Mineral dust is considered a sensitive natural tracer of modern and past environmental conditions and a paleoclimatic proxy that reflects the variations in the atmospheric circulation during different climatic periods and their transitions. The type and intensity of dust emissions are a consequence of different conditions at the source (sediment supply and availability) and aeolian processes (deflation, transport, deposition)1. Once dust is emitted into the atmosphere, it remains in suspension until physical and chemical conditions favor its deposition. The deposited dust is archived in different proxy records either, over the oceans or onto continental land-masses within ice cores, deep-sea sediments, loess, peat bogs, etc. For the Southern Hemisphere, reconstructions of dust buried in Antarctic ice cores are considered an excellent proxy for the paleo-record of wind intensity and circulation. In this regard, these records indicated that dust fluxes were up to ~25 times higher during glacial periods due to dust production strengthening at their source areas, compared to warmer periods when dust concentrations were extremely low2.

The term Potential Source Areas (PSAs) denotes either a primary source of mineral dust, derived directly from the physical and/or the chemical weathering of the parent material or a secondary source for dust, where airborne particles have already been subjected to a phase of aeolian and/or fluvial transport3. For the Southern Hemisphere, Southern South America (SSA), Australia (AUS), New Zealand (NZ), Southern Africa (SAF), and the Antarctic Dry Valleys have all been identified as PSAs. The first study to trace the dust origin in Antarctica identified SSA as the source area from which dust from the Last Glacial Maximum (LGM) section of the Dome C ice core was derived4. In agreement with the Sr-Nd isotope data presented in that study, all the studies since then have demonstrated that the main dust contributor to the Southern Ocean (SO) and East Antarctica (EA) during glacial (MIS2-MIS20) and interglacial periods was SSA, mainly Patagonia3,5,6,7,8,9,10. Despite the agreement on the dominance of a Patagonian dust origin during glacial periods, there is currently a debate regarding the Holocene dust PSAs with the following scenarios of dust provenance proposed: (i) a weakened Patagonian source complemented by a lower latitude source region (Southern Central West Argentina, SCWA7,10,11); (ii) a Puna–Altiplano dust source transported via the southern tropical jet stream7,9,10,12 and, (iii) a general contribution from Southern South America plus a secondary input from AUS13,14,15,16,17 or NZ17. Some of these geochemical and mineralogical interpretations were partially supported by Atmospheric General Circulation Models (AGCMs), suggesting that while SSA dust dominates deposition over the Atlantic Southern Ocean and East Antarctica, AUS modern dust prevails over the Pacific SO and towards the western coastal regions of Antarctica18,19,20. Additionally, two important recent studies reviewed and evaluated the role of AUS as a dust source during modern and Late Quaternary times showing that Australian dust likely reached certain sectors of Antarctica, such as Taldice and Taylor Glacier, predominantly during interglacials periods21,22. For NZ, back-trajectory modeling showed that the transportation of aeolian material from this region to the southern high latitudes is possible23. However, a recent study carried out on sediments from the major present-day dust and sediment producing regions of the South Island in NZ, has indicated this area was not a significant dust supplier to EA24.

AGCMs have shown that after SSA and AUS, Southern Africa is the third main dust supplier in the Southern Hemisphere with dust depositions mainly taking place in the South Atlantic and Indian Oceans18. Even though this region is a significant dust emission source, little is still known about it being a Potential source area as the geochemical and modeling data available for this region is extremely sparse. Thus, it is questionable if the role of SAF as a dust contributor to the SO and EA has plausibly been underestimated. For many years, only the Makgadikgadi Pan in Botswana and the Etosha Pan in Namibia were recognized as persistent dust sources in SAF18,25,26,27, neglecting the coastal Namibian sources. In fact, it was only until recently that the Namib Desert coast has been identified as one of the largest sources of dust in Southern Africa28,29. Using remote sensing satellite images, these authors have found that along the southwestern coast of SAF, the Kuiseb, Omaruru, and Huab Rivers beds are the three main areas of greatest dust emission located on recently deposited fluvial surfaces. Revisiting the literature, it became apparent that while most of the studies have been focused on the detailed study of SSA, AUS, and most recently NZ dust sources9,10,15,21,22,24, other PSAs, such as SAF, have been overlooked. Thus, there arises a compelling need for more extended geochemical studies to fill the current gap in understanding the past and present long-range transportation of mineral dust particles to the high-latitude environments.

Responding to this need, this study aims to investigate the role of SAF as a PSA and provides the first comprehensive north-to-south geochemical characterization of riverbed sediments collected along the major dust-producing river systems in the Namib desert (Huab, Omaruru, Kuiseb and the Namibia Sand Sea) (Fig. 1a). It is nevertheless noteworthy that many of these river catchments have been (and still are), subject to extreme anthropogenic activities, mainly because the alluvial deposits of the rivers hold the groundwater that supports urban and industrial development in the region30. Such activities can substantially modify these hydrographic systems affecting the natural dust emission sources and increasing, as a consequence, the dust production in certain areas (see more details in Supplementary Note 1).

