Southern Hemispheric Westerlies control sedimentary processes of Laguna Azul (south-eastern Patagonia,Argentina)

Introduction

The intensity and latitudinal position of Southern Hemispheric Westerlies (SHW) are fundamental for global carbon cycling. A more southern position causes degassing of CO₂ from the deep ocean to the atmosphere, whereas a more northern position decreases ocean-water ventilation resulting in additional atmospheric CO₂ transfer to intermediate water levels (Lovenduski et al., 2008). Today, southward shifting and intensifying winds are increasingly reducing the efficiency of the Southern Ocean as a carbon sink (Le Quere et al., 2007; Lovenduski et al., 2015). This influences the rate of CO₂ accumulation in the atmosphere (Fischer et al., 2010), modifies the natural greenhouse effect and causes droughts on Southern Hemispheric continents (Fletcher and Moreno, 2012). Aside from natural climate variability (Jones and Widmann, 2004), this is related to a mid-latitude temperature increase (Shindell and Schmidt, 2004) and reduced ozone concentration across Antarctica (Kang et al., 2011). Today, SHW are responsible for more than 25% of the interannual global atmospheric CO₂ flux variability (Takahashi et al., 2012). Therefore, it is important to understand the mechanisms that drive SHW and explain their response to global climate change (Toggweiler et al., 2006).

Against this background we investigate a sediment record from Laguna Azul in south-eastern Patagonia. This lake without surficial inflow and outflow is highly sensitive to hydroclimatic shifts and strongly dependent on the precipitation/evaporation ratio, which in turn affects aquatic biota and lacustrine geochemistry (cf. Schindler, 2009). Such lakes typically respond rapidly to environmental variability with changes in lake levels, redox conditions and salinity (Valero-Garces et al., 1996).

A first study of sediments from Laguna Azul focused on the last millennium and recognized a temporal equivalent of the Northern Hemispheric ‘Medieval Climate Anomaly’ (MCA) and the ‘Little Ice Age’ (LIA) as well as recent anthropogenic disturbances (Mayr et al., 2005). In the same region, it was demonstrated for Laguna Potrok Aike that wind-related evaporation is a controlling factor of its hydrological balance (Mayr et al., 2007a; Ohlendorf et al., 2013). Overall, regional environmental variations are linked to SHW, the dominant atmospheric circulation pattern in these latitudes (Garreaud et al., 2013; Toggweiler et al., 2006). SHW are controlled by insolation with complex feedbacks to the ocean-atmosphere-cryosphere system. Negative (positive) correlations between zonal wind and the amount of precipitation are documented for eastern (western) Patagonia (Garreaud et al., 2013; Moreno et al., 2014). Thus, rainfall decreases with intensified SHW east of the Andes and increases west of the Andes. Moreover, strong SHW prevent moist Atlantic low-pressure systems from entering the continent amplifying aridity in eastern Patagonia (Garreaud et al., 2013; Mayr et al., 2007b). In Argentina, droughts reduce the amount of rainfall and thus the availability of water for drinking, agriculture, industry and hydropower production causing socio-economic damage (Berman et al., 2012). Nowadays, permanently strong SHW with little precipitation are one cause for desertification and wind erosion in arid regions of south-eastern Patagonia. Therefore, understanding SHW variability and its forcing provides information to deal with future climate change scenarios in this part of South America.

Despite its importance, little knowledge exists about past SHW variations. Available studies often provide controversial conclusions (Fletcher and Moreno, 2012; Kilian and Lamy, 2012; Moreno et al., 2014) because of different sensitivities of SHW recording, different suitability of proxies for SHW reconstruction and differences in temporal resolution with associated chronological issues. Moreover, it is likely that reconstructions relying on precipitation-related proxies like pollen-based vegetation changes (e.g. Moreno and Videla, 2016) or runoff changes related to minerogenic sediment transfer (e.g. Lamy et al., 2010) are much less pronounced in semi-arid eastern Patagonia compared with west of the Andes, because the amount of rainfall is lower by a factor exceeding 10, east of the Andes (Garreaud et al., 2013). In any case, interpretation of palaeoenvironmental data in relation to zonal wind fields is not straightforward and involves a chain of assumptions (Kohfeld et al., 2013).

Here, we present a high-resolution multiproxy study to assess intensity and position of the SHW throughout the Holocene. Furthermore, we compare our findings with regional hydroclimatic fluctuations to gain a better understanding of past and future climate changes.

Site description

Laguna Azul, a crater lake of the Pliocene to late-Quaternary Pali Aike Volcanic Field (D’Orazio et al., 2000), is one of the few permanent lakes in the semi-arid steppe of extra-Andean south-eastern Patagonia (52°04.7’ S, 69°35’ W, 100 m a.s.l.; Figure 1). Its rugged morphology and the well-preserved pyroclastic ring wall point to a young monogenetic volcanic system (cf. Nemeth and Kereszturi, 2015) supported by ⁴⁰Ar/³⁹Ar ages (0.01 ± 0.02 Ma) of a related lava flow (Corbella, 2002).

OPEN IN VIEWER

Figure 1.

Study area in southern South America with locations mentioned in the text. Laguna Azul is site 18 and indicated by a black square (left). The expanded view of Laguna Azul (right) includes bathymetry with exposed lake-level terraces, coring sites, sampling locations and surrounding topography.

The lake is elliptical in shape (560 m × 240 m; Figure 1) and exhibits overlapping sub-basins with a maximum water depth of 56 m in 2003. Lake shores consist of basaltic rocks except in the west, where sand and lapilli-sized pyroclastics prevail. The catchment area (0.24 km²) is restricted to the inner crater walls and is small in relation to the modern lake surface area (0.15 km²). Laguna Azul is dimictic, holomictic and groundwater-fed without tributaries or outlets (Zolitschka et al., 2006). Continuous logging of limnological parameters from March 2002 to February 2005 reveals a stratified water body with a well-developed metalimnion from 17 to 25 m water depth during southern summers. The lake is classified as mesotrophic to eutrophic with epilimnetic pH values between 8.5 and 9.0, mean electric conductivity of 443 µS cm⁻¹ and salinity of 0.3‰ (Mayr et al., 2005; Messyasz et al., 2007; Zolitschka et al., 2006).

