three samples with the JNdi-1 standard (143Nd/144Nd = 0.512115 +/− 0.0000007; Tanaka et al.39). Nd isotopes ratios were normalized to the Nd isotopic composition of bulk earth obtained from chondrites by the following Eq. (1)40:

$${{\rm{\varepsilon }}}{{\rm{Nd}}}=\left({\frac{{143}_{{Nd}}}{{144}_{{Nd}\,}}}_{{sample}}\Bigg/{\frac{{143}_{{Nd}}}{{144}_{{Nd}\,}}}_{{chondritic}}\,-1\right){{\rm{\times }}}{10}^{4}$$
(1)

Both NIST SRM 981 and JNdi−1 standards were analyzed within the same concentration range as the sample aliquots (5–25 ppb). 10 replicate samples were analyzed for Pb and Nd. The average reproducibility (2σ standard error) of the replicates for Pb were 0.03, 0.003, and 0.01 for 206Pb/204Pb, 207Pb/204Pb, and 208Pb/204Pb, respectively. Average external reproducibility (2σ standard error) of the JNdi-1 standards was 0.19 ε units. Procedural blanks were measured such that blank corrections were unnecessary.

Compilation of dust source data

Major dust sources to the EEP were chosen from the eight main global dust source regions following the work of Kok et al.8. Their model showed that the two main dust sources to the EEP east of the Galapagos today are South America and North Africa. Prospero et al.41 found that the main dust source areas on the South American continent broadly include the Bolivian Altiplano, Western Argentina, and Patagonia. More recent work by Gili et al.21,22 and Briceño-Zuluaga et al.42 refined the list of South American Potential Source Areas (PSAs) to five arid regions of the Andean margin extending from southern Peru to the tip of the continent: the Paracas High (Paracas; ~13°–15°S), the Puna Altiplano Plateau (PAP; ~15°–27°S), Central Western Argentina (CWA; ~27°–38°S), Patagonia (~38°–52°S) and Tierra del Fuego (~53°–55°S). Based on prevailing winds, we limited South American PSAs to 17JC to those north of 38°S, as those south of this boundary should not physically reach the EEP and therefore should not contribute significantly to the South American dust budget at 17JC. This has been further corroborated by dust provenance studies that have shown Patagonia and Tierra del Fuego are significant sources of dust reaching the Southern Ocean and Antarctica21,22,43 and not the equatorial Pacific15.

Nd and Pb isotopic data for the Paracas High and PAP were obtained from Gili et al. (2016; 2017)21,22 and by filtering Central Volcanic Zone rock data from Mamani et al.23 to the latitudinal and longitudinal ranges of the Paracas, Puna, and Altiplano, respectively24. Puna and Altiplano data from Mamani et al.23 were further filtered by sample/rock type to only include ignimbrites (South Puna and South Altiplano) and gneisses (North Puna) in order to best capture the isotopic signatures of the deflatable material from these PSAs, following the work of Gili et al.21,22.

Nd and Pb isotopic data for CWA were obtained from Gili et al.21,22 and from the GEOROC database24 (https://georoc.eu/georoc). Data from the latter database were downloaded on 16 May 2023 using the following parameters: geography = latitude 27°–38°S; longitude 65°–70°W: All samples: All types of material. This yielded mostly volcanic whole rock data from the Central and South Volcanic Zones (CVZ; SVZ) of the Andes. Gili et al.21,22 defined three regions of the CWA based on the rare earth and isotopic composition of deflatable material found there: the North CWA (NCWA; ~27°–29°S), the Middle CWA (MCWA; ~29°–33°S), and the South CWA (SCWA; ~33°–36°S). NCWA and MCWA deflatable material are most similar in Nd and Sr isotopic composition to granitoids and metamorphic rocks of the Sierras Pampeanas whereas that of deflatable material from the SCWA is most similar to rocks of the Northern South Volcanic Zone (NSVZ;~33–36°S)32. Unfortunately, there is no Pb isotopic data available for Sierras Pampeanas granitoids and metamorphic rocks or NCWA/MCWA sediments. Consequently, we filtered our GEOROC CVZ/SVZ volcanic rock dataset to best match the latitudinal ranges and/or source areas of the CWA regions defined by Gili et al.21,22, and which we named the following: NCWA_Volcanics (27°–29°S), Sierras Pampeanas (SP)_Volcanics (~31°S), and NSVZ-SCWA (33°–36°S).

North African PSAs and their isotopic signatures were taken from Guinoiseau et al.26. These authors divided the PSAs into several groups which include Libya-Algeria-Mali (PSA-LAM), Libya-Egypt (PSA-LE), the Bodélé Depression (PSA-BD), Mali Center (PSA-MC), and the West African Coast (PSA-WAC). Outliers in 206Pb/204Pb, 208Pb/204Pb, and εNd were identified for each South American PSA using Rosner’s test in the EnvStats package in R12 and removed from their respective datasets. Outlier identification and removal was performed by similar means for Pb and Nd isotope data of the North African PSAs by Guinoiseau et al.26.

It is important to note that the PSA data from the literature may not represent perfectly the dust emitted from these areas and delivered to the eastern Pacific. For example, for Pb isotopes grain size sorting during dust transport may marginally influence the Pb isotope ratio26. Nd isotopes are less likely to undergo a grain size fractionation effect during transport44. When possible, we used only samples for which both Pb and Nd isotope ratios were measured for the detrital deflatable material in the source region. Although this was not always possible, our attempt at creating a database of PSAs is a good first estimate or starting point given the lack of detrital source data from the South American continent.

Isotope mixing models

The MixSIAR package in R10,11 was used to estimate the proportions of dust source contributions to the 17JC dust load through time, following the work of Longman et al.45 Dietrich et al.46 and Lopez et al.47 on environmental Pb source modeling with Pb isotope ratios. Bayesian mixing polygon simulations10,11 were first run in R to determine which PSAs and which combination of Pb and/or Nd isotope ratios (206Pb/204Pb, 207Pb/204Pb, 208Pb/204Pb, εNd) best explained the isotopic signature of 17JC data. The results of these simulations were then used to construct two two-isotope models using 206Pb/204Pb and εNd and 208Pb/204Pb and εNd. Summary statistics (means, variance, sample size) were used for source data input in the model as the source isotopic datasets were found to be neither consistently univariate nor multivariate normal for all PSAs (results of Shapiro-Wilk, Madia, Henze-Zirkler, and Royston statistical tests; see Supplementary Data Tables 2–5). All model iterations were run with no continuous effects, no concentration dependence, and no trophic enrichment factors (means and standard deviations equal to zero). Further, all models were run with residual error only and no process error. For iterations that evaluated source contributions over the whole 17JC record, 17JC sample age was used as a fixed effect for the mixture data. Dust contributions during distinct climate events of the past 30,000 years were obtained by grouping the model results of dust contributions over the entire core record a posteriori and calculating the average contribution of each source during each climate interval. MCMC chain length of each model iteration was optimized to reach convergence based on the results of post-run diagnostic tests offered in MixSIAR, the Gelman-Rubin and Geweke tests. For most models, we used the “very long” (chain length = 1 million) or “extreme” (chain length = 3 million) MCMC run options in MixSIAR. We report results from models for which all MCMC chains passed the Gelman–Rubin test (Gelman–Rubin reduction factor <1.05) and at least 2/3 chains passed the Geweke test (z-score = +/− 1.96) (see Supplementary Data Tables 2–6).