Fig. 1: Geographical and climatological characterization of Southern Africa Potential dust source areas.
figure 1

a satellite image of dust plumes originated off the coast of Namibia captured in May 2013 by The Moderate Resolution Imaging Spectroradiometer (MODIS) on NASA’s Terra satellite showing the study areas of Huab, Omaruru, Kuiseb and the Nambia Sea Sand. The hydrographic Huab, Omaruru and Kuiseb systems are indicated by the blue lines. b map of southwestern Africa showing the major features of the regional climate system. Main oceanic currents are indicated in bold green arrows and atmospheric circulations in red (south easterly trade and westerlies winds) and yellow (Berg winds). BC: Benguela Current; ACC: Antarctic Circumpolar Current; AC: Aghulas Current. The map in Fig. b was built using Google Earth Pro, Image Landsat/Copernicus; Data: SIO, NOAA, U.S Navy, NGA, GEBCO, 13 December 2015.

The Namib desert extends N-S for approximately 2000 km along the western coast of southern Africa between the Kunene and Orange Rivers31. This massive and arid region of SAF, consists of a complex system of different geomorphological structures (deltas, ephemeral rivers, pan complexes, dune fields, gravel plains) and each of these structures compromises several landforms (floodplain terraces, river channels, nebkhas, basins, margins of playas and sabkhas, stone pavement) that have proven to be accountable for sediment entrainment and wind erosion29,32. The eastern border for the Namib desert is delimited by the Great Escarpment (~2000 msal) running parallel to the coast and 100–150 km inland. Traversing the escarpment, between the Kunene and the Orange Rivers, twelve aligned ephemeral river valleys extend sub-parallel east-west along the Atlantic Coast30. The rainfall in the headwaters of these river catchments (250–350 mm/year) generates floods that can last a few days flowing through the Namib Desert eastward towards the coast. However, because that rainfall becomes insignificant towards coastal areas (<25 mm/year), the limited floods only reach the ocean occasionally33. Thus, the water that dissipates due to infiltration and evaporation along with a decrease in vegetation, makes the deposited sediments at the terminal stages of the river valley channels available for aeolian entrainment and transport. In addition, the presence of erosion agents, such as salting sands from the Namib Sand Sea region, sediment mineralogy and grain size, vegetation roughness, and soil moisture are all parameters responsible for controlling the dust emission rate along the SAF coast34. Anthropogenic disturbances of certain dust production areas are also worth mentioning. Especially the hydrological systems of Omaruru and Kuiseb, where the increased demand for water extraction and diversion in combination with population growth has played a role in altering some of the main landforms29,35. The arid and stable climate of the Namibian coast is produced by the quasi-stationary subtropical South Atlantic High Pressure (anticyclone) system and the cold northward-flowing Benguela current36 (Fig. 1b). This upwelling system is wind-driven by the regional alongshore Southeast trade-winds (Fig. 1b) and is recognized for being a region with the highest rates of primary production in the ocean where dust with high nutrient content deflated from the continent is discharged offshore fertilizing the ocean waters37. Although minimum annual variations in wind intensities occur, the Benguela upwelling system is often separated into three zones; north, central, and south38. The north and south Benguela upwelling zones are respectively bounded by the Angola and Agulhas Currents. The central zone presents the least fluctuating regime and is the most relevant zone for our research as it lies contiguous to the main dust source region investigated here between 22° and 28°S. Another important atmospheric agent of aerosol transport and deposition off the Namibian coast are the so-called Berg winds (Fig. 1b). These are katabatic winds, generally blowing from the east and northeast, supplying large amounts of aeolian dust into the SE Atlantic Ocean due to their high velocities and subsidence offshore39. Common from April to August, associated with high daily evaporation rates and low monthly fog–water precipitation values, the Berg winds are locally intensified by topographic features such as river valleys and are shown to be responsible for >60% of the offshore sediment transport with dust plumes extending ≥100 km into the Southern Ocean. In addition, with an estimated annual dust input of 4.4 tons into the South Atlantic Ocean, these Berg winds are in charge of supplying nutrients to the highly fertile Benguela upwelling system28,40.

In this work, we combine rare earth elements (REE) concentrations and the Sr-Nd-Pb isotope compositions to evaluate the contribution of Southern African dust to the high latitude environments of the Southern Hemisphere. This is, to the high elevation sites of East Antarctica with a special focus on the coastal sites and the Atlantic sector of the Southern Ocean. We compare the geochemical signature of bulk and resuspended aerosol fractions from SAF, with the dust fingerprint reported for Antarctic ice cores (Vostok and EDC) during glacial (MIS2-MIS20) and interglacial periods (Holocene and MIS 5.5) as well as present-day dust from Berkner Island41. In addition, we also include a re-analysis of the isotopic composition of SAF samples previously reported in3,8. The results presented will demonstrate that the SAF dust signature can be an important additional end-member in explaining the dust fingerprint found in the South Atlantic Ocean and coastal regions of East Antarctica.


Geochemical characterization of Southern Africa potential source areas

The complete REE concentrations and Sr-Nd-Pb isotopic composition database of the samples presented in this study are presented in Supplementary Data 1. The Southern Africa PSAs signature of the windblown samples presented in this study reflect the compositional variability of the locally outcropping rocks and sediments: a chemical and isotopic fingerprint that is the result of the mixing between the Damara Orogenic system including 2.0–1.2 Ga gneissic basement overlaid by Neoproterozoic to Paleozoic metasediments intruded by 840–460 Ma granitoid plutons, and the Orange River sediments42,43 (see more details in Supplementary Note 1 and Supplementary Fig. 1).

