Dust deposition tracks late-Holocene shifts in monsoon activity and the increasing role of human disturbance in the Puna-Altiplano,northwest Argentina

Mineral content and magnetic susceptibility

The ash (mineral) content of the core, as determined from loss on ignition (LOI), varied substantially throughout the core (Figure 2a), broadly reflecting the four zones identified in the core image and indicating distinct periods of differential mineral input and/or organic accumulation of the studied peat mire. Zone 1 of the core has a consistently high mineral content (mean = 57% by weight). Conversely, in Zone 2 of the core the mineral content is lower (mean = 37%) but more variable. Variability in mineral content increases further in Zone 3, ranging from 11% to 79 %, while the deepest section of the core (Zone 4, 445–585 mm) contained the highest average mineral content of the core (67%).

Magnetic susceptibility κ varied with depth. Values were relatively high and variable in Zones 1 and 4 of the core, with peaks occurring at 22 mm, 52 mm, 117 mm and 457 mm. In Zones 2 and 3, covering the middle of the core, κ was reduced, often negative, and less variable, with the exception of a small peak centred at 200 mm (Figure 2b). Negative κ is likely associated with greater abundance of diamagnetic minerals, for example, carbonate, calcite and quartz, and/or heavy metals (such as Cu and Pb, for example see Figure 8; Maher, 2016).

Grain size

Average median grain size through the core was 41 µm, although varied by (approximately) a factor of two (Figure 2c). In general, relatively coarser grained material was deposited in core Zone 2 and in the bottom half of Zone 3. In Zone 1, the medium grain size ranged between 32.4 and 36.8 µm, whereas Zone 2 contained both larger median grain sizes and a larger size range (44.1–59.7 µm). Zone 3 contained the largest grain size range (31.3–77.5 µm) within the core, while Zone 4 also a displayed a comparatively large grain size range (21.6–49.5 µm), although median grain sizes were generally smaller. Interestingly, there appears to be an inverse relationship between grain size and ash content. That is, Zones 1 and 4 of the core have lower medium grain sizes but higher mineral (ash) content. Conversely, Zones 2 and 3 contain more variable grain sizes and regions of larger grain sizes, in combination with a generally lower but also more variable mineral content. Peaks in mineral content in these two zones are not coincident with grain size peaks.

ITRAX μXRF

The Ti/Mo coherent (coh) ratio (Figure 2b), as determined from Itrax μXRF elemental counts, is commonly used as a measure of allochthonous sediment input in peat and lake cores (e.g. Ohlendorf et al., 2014; Schittek et al., 2016). In SV1, it largely matches ash content (Figure 2a and b), that is Ti/Mo_coh is high in Zone 4, then decreases in Zone 3 before increasing and becoming more variable in Zones 2 and 1. Other high field strength elements (HFSE), for example, Zr/Mo_coh, show a largely identical profile to Ti (data not shown) implying the ash is largely comprised by mineral flux, that is, dust deposition. However, there are some differences between Ti/Mo_coh and ash content, most notably there is a prominent spike in ash content at 288 mm not recorded by Ti/Mo_coh and there is higher variability in Ti/Mo_coh in Zone 1 which is not reflected by the ash data. These differences may represent changes in provenance of mineral matter (dust) deposited in the core.

Because the HFSE are effectively immobile during weathering (e.g. see Babechuk et al., 2015), changes in the ratio between HFSE measured by the ITRAX μXRF data may be expected to provide prima facie evidence for changes in dust provenance in SV1. The structure of the Zr/Ta ratio (Figure 2b) through SV1 broadly matches the ash content, indicating the four zones of the core are potentially characterised by changes in dust provenance (we note however, heavy mineral winnowing can affect Zr concentrations in aeolian systems; Marx et al., 2014a, 2005b). Zones 1 and 4 are characterised by higher variability in Zr/Ta, implying potentially more variability in contributing dust sources at this time. Notably the prominent ash spike at 288 mm does not coincide with a corresponding change in Zr/Ta. Although ITRAX μXRF data provide a useful approximation of changes in SV1, elemental counts can be affected by a number of factors including dilution by the high organic content of the peat (Longman et al., 2019), therefore changes in chemistry/provenance need to be additionally verified.

Ultra trace elements

Differences in geochemical composition of the core can be explored in detail using the ultra-trace element composition of mineral matter in the core. Changes in relative concentrations of conservative trace elements, indicative of changing provenance (Marx et al., 2005b), are most easily identified when the chemistry of core samples are normalised, in this case against Upper Continental Crust (UCC) (Taylor and McLennan, 1995). The UCC normalised REE concentrations indicate there are two main distinctive REE patterns within the core (Figure 3). These broadly conform to the previously identified four core Zones (i.e. see Figure 2). REE concentrations in samples from the top 150 mm and sub 445 mm sections of the core (Zones 1 and 4) are enriched in light rare earth elements (LREE) over heavy rare earth elements (HREE) and display a positive Eu anomaly (Pattern a; Figure 3a). By comparison, samples from the middle Zones (2 and 3; 150–445 mm) of the core have HREE/LREE, and a pronounced positive Gd anomaly (Pattern b, Figure 3b). It is noteworthy, that the chemistry of sediment at the base of core has a similar REE pattern to that in Zones 2 and 3.

OPEN IN VIEWER

Figure 3.