Laguna Azul is situated in the dry Magellanic grass steppe (León et al., 1998). Nearest stands of subantarctic deciduous forests, dominated by southern beech (Nothofagus pumilio and N. antarctica), occur along the eastern foothills of the Andes about 160 km to the west. The climate of south-eastern Patagonia is dominated by westerly winds and a rain-shadow east of the Andes. Correlation fields between local precipitation and zonal wind speed document a low (c. −0.3; Garreaud et al., 2013) to high (c. −0.8; Moreno et al., 2014) negative correlation for the study site. In addition, the weather station at Laguna Potrok Aike (~60 km west of Laguna Azul) shows that precipitation event frequency is dominated by westerly winds. However, occasional events from the east provide very high rainfall (Mayr et al., 2007b). Altogether, these factors result in a cool-temperate semi-desert climate with extremely windy and highly evaporative conditions (Garreaud et al., 2013). At Laguna Azul the sum of annual precipitation is in the range of 200–300 mm with a mean annual temperature of 6–7°C (Oliva et al., 2001).

Materials and methods

Dating

Nineteen AMS ¹⁴C samples were dated at the Poznań Radiocarbon Laboratory, Poland. Of the 15 samples from the composite profile and two from the gravity cores AZU 02/4 and AZU 02/11 (Figure 1), only four contained wood or terrestrial plant remains. Therefore, dating was extended to aquatic material (Table 1). In addition, a sample of aquatic macrophytes was obtained from lake sediments exposed 2 m above the present lake level (Figure 1). To test for a potential reservoir effect, living aquatic macrophytes were also dated (Mayr et al., 2005).

OPEN IN VIEWER

Table 1.

AMS radiocarbon dates from Laguna Azul calibrated with Clam 2.2 (Blaauw, 2010) and the Southern Hemisphere calibration curve SHCal13 (Hogg et al., 2013). Modelled ages (n.d.: not determined) were obtained with the smoothed spline option of Clam 2.2. Samples above the upper dashed line are not from sediment cores. Samples below the lower dashed line are artificially aged by volcanogenic CO₂.

Core section or sample number	Core section depth (cm)	Mean composite sediment depth (cm)	Sample type	Lab. no.	Radiocarbon age (14C BP ± 1σ)	δ13C (‰)	Calibrated 2σ minimum age (cal. BP)	Calibrated 2σ maximum age (cal. BP)	Modelled age (cal. BP)	Error of modelled ages (cal. BP)
PAIS-48 a	Outcrop		Aquatic macrophytes	Poz-5071	1325 ± 30	−10.5	1090	1290	n.d.	n.d.
PAIS-45 a	Modern		Potamogeton sp. b	Poz-3574	104.9 ± 0.4 g	−8.8	−6 to −8 h	−57 to −61 h	−53	−10/+10
AZU-02/4 a	7.0–9.0	29.0	Twig of Berberis c	Poz-3575	157 ± 0.5 g	−25.4	−20	−10	−10	−10/+30
AZU 02/11 a	64.0–66.0	46.0	Charred plant remains c	Poz-8445	145 ± 30	−27.4	0	270	80	−60/+50
AZU 03/5 a	42.0–45.0	53.3	Charred plant remains c	Poz-8447	215 ± 35	−27.1	0	300	120	−80/+70
AZU 03/6-1	41.0–42.0	156.5	Coarse fraction d	Poz-8501	900 ± 30	−20.3	690	900	850	−90/+180
AZU 03/6-1	63.0–64.0	178.5	TIC fraction e	Poz-8624	1360 ± 50	−14.2	1090	1310	1100	−150/+120
AZU 03/4-3	64.0–65,0	261.5	Fish bones	Poz-6099	1700 ± 30	−18.4	1480	1700	1560	−80/+80
AZU 03/6-3	15.0–16.0	335.5	Coarse fraction d	Poz-8553	1830 ± 50	−19.0	1590	1830	1760	−90/+130
AZU 03/6-3	88.5–89.3	408.9	Fish bones	Poz-5976	2410 ± 40	−15.0	2320	2690	2340	−160/+110
AZU 03/6-4	43.0–44.0	464.5	Fine fraction f	Poz-8449	2840 ± 35	−15.1	2790	3000	2970	−80/+160
AZU 03/6-5	12.5–13.5	538.0	Fine fraction f	Poz-8554	4120 ± 40	−23.4	4440	4810	4670	−130/+130
AZU 03/6-5	12.5–13.5	538.0	Coarse fraction d	Poz-8556	4270 ± 40	−27.0	4620	4860	4670	−130/+130
AZU 03/4-7	22.0–23.0	583.0	Bulk sediment	Poz-5975	5280 ± 50	−23.7	5910	6130	6090	−110/+130
AZU 03/4-7	75.0–76.0	638.5	Wood c	Poz-11585	7500 ± 80	−35.1	8050	8410	8220	−290/+120
AZU 03/4-7	80.0–81.0	643.5	Ruppia seed b	Poz-8446	16,710 ± 100 i	−27.3	19.830	20.420	8420	−310/+120
AZU 03/4-8	31.0–32.0	694.5	TIC fraction e	Poz-8552	16,170 ± 110 i	−0.3	19.140	19.780	10.460	−560/+280
AZU 03/4-8	62.2–63.2	725.7	Fine fraction f	Poz-8500	26,420 ± 350 i	−33.2	29.800	31.120	11.710	−710/+380
AZU 03/4-8	63.2–64.2	726.7	Bulk sediment	Poz-5191	22,660 ± 190 i	−32.9	26.460	27.370	11.750	−720/+390

a

Mayr et al. (2005), modified.

b

Aquatic macrophyte.

c

Terrestrial plant remains.

d

Coarse fraction (>100 µm) of bulk sediment.

e

Total inorganic carbon fraction of bulk sediment.

f

Fine fraction (<100 µm) of bulk sediment.

g

Percent modern carbon (pMC).

h

Two probability ranges because of two well-defined intersections with the calibration curve.

i

Date excluded from age-depth model: effected by volcanogenic CO_2.