In Fig. 2, REE patterns for the Southern Africa PSAs normalized to the upper crust (UCC44) are plotted. This figure indicates that no significant difference in the REE composition prevails among all four areas characterized in this study and a regional differentiation is difficult. In general, samples show flat REE patterns with a minor decrease in the heavy REE abundances (Fig. 2a). Note that for the Omaruru catchment only the REE data for three out seven samples are available. Both the bulk and the fine fractions from Huab and Kuiseb exhibit Eu/Eu* < 1, with Omaruru displaying slightly higher values (Fig. 2b) for this ratio. For the Namib Sand Sea region, only one data point for a bulk sample is available in terms of REE and its composition is similar to the Huab area. The Eu anomaly (Eu/Eu*) was calculated with the equation Eu/Eu* = ¼ [Eu]/(([Sm] + [Gd])/2). Figure 2b also shows that the REE composition of the bulk and the fine fractions from the Huab area are similar, suggesting bulk samples are representative of the fine particles and that Southern Africa REE signature is independent of the grain-size fraction of sediments. For comparison purposes, soil samples from the Walvis Bay in Namibia have also been included (Fig. 2a, b), whose source has been attributed to the local dust blown from the nearby Namibia Sea Sand45.

Fig. 2: Rare Earth Elements (REE) data of surface sediments representing the Potential source areas of Southern Africa.
figure 2

a Mean REE patterns. b Relationships between Eu/Eu and La/Ybucc. REE data has been normalized to upper continental crust44. Errors bars are smaller than symbols.

Compared to REE data, the Huab, Omaruru, Kuiseb and Namib Sea Sand samples can be better discriminated based on their isotopic compositions (Fig. 3). Lead isotope compositions span within a relatively large range of values: 206Pb/207Pb: 1.15–1.28, 208Pb/207Pb: 2.43–2.62 defining variable Pb isotope fields (Fig. 3a). For the Huab valley, the nebkhas in both the bulk and fine fractions display intermediate lead values (206Pb/207Pb: 1.205–1.228, 208Pb/207Pb: 2.482–2.518). This is, between the bulk fraction of the sand dunes from Kuiseb and the Namib Sand Sea regions which represent the highest and lowest (respectively) end-members in terms of Pb isotopic compositions. In this sense, the Kuiseb related samples have a more Th-rich signature (206Pb/207Pb: 1.23–1.277, 208Pb/207Pb: 2.495–2.612) indicating derivation from sources with higher Th/U ratios. Contrarily, the Namib Sand Sea area shows comparatively the most unradiogenic Pb isotopic values (206Pb/207Pb: 1.159–1.206, 208Pb/207Pb: 2.425–2.468), with a clear distinction between the bulk and the fine fractions. Interestingly, nebkhas samples associated with the Omaruru valley define two isotopic subfields: one overlapping with Huab and a second field with less radiogenic isotopic ratios (206Pb/207Pb: 1.181–1.212, 208Pb/207Pb: 2.439–2.488) showing a similar fingerprint to the fine fraction of the Namib Sea Sand (Fig. 3a). We also observe a Pb isotope offset between our bulk and fine fraction for one sample from the Kuiseb catchment and the previously published values on the same sample8. Our Pb values for that sample are more similar to the Namib Sand Sea dunes (Fig. 3a).

Fig. 3: Sr-Nd-Pb isotopic compositions of Southern Africa sediments.
figure 3

a 208Pb/207Pb vs. 206Pb/207Pb. b 87Sr/86Sr vs. εNd(0). Grain size effects on Sr-Nd-Pb isotopic compositions for the Huab and Namib Sand Sea are also shown and indicated by symbols in gray (fine) and open symbol (bulk).

As a whole, the Sr–Nd isotopic ratios from SAF samples measured in this study in both the bulk and fine fractions (Fig. 3b) exhibit two main groups, indicating plausible dust provenance from different source end-members coherent with the geological history of this part of the continent. The isotopic variability in the bulk fraction for sediment samples derived from the Omaruru and the Kuiseb river systems is large (87Sr/86Sr = 0.7208–0.7384 and εNd(0) = −12.80 to −10.17). In contrast, the bulk and fine fractions of the Namib Sand Sea dune samples depict an even broader range of isotopic values: 87Sr/86Sr = 0.7128–0.7305 and εNd(0) = −10.39 to −9.84, reflecting provenance from the granitoid intrusions of the Damara belt46. The nebkhas from the Huab catchment, which drains the Paleoproterozoic basement of the Kaoko belt43, display a narrower range of isotopic values for both the bulk and fine fractions: 87Sr/86Sr = 0.725–0.728 and εNd(0) = −18 to −15. We also observe a Nd isotopic trend with values becoming progressively less radiogenic from north to south. This is, the Huab region (in the north of the Namib coastline), exhibits the less radiogenic Nd composition with εNd(0) between −18 and −15, continuing towards the Omaruru and Kuiseb valleys with intermediate Nd compositions before reaching the southernmost location of the Namib Sand Sea region which exhibits εNd values between −10.5 and −8. We also observe a Sr isotopic offset between our fine fraction samples from the Namib Sand Sea dunes and the previously published values on the same samples3 with our dataset showing higher 87Sr/86Sr values (>0.725, Fig. 3b).