(a)–(c) Rare Earth Elements (REE) in SV1 sediment samples normalised against Upper Continental Crust (UCC) (Taylor and McLennan, 1995). (d) image of the SV1 core and core ash content. The four main Zones in SV1 are indicated on the panel. Coloured bars overlying the core image and letters on panel (d) indicate the depths of samples plotted in panels (a–c). That is, (a) displays SV1 sediments from Zones 1 and 4. The position of these samples is denoted by ‘a’ on panel (d). (b) contains SV1 samples from Zones 2 and 3 which are denoted by the ‘b’ on panel (d), while in (c) a distinctive sample from 288 mm depth is plotted (indicated by ‘c’ in panel (d)) alongside SV1 samples that have intermediate REE patterns between those plotted in panels (a) and (b). These are indicated by ‘d’ on panel (d)). Note that panel (b) has an adjusted scale.

The REE composition of some samples from Zones 2 and 3 of the core depart from the generalised pattern shown in Figure 3b (as indicated in Figure 3d). This included a sample from 193 mm depth which was associated with a mineral content peak and more closely resembles pattern ‘a’ (Figure 3a). In addition, one sample from 288 mm depth has a distinctive LREE > HREE pattern (Figure 3c). Ash from the Zone 2 to Zone 3 transition plot with intermediate REE patters between that of patterns a and b, while samples from Zone 3 containing < 70% ash also plot with intermediate patterns (Figure 3c).

n-alkane hydrogen isotopes

Carbon preference indices (CPI) of C₂₂–C₃₄ n-alkanes (Bray and Evans, 1961), which are indicative of the source of n-alkanes, are in a range typically observed in higher vascular plants, indicating that n-alkanes in the peat samples originate from local vegetation in the wetland. δD values of plants can provide information on the climatic conditions affecting vegetation growth. In general, more negative δD values represent wetter or colder conditions while less negative δD values suggest dryer or warmer conditions (Seki et al., 2011, 2012). δD values of C₃₁ n-alkanes vary with less negative δD of around −200, indicating dryer/warmer conditions in the bottom 445–585 mm section of the core (Zone 4; Figure 2e). Above these depths, δD becomes more negative reaching values between −210 and −220, implying wetter/colder conditions in Zones 2 and 3 (150–445 mm), with the exception of a peak of less negative δD values (−205) at 170 mm. In Zone 1 of the core, above 150 mm, δD values are initially between −210 and −200, before becoming increasingly less negative above 70 mm depth and peaking at the top of the core where δD = −185.

Pollen counts

The ratio of grasses to shrubs, that is, Poaceae (grasses) versus Asteraceae (shrubs), has previously been shown to indicate changing conditions (moisture availability) in the tropical Andes (Liu et al., 2005; Schittek et al., 2016). This is possible because the Poaceae/Asteraceae (P/A) ratio reflects modern precipitation gradients (Kuentz et al., 2012). Consequently, in this study increased Poaceae pollen abundance is interpreted as indicating a wetter environment, while the greater occurrence Asteraceae pollen signifies a more stressed environment, due to reduced moisture availability or disturbance (Schittek et al., 2015; Torres et al.,2018, 2019). Figure 2e shows Asteraceae pollen is more abundant in Zones 1 and 4 and of the core. This pattern is similar to the ash content and n-alkane δD, although notably the δD/Asteraceae relationship breaks down in the top 150 mm of the core. The P/A ratio (Figure 2e) exhibits two substantial peaks in the middle of the core at 256 mm and 404.5 mm depth, suggesting enhanced moisture availability, matching more negative n-alkane δD values. Conversely, in the top and bottom parts of the core the P/A ratio remains low.

²¹⁰Pb and fallout radionuclides

Unsupported ²¹⁰Pb activity measured in the top of the SV1 decreases steadily with depth from 199 Bq/kg to 48 Bq/kg between 2.5 mm and 42 mm (n = 5, Supplementary Table 5, available online). The activity of samples below 64 mm depth (n = 3) was within counting error, therefore these samples are excluded from further consideration. Lead-210 chronologies were calculated for CRS and CIC age models (Supplementary Table 5, available online). The CRS model is more routinely applied in peat and lake studies (Marx et al., 2016), and is generally regarded as more reliable (Appleby, 2001). It is therefore used in this study, although its appropriateness is evaluated by comparison to atmospheric ^{239 + 240}Pu and ²³⁶U fallout.

The appearance of the fallout radionuclides ^{239 + 240}Pu (from atmospheric nuclear weapons testing) in the environment occurs from 1952 CE, reaching maximum concentrations in 1963 CE (Crusius and Anderson, 1995; UNSCEAR, 2000). Likewise, concentrations of ²³⁶U, which is also a fallout radionuclide, show the same temporal pattern as ^{239 + 240}Pu (albeit at lower concentrations) and have begun to be used as chronostratigraphic markers (Ketterer et al., 2013; Srncik et al., 2014; Wendel et al., 2013a, 2013b). In SV1, both ^{239 + 240}Pu and ²³⁶U show a characteristic fallout concentration structure, exhibiting a well-defined peak which corresponds with the date ~1963 CE at 30 mm depth (Figure 4a). The onset of fallout accumulation in SV1 occurs between 45 and 51.5 mm depth, indicating this depth equates to the early 1950s CE.

OPEN IN VIEWER

Figure 4.

(a) The CRS ²¹⁰Pb ages (brown line) and ^{239 + 240}Pu (dotted blue line) and ²³⁶U (dotted green line) concentrations plotted through SV1. The depth of the years, post-1952 and 1963 CE, as indicated by the chronostratigraphic position of ^{239 + 240}Pu and ²³⁶U concentration changes, are indicated on the figure, (b) The Bacon derived Bayesian age model for SV1.

