We obtain HYSPLIT NCEP trajectories for December, January, February (austral summer; DJF) and June, July, August (austral winter; JJA) from 1980–2022 (inclusive) hourly for 10 d over a 90 d period. We run the model at an altitude of 1000 m following Schwanck et al. (2017) and use the default vertical motion for the model. We then calculate air mass origin percentage using a 2×2° grid to identify where air masses originate 10 d prior to reaching the South Pole.

We also explore changes in atmospheric temperature, wind speed, and precipitation using CMIP6–PIMP4 model outputs. We use 11 different models (Table 1) to assess model agreement between mid-Holocene runs (6000 years ago) relative to pre-industrial control (piControl). We calculate model difference by subtracting the mid-Holocene from pre-industrial to assess change in each climate variable from the mid-Holocene to the pre-Industrial. We replace all positive values with a value of 1, which signifies an increase with time, and replace all negative values with a 0 and sum all model runs. Our method allows us to assess if the models agree to an increase/decrease in each climate variable rather than assessing the magnitude of change between each model. We test statistical significance by selecting data in the mid-Holocene and piControl runs using a 10×10° grid. We use a t-test to identify significant difference between the two periods for each model run (p-value < 0.05). If the two periods are significantly different, then replace the values with a 1 and sum the totals between models, similar to our model agreement analysis.

The number concentrations of particles in the size range 1.1–5.1 µm comprise on average ∼ 85 %–96 % of all particles measured (median ±1σ; Fig. 1a and b). The total dust number concentration and CPP (Fig. 1a–c) abundance have opposing trends, yet the timing of change in both trends begins about ∼ 7000 years BP. This inverse relationship is most apparent during the mid-Holocene (∼ 7000–3000 years BP). Because trend changes occur around the mid-Holocene, we divide the Holocene into early (11 300–7000 years BP), middle (7000–3000 years BP) and late (3000–250 years BP) periods. During the early Holocene, dust concentration remained relatively stable (∼-1.7 particles mL⁻¹ kyr⁻¹) followed by a significant decrease (∼-50 particles mL⁻¹ kyr⁻¹) until the Late Holocene where particle concentrations continued to decrease but at half the rate of the mid-Holocene (∼ 24 particles mL⁻¹ yr⁻¹; Table S1). Dust concentration distributions in all three time periods are statistically different (Fig. 1d; p-value < 0.01; Table S2a), further indicating a significant reduction in Holocene dust deposition. During the mid-Holocene, there is a bimodal distribution highlighting the transition from early to late Holocene (Fig. 1d). During the early and late Holocene, the dust concentration is more normally distributed.

The number concentration of coarse particles (Fig. 1b) has a decrease similar to the total particle concentration (Fig. 1a), but the coarse particle decrease is an order of magnitude smaller (∼-4 # particles mL⁻¹ kyr⁻¹) during the mid-Holocene. The decrease in total dust concentration from the Early Holocene to Late Holocene is about 40 % and about 35 % for coarse dust concentration. This 5 % difference results in an increase in the CPP starting at ∼ 7000 years BP, where over the Holocene, CPP significantly increases from 8.6 % to 9.4 % (two-sided t-test; p-value < 0.01; Table S2b). Although the coarse dust concentration also decreases slightly throughout the Holocene, the decrease is much less than that of the total dust concentration slope such that the proportion of coarse particles rises, especially during the middle Holocene. CPP has a weak statistical relationship with dust concentration during the Holocene (Spearman r=-0.28; p<0.01), suggesting that their variability may be driven by different and/or independent factors. Because total dust concentration was used in the creation of the CPP metric, we test this relationship with CFR and identify a statistically weaker relationship (r=-0.15, p<0.05). The statistically weaker relationship further supports the above interpretation of independent factors driving their variability.

South Pole Holocene dust flux (0.45±0.23 mg m⁻² yr⁻¹; 2σ; 50-year resampled) has a similar magnitude and trend to other East Antarctic Plateau ice core sites (Fig. 2a; following Koffman and Kreutz, 2014). The SPC14 Holocene median dust flux is within 2σ of the EPICA Dome C (0.33±0.23) and Vostok dust flux records (0.63±0.21), while non-plateau sites (i.e., TALDICE [∼ 1.31 mg m⁻² yr⁻¹] and WAIS Divide [∼ 1.36 mg m⁻² yr⁻¹; particles <5.1 µm]) are significantly higher (Fig. 2b; Koffman et al., 2014; Albani et al., 2012; Lambert et al., 2012; Petit et al., 1990).