An age-depth model was constructed with the R package clam 2.2 (Blaauw, 2010). All ¹⁴C-ages were calibrated with the Southern Hemisphere calibration curve SHCal13 (Hogg et al., 2013) and the post-bomb calibration curve SH1–2 for postmodern ¹⁴C dates (Hua et al., 2013). The sediment/water interface (AD 2003) was used as additional tie point. For age modelling the smoothed spline option (type = 4 with spar 0.3) of clam 2.2 was applied to the 13 radiocarbon ages between the dashed lines in Table 1. Beyond the oldest dating point, ages were extrapolated. All calibrated and modelled radiocarbon ages were rounded to decadal values.

Two macroscopically visible volcanic ash (tephra) layers were analysed petrographically and geochemically (cf. supplementary material for more information).

Stable isotopes

Sub-samples for isotopic analyses were freeze-dried, homogenized with a spatula and sieved (200 µm) to eliminate macro remains. Nitrogen stable isotope ratios (δ¹⁵N) were determined on bulk sediment weighed into tin capsules and combusted at 1080°C in an elemental analyser (EuroEA, Eurovector) with automated sample supply linked to an isotope-ratio-mass spectrometer (Isoprime, Micromass). For analyses of stable isotope ratios of organic carbon (δ¹³C_org) samples were decalcified with HCl (5%) for 6 h in a water bath at 50°C to remove calcite and siderite, centrifuged, rinsed repeatedly with deionized water to neutral pH and freeze-dried. Organic carbon-isotope ratios were determined using a Carlo Erba elemental analyser linked to an Optima isotope-ratio mass-spectrometer or with the system described above for nitrogen isotope analyses. Isotope ratios are reported as δ values in per mil according to the equation

δ = ( R s / R s t − 1 ) * 1000 ,

where R_s and R_st are the isotope ratios (¹³C/¹²C, ¹⁵N/¹⁴N) of the samples and international standards (VPDB for carbon, AIR for nitrogen). Analytical uncertainty (1 σ) is 0.14‰ for δ¹⁵N and 0.08‰ for δ¹³C_org.

Results and interpretation of the multiproxy record

Lithology

Based on CONISS, the record from Laguna Azul is subdivided into five lithological units (A–E in Figure 2); Basal unit A (754–728 cm) consists of graded basaltic scoria with grain sizes ranging from volcanic gravel up to ~1 cm in diameter to coarse ash. Because of the minerogenic composition of unit A, analyses were not carried out for pollen, diatoms and stable isotopes. Compared with minerogenic unit A (1.6% BioM), the other units have distinctly higher contents of BioM (15.8–82.7%) and classify as diatomaceous ooze (Table S2, available online). Organic sediments start with unit B (728–686 cm; 15.8% mean BioM) characterized as dark grey biogenic sediment with intercalations of dark and light layers. Unit C (686–641 cm; 37.7% BioM) is black and laminated while unit D (641–588 cm; 42.7% BioM) is characterized by fine laminations changing in colour from dark olive and black to dark grey. The uppermost highly organic lithological unit E (588–0 cm; 65.2% BioM) makes up 79% of the entire record and consists of reddish-brown to brown and faintly layered to homogeneous diatomaceous ooze. CONISS separates it into four subunits (Ea–Ed). Lowermost subunit Ea (588–563 cm; 40.9% BioM) contains two visible volcanic ash layers between 586 and 580 cm (Figure S2, available online) and the first of two dark grey to black layers with intercalated coarse sand grains labelled as sandy layer S1 (568–563 cm). The succeeding subunit Eb (563–466 cm; 62.8% BioM) ends after the second sandy layer (S2; 476.5–474.5 cm) followed by subunits Ec (466–201 cm; 74.2% BioM) and Ed (201–0 cm; 82.7% BioM; Figure 2).

OPEN IN VIEWER

Figure 2.

Lithology of the composite record from Laguna Azul with lithological units (A–E) and subunits (Ea to Ed) versus depth displaying dry bulk density (DBD), magnetic susceptibility (MS), titanium (Ti), total organic carbon (TOC) and the age-depth model with radiocarbon dates (error bars are smaller than the triangle symbols) and modelled age (thick line) with 2σ error margins (thin lines). Also indicated are tephra layers AZU-T1 and AZU-T2 (T1,2) and sandy layers S1 and S2.

Minerogenic sediment components

For minerogenic unit A, the elements Ti and Ca as well as MS and DBD show the highest values of the record. Compositionally, they correspond to the alkali-basaltic bedrock. At the base of lithological unit B, Ti is high (230 cps), then decreases to ~100 cps and finally increases again to 260 cps (Figure 2). The mean value of Ti for unit B is 157 cps. A distinct drop in Ti marks the transition to unit C (mean: 97 cps). At the start of unit D Ti increases to 210 cps. One peak reaches >300 cps but the mean for unit D is 149 cps. From unit D towards the top of the profile Ti steadily decreases to 35 cps. Sandy layers S1 and S2 interrupt this trend with higher values

Total inorganic carbon (TIC) summarizes different carbonate minerals, for example, calcite, rhodochrosite and siderite. The variability of this proxy is similar for units B–D. At the beginning of each unit, TIC values increase to >0.5% (labelled TIC1–TIC3 in Figure 3) and thereafter diminish to <0.2%. This pattern changes to a higher frequency (TIC4–TIC10) throughout unit E (Figure 3). Two of these maxima (TIC4, TIC7) coincide with sandy layers S1 and S2. The last three maxima (TIC8–TIC10) reach higher values of up to 0.9 wt.% TIC (Figure 3).