The depth at which maximum ^{239 + 240}Pu and ²³⁶U fallout occurs (i.e. 1963 CE) matches the position of the year 1963 CE derived from the ²¹⁰Pb CRS model, implying the age-structure of that section of core is well constrained. By contrast, neither the CRS nor CIC (not plotted) ²¹⁰Pb modelled ages match the position of 1952 CE as suggested by ^{239 + 240}Pu and ²³⁶U, with ²¹⁰Pb ages older by ~50 years (Figure 4; Supplementary Table 3, available online). This age discrepancy implies the assumptions of the CRS and CIC ²¹⁰Pb model are invalid, or there is mobility in either ²¹⁰Pb (Urban et al., 1990) or ^{239 + 240}Pu (Kaplan et al., 2006) and ²³⁶U in the peat profile, as has been demonstrated for peats, like SV1, with low pH and high organic content (Lee et al., 1997; Sohlenius et al., 2013). Another possible explanation is that the difference between the ²¹⁰Pb model and the ²³⁶U and ^{239 + 240}Pu concentration profile could be caused by slight bioturbation of the surficial samples by rootlets. Alternatively, this discrepancy could result from a change in the provenance of dust deposited to the SV1 site, with ²¹⁰Pb concentrations in dust a function of trajectory length/atmospheric residence time (Marx et al., 2005a). Consequently, there remains some uncertainty in the ages between 30 and 50 mm depth.

Radiocarbon ages

As previously described radiocarbon ages were calibrated using the INTCAL13 curve (Reimer et al., 2013) in the Bacon package in R (Blaauw and Christen, 2011). Despite the study site being in the Southern Hemisphere, INTCAL13 was used to calibrate the ¹⁴C ages as the southerly position of the ITCZ over South America during the austral summer (the growing season) delivers Northern Hemisphere CO₂ to the tropical Andes (Schittek et al., 2016). Use of the Northern Hemisphere calibration curve is therefore consistent with previous research in this region (e.g. Abbott et al., 2003; Wolfe et al., 2001).

Radiocarbon ages increased with depth between 150 mm and 585 mm (Supplementary Table 4, available online). The oldest date, at the base of the core (585 mm) returned an age of 4040 ± 58 cal. yr BP, indicating peat growth initiated around this time. Returned ages from Zone 1 contained modern radiocarbon (Supplementary Table 4, available online), despite occurring below the depth (50 mm) where modern peat would be expected based on the ²¹⁰Pb chronology. These samples are also below the depth of fallout derived ^{239 + 240}Pu and ²³⁶U in the peat, that is, they occur before the onset of bomb fallout. Therefore, it appears these ages have been contaminated by modern C, most likely in the form of roots introducing younger C. They were subsequently excluded from the age model (Figure 4b). The six other ¹⁴C ages between 254 and 585 mm were obtained on leaf fragments and are therefore not affected by the introduction of younger C by roots. Unlike the three modern samples, the samples between 254 and 585 mm date sequentially with depth and suggest distinct periods of different accumulation rates within the core (Figure 4b).

Age model construction

An age model for SV1 was constructed using the five ²¹⁰Pb ages and six selected ¹⁴C ages using the Bayesian programme Bacon in R (Blaauw and Christen, 2011). The Bacon package is designed specifically for use with lake and peat sediment cores and uses a Markov chain Monte Carlo (MCMC) sampling algorithm with prior information about accumulation rates to create an age model with probability intervals based on input dates. The age model was run using the default settings of Bacon (not shown). This captured changes in accumulation rate apparent from the ¹⁴C dates and indicated a mean accumulation rate of 50 yr/cm. Ages for the section of core between the deepest ²¹⁰Pb age and the shallowest ¹⁴C date (42–254 mm) were poorly constrained due to the lack of age control for this part of the core.

As previously discussed, there are four distinct zones within SV1 denoted by changes in organic content and peat structure in addition to mineral content and geochemistry (Figures 2 and 3). They are also evident (albeit more subtly) within pollen concentrations and n-alkanes results, as well as changes in grain size (Figure 2). Therefore, it is likely the zones represent different palaeo-environmental conditions, resulting in different peat growth rates. Consequently, breakpoints were inserted between each zone of the core; namely at 150 mm, 240 mm and 455 mm representing Zones 1 and 2, 2 and 3, and 3 and 4, respectively. Mean accumulation rates based on dated samples were used as inputs for each zone as prior knowledge of accumulation rates is crucial to the successful functioning of Bacon (Blaauw and Christen, 2011). In addition, the Bacon model’s memory function was reduced to reflect the likely significant changes in accumulation rate between zones, implied by peat structure and physical and geochemical data. The resulting age model including breaks (Figure 4b) is broadly similar to the default age model (not shown), although more (visually) organic core zones exhibit faster growth rates, as would be expected. Consequently, Zone 2 exhibits a slower accumulation rate in the breaks model. In addition, the faster growth rate of the upper core is extended deeper, that is, to 15 cm (note growth rates in Zone 2 are inferred as no chronology was obtainable for this zone). Below 240 mm depth (Zones 3 and 4), the difference between the two age models is low (<5%), whereas between 60 and 240 mm depth (Zones 1 and 2) model ages diverge by up to 54% (at 150 mm), although median ages are within error. Overall, the breaks model is preferred as it more closely resembles sedimentary zones of the core, which likely represent different peat growth rates.