Following Albani et al. (2012), we compare our Holocene dust flux record against δ18O, a proxy for local atmospheric air temperature, to assess whether there are primarily remote or mixed dust source (Steig et al., 2021). Albani et al. (2012) compared the TALDICE dust-δ18O profile against EPICA Dome C as EPICA Dome C is considered to represent remote dust sources. We define remote dust sources similarly, as those from the mid-latitudes (i.e., southern South America and Australia; Delmonte et al., 2020), local sources from Antarctica (i.e., Aarons et al., 2019), and mixed being a combination of the remote and local. The Holocene South Pole dust record and δ18O have a weak, albeit significant relationship (r=-0.17, p=<0.01; Fig. 2b), which suggests a primarily remote dust source signal. Albani et al. (2012) interpreted the close relationship between dust and δ18O trends at Talos Dome as evidence of a local/regional climate influence. Conversely at Dome C during the Holocene, the dust and δ18O did not vary coevally, which is interpreted to reflect mid-to-high latitudinal processes (Lambert et al., 2008; Albani et al., 2012).

Our NOAA HYSPLIT results from austral summer (DJF) and winter (JJA; 1980–2022) both highlight a dominant Atlantic sector signal (Fig. 3; Rolph et al., 2017; Stein et al., 2015). Austral summer air masses at their 10 d origin are from the Weddell Sea, Dronning Maud Land, Enderby and Kemp Land (i.e., Atlantic Sector; Fig. 3a), while austral winter air masses origins are more latitudinally and longitudinally diverse, with a greater percentage of air mass originates from the Atlantic and Pacific sectors of the Southern Ocean (Fig. 3b). While both the longitude and latitude seasonal origins differ, the change is most pronounced in the latitudinal variability (Fig. 3c and d). A primary austral summer peak was observed in previous South Pole dust and meteorological records (Lazzara et al., 2012; Carrasco et al., 2003; Parungo et al., 1981; Hogan et al., 1984; Bodhaine et al., 1986). Our HYSPLIT analysis identifies DJF 10 d air mass origin locations with an average coordinate of 70.9 (±9.4)° S, -1.4 (±80.03)° W, indicating a primarily Weddell Sea/Atlantic Sector modern signal. The similarity results from other East Antarctica sites (i.e., Dome C and Vostok; Fig. 2b), the modern back-trajectory modelling (Fig. 3), and the dust-water isotopes relationships all suggest that South Pole represents a site typical of Antarctic plateau locations, with the majority of dust likely sourced from remote regions rather than mixed remote-local dust sources, facilitating comparisons (Fig. 2).

The South Pole receives air masses from multiple different Antarctic regions, including Western Antarctica (Nicolas and Bromwich, 2011), the Weddell Sea (Winski et al., 2021; Lazzara et al., 2012; Arimoto et al., 2008), and downslope from the East Antarctic Plateau (Lazzara et al., 2012). The dominant advection of remote air masses are from Weddell Sea and WAIS with a minor source from EAIS (Harris, 1992). Our HYSPLIT data analysis then is reflective of past studies and highlights modern seasonal latitudinal changes in wind strength and position of 10 d air mass origins. While we recognize that some air masses and dust particles can be sourced from the East Antarctic Plateau, this area would not be the primary source of reduced dust emission during the mid-Holocene since southern South America and Australia are the primary sources of dust during the Holocene and modern period (Wegner et al., 2012; Vecchio et al., 2024; Delmonte et al., 2020). Given the multiple trajectories of air masses to the South Pole, we do not specify a specific dust source contributor but rather limit our interpretation to local or remote dust sources.