OPEN IN VIEWER

Figure 3.

Minerogenic sediment parameters versus time displaying titanium (Ti), total inorganic carbon (TIC) as a 7-point running mean (7PRM), calcium (Ca) and selected elemental ratios. Lithological units and subunits (A–Ed) are labelled and indicated by dashed horizontal lines. Also indicated are tephra layers AZU-T1 (T1) and AZU-T2 (T2) and the sandy layers (S1, S2). TIC maxima (shaded in blue) are numbered as TIC1–TIC10 and five periods with wave erosion (WE1–WE5) are labelled.

Compared with TIC, the Ca/Ti ratio is rather constant along units B–D with values around 2 (Figure 3). TIC1 is not evident from this parameter, while TIC2 and TIC3 are weakly represented. Subunit Ea has low Ca/Ti ratios, except for both tephra layers. Subunit Eb separates into two parts: a lower half with values <2 and little variability and a variable upper half (values >4). Sandy layers are not manifested by this parameter.

Potassium is associated with silicates and regarded as additional proxy for minerogenic sediment. However, K/Ti shows constant low values throughout the record with three exceptions: both tephra layers and a less pronounced peak in unit C (Figure 3). As K is not a major component of basaltic rocks (cf. low values in unit A and for both sandy layers), it is attributed to precipitation from the water column and thus not a proxy for allochthonous minerogenic sediment.

The elements Fe and Mn display a different pattern compared with proxies for minerogenic components. Both are redox-sensitive and regarded as transitional between minerogenic and organic records in lacustrine sediments (Figure 3). The Fe/Mn ratio is indicative for reducing conditions in lake sediments. Increasing Fe/Mn links to more prominent anoxia and can be applied as a palaeo-redox indicator (Haberzettl et al., 2007). During early lithological units B and C the Fe/Mn ratio peaks and decreases thereafter (Figure 3, Table S2, available online). From unit D to subunit Eb it is stable at low values. However, two peaks interrupt subunits Ea and Eb linked to sandy layers S1 and S2. With the onset of subunit Ec the Fe/Mn ratio increases to a maximum and then decreases towards subunit Ed.

Because of the highly organic character of the lacustrine deposits, XRD analyses detected only few minerals (Table S3, available online): quartz and plagioclase are present in units C–Ea, pyrite in units B–C, vivianite in unit D and calcite in unit Ec.

Organic sediment components

TOC comprises autochthonous and allochthonous organic matter. Lithological unit B has rather low and almost constant values (mean: 2.3%). A distinct increase to a mean of 5.5% characterizes unit C. From the onset of unit D, TOC constantly increases towards 16% at the top of subunit Eb (Figure 4, Table S2, available online). Subunit Ea has three and subunit Eb one excursion to lower values linked to tephras and sandy layers. Finally, TOC drops to 12% at the onset of subunit Ec and increases again to 15% in the topmost subunit Ed.

OPEN IN VIEWER

Figure 4.

Organic sediment parameters versus time displaying total organic carbon (TOC), biogenic silica (BSi), total organic carbon to total nitrogen molar ratio (TOC/TN_molar), total sulphur (TS), stable isotope records of nitrogen and carbon (δ¹⁵N, δ¹³C) and the pollen records of Poaceae and Nothofagus. Lithological units and subunits (A–Ed) are labelled and indicated by dashed horizontal lines. Also indicated are tephra layers AZU-T1 and AZU-T2 (T1, T2), sandy layers (S1, S2) and periods of meromixis (shaded in green).

A general trend towards higher lacustrine productivity with time (Table S2, available online) is documented by TOC and confirmed by BSi. Both parameters correlate positively (r = 0.87, n = 96). BSi documents opal mainly from diatoms. Like TOC, BSi values drop for sandy and tephra layers. However, volcanic ash layers are veiled in our BSi record because of its lower resolution.

Total sulphur in lake sediments is bound to the authigenic mineral pyrite (FeS) formed under anoxic conditions in the presence of iron. The TS record (Figure 4, Table S2, available online) shows 1.9% for unit B and increases distinctly to 4.5% in unit C. Units D–E are uniform with much lower values (0.4–0.6%). TS is positively correlated with Fe/Ti (r = 0.75, n = 736). At the transition from subunit Eb to Ec, two peaks >1.5% TS occur. The older is contemporaneous to sandy layer S2 (Figure 4).

Lacustrine signal

Biogenic matter

The amount and chemical composition of biogenic matter and the diatom record are used to infer Holocene environmental conditions in the lake. TOC is a mixture of lacustrine algal biomass, aquatic macrophytes and terrigenous influx (Mayr et al., 2005). Therefore, lacustrine biomass is best archived by BSi documenting a steady increase from 8% (base of unit B) to 61.2% (subunit Ed; Figure 4, Table S2, available online). Assuming minimum effects of dissolution for biogenic opal, documented by good diatom preservation, we consider an increase of the trophic state throughout the Holocene. This process is interrupted by the sandy layers probably introducing additional nutrients into the lake and causing pronounced increases in BSi following both events (Figure 4). Moreover, changes in diatom-species composition as well as in absolute diatom abundance during and after these events are evident (Figure 5).

OPEN IN VIEWER

Figure 5.

Percentage diagram of relevant diatom taxa versus time (Arc. platti = Arcanodiscus platti, Cocc. eug. = Cocconeis euglypta, Cycl. men. = Cyclotella meneghiniana, Pseudo. subs. = Pseudostaurosira subsalina, Stephanod. parvus = Stephanodiscus parvus, Staur. pin. = Staurosirella pinnata, Thal. pata. = Thalassiosira patagonica). Diatom zones (DZ1–DZ4) are labelled and marked by shaded horizontal bars and lithological units and subunits (A–Ed) by dashed horizontal lines. Also indicated are sandy layers (S1, S2). Trophic preferences (m: mesotrophic, e: eutrophic), diatom life forms and absolute diatom abundances are provided as well.