Ash composition in SV1

The topology of the mineral content and conservative element concentrations/counts are largely identical through the core (Figure 2), indicating mineral content is largely controlled by the input of material from the same general source, most likely aeolian dust. A notable exception occurs at 288 mm depth, where a spike in the ash record (Figure 2a) is not matched by a corresponding change in some conservative elements (e.g. Ti/Mo coh; Figure 2b), and characterised by a unique REE pattern (Figure 3b), implying the spike has a distinctive provenance. Inspection of the ash by SEM imaging revealed the presence of numerous volcanic glass shards (Figure 5a) confirming the presence of a cryptotephra at this depth.

OPEN IN VIEWER

Figure 5.

(a) SEM photomicrographs showing volcanic glass shards within the ash content of the SV1 at 286 mm depth, (b) A large number and diverse range of diatom species are also present in the core.

The cryptotephra, which forms a broad peak in the core ash content, represents a ~200-year period in the age model (Figure 6a), although ages bracketing the cryptotephra are within error (Supplementary Table 4, available online) indicating it represents a short duration high magnitude event. A number of eruptive events between 1200 cal. yr BP and 2000 cal. yr BP (Fauqué and Tchilinguirian, 2002; Ginibre and Wörner, 2007; Hermanns and Schellenberger, 2008; Sampietro-Vattuone and Peña-Monné, 2016; Wörner et al., 2000) could have deposited the cryptotephra, however their timing is poorly constrained, thus the source of the cryptotephra remains unknown at present. Although there were no other pronounced geochemical perturbations indicating primary cryptotephra in the core, SEM imaging revealed occasional glass shards within the ash samples. These are thought to represent secondary tephra, entrained alongside dust, for example, aeolian tephra (e.g. Hadley et al., 2004).

OPEN IN VIEWER

Figure 6.

(a) Continuous ash derived Dust flux (DUST_ACC) to SV1. (b) Dust flux to SV1 estimated using REE concentrations (DUST_REEC) (purple line) plotted alongside DUST_ACC flux (blue line). Note DUST_ACC has been temporally rescaled to match DUST_REEC to aid comparison. (c) The DUST_REEC/DUST_ACC ratio through SV1. (d) Pb Flux to SV1. Please refer to the Online version for the color figure.

SEM imaging revealed diatom frustules, consisting of a diverse assemblage of genera, are a dominant constituent of core ash (Figure 5). Diatoms can be transported as dust (Bristow et al., 2009) and the large endorheic playas on the Puna-Altiplano contain significant deposits of diatomaceous sediments (e.g. Achem et al., 2014; McGlue et al., 2013; Tchilinguirian et al., 2014), implying they are component of dust. Diatoms could alternatively live in situ within the peat matrix, although the low pH of SV1 (generally < 4) would likely restrict diatoms to a few tolerant genera (mainly Eunotia, Nitzschia and Pinnularia; Denicola, 2000). The diverse diatom assemblage found in the core ash is more likely to flourish in the more neutral/basic conditions common to the endorheic basins of the Puna (López Steinmetz and Galli, 2015; McGlue et al., 2012).

As diatom frustules are comprised almost exclusively of SiO (Supplementary Figure 1c and d, available online), the ash content (containing diatoms) should vary from the trace element concentration (mineral dust only) if they have grown in situ within the peat. Therefore, their origin can be explored further by comparing dust flux derived from the total ash content ( D u s t A C C ; ash content/divided by age), with dust flux estimated using the REEs ( D u s t R E E C ; Eq. 1) (after Vanneste et al., 2015).

D u s t R E E C = R E E A R i × ∑ [ R E E ] U C C × 10 , 000

(1)

where R E E A R i is the REE accumulation rate (μg/m²/yr):

R E E A R i = ∑ [ R E E ] i × A R P i × d i

(2)

∑ [ R E E ] i , the sum of REE concentrations; A R P i , the peat accumulation rate (cm/yr), determined using the age model; d i is sample density (g/cm³); ∑ [ R E E ] U C C , the sum of REE in UCC (Taylor and McLennan, 1995).

D u s t A C C and D u s t R E E C flux show the same generalised pattern through SV1 (Figure 6b), implying that diatoms largely represent a part of the dust load rather than growing in situ at the study site. Minor variability between the two approaches is apparent and is best visualised by plotting D u s t R E E C and D u s t A C C ratio (Figure 6c), with highest ratios occurring when dust flux is lowest between 2600 cal. yr BP and 400 cal. yr BP (Zones 2 and 3). Diatom frustules form a lower component of the ash at that time, indicating less dust derived from the diatomaceous rich playa lakes. For simplicity, the more conservative D u s t A C C flux calculation is referred to in the remainder of the discussion.

Dust flux and dust provenance in SV1

Variability in dust flux in SV1 generally matches the four identified zones in the core (Figure 2). That is, dust flux between 4200 cal. yr BP and 2600 cal. yr BP (Zone 4) was relatively high (>15 to > 25 g/m²/yr), with peak flux occurring at the Zone 4 to Zone 3 transition. In the middle of the core at 2600−400 cal. yr BP (Zones 2–3) flux reduced to < 10 g/m²/yr, with exception of the deposition of the cryptotephra (approx. 1750 cal. yr BP). After 400 cal. yr BP (Zone 1), dust flux increased significantly to > 50 g/m²/yr and became more variable. The four core Zones are demarked more clearly by changes in geochemistry (Figures 2 and 3), with REEs (Figure 3), Zr/Ta ratios and magnetic susceptibility (Figure 2) all implying dust deposited before 2600 cal. yr BP (Zone 4) and after 400 cal. yr BP (Zone 1) has a different provenance to dust in Zones 2–3. For example, increased κ in Zones 1 and 4 indicates deposition of minerals with high magnetic susceptibility including Fe oxides (strong, positive charge) and Fe-bearing silicates (weak, positive charge) (Maher, 2016).