We suggest that the SPC14 dust concentration and CPP records indicate a mid-to-high latitude (>∼ 51° S) strengthening and southward shift of the SHWW during the mid-Holocene (∼ 7000–3000 years BP), consistent with other proxy-based reconstructions of SHWW migration and CMIP6-PMIP4 model output (Fig. 4). Simulations from an ensemble of CMIP6-PMIP4 models (Table 1; Fig. 4b), all show a clear pattern of decreased and significant surface wind strength over South American and mid-latitude (40–50°) oceans concurrent with increasing surface wind strength over the Antarctic periphery between the mid-Holocene (6000 BP) and pre-Industrial (piControl; Fig. 4b and e). The weakening of the SHWW over lower latitudes <∼ 51° S is consistent with the observed decrease in the SPC14 dust concentration, indicating a reduced entrainment of fine remote dust sources in surface winds. We note broad agreement spatially with Southern Hemisphere temperature and precipitation increase between the mid-Holocene and pre-Industrial periods over common dust source regions (i.e., South America, Australia, and Antarctica; Fig. 4a, c, e, and f). A southward shift in the core of the SHWW can simultaneously explain the opposing trends in CPP and total particle concentration at the South Pole while being consistent with proxy records reflecting changes in winds, precipitation and/or glacier retreat across the mid-to-high-latitude Southern Hemisphere during the mid-Holocene.

More northern proxy records corroborate a poleward shift in the SHWW during the mid-Holocene. Lacustrine sediments from Emerald Lake, Macquarie Island (∼ 55° S, ∼ 159° E) reflect a weakening of the SHWW between 11 200–9200 years BP, followed by a period of increased variability between 9200–5300 years BP, and sustained and strengthened winds between 5300–200 years BP (Fig. 5f; Saunders et al., 2018). In southern South America, Moreno et al. (2010) also suggested a progressive southward shift in the SHWW, which they interpreted from precipitation dynamics via plant community changes beginning ∼ 7800 years BP (Fig. 5e; Lago Guanaco and Lago Condorito; 51° [Fig. 5e] and 41° S, respectively, ∼ 72° W). We discuss the impact of precipitation changes in Sect. 4.3. These changes were interpreted to reflect a relatively weak SHWW influence during the Early Holocene followed by a strengthening of the SHWW influence during the middle and late Holocene, which is consistent with Saunders et al. (2018; Fig. 5f).

The timing of inferred changes in the SHWW are not zonally consistent, suggesting regional variability. In the eastern Pacific Ocean in New Zealand (43° S, 171° E), progressive glacier retreat and southward shift in the ITCZ and SHWW occur between ∼ 9800 and ∼ 6900 years ago, suggesting a warming early Holocene (Fig. 5c; Putnam et al., 2012). Whereas in South America, the changes located at Lago Condorito and Lago Guanaco suggest southward migration starting at the earliest at 7800 years ago (Fig. 5e; Moreno et al., 2010). While southern South America might suggest a relatively linear change in the SHWW (Fig. 5e) this does not negate a non-linear shift in the SHWW belt in other regions. The offset between these two records would suggest zonal variability across the Pacific, in agreement with Kaplan et al. (2016).

The regional inconsistency is reflected in the global mean surface temperature (GMST) reconstruction in Osman et al. (2021). We also compare our total dust number concentration results using the GMST reconstruction record throughout the Holocene (11 300–100 years ago) from the Western and Eastern Pacific Ocean 30–55° S similarly to Lamy et al. (2010) to assess temperature (wind) gradients (Fig. S2). We find that average temperature gradients steepen, which we interpret to reflect a strengthening of the winds throughout the Holocene. Statistically, the dust concentration record has a significant anticorrelating relationship to both temperature gradients (p-value < 0.01) for both the western and eastern mid-latitudinal Pacific Ocean. However, the South Pole dust concentration has a stronger anticorrelating relationship to the western Pacific Ocean compared to the eastern Pacific Ocean (r=-0.76 and -0.65, respectively). We interpret this relationship to reflect changes in the Westerly Winds shifts and warm water circulation into the south Pacific gyre, which was also hypothesized by Denton et al. (2021) to explain the onset of mountain glacier retreat during Termination I. Similar atmospheric dynamics may also occur during the Holocene. Furthermore, regional asymmetry in the SHWW can explain our record of a southward shift and strengthening of the winds and cooling of coastal South American temperatures throughout the Holocene (Lamy et al., 2010; Kaplan et al., 2016), and upwelling in Pine Island Bay during the early Holocene (Fig. 5g; Hillenbrand et al., 2017).