Decomposition during and after deposition mainly affected algal organic matter (TOC values). The TOC record starts with a mean of 2.3% during unit B (Figure 4, Table S2, available online). An abrupt increase at the transition to unit C leads to maximum values of 9% (mean: 5.5%) during unit C and a subsequent decrease to 2.5% at the onset of unit D. Since then, TOC increased steadily to 16.3% at 3000 cal. BP (top of subunit Eb; Figure 4) with evidence for dilution by tephra and sandy layers as in the BSi record. A marked drop in TOC to a mean of 12.5% occurred at 2870 cal. BP, after which it started to increase again towards the top with a maximum of 18% in subunit Ed at AD 1100 (Table S2, available online). In a similar way to BSi, pulses of higher TOC follow the sandy layers, likely linked to eutrophication. This trend is less pronounced for S1 than for S2 (Figure 4).

Information about the composition of organic matter and its terrestrial versus lacustrine source is derived from TOC/TN in combination with δ¹³C values (Figure 4). Mean TOC/TN ratios vary between 4.2 and 11.1 (Table S2, available online). In general, TOC/TN ratios <10 are interpreted as originating from lacustrine organic matter (Meyers, 2003). Higher values obtained during a monitoring study at Laguna Azul (Mayr et al., 2005) suggest either an admixture of reworked littoral sediments (TOC/TN: 8–11), submerged aquatic macrophytes (26–38), terrestrial plants (23–142) or organic matter from terrestrial soils (10–14). This can be resolved by using δ¹³C values (Figure 4), which indicate that modern aquatic macrophytes of Laguna Azul and littoral sediments are characterized by δ¹³C values >−22.5‰, while values <−22.5‰ are typical for terrestrial plants, soils and profundal sediments of algal origin (Mayr et al., 2005). The δ¹³C values of the sediment record are always <−23‰ (Figure 4). They document a steady increase from the base of unit B (−29.3‰) to the top of subunit Eb (−26.0‰) with one distinct interruption during unit C, where δ¹³C values decrease to a mean of −31.2‰. Only during this lithological unit TOC/TN ratios >9.4 occur (Table S1, available online). A traditional interpretation (e.g. Meyers and Lallier-Verges, 1999) would conclude that such TOC/TN ratios indicate higher admixtures of vascular plants. However, recent studies have shown that algal organic matter in crater lakes with natural eutrophication can also exhibit TOC/TN ratios >10 because of N-limiting conditions (Heyng et al., 2012). Eutrophication and/or meromictic conditions can lead to δ¹³C values below −30‰ because of microbial degradation of organic matter (Braig et al., 2013; Heyng et al., 2012; Hollander and Smith, 2001) explaining the observed values in unit C. The lack of correlation between TOC and BSi data during this unit further confirms that sedimentary organic matter is not only controlled by photosynthetic productivity but also by microbial sources and processes of decomposition.

The nitrogen isotope ratio (δ¹⁵N) also reflects the source of organic matter, the isotopic signature of dissolved inorganic nitrogen (DIN) and lake internal processes such as denitrification. Values between 2‰ and 6‰ characterize modern terrestrial plants in the catchment area of Laguna Azul (Mayr et al., 2005). The δ¹⁵N record (Figure 4) marks a trend from low mean values (2.8‰) in unit B to more positive values in unit C (mean: 5.9‰). Then δ¹⁵N mean values decrease to 5.1‰ (subunit Eb). At the transition to subunit Ec, a negative shift to a mean of 4.1‰ is recorded. Like the TOC/TN ratio, this value remains almost constant towards the top (Figure 4, Table S2, available online). For the base of lithological unit B, low δ¹⁵N values indicate little nitrate utilization and denitrification, which would cause a shift towards more positive values. TOC/TN ratios suggest predominance of lacustrine production. This indicates a lake at an early stage of evolution with low availability of nitrate or less intense denitrification compared with unit C. In this unit, TOC/TN ratios increase possibly because of loss of nitrogen related to enhanced microbial denitrification, which increases δ¹⁵N values as evidenced by positive δ¹⁵N excursions (Braig et al., 2013). This interpretation agrees with the strong decline of δ¹³C and high concentration of TS (Figure 4). Most likely, this is the result of methanogenesis and an expression of pronounced meromixis during unit C. Related anoxic conditions favour denitrification, increase the TOC/TN ratio and are indicated by high amounts of TS. The δ¹⁵N values decrease from unit D to the Eb/Ec transition. Since 2800 cal. BP they remain almost constant, documenting a decrease in nitrate utilization by phytoplankton and denitrification (Figure 4).

Redox conditions

In the hypolimnion of lakes with seasonal or permanent anoxia (meromixis) reducing conditions are often observed under eutrophic conditions causing oxygen consumption (Cohen, 2003). This modifies the primary sediment composition and produces distinctive secondary minerals with redox-sensitive elements such as Fe, Mn and S. After all the oxygen is depleted, slower anaerobic decomposition starts with processes like methanogenesis, denitrification and sulphate reduction. The latter causes the production of hydrogen sulphide (H₂S) which reacts with Fe to form pyrite (FeS). This mineral is well preserved and causes, together with high organic carbon content, the black sediment colour.

Considering black colour as a first indication for anoxia, lithological units B and C discriminate as such from the entire record (Figure 2). Moreover, TS values and Fe/Ti ratios are elevated for both units and suggest the presence of pyrite (Table S2, available online), which has been detected by XRD analyses for both units (Table S3, available online). From unit D onwards, an oxic type of sediment composition prevails (Figure 2). The lack of pyrite is confirmed by low TS values, low Fe/Ti ratios and olive to brown sediment colours (Figure 2). Distinct peaks in TS are recognized only in relation to sandy layer S2 (3200 cal. BP) and at 2800 cal. BP (Figure 4).