As previously discussed, the endorheic basins of the Puna-Altiplano are the most likely source of the dust to SV1, however, source areas to the south and east could also contribute. The geochemistry of potential regional dust source sediments (PSS) has been mapped by Gaiero et al. (2013) and Gili et al. (2017) who analysed the REE composition of PSS from the Puna-Altiplano, and northern and central western Argentina. Conservative trace elements, such as REEs have been used to determine dust provenance in a number of studies (see Marx et al., 2018), and therefore the REE chemistry of these PSS can be compared with the sediments of SV1 to examine their provenance (Figure 7). SV1 sediments from Zones 1 and 4 were marked by MREE enrichment, positive Eu anomalies and LREE > HREE by comparison to UCC (Figure 3). Existing PSS from the Altiplano largely come from the Salar de Uyuni, some of which exhibit an REE composition similar SV1, although have less pronounced MREE enrichment and plot with a more sub-parallel pattern with respect to UCC (Figure 7). PSS from central western Argentina are largely south of the study site and plot with HREE > LREE. The closest matching PSS, from Salina La Antiga, has reduced HREE > LREE but lacks Eu and Gd enrichment of SV1 sediments. By comparison, PSS from the Puna match the SV1 sediments much more closely, including sediment from Laguna De Pozuelo, a perennial lake within a large endorheic basin proximal to SV1 (~80 km west), and Salar de Arizaro, a large salt playa 350 km southwest. Both these lakes are surrounded by significant fans and deltaic sediments from which dust is likely to be entrained. Therefore, the endorheic basins are most likely source of dust deposited in Zones 1 and 4 of the core.

OPEN IN VIEWER

Figure 7.

(a)–(c) The REE concentrations of potential dust source sediments (PSS) from the Altiplano, the southern and northern Puna, and northern and middle central western Argentina (sensu Gili et al., 2017) plotted normalised to UCC (Taylor and McLennan, 1995). Samples where the normalised REE patterns more closely match those of SV1 sediments are highlighted by colour on the three panels. (d) the closest matching PSS plotted alongside SV1 sediments from Zones 1 and 4. PSS data sample locations from Gaiero et al. (2013) and Gili et al. (2017), are provided in those publications, however, the location of key PSS samples is also indicated in Figure 1. Note the superscript 1 refers to samples sieved to < 63 μm, superscript 2 to alluvial samples and superscript 3 to pan samples. Please refer to the Online version for the color figure.

SV1 sediments from Zones 2 and 3 (Figure 3) have HREE > LREE and a more pronounced Gd anomaly. None of the existing analysed PSS samples display this pattern (Figure 7), although their composition is again most similar to samples from the Puna. The REE composition of this sediment closely resembles sediment from the base of the core (Figure 3) implying it is either locally sourced dust, potentially derived from the arid inner montane basin directly below the study site to the west, that is, the high elevation intra-montane basin south of La Quiaca. In addition, or alternatively, these sediments may also contain a component of very localised alluvial material from the cirque itself. A further possibility is that there is greater dust transport from the lowlands (the Chaco) directly east of the SV1 site. Increased sampling of PSS is required to test these scenarios however.

High dust flux in response to reduced SASM effectiveness; 4200–2600 cal. yr BP

Between 4200 cal. yr BP and 2600 cal. yr BP (Zone 4), SV1 recorded high to moderate dust flux (15–30 g/m²/yr; Figure 6a) and relatively small grain size (Figure 2), while the REE chemistry of core sediments resembled that of the endorheic basins west of the study site. Collectively, this suggests the playa lakes and other dust producing landforms (fans, river channels, etc) were active, implying overall low lake levels, that is, sufficient subaerial exposure for deflation, and possibly occasional recharge events.

In addition to dust flux, C₃₁ n-alkane δD in SV1 provides further information on climate at this time. More negative δD is generally interpreted as wetter or colder conditions (Seki et al., 2011, 2012), although in the Central and Tropical Andes, it has been interpreted to dominantly reflect precipitation (Bird et al., 2011a; Fornace et al., 2014), as values are highly correlated to precipitation (Sachse et al., 2012). Less negative δD between 4200 cal. yr BP and 2600 cal. yr BP (Figure 2e) therefore implies drier, although also potentially warmer conditions at the core site, as also implied by greater dust flux. Further evidence of dry conditions is provided by low Poaceae/Asteraceae (P/A) ratios (Figure 2e). P/A reflects aridity as shrubs in the Asteraceae family are xerophytic with deeper roots allowing them to proliferate in drier conditions relative to Poaceae (grasses) (Liu et al., 2005). Interestingly, peaks in Asteraceae pollen lag behind δD (Figure 2e), implying mire vegetation is more sensitive to moisture stress by comparison to vegetation over the broader landscape.

Although there are few complimentary records to SV1 from the Puna itself, dry conditions are indicated by reduced geomorphic activity and fluvial sedimentation from 4700–2900 cal. yr BP (Tchilinguirian et al., 2014). Similarly, the accumulation of loess in the sub-montane Quebrada de Humahuaca river valley (at ~24°S), was used to infer dry conditions between 6400 and 1400 ¹⁴C years (Alcalde and Kulemeyer, 1999). More definitive records of regional climate at this time come from the Altiplano and Central Andes north of the Puna, this includes from wetlands in the Bolivian Andes which indicated drier conditions after 4 ka (Servant and Servant-Vildary, 2003) and from multiple lake records which indicate the onset of pronounced aridity after 7000–6000 cal. yr BP, as inferred from δ¹⁸O from cellulose in sediments (Abbott et al., 2003; Wolfe et al., 2001). Overall therefore, there is significant evidence of regionally dry conditions across the Puna-Altiplano concomitant with increased dust flux in SV1 between 4200 cal. yr BP and 2600 cal. yr BP.