Grain-Size isotopic fractionation

Attributable to different mineral contributions, mainly clays and feldspars, that are ultimately controlled by mineral sorting processes47, it is observed that the fine-grained fractions generally show a more-radiogenic signature compared to the bulk of the same samples. In the present study, the set of samples from the Huab (n = 6), Kuiseb (n = 1), and the Namib Sand Sea (n = 4) regions analyzed in the bulk and the fine fractions, conforms to the longstanding observation that there are statistically significant differences in 87Sr/86Sr, but much less so for εNd(0) and Pb isotopic values (Fig. 3a, b). While the dunes from Kuiseb and the Namib Sand Sea show the highest Sr isotopic differences (Δ 87Sr/86Sr ~0.012 units), the variability in nebkhas from the Huab region is limited to only ~0.002 units (Fig. 3b). Furthermore, for the three areas the mean difference between εNd(0) values for the fine and bulk fractions is <1ε units and not significant. For Lead, these differences are also small (Δ 206Pb/207Pb and 208Pb/207Pb ~0.022 units for Kuiseb, and Namib Sand Sea samples and, even smaller Δ for the Huab samples: ~0.01 units) meaning that the sorting by size, although not a zero effect, does not play a significant role that could complicate attempts to relate the Pb isotope data in remote locations to PSAs. Overall, our results agree with the observed small differences between clay and silt for grain-size fractions in river sediments48. The new data presented here for the finer fraction in dunes samples from the Namibia Sand Sea, agrees with the previous Pb isotopic ratios (n = 2) reported for the same region8. As mentioned above, the strontium isotope ratios previously presented for this area3, are slightly less radiogenic in 87Sr/86Sr compared with the new values obtained here for the same samples (Fig. 3b). Despite the disparity, this does not affect our interpretations; any difference in the Sr isotopic data between the bulk and the fine grain-size fractions of sediments is small compared to the variability of the whole Southern Africa region, reinforcing the hypothesis that the isotopic signatures can be used to fingerprint provenance.


For over three decades, Sr-Nd-Pb isotopes in combination with chemical, physical, and modeled studies have focused on the provenance of dust recovered from East Antarctica ice cores. Compelling evidence has attributed Southern South America as the primary source of dust to Antarctica during both, cold and warmer periods3,5,9,10,18,49,50. Nevertheless, it is also well known that significant overlaps between the different Sr-Nd-Pb isotopic fields defining key dust areas of SSA, AUS, and NZ occur. The isotopic data presented in this study for the Southern Africa PSAs is not the exception to the imbrication issue because SAF shares some similarities in chemical/isotopic composition with some SSA PSAs (Fig. 4).

Fig. 4: Comparison of Southern Africa (SAF) with Southern South America (SSA).
figure 4

a Eu/Eu vs. εNd(0). b 206Pb/204Pb vs. 143Nd/144Nd. SSA data from9,10.

Nevertheless, the SAF and SSA signatures can be partially discriminated by combining REE with isotope data. Overall, SAF dust sources evidence very low εNd(0) values (lower than −10), and more radiogenic values of 206Pb/204Pb and 207Pb/204Pb, reflecting the signature of the Damara basement (Fig. 4a, b). In terms of REE, Huab and Kuiseb sediments exhibit an average Eu/Eu* < 1, while Omaruru appears divided in two groups of samples; one following the same trend with Eu/Eu* < 1 and another group with scattered data, showing higher values of Eu/ Eu* (Fig. 4a). Contrarily, SSA exhibits a fingerprint that varies from more “mafic-like” compositions from Patagonia, reflected by Eu/ Eu* values >1, higher 143Nd/144Nd and lower 206Pb/204Pb, in contrast with more “crustal-like” compositions of material from the Puna–Altiplano Plateau area9,10 (Fig. 4a, b). In terms of 143Nd/144Nd and 206Pb/204Pb isotope ratios, the Namib Sand Sea data reveals a similarity with some areas of SSA, overlapping with the Southern Altiplano and Northern Puna isotopic composition fields (Fig. 4b).

The similar geochemical composition exhibited by the surface sediments of these two areas (i.e. SAF and the PAP) could be related to the comparable climatological conditions both, the Puna–Altiplano Plateau and Southwestern Africa share from their tropical to subtropical location (between about 17° and 30°S latitude). These regions are influenced by atmospheric circulation dominated by subsidence air zones (SE Atlantic-Namibia- and SE Pacific -PAP-), promoting extreme arid conditions and producing the world’s desert lands. Reinforcing the aridity in these areas is the presence of an upwelling, north-flowing, cold oceanic currents; the Humboldt current along the west coast of South America and the Benguela current along western South Africa. Precipitation is the result of tropical convection and rainfall patterns are controlled by the South American summer monsoon (SASM) and East African summer monsoon (EASM), respectively. Both areas have also shown important climatic variability during the late Quaternary, indicating altering cycles of wet/dry conditions, however interpretations about the main factors determining these changes are still controversial. Some authors have pointed to a precessional orbital control after finding a correlation between periods of Southern Hemisphere maximum insolation and enhancement of moist air on land, increasing precipitation delivered by the SASM and EASM51,52,53. Others have suggested that southwestern Africa responds to variations in phase with the Northern Hemisphere summer insolation54. Despite the debate of these complex mechanisms, there is persuasive evidences supporting the important role of both the PAP and SAF regions as dust suppliers during the late Quaternary. The arid and dry environmental conditions in Namibia have been similar to what it is today for the last 50 Ky, with the late Holocene being the driest period54. In addition, a pollen marine record indicated two arid peak periods (14–12 Ky and 11–9 Ky) for the early Holocene in this part of the Southern African continent55. Authors suggested a northward migration (from its present position) of the intertropical convergence zone (ITZC) resulted in a reduction of the water vapor supply in southern subtropical Africa during the early Holocene. A marine grain-size record and a collection of terrestrial records also suggested arid conditions in southern Africa during interglacial stages due to meridional shifts in the atmospheric circulation system56,57. In line with this, a more recent pollen record from the southern Namib Desert obtained from rock hyrax middens have shown the Holocene was a time were high temperatures and evapotranspiration resulted in increased aridity and an expansion of the desert vegetation58. These environmental conditions of dryness during interglacial times are also indicated for the PAP area10, especially during the Holocene59,60 and MIS 5.561. In addition, a combined grain-size study of two marine cores taken off the coast of southwestern Africa (~20 S) and western South America (~27 S), observed that both sites showed arid conditions for at least the last 35,000 Ky62.