Our interpretation of SHWW southward shift during the Holocene should correspond with changes in precipitation. Koffman et al. (2014) used Patagonian and Southern South American precipitation records in conjunction with WAIS Divide dust particle records to infer wind changes. Although the South Pole dust records are consistent with a southward shift in the SHWW, a corresponding increase in precipitation would further suggest such a change throughout the Holocene, rather than a possible decrease in glacial activity and/or vegetation change. Therefore, we assess our inferred dynamics of the SHWW within the context of precipitation variability over the Southern Hemisphere.

From 7000–3000 years BP, South Pole dust concentration decreased by ∼ 40 % (Table S1). While we have shown this decrease is qualitatively consistent with a reduction in Patagonian surface winds, possibly related to SHWW variability (<∼ 51° S; Fig. 5a), increased precipitation in the remote dust source regions (Figs. 4 and 5d and e; Tonello et al., 2009; Moreno et al., 2010) or during transport may also have contributed to this change. Precipitation in Patagonian source regions is closely related to changes in surface wind speed on the Andes western slope with an opposing pattern between wind speed and precipitation on the eastern slope, due to increasing cyclonic activity and the presence of the mountains causing a rain shadow effect (Berman et al., 2012; Moreno et al., 2010; Garreaud, 2007). Precipitation trends in South America vary across regions north and south during the Holocene with icefields in Patagonia (47–48° S, 73.4° W) and Andes experiencing arid conditions during the early (9400–6300 cal yr BP) and late (2400–1600 cal yr BPP) Holocene, while regions in Southern Patagonia (∼ 50° S; ∼ 72° W) received increased precipitation during the mid-late Holocene (Fig. 5d and e; Glasser et al., 2004; Tonello et al., 2009; Moreno et al., 2010). However, both regions show increased in precipitation during the mid-Holocene (∼ 7000–>3000 cal yr BP; Fig. 5d and e; Glasser et al., 2004; Tonello et al., 2009; Moreno et al., 2010; Bamonte et al., 2015). At lower latitudes (∼ 44° S), there was a relative increase in effective moisture during the Holocene (∼ 9500–4200 years BP; de Porras et al., 2014). We note that the CMIP6-PIMP4 output models suggest a broad increase in precipitation over most of South America, with disagreement in the models over the southern South America (Fig. 4a). Paleoproxy lake records (Fig. 5d) also support a progressive increase in precipitation throughout the Holocene suggesting southward migration of the SHWW belt (Tonello et al., 2009).

Reduced dust entrainment in the mid-latitudes may be a result of either increases in vegetation and/or a southward shift in the SHWW. Modelled and paleoproxy increases in precipitation are also observed with increases in Nothofagus compared to grassland vegetation cover (Fig. 5e; i.e., Moreno et al., 2010). The increase in vegetation cover and precipitation can reduce soil erosion and dust entrainment, while increasing dust fallout (i.e., Mahowald et al., 2005; Tegen and Schepanski, 2018). Therefore, the Holocene decrease in South Pole dust may not merely be related to a southward migration of the SHWW away from the dust sources, but also to a rise in precipitation resulting from this shift that also suppressed dust mobilization.

Previous studies have linked CPP variability with changes in air mass advection and/or activation of local dust sources (Delmonte et al., 2020; Koffman et al., 2014; Albani et al., 2012; Wegner et al., 2012; Aarons et al., 2017; Gabrielli et al., 2010). We explore the changes in the SPC14 CPP within the context of the coarse particle concentration. The coarse particle concentration has a similar trend throughout the Holocene as the total particle concentration (Fig. 1a and b), suggesting predominantly similar dynamics between coarse and total particle concentration change. However, because the CPP increases throughout the Holocene (Fig. 1c) there must be an additional forcing mechanism(s) acting upon the coarse particles.