Because precipitation in the Puna-Altiplano is dominated by the SASM (Bird et al., 2011a; Garreaud et al., 2003), drier conditions between 4200 cal. yr BP and 2600 cal. yr BP imply the SASM was less effective over the Puna at this time, that is, either monsoon activity was diminished or the SASM was located further north. This is consistent with the interpretation of lake records from the Central Andes, which indicate diminished glaciers and moisture implying reduced SASM effectiveness during the mid Holocene (Abbott et al., 2003; Wolfe et al., 2001).

A pronounced spike in dust flux at the end of Zone 4 (~2700–2600 cal. yr BP; Figure 6a) coincides with reduced C₃₁ n-alkane δD, which would indicate wetter conditions (Figure 2e). The reasons for these apparently contradictory conditions are uncertain. In some arid environments, such as Australia, dust has been shown to be limited by sediment availability and is therefore partly related to fluvial recharge events (Bullard and McTainsh, 2003; Marx et al., 2005a, 2009, 2018), that is, the relationship between dust emissions and aridity is non-linear (Marx et al., 2018). In the Puna, increased frequency of episodic intense rainfall interspersed with drier conditions may also result in increased sediment availability and enhanced dust emissions. This peak also coincides with increased variability in dust geochemistry and grain size (Figures 2 and 3), implying changes in dust provenance associated with rapidly changing landscape conditions. The period around 2800–2600 cal. yr BP is also associated with rapid drying and wetting in peat bogs in both Europe and in Tierra del Fuego ascribed to a solar minimum (see Chambers et al., 2007), implying this was a period of significant climate variability, which appears to have resulted in increased dust flux to SV1.

Reinvigoration of the SASM and suppressed dust emissions; 2600–1700 cal. yr BP

After 2600 cal. yr BP conditions in SV1 change substantially. Peat accumulation increases as does organic content (Figure 2) and dust flux reduces < 10 to <5 g/m²/yr (ignoring the effect of the cryptotephra at 1860 cal. yr BP; Figure 6a), indicating more humid conditions in the Puna. Biological proxies in SV1 also imply wetter conditions. C₃₁ n-alkane δD decreases to approximately −220 (the lowest in SV1), Asteraceae pollen counts decrease, while peaks in the P/A ratio occur (Figure 2). A corresponding increase in median grain size suggests dust was sourced from more proximal locations (although we note increased grain sizes can reflect both transport distance and wind speed (Tsoar and Pye, 1987)). Changes in dust chemistry also indicate a change in dust source areas after 2600 cal. yr BP (Figure 3), with REE patterns showing a lower degree of match with existing PSS from the Puna (Figure 7), while similarly Zr/Ta ratios and magnetic susceptibly κ decrease (Figure 2b).

The return of more humid conditions at the study site mirrors the findings of other studies undertaken across the Puna-Altiplano. Lakes in the Altiplano recorded the return of moisture and glacial ice around this time (Abbott et al., 2003; Wolfe et al., 2001). Interestingly, drier conditions were found to persist longer at more southerly sites (closer to the SV1), indicating reinvigoration of the SASM from north to south, with the southernmost lake, Laguna Potosi at ~20°S, becoming wetter after 2500 cal. yr BP (Abbott et al., 2003), similar to SV1. Glaciers, which had disappeared from many catchments by 10,000 BP also returned progressively from north to south, returning to the catchment of Laguna Paco Cocha at 14°S by 4800 cal. yr BP, and catchments at 16°S by 2300 cal. yr BP (Abbott et al., 2003). Further evidence of increasing humidity is provided by re-establishment of wetlands and increased storminess (indicated by sedimentation) (Servant and Servant-Vildary, 2003), while to the South of SV1 on the Puna, evidence of increased humidity occurred from ~1700 cal. yr BP to 1400 cal. yr BP (Alcalde and Kulemeyer, 1999; Schittek et al., 2016).

Collectively, evidence from the SV1 and other records indicate more moist conditions after ~2600 cal. yr BP. This was likely associated with increased southward penetration and/or reinvigoration of the SASM over the Puna-Altiplano (Figure 8c), which has been attributed to increased summer insolation and associated wet season convection (Abbott et al., 2003; Bird et al., 2011b). The increase in humidity at this time significantly reduced dust emissions from the Puna, presumably as a result of increased flooding of ephemeral lakes and increasing vegetative cover reducing dust entrainment from other environments. As previously discussed, the REE chemistry of sediment in SV1 from this time matches the basal core sediment implying more local dust and/or alluvial input to the mire.

OPEN IN VIEWER

Figure 8.

Schematic representing possible prevailing climatic and environmental conditions for each zone of the core: (a) Zone 1 (Present–400 cal. yr BP), (b) Zone 2 (400–1700 cal. yr BP), (c) Zone 3 (1700–2600 cal. yr BP), and (d) Zone 4 (2600–4200 cal. yr BP).