The Fe/Mn ratio combines the two redox-sensitive elements Fe and Mn, where higher values indicate increased anoxia (Figure 3). We distinguish five maxima of two different types: three maxima are >1000 years long and centred on 11,600, 10,000 and 1500 cal. BP while two ~300 year-long maxima are centred on 5500 and 3100 cal. BP (Figure 3). Evident from Fe/Mn and black colour, unit C is the result of most pronounced anoxic conditions. This is supported by the highest values for TS, high Fe/Ti ratios and most negative δ¹³C values (Table S2, available online). The latter are indicative for methanogenesis and suggest meromictic conditions. Furthermore, very high δ¹⁵N values during unit C indicate denitrification (Figure 4), which also explains the elevated TOC/TN ratio with a mean of 11.1 for unit C.

After unit C (mean Fe/Mn: 69.7), this redox-indicating ratio decreased towards subunit Eb with a mean of 26.3. At the same time, the Mn/Ti ratio increased in unit D (Figure 3, Table S2, available online). Such redox conditions are characteristic of oligotrophic lakes without pronounced anoxia. At 3000 cal. BP the situation changed again and Fe/Mn increased to >45 for subunits Ec and Ed, possibly related to re-establishment of anoxia in the hypolimnion because of increasing trophic conditions.

Diatoms

Almost all diatoms, 85% of our record, are alkaliphilous or alkalibiontic species typical of neutral or basic lakes in a semi-arid climate. This is in agreement with monitored pH values of 8.5–9.0 (Zolitschka et al., 2006). The diatom record subdivides into four diatom zones (DZ1–DZ4; Figure 5). DZ1 is dominated by Thalassiosira patagonica (maximum: 85.1% at 9020 cal. BP), a very small planktonic species that tolerates high conductivities (Wille et al., 2007). Additionally, the pioneer epiphytic species Pseudostaurosira subsalina is present with up to 15.3%. Both species are indicative of saline conditions, which can be related to a negative water balance with a low lake level.

In DZ2 the diatom assemblage changes to dominance of epiphytic Arcanodiscus platti (48%) and Cocconeis euglypta (17%). Both species have wide ecological tolerances and point to freshwater conditions with higher lake levels supporting the growth of macrophytes in flooded areas. DZ3 is dominated by periphytic Staurosira venter (>90%), which tolerates fluctuations of lake levels and salinities indicating a higher trophic status. The diatom assemblage from the exposed lake-level terrace dated to 1220 +70/−130 cal. BP (Figure 1; Table 1) is entirely composed of Staurosira venter suggesting formation during DZ3. Uppermost DZ4 (since AD 1780) is dominated by planktonic Stephanodiscus parvus indicative of eutrophic conditions. The final step of eutrophication in the 20th century is marked by an exponential increase of absolute diatom abundance (Figure 5). Changes in habitats from benthic to planktonic indicate a loss in littoral environments in response to a lake-level lowering. Simultaneously, a change of nutrient preferences from mesotrophic to eutrophic occurred. Lower lake levels caused eutrophication as well as calcite precipitation (TIC10) and likely were triggered by the post-LIA temperature increase of the ‘Current Warm Period’ (CWP) documented as +2.5°C (summer) and +1.4°C (annual) temperatures for 1931–1990 at Rio Gallegos (Villalba et al., 2003). Since we observed a c. 11 m lake-level drop at close-by Laguna Potrok Aike since the 1890s (Haberzettl et al., 2005; Ohlendorf et al., 2013), the most conclusive regional cause for modern lake-level lowering is climatic.

Hydrological signal

Under the prevailing semi-arid climate at Laguna Azul direct rainfall is of little influence on lacustrine hydrology. As the lake has no tributary, groundwater inflow, favoured by rapid infiltration through porous and fissured volcanic rocks, and evaporation, mainly controlled by SHW intensity, determine the lake’s water budget. As no inflow exists, we regard clastic sediment transfer into the lake as a hydrological proxy related to wave erosion (WE) of Ti- and Ca-bearing alkali-basaltic rocks at the shoreline.

Periods of increased shoreline erosion are evident via elevated magnetic susceptibility as well as by high Ti and Ca contents at the end of lithological unit B (WE1: 10,400–10,100 cal. BP) and in early unit D (WE2: 8300–7500 cal. BP). Shorter and event-like erosive periods follow with sandy layers S1 (WE3: 5500 cal. BP), S2 (WE4: 3200 cal. BP) and WE5 centred at 1300 cal. BP. We interpret this mobilization of littoral sediment as a consequence of rising and falling lake levels.

The signature of lithological unit A (pyroclastic material) is regarded as the external geochemical background signal (high Ti, Ca, DBD and MS) for wave erosion. During WE1, these para-meters are relatively pristine as the lake is still at an early stage of lacustrine evolution. This changed since WE2, when increased TIC values document the presence of carbonates, for example, linked to the formation of lacustrine tufa. Lacustrine tufa (beach rock) is observed on shoreline rocks of Laguna Azul formed by photosynthetic algae and cyanobacteria.

In addition to lakeshore erosion, carbonates also precipitate from the water column as calcite (CaCO₃). This process intensifies when the lake level lowers because of enhanced aridity causing higher concentration of soluble elements in the water body. As both TIC and the Ca/Ti ratio are influenced by calcite precipitation (Figure 3), comparing them allows CaCO₃ to be distinguished from siliciclastic Ca that entered the sediment, for example, as plagioclase (cf. Table S3, available online). The latter has a constant Ca/Ti ratio (cf. WE1) unless other minerals are added. This effect is documented by the two Ti-depleted tephra layers causing peaks in Ca/Ti but no comparable peaks for Ti and TIC, suggesting that only siliciclastic Ca is added to the system. If Ca/Ti increases parallel to TIC, autochthonous calcite precipitates. However, if TIC is peaking without response of Ca/Ti, other carbonaceous minerals are responsible for the additional inorganic carbonate. This can be related to siderite (FeCO₃) under anoxic conditions, to rhodochrosite (MnCO₃) under less anoxic conditions and to potassium carbonate or potash (K₂CO₃) under extremely saline conditions (Cohen, 2003).