Subsequently, how the SAF fingerprint compares respect to AUS and NZ signatures is necessary to better understand the provenance of dust reaching the SO and EA. Recent studies have reviewed in detail the location, geology, atmospheric circulation patterns, and geochemical signature of all different AUS and NZ Potential Source Areas21,22,24. These studies have also thoroughly examined the geochemical fingerprint of both Australian and New Zealand sediments with the signature of Southern South America samples9,10, concluding that AUS and NZ sources cannot easily be distinguished from the dust sources regions of SSA. Here, for comparison against the Southern Africa signature, we use the Sr-Nd-Pb isotope ratios of the Australian and New Zealand PSAs that are relevant for comparison with ice cores records from E. Antarctica. These are: the Kati Thanda-Lake Eyre region, and the Darling Riverine Sub-Basin in Australia and New Zealand’s South Island21,22,24. As previously mentioned, Southern Africa’s diverse geology leads to a large range of Sr-Nd-Pb isotope ratios. However, these clearly differ from the AUS and NZ values and they can easily be distinguished from one another as end-members (Fig. 5a, b). Southern Africa sediments have higher 87Sr/86Sr (>0.719), lower Nd(0) values (<−8) and higher 208Pb/207Pb (>2.47) compared to AUS and NZ samples21,24.

Fig. 5: Comparison of Southern Africa (SAF) with Australia (AUS) and New Zealand (NZ) geochemical signatures.
figure 5

a 87Sr/86 Sr vs. εNd(0). b 208Pb/207Pb vs. 206Pb/207Pb. Lake Eyre Basin, AUS data from13,21,24; Darling Sub-basin, AUS data from8,84; NZ data from3,24.

Based on major/trace and Sr-Nd-Pb isotope ratios, New Zealand was indicated as a not significant dust contributor to E. Antarctica24. In contrast, these authors also found that the composition of marine sediments in the Pacific sector of the Southern Ocean (and by extension to West Antarctica) during the Holocene can be explained by a mixing of New Zealand South and North Islands with potential contributions from Lake Eyre Basin in Australia, if dust from this latter region is able to reach the Pacific sector independently of other Australian sources.

Regarding Australia, the potential of this region to reach the Antarctic continent is still a matter a debate mainly due to that both AUS and some SSA PSAs share a similar isotopic signature with the Antarctic dust fingerprint. In this regard, a recent study has reviewed this specific subject in detail concluding that while the Southern South America signature is the predominant signature found in Antarctica, the two AUS PSAs that could have contributed dust to Antarctica are the Kati Thanda-Lake Eyre region, and the Darling Riverine Sub-Basin with the western Antarctic sites being more in favor (due to their position respect to the source)22. These two main dust regions in AUS (Supplementary Fig. 2) share a similar isotopic composition with those SSA areas identified as main dust suppliers to E. Antarctica during glacial/interglacial periods (Patagonia, Southern Central Western Argentina and the PAP) and, also show some correspondence with a few E. Antarctica glacial/interglacial samples. Moreover, only samples from the Kati Thanda-Lake Eyre region could explain the composition of a few Antarctic samples. Minor contributions from Lake Eyre have been previously suggested13,63, although the mechanism for the long-range dust transportation to Antarctica was not given. In fact, based on HYSPLIT models, there is only one study (to the best of our knowledge) that showed that during spring months when dust from Lake Eyre is entrained by warm and dry northwesterly winds, could potentially be transported to certain Antarctic sectors such as, George V Land, and to a lesser extent, the peninsula within eight days64. However, the isotopic composition of Lake Eyre doesn’t match with the Sr-Nd isotopic ratios of modern dust measured at Berkner Island, Antarctic Peninsula41 (Supplementary Fig. 2). Furthermore, AGCMs have shown that during both climatic periods present-day and the LGM, AUS ‘tracers’ are/were highly concentrated in West Antarctica sectors, particularly in Mary Byrd Land19. Although, a description of which AUS sources were used in the study was not provided. It should also be noted that due to the geographic distribution, Australian dust gets well-mixed before reaching the southeast Australian coastline. This is; the AUS PSAs come across a narrow latitudinal strip but a broad longitudinal area which has important implications for the dust mobilization in the region mainly associated with easterly fronts within the westerly wind belt13. Hence, when dust events occur, different AUS PSAs can be picked up as a whole and the result is a dust plume with a very similar and comparable Australian fingerprint13,65,66. Thus, while it is clear that based on the Sr-Nd-Pb combination approach we cannot dismiss the possibility of a minor contribution of Australian dust to Antarctica, especially to West Antarctica, it is also true that the processes on how dust from the Lake Eyre is/was transported to East Antarctica is not fully understood yet and more research in this topic is needed. With these observations made above, is that this study does not consider AUS as an important end-member contributing dust to E. Antarctica.