The increase in SPC14 CPP begins ∼ 8000–6000 years BP (Figs. 1 and 5a) coinciding with the timing of glacial retreat and dust source exposure in lower-elevation regions of Antarctica between ∼ 7500–6000 years BP (Fig. 5h; Jones et al., 2020, 2015; Stutz et al., 2021; Stone et al., 2003; Johnson et al., 2014; Hein et al., 2016; Kawamata et al., 2020). For consistency, we recalibrate ¹⁰Be surface exposure ages using production rates from Putnam et al. (2010) and Kaplan et al. (2011; Sect. S1 in the Supplement). Newly exposed source areas in the Weddell Sea sector are plausible origins of dust in SPC14 given prevailing low-level wind patterns (Lazzara et al., 2012; Winski et al., 2021; Clem et al., 2020). However, because the South Pole has multiple air sources, either moving into the East Antarctic plateau via in the Weddell Sea (Winski et al., 2021; Clem et al., 2020; Lazzara et al., 2012) or across West Antarctica (Nicolas and Bromwich, 2011), we interpret the changes in coarse particles to reflect broader scale Antarctic changes rather than to indicate activation of a specific local source (Figs. 2 and 3).

Holocene icesheet thinning, leading to exposure of Antarctic dust sources, is primarily related to (1) ocean warming (Jones et al., 2020; Stutz et al., 2021), (2) change in glacial flow patterns (Hein et al., 2016), and/or (3) reduced buttressing (Stutz et al., 2021). Strengthening and/or southward latitudinal shifts in the SHWW have been shown to increase surface upwelling of relatively warmer circumpolar deep water (CDW) and increase air mass incursions driven by cyclogenesis into the Antarctic plateau (Hillenbrand et al., 2017; Carrasco et al., 2003). It seems likely that contracting, southward shifting, and accelerating SHWW, leading to warming ocean and atmosphere, is linked to the mid-Holocene glacial retreat (i.e., Albani et al., 2012; Koffman et al., 2014) and would increase Antarctic dust sources and the increase in CPP observed in SPC14. However, the coarse dust concentration has a slight decrease and a strong running correlation until ∼ 4000 years BP, suggesting a similar dust source and/or dynamics, which could indicate either (1) deflation of local Antarctic dust sources, similar to Holocene dynamics at Talos Dome (i.e., Albani et al., 2012) or (2) transport from mid-latitudinal sources (Adebiyi and Kok, 2020). Furthermore, the timing of surface elevation decrease around Antarctica corresponds to the SPC14 CPP increase, a southward shift in the Atlantic ITCZ, increasing rainfall in southern South America (Lago Guanaco), all of which can be explained by a southward shift in the SHWW due to increasing Southern Hemisphere temperatures (Fig. 5a, c, and e; Haug et al., 2001; Moreno et al., 2010; respectively). We suggest the CPP increase is primarily related to changes in remote dust contributions in the mid-latitudinal regions and also secondarily related to Antarctic dust source deflation.

A southward migration of the SHWW can simultaneously explain the decrease in South Pole dust concentration and the increase in CPP, which is supported by our CMIP6-PMIP4 model trends and Southern Hemisphere proxy analysis. Southward migration of the SHWW is linked to Southern Ocean CO₂ outgassing (i.e., Moreno et al., 2010; Anderson et al., 2009; Toggweiler et al., 2006). The South Pole dust concentrations and CPP have significant relationships (r=-0.76 and 0.52, respectively, p<0.01) with the composite Antarctic CO₂ record throughout the SPC14 Holocene record (Fig. 5j; Bereiter et al., 2015). However, based on our data, this is outside the scope of this paper.

We suggest that the southward migration of the SHWW was in-part driven by increases in Southern Hemisphere temperatures related to increasing insolation (Fig. 5b and i; Das and Alley, 2008; Laskar et al., 2004). Our SHWW model and proxy-based latitudinal reconstructions suggest that the latitudinal position and/or strength can reflect changes in regional temperature variability over millennial to centennial timescales consistent with previous studies (i.e., Jiang and Yan, 2020; Thompson and Solomon, 2002; Toggweiler et al., 2006). High-latitude zonally averaged Holocene temperature reconstructions (60–90° S latitude) differ, where Kaufman et al. (2020) identify a relatively stable Holocene temperature of 0.04±0.22Δ °C (slope =<0.01Δ °C kyr⁻¹; relative to pre-industrial temperature) and Osman et al. (2021) suggest a more insolation-like increase (slope = 0.11Δ °C kyr⁻¹; relative to last two millennia). The insolation-like structure in the South Pole dust record seems most consistent with the reconstruction of Osman et al. (2021), suggesting progressive, insolation-style, Southern Hemisphere warming and corresponding southward shift in the SHWW.