The record of Laguna Azul is characterized by 10 distinct periods of elevated TIC (TIC1–TIC10) of which 7 link to calcite precipitation. TIC1 (11,600–10,800 cal. BP) is not associated with any change in Ca/Ti (Figure 3), thus the presence of calcite is unlikely. Based on elevated Fe/Ti ratios, which point to an additional source of Fe, and in combination with an Fe/Mn peak indicating anoxic conditions, the formation of siderite is regarded as the cause for TIC1. XRD analyses confirm the presence of pyrite for TIC1 (Table S3, available online), supporting the presence of anoxic conditions. However, no evidence of siderite or other carbonate minerals was detected, presumably because they occur only with minor contributions.

Also, TIC2 (10,000–9600 cal. BP) does not coincide with siliciclastic input (Figure 3). However, a peak of Ca/Ti suggests a link to calcite precipitation. In addition, and according to the pronounced maximum of Fe/Ti, siderite has to be regarded as present (Figure 3), which is supported by Fe/Mn and the presence of pyrite (Table S3, available online). Moreover, Mn/Ti and K/Ti show peaks, which may be related to carbonate minerals like rhodochrosite and potash. Especially the latter needs highly saline conditions close to desiccation (Cohen, 2003), which would be supported by the salt-tolerant diatom Thalasiosira patagonica (Figure 5). Because of their low concentrations, neither of these minerals could be verified mineralogically. TIC3 (8300–7600 cal. BP) occurs with a pulse of minerogenic matter as Ti and Ca increase together (Figure 3). As Ca/Ti responds positively, we conclude that calcite is present. For the second half of TIC3, the presence of rhodochrosite is possible because Mn/Ti increased. TIC4 (5500 cal. BP) coincides with sandy layer S1 (WE3), pronounced Ca and Ti maxima but no response in Ca/Ti. Therefore, carbonates are lacking and eroded volcaniclastics are regarded as the source for TIC4 (Figure 6). TIC5 (4700 cal. BP) has no corresponding responses for Ca, Ti or Fe. Elevated Mn/Ti ratios suggest rhodochrosite as carbonate mineral (Figure 3). With TIC6 (3800 cal. BP) lacustrine conditions changed. Since c. 4000 cal. BP, calcite became the dominant carbonate mineral under increasingly eutrophic conditions (change from DZ2 to DZ3) with anoxic bottom waters as indicated by rising Fe/Mn (Figure 3). In addition, non-silicate Ca is documented in the system by a step-like increase of Ca/Ti. Still, rhodochrosite seems to contribute to carbonates as indicated by a peak of Mn/Ti related to TIC6. TIC7 (3200 cal. BP) is linked to sandy layer S2 (WE4). A negative amplitude of Ca/Ti relates to siliciclastic deposition without carbonaceous Ca. Like for TIC4, the explanation is wave erosion. TIC8 (2200 cal. BP) is distinctly assigned to lacustrine calcite precipitation, as confirmed by XRD analyses at 2180 cal. BP (Table S3, available online). Calcite is also the reason for TIC9 (1300–600 cal. BP) and TIC10 (past 50 years; Figure 3).

OPEN IN VIEWER

Figure 6.

Synthetic diagram of climatic and environmental proxies from Laguna Azul displaying selected parameters with environmental and climatic implications. Titanium (Ti) indicates minerogenic sediment flux from the crater slopes. Lacustrine biomass is represented by total organic carbon (TOC). Total inorganic carbon (TIC) is indicative of evaporation and representative for the lake level. δ¹³C indicates methanogenesis and is related to an anoxic hypolimnion. The diatom Thalassiosira patagonica (Thal. pata.) is a proxy for salinity. Fe/Mn corresponds to redox conditions and Nothofagus pollen amounts to SHW intensity. For comparison, the chironomid-based mean annual temperature (MAT) from Laguna Potrok Aike is shown on its own independent timescale (Massaferro and Larocque-Tobler, 2013). Arrowheads point towards higher values. Lithological units and subunits (B–Ed) are labelled and indicated by dashed horizontal lines. Also indicated are tephra layers AZU-T1 and AZU-T2 (T1, 2) and sandy layers (S1, S2). Furthermore, five periods with wave erosion (WE1–WE5) and 10 with maxima in total inorganic carbon (TIC1–TIC10) are labelled. The warm/dry periods of the ‘Roman Climate Anomaly’ (RCA), the ‘Medieval Climate Anomaly’ (MCA) and the ‘Current Warm Period’ (CWP) are labelled and marked by a horizontal bar shaded in red. The cold/moist ‘Dark Age Cold Period’ (DACP) and the ‘Little Ice Age’ (LIA) are labelled and marked by a horizontal bar shaded in blue. In addition, the initial lake phase and two meromictic periods are shaded in pink and green, respectively.

Summarizing the hydrological signal of Laguna Azul, there are seven phases that relate calcite with lower lake levels while three (TIC1, TIC4, TIC7) are unrelated. The most prominent low lake-level event (TIC10) occurred during recent decades and is attributed to the CWP. The second and third major low lake levels date to 2200 (TIC8) and 1000 cal. BP (TIC9), corresponding by age to the ‘Roman Climatic Anomaly’ (RCA) and the MCA documented for European records, respectively.

The lacustrine circulation type is another link to the hydrological setting. During units B and C as well as during subunits Ec and Ed, Fe/Mn suggests anoxic conditions. To achieve such conditions at least seasonally, the water column needs to be stratified. A surplus of organic matter provided by higher lacustrine productivity provokes oxygen consumption during decomposition, eventually leading to endogenic (biogenic) meromixis. Such processes are considered responsible for anoxia of the past 3000 years. However, these biogenic anoxia are less pronounced compared with those of unit B and especially unit C. During the early lake history, permanent anoxia were established under fully meromictic conditions accompanied by methanogenesis despite low lacustrine productivity (oligotrophic–mesotrophic); a development explained by ectogenic meromixis (Hakala, 2004). Presumably triggered by a distinct increase of climate-induced evaporation, the lake level had fallen and salinity increased. Thus, density of the bottom water increased and stabilized the water column to form a permanent stratification. Without holomixis, strong anoxia established in the hypolimnion and caused ectogenic meromixis.