Based on Sr-Nd-Pb isotopes, only two earlier studies3,8 suggested the possibility of SAF as a PSA to the polar Antarctic regions. However, due to the limited number of sediment samples and their distinct geochemical signature with respect to EA, they concluded isotopic data was not conclusive enough to support a dust contribution from SAF, excluding this region from the provenance puzzle. Contrarily, the new and expanded dataset of Southern Africa Sr-Nd-Pb isotopic ratios and REE concentrations presented in this study indicate that, SAF emerges as PSA that together with high-latitudes sources in Southern South America (i.e, Patagonia, Southern Central Western Argentina, and the Puna–Altiplano Plateau) can explain the glacial/interglacial dust signature observed in East Antarctica ice cores, as well as the signature of marine sediments from the South Atlantic sector of the Southern Ocean (Fig. 6).

Fig. 6: Comparison of Southern Africa (SAF) and Southern South America (SSA) isotopic signatures with the glacial/interglacial fingerprint of dust deposited in East Antarctica and the South Atlantic sector of the Southern Ocean.
figure 6

a, b 208Pb/207Pb vs. 206Pb/207Pb. The orange dotted lines in Fig. 6b represent compositional trend lines necessary to explain EA dust compositions. c, d 87Sr/86 Sr and εNd(0). The black lines in Fig. 6d represent the calculated tri-mixing hyperbola between different PSAs of SAF (dotted line) and SSA (fill line). End-member values used to calculate Sr–Pb mixing curves can be found in Supplementary Table 1. Local Antarctic Volcanoes data in Fig. 6a is from8. SSA data from9,10. East Antarctica glacial/interglacial data from3,5,8,63,70,85. Berkner Island data from41.

In general, the 87Sr/86Sr and εNd(0) values are more diagnostic than the 206,208Pb/207Pb for identifying the glacial/interglacial dust provenance in paleo-archives. This means that even though the Pb isotopic data is usually less indicative because of the Pb input related to local (Antarctic) volcanoes in comparison to the dust related Pb8, the Pb data in the present study add critical support to our interpretations. This is, the new Pb isotope data of SAF samples define a field that, together with Southern South America PSAs can better explain the entire range of Pb isotopic compositions observed in the East Antarctic dust from Vostok and EDC (208Pb/207Pb values from 2.44–2.50) (Fig. 6a). It should also be noticed that the very radiogenic Pb data similar to Kuiseb envelope are also those of the Antarctic local volcanoes (Mt. Erebus); contribution that has been recognized in previous studies8,9,67 and cannot be denied. Thus, our results show that EA dust samples show two main compositional trends that comprise different end-members (orange dashed lines in Fig. 6b). This is, a common end-member with low values in both 208Pb/207Pb, and 206Pb/207Pb elucidated by either the Namib Sand Sea (fine fraction) or the Patagonia/Southern Central Western Argentina sources in combination with a more radiogenic end-member, that is, either Kuiseb PSA or the local Antarctic volcanism with a contribution from the Puna Altiplano Plateau (PAP) and the Huab region, especially during interglacial times (Fig. 6b).

This is also in part supported by Sr and Nd isotopic data (Fig. 6c, d), where the new SAF dataset defines a compositional field that turns out to be a key end-member for describing the composition of dust deposited in the South Atlantic and East Antarctica. In general, during the last glacial climatic periods of the late Quaternary, EA data is very uniform converging to an unequivocal signature from Patagonia with additional inputs from other lower latitude areas in SSA, such as, the Southern Central Western Argentina (SCWA) and Southern Puna10 (Fig. 6c). On the other hand, the EA interglacial dust fingerprint, as well as the glacial/interglacial signature of marine sediments from the South Atlantic are marked by a variable mix of sources (Fig. 6c). An important finding of this study is that a tri-mixing hyperbola with two common end-members can be drawn and help to decipher this variability. This is, the best candidates to elucidate the origin of dust during the Holocene and likely during the last 15 ka are a more mafic-like end-member represented by Patagonia and/or SCWA, in addition to a second crustal-like end-member, defined by the Southern Altiplano, plus a third end-member, which can either be the Northern Puna (black solid line) and/or Southern Africa region as a whole (dotted black line) (Fig. 6d, Supplementary Table 1). This is also in agreement with the isotopic composition of present-day dust deposited on the Berkner Island in the Atlantic sector of coastal Antarctica, which can clearly be ascribed to a mixing isotopic composition from PAP10,41, and SAF PSAs (Fig. 6d). It was also suggested that neither of the PSAs in SSA could account for the entire Sr-Nd isotopic range of the dust deposited on Berkner and that the contribution of another secondary source with more negative εNd (0) was missing41. The Sr and Nd signature of the SAF samples presented here can account for that ‘missing’ end-member, although local sources cannot be dismissed. Alike, it was inferred that the Holocene signature identified in ocean sediments from the Atlantic sector of the Southern Ocean is associated with material derived from SAF, but also with compositions similar to the Northern Puna68 (Fig. 6c, d). Moreover, detrital Nd isotopic composition of marine sediments recovered in the South Atlantic region, right off the Namibia coast, have shown εNd (0) values between −10 and −22 that have been attributed to a Southern Africa origin69, which is in accordance with the new Neodymium isotopic data presented in this study. Authors associated trade-winds as the main mechanism responsible of transporting detrital material to the South Atlantic Ocean. Likewise, another marine core recovered from the Walvis Ridge pointed the Berg winds as another possible transport mechanism for the aeolian dust sweeping away from the Namibia coast56. The same study revealed that SAF climate was relatively arid during the interglacial stages and relatively humid during the glacial stages. During glacial stages, the polar front shifted equatorward resulting in a northward displacement of the westerlies belt, causing increased rainfall in southwestern Africa55. This could help explain why the signature of SAF PSAs may have had a lesser role during cold periods in comparison with warmer periods when the activation of other sources, such as the Puna–Altiplano Altiplano Plateau and SAF prevail. Additionally, the strong Southern Central Western Argentina fingerprint seen in glacial dust deposited in EA agreed with an overall equatorward displacement of the mid-latitude westerlies during glacial times10.