Low dust flux and increasing snow cover(?) indicating enhanced SASM activity: 1700–400 cal. yr BP

After 1700 cal. yr BP, dust flux remained low (<5 g/m²/yr) until 400 cal. yr BP, with the exception of a 300-year period from 1000 cal. yr BP to 700 cal. yr BP when dust flux increased to ~10g/m²/yr (Figure 6a). Sustained negative C₃₁ n-alkane δD values (−210 to −220) (although δD rose after ~1000 cal. yr BP), low Asteraceae counts (Figure 2e) and a prominent spike in the P/A ratio suggests continued humid conditions. Although, both dust flux and biological proxies remained similar to Zone 3 (2600–1700 cal. yr BP), the physical structure of the core changed after 1700 cal. yr BP, becoming darker grey and more humified (Figure 2), while the peat accumulation rate decreased (Figure 4b). This implies that despite broadly similar regional climate conditions, environmental conditions at the core site itself are different. Together, the change in peat composition and reduced accumulation rate imply cooler conditions possibly associated with increased snow cover at the SV1 site.

Presently in the Eastern Cordillera, the 0°C isotherm is at ~4000 m a.s.l. in the free troposphere (Garreaud, 2009). This is lower than the SV1 site (4300 m a.s.l.), implying snow cover at the study site would be sensitive to small changes in temperature and/or precipitation. The site position is conducive to the persistence of snow late in the season; the cirque has a headwall to its north and an easterly aspect, limiting direct solar radiation. Therefore, changes in snow cover/persistence could have driven the changing peat structure after 1700 cal. yr BP. Indeed, it is likely that the cirque aspect and its effect of limiting evaporation/sublimation facilitated peat development, which is otherwise sparse in the Central Andes.

The interpretation of cooler conditions between 1700 cal. yr BP and 400 cal. yr BP is supported by records of glacial extent. Glaciers re-established in the Central Andes north of 16°S, from ~2300 cal. yr BP (Abbott et al., 2003) and reached maximum late-Holocene extent at ~600–300 cal. yr BP (Jomelli et al., 2009), concomitant with the period of changed peat structure. Similarly, increased snow persistence is also consistent with records of the establishment of, and increased sedimentation in, wetlands in river basins in the Bolivian Andes, including around Lake Titicaca (Servant and Servant-Vildary, 2003). That study suggested increased precipitation between 2700 cal. yr BP and 500 cal. yr BP was driven by a greater frequency of southerly and westerly airmass incursions resulting in more even annual precipitation. However, that conclusion appears at odds with increasing SASM activity from the north after ~2300 cal. yr BP (Abbott et al., 2003) and present-day conditions in which increased westerlies coincide with drier conditions on the Altiplano (Garreaud et al., 2003) and increased frequency of El Niño events (Moy et al., 2002), also linked to dry conditions (Garreaud et al., 2003). Overall, the exact mechanisms for the change in peat structure and sedimentation at the site remain unclear, however, the hypothesis of increased snow cover at the study site is not inconsistent with existing regional evidence or a shift to a period of enhanced SASM activity.

The REE chemistry of dust deposited in the SV1 between 1700 cal. yr BP and 400 cal. yr BP is similar to that deposited between 2600 cal. yr BP and 1700 cal. yr BP, and considered to represent increased relative input of local sediment. However, a change in REE patterns occurs between 1000 cal. yr BP and 700 cal. yr BP concurrent with increased dust flux. These sediments had LREE > HREE and a pronounced positive Eu anomaly similar to dust deposited in Zones 1 and 4 of the core (before 2300 cal. yr BP and after 400 cal. yr BP) (Figure 3). This occurred alongside changing HFSE ratios (i.e. increased Zr/Ta ratios) and increased magnetic susceptibility κ. C₃₁ n-alkane δD also increased, albeit slightly after the dust peak, implying a shift to more arid conditions in the Puna coinciding with the MCA, and mirroring other studies from the Central Andes (Bird et al., 2011b; Schittek et al., 2016). However, more arid conditions during the MCA have also been linked to the intensification of landuse, and regional-scale social reorganisation (Kulemeyer, 2005; Lupo et al., 2007; Morales et al., 2009; Schittek et al., 2016). Similarly, increased dust flux and Hg contamination in ice and lake cores at this time was attributed to pre-Incan human activity (Cooke et al., 2011; Thompson et al., 1988). It is possible that both a changing climate and human activity contributed to the doubling in dust flux between 1000 cal. yr BP and 700 cal. yr BP, demonstrating the sensitivity of dust emission from the Puna to changes in aridity/landuse.

The increasing impact of human activity in the Puna-Altiplano and unprecedented dust flux after 400 cal. yr BP

After 400 cal. yr BP, dust flux increased markedly to > 60 g/m²/yr, reaching a peak of 173 g/m²/yr in the late 20th century (Figure 6). This is approximately five times higher than maximum flux recorded over the previous 4200-year period. Enhanced dust flux is paralleled by a reduction in grain size (Figure 2), potentially representing deposition of more distally sourced dust, while REE of core sediments again match PSS from endorheic lakes on the adjacent Puna more closely (Figure 3).

Elevated dust flux after 400 cal. yr BP could result from an increase in ephemeral conditions on the Puna with n-alkane δD indicating a return to drier/warmer conditions (Figure 2e). Drying of lake basins following ~2200 years of wetter conditions and active sediment recharge would be expected to expose fresh sediment deposits to the airstream. Increased climate variability has been hypothesised to drive increased dust flux due to enhanced sediment recharge during wet phases and dust entrainment during dry phases (Marx et al., 2009). In SV1, C₃₁ n-alkane δD implies enhanced climate variability after 400 cal. yr BP. In general, δD became less negative, indicating drier/warmer conditions after 400 cal. yr BP, however, between 320 cal. yr BP and 130 cal. yr BP δD become more negative (centred at ~1750 CE) implying a short return to wetter/cooler conditions, matching cooler conditions in the Quelccaya and Sajama Ice Caps (Thompson et al., 2013; Vimeux et al., 2009), glacial expansion in the Tropical Andes (Jomelli et al., 2009) and coinciding with the timing of the LIA. This is followed by δD reaching values as low as −185 in the top of the core, indicating significant drying/warming during the 20th century, as also recorded in the Quelccaya Ice Cap (Thompson et al., 2013).

Although natural climate variability may have contributed to increased dust flux in SV1 after 400 cal. yr BP, dust flux after that time is unprecedented by comparison to the period before 2600 cal. yr BP when conditions were also variable (Figure 6a), suggesting other factors may have also contributed, such as human activity, which is known to influence dust emissions (Hooper and Marx, 2018; Webb and Pierre, 2018). There is a long history of human activity in the Puna-Altiplano with evidence of a pastoral society involving the domestication of llamas from as early as 2600 cal. yr BP (Morales et al., 2009). Despite this, early agriculture did not appear to result in significant landscape disturbance or increased dust flux, although, as previously discussed, pre-Incan activity may have also contributed to increased dust flux during the MCA. Expansion of the Inca Empire into the Puna between 1450 CE and 1570 CE resulted in increased mass accumulation rates (MARs) in regional lakes, although the environmental impact was limited in comparison to the Spanish Colonial period that followed (Lupo et al., 2006). The period beginning with the Spanish arrival was associated with unprecedented environmental degradation between the 1600s and late 1900s CE. This included enhanced erosion, valley incision and increased sedimentation in lakes and bogs resulting from landuse intensification and the introduction of cattle (Lupo et al., 2006; Sampietro-Vattuone and Peña-Monné, 2016; Schittek et al., 2016). These changes appear to have driven a substantive increase in regional dust emissions from the Puna, as recorded in SV1. It is noteworthy that similar to this study, elevated dust flux attributed to human landuse has been recorded in a number of regions globally since the 18th century CE (Hooper and Marx, 2018).

In addition to enhanced dust flux, SV1 records change to biological/climate proxies attributable to human impact. Asteraceae pollen and n-alkane δD covaried, albeit with a lag, over most of the length of SV1, indicating Asteraceae increases during drier conditions. However, the two proxies diverge after 400 cal. yr BP (Figure 2e), that is while δD indicates humid conditions during the LIA, increasing Asteraceae pollen counts imply increased aridity. This suggests a decoupling of the climate control on vegetation. Shrubs are known to succeed in environments where soil erosion and degradation has resulted in reduced grass cover (Chartier and Rostagno, 2006), as would be expected following the introduction of hard-hooved livestock grazing and associated removal of stabilising vegetation and the break-up of soil crusts (Okin et al., 2006). It is probable that increasing anthropogenic activity in an environment experiencing changes in climate from wetter to drier conditions combined to make the landscape more susceptible to wind erosion, resulting in the elevated high dust flux recorded in SV1 since the late 16th century CE.

As well as changing dust flux, increased human activity in the Puna-Altiplano region is also implied by increasing Pb flux to SV1. Pb flux increased from approximately 1 mg/m²/yr before 400 cal. yr BP to > 4 mg/m²/yr after this time, likely due to Incan and Spanish metallurgy. This matches records of increasing human disturbance in the Puna (Lupo et al., 2006; Sampietro-Vattuone, Peña-Monné, 2016; Schittek et al., 2016), which in combination with increasing dust flux and vegetation/climate relationships in SV1 all suggest substantive human landscape disturbance after 400 cal. yr BP. It is noteworthy that concomitant changes in dust flux and metal contaminants have been similarly observed in other settings, including Australia, where both increasing dust flux and environmental metal accumulation track landscape disturbance following European settlement (Marx et al., 2014b). After the mid-20th century, Pb flux in SV1 increased further, reaching 11 mg/m²/yr, which is assumed to reflect increases in regional Pb and Ag mining and smelting (Hong et al., 2004; Lupo et al., 2006; Uglietti et al., 2015).

Despite the increase in Pb flux in SV1 after 400 cal. yr BP matching the timing of known increases in regional metallurgy, the anthropogenic origin of increased Pb flux cannot be demonstrated unequivocally. Elevated Pb flux also occurred lower in SV1 between 1800 and 2300 cal. yr BP. This predates the earliest onset of Pb contamination in Bolivia by approximately 700 years (Cooke et al., 2008; although there is significant local variability in the onset of Pb contamination in the Central Andes (Marx et al., 2016)). Elevated Pb flux at the time could alternatively result from the depositing of dust with high natural Pb concentrations, for example, high natural Pb concentrations were recorded in dust in the Sajama ice cap to the north of the study area (Hong et al., 2004). Therefore, changes in dust sources could be contributing to the elevated Pb flux after 400 cal. yr BP. The broad peak in Pb flux between 1800 and 2300 cal. yr BP also coincides with the timing in deposition of the cryptotephra, although the cryptotephra peak itself does not appear to contain elevated Pb concentrations. Calculation of an enrichment factor is a standard approach for identifying the extent of environmental metal contamination, this involves comparing the ratio of a suspected contaminant to a conservative element in the pre-anthropogenic portion with the anthropogenically influenced section of a core (see Marx et al., 2016). In SV1, the calculation of enrichment factors is complicated by high Pb concentrations between 1800 cal. yr BP and 2300 cal. yr BP. Thus, the origin of Pb in SV1 may only be determinable through the use of Pb isotope fingerprinting.