While the geochemical evidences presented here demonstrate that SAF sediments and interglacial/modern dust deposited in different environments of the Southern Hemisphere (SH) share a similar isotopic signature, the concept about how the transport occurs from the Namibian coastal regions to Antarctica is still not well known. The deposition of dust around Antarctica can happen by either a scenario whereby it is deposited in the interior high elevation sites or in the periphery low elevation sites of the Antarctic plateau70. Mineral dust reaching the EA plateau is windborne long-range transported from the continental land-masses carried through the mid-to high troposphere. For dust sources that are not located at high latitudes, such is the case of SAF, there is need of strong convective uplift of particles up to 200–400 mbar to reach the interior plateau and the presence of SAF particles has mostly been reported in the lower troposphere71. Another possibility is that the source itself is already elevated in altitude allowing the injection of dust particles directly in the high troposphere. This is the case of the Puna–Altiplano Plateau in SSA, where the subtropical jet stream acts as main mechanism mobilizing large amounts of dust that ultimately, get deposited onto the polar plateau7,10,12 and whose signal has also been found in peripheric areas of Antarctica in the Berkner Island41. Modeling data have suggested that material from SAF only contributes with ~5% of the total dust burden over Antarctica and a southward transport to the inner part of the Antarctic continent would be unlike18,20. However, modeled dust emission fluxes from the different PSAs of the SH and their subsequent deposition driven by the local trade winds, can vary by as much as a factor of four or more, which could mislead to controversial interpretations23. This has important implications especially at ice core sites located on the high-elevated plateau areas (~3000–4000 m) that can be particularly isolated from extra-Antarctic dust sources, and where AGCM experiments struggle to reproduce dust concentrations in ice core samples18,20,72. While recognizing the limitations of modeling in representing the global dust cycle, a more recent study introduced a novel tool known as DustCOMM (Dust Constraints from joint Observational-Modelling-experimental analysis)73; a continuously expanding dataset of constraints on the global dust cycle that combines observational and experimental analyses on dust properties and dust abundance with an ensemble of global aerosol model simulations. DustCOMM showed that while SSA accounts for the large majority (70%) of the dust deposited in almost the entire Antarctic continent, smaller contributions from Australia (20%) and southern Africa (10%) are also observed73. The work done by Neff and Butler (2015) also indicated that the dust contribution from SAF sources located in the arid regions of Namibia, is significant in Southern Ocean areas off the Antarctic coast near Wilkes Land, E.A, and at Law Dome. Using HYSPLIT, a trajectory model, another study showed that dust emissions from the Southern Kalahari commonly reach the Southern Ocean74, with most of the trajectories traveling beyond 40°S during the dry period (April-September) when conditions favor the dust production and transportation75.

We note however, that most of the regional to global dust modelings regarding Namibian sources are solely based on the atmospheric circulation of the trade winds to emit dust. While in reality, it is the Berg winds, a local southwestern Africa topographically enhanced thermal wind that represents a more important driving force responsible of dust mobilization and deposition over the ocean28,76. These hot and dry winds are also more common in winter and they are the result of adiabatic heating; a process that occurs when air coming from the escarpment and plateaus descends towards the Namibian coastline, getting compressed and heated. Multiple dust streams and dust storms originated in this part of the African continent evidencing the substantial transport of material to the Southern Ocean as a result of Berg winds have been spotted by different satellite platforms28,29. This has important implications for the transportation of material to the marginal and coastal sites of Antarctica, which can receive an unquantified amount of SAF dust. On this matter, more recent studies have showed strong support that highlights Southern Africa as a secondary and significant dust supplier to East Antarctica. Sr, Nd and Pb isotope ratios of Antarctic modern dust recovered from snow samples collected along coastal sites of the Dronning-Maud Land area (DML-East Antarctica), have shown similar composition to the Southern Africa sediments77. This data is consistent with the results observed using back trajectory FLEXPART particle dispersion transport model where a mix of modern Southern Africa and Southern South America dust sources was clearly spotted in the coastal area of East Antarctica78. Likewise, using a statistical analysis a study revisited the only EDML ice core rare earth elements dataset available from17, and observed that the Southern Africa fingerprint appears at ~15 ky as an important mixing end-member that together with SSA, can explain the East Antarctica interglacial dust signature79. Ultimately, another geochemical study carried out in a core from the Southern Indian Ocean, has also proposed a mixing model where SSA and SAF have been found to be the main distal dust sources through most of the Holocene80.

In resumé, this study agrees with previous findings suggesting that SSA is the best candidate to explain the dust signature recorded in the SO and EA during cold and warm stages. For warm periods, this study proposes a scheme where the typical isotopic variability observed in interglacial and present-day dust depositions in the South Atlantic Ocean and coastal regions of East Antarctica can be related to the contributions of two main PSAs: Southern South America, through westerlies dominating the atmospheric circulation over Patagonia and Southern Central Western Argentina and, the jet stream over the Puna–Altiplano Plateau, in combination with Southern Africa in response to the plausible transport by strong Berg winds. We rely in part on an understanding of modern wind-dust activities showed by satellite images to conclude the possible contribution of Southern Africa dust emissions during interglacial times. However, modeling studies focusing on the interplay between the southeast trade-winds, and Berg winds that dominate dust transportation off the Namibian coast and the extent of this transport to higher latitude areas, are immensely needed.


Acknowledging the location of the main dust sources of the areas investigated in this study, nebkhas and sand dunes were sampled across the lower valleys of the Huab, Omaruru, and Kuiseb river systems and the Namib Sand Sea (Fig. 1a, blue diamonds. See Supplementary Table 1 for more details). Aiming to evaluate the provenance and long-range transportation of dust from these Southern Africa PSAs and depositions in remote areas, a total of 22 bulk samples (nine samples from Huab, seven from Omaruru and six from Kuiseb) were analyzed in terms of their REE concentrations and Sr-Nd-Pb isotope compositions. Additionally, measurements of Sr-Nd-Pb isotope ratios were carried out in five bulk and five fine (<5 μm) fraction samples from the Namib Sand Sea, to produce a comparable result to the bulk samples previously analyzed in3,8. Furthermore, to assess a possible change in the geochemical fingerprint due to grain-size effects, six bulk samples from the Huab valley were selected to generate airborne aerosols dust in the fine fraction (~<11 μm) using an atmospheric simulation chamber (see section 4.1 and supplementary materials for further details). The aerosol fraction investigated in this study has also important implications for provenance studies as only certain particle sizes can be transported long distances. Hence, analyzing the right size fraction in the source region is crucial to identify the origin of dust reaching the Southern Ocean and Antarctica.

Sample collection

According to remote sensing analyses, the most persistent dust sources in the Huab region are concentrated upstream of the river catchment; this is, in the river flood channel itself, in a playa and a delta. Dust sources are associated with fluvial silt crust terraces covered with nebkhas. In the case of the silt crust, these surfaces will only emit dust if an erodible material such as sand is present to initiate the saltation. The nebkhas in the delta section of the Huab River, present almost no vegetation cover making them a source of aeolian sediments that is picked up not only by the high magnitude north-easterly Berg winds, but also during the predominant southerly winds28,29. In Omaruru, most of the dust emissions (>80%) happen downstream within the floodplain channel of the river as a result of the high-magnitude northeast winds. Nebkhas and degraded silt crusts are the two main landforms in the terrain, with the nebkhas being the major source for dust deflation due to the sparse vegetation they present which results in poor protection against the wind29. For the Kuiseb, the silt-clay fines are deposited in the lower sections of the river becoming soil crusts that are the dominant source of dust through saltation, erosion and suspension. As for the Huab river, saltation is the mechanism responsible of dust entrainment and emission28,29.

With this in mind, small vegetated dunes (nebkhas) were targeted to be a representative sample of that particular basin for currently active transport at a distance from the ocean. For the Huab and Omaruru, the sites were uniform in terms of landform. These were nebkhas with loose sediment around their bases that the surface sediment was sampled (‘last remaining dust stores’). In the case of the Kuiseb, silt-fine material was sampled. These vegetation-free Nebkhas are located where the Namib Sand Sea and the silt terraces from the river meet, and where the processes involving these two sediment sources (but most importantly the presence of the silt terraces from the river) are what makes this area so dusty. Thus, bulk surface samples from Huab, Omaruru and Kuiseb riverbeds were collected in a field camping in August 2017, using a trowel to obtain approximately 200 g of sediment from the surface. Sub-samples of ~350 mg were stored in pre-cleaned centrifuges tubes and shipped to Université Libre de Bruxelles (ULB), Belgium for analysis. From the bulk of six samples from Huab, dust aerosols (fine fraction) were generated into the CESAM atmospheric simulation chamber, to mechanically reproduce the saltation and sandblasting processes responsible for the release of mineral dust in natural conditions81. Briefly, dust aerosols are generated from the bulk sediments by a mechanical shaker and the aerosols are injected into the CESAM chamber82, a stainless-steel reactor with a volume of 4.2 m3 where dust is dispersed, left in suspension for ~3 h, and ultimately recovered on acid-cleaned 0.2 µm Teflon filters. Additional details about the dust generation in the CESAM are provided in the Supplementary Methods. Samples from the Namib Sand Sea correspond to sand dunes previously analyzed for the Sr-Nd3 and Pb8 isotopic composition. Here, we have re-analyzed the bulk and the fine fraction of those samples, where the fine fraction was separated using the Stoke’s Law3.

Geochemical analyses

Sample preparation and analyses for REE elemental and, Sr-Nd-Pb isotope compositions were carried out on bulk and fine fraction samples. Sediment samples were dissolved in a concentrated HF + HNO3 mixture followed by an HCl bath in Savillex vials and Sr-Nd-Pb were separated through chromatography using ion-exchange resins following the procedure reported in83. Isotopes were measured at the G-Time Lab, Universite Libre de Bruxelles (Belgium) using a High-Resolution MC-ICP-MS Plasma II (Nu Instruments) in dry mode. The REE concentrations were determined using a Quadrupole ICP-MS Agilent 7700. Additional details of the analytical methodology are provided in Supplementary Methods.