Evolution of a young crater lake

During the early stage of lake evolution, the inner slopes of the crater stabilized by mass wasting causing deposition of volcaniclastic sediments (unit A) in the central lake basin. The Early-Holocene onset of lacustrine sedimentation (unit B) is transitional from clastic to organic sediments. According to the SHW intensity proxy (Nothofagus) and chironomid-based mean annual temperature (MAT) reconstructed at nearby Laguna Potrok Aike (Massaferro and Larocque-Tobler, 2013), the first half of unit B has low Late-Glacial temperatures and SHW intensity was high. This is supported by an oligotrophic, shallow and meromictic lake (Figure 6), where meromictic conditions are indicated by higher Fe/Mn and saline conditions (halophilous diatoms in DZ 1).

At 11,400 cal. BP TIC1 is linked to elevated Fe/Ti (Figure 3) and indicates siderite. The end of TIC1 is defined by decreasing Nothofagus pollen amounts pointing to less intense SHW. Thus, less wind-induced evaporation caused the water level to rise, probably supported by increased precipitation. The lake freshened and anoxia disappeared. Deposition of minerogenic components indicates a rising lake level mobilizing pyroclastic lakeshore sediments via wave erosion (WE1; Figure 6). High counts for Ti and Ca correspond geochemically to volcaniclastica of unit A.

With the onset of unit C, MAT reached its Early-Holocene maximum coinciding with an increase in Nothofagus pollen amounts and the maximum saline conditions documented by halophilous diatom Thalassiosira patagonica. A lake-level drop is documented by precipitation of calcite, siderite, rhodochrosite and potash (TIC2), the latter associated with the highest salinity or desiccation (cf. Zhang et al., 2013). Simultaneously, stratification of the water column established after solutes increased in the hypolimnion (maximum of halophilous diatoms) initiating ectogenic meromixis (cf. Hakala, 2004), evidenced by presence of pyrite (Table S3, available online) with high TS and Fe/Mn (Figure 3). Furthermore, negative δ¹³C values and a positive shift in δ¹⁵N indicate methanogenesis and denitrification, respectively (Figure 4). Altogether, this points to increased zonal winds accompanied by higher evaporation under increased MAT.

The transition to unit D is characterized by a salinity decrease indicated by diatoms in response to termination of meromixis indicated by lower Fe/Mn. Higher minerogenic contributions to the sediment (Ti, Ca) document lakeshore erosion (WE2; Figure 6) and flooding of the littoral zone (Figure 1). Simultaneously, TIC3 marks the onset of unit D resulting from high Middle-Holocene temperatures and increased lacustrine productivity (Figure 3). In the second half of unit D, minerogenic matter and carbonates decline while lacustrine productivity continues to increase (Figure 4). These changes are interpreted as response to decreasing SHW intensity.

A general change from blackish to brownish sediment characterizes the onset of unit E (Figure 2). Subunits Ea and Eb continue with low Nothofagus pollen amounts and high MAT (Figure 6). TIC4 occurred simultaneous with WE3. Thus, the majority of Ca is bound to minerogenic matter (Ca/Ti; Figure 3). The onset of subunit Eb is characterized by a distinct increase in lacustrine productivity (Figure 4, Table S2, available online). Diatoms indicate an increasing trophic state from subunit Eb towards the top. TIC5 is not linked to calcite but probably to rhodochrosite (Mn/Ti; Figure 3). We interpret this as related to low SHW intensity with a positive water balance for subunits Ea and Eb.

At 3800 cal. BP during ending subunit Eb, a drop in MAT to its Late-Holocene mean occurred (Figure 6). This transition is evident in the diatom record with a prominent shift from planktonic and epiphytic (DZ2) to benthic (DZ3) life forms (Figure 5), and in Ca/Ti with a lower mean of 1.9% before and 3.0% after 3800 cal. BP (Figure 3). The now prevailing benthic diatom Staurosira venter tolerates frequent lake-level fluctuations and is adapted to higher trophic states (Figure 5). We interpret this step-like increase of Ca/Ti as a consequence of higher lacustrine productivity, explaining the first calcite-dominated TIC6 probably with admixture of rhodochrosite (Figure 3). TIC7 coincides with sandy layer S2 (WE4); extra Ti and Ca of basaltic origin links to higher lake levels and wave erosion (Figure 6).

Following the temperature drop at 3800 cal. BP, Nothofagus pollen amounts started to increase heralding the SHW intensity change at 3000 cal. BP, which marks the transition to subunit Ec (Table S2, available online; Figure 6). Interpreted as the result of SHW intensity increase, windier conditions with more evaporation and less precipitation caused the lake level to drop culminating in calcite-dominated TIC8. A falling lake level further supported higher productivity as indicated by diatoms of DZ3 causing seasonal anoxia because of intensified decomposition of organic matter at the lake floor mirrored by a positive shift in Fe/Mn. For TIC8 (RCA) and the following warm/dry events related to TIC 9 (MCA) and TIC10 (CWP), we use names established for contemporaneous climatic events of the Northern Hemisphere.

The transition to subunit Ed is marked by WE5 caused by a higher lake level at 1300 cal. BP coinciding with the subaerial lake-level terrace composed of Staurosira venter and dated to 1190 ± 100 cal. BP (Figure 1, Table 1). This period corresponds to the ‘Dark Age Cold Period’ (DACP) of the Northern Hemisphere (Helama et al., 2017). Immediately after WE5, TIC9 occurred with eutrophic conditions as documented by appearance of Staurosirella pinnata (Figure 5). This is related to a more negative water balance with concentration of nutrients and calcite precipitation during the MCA (TIC9)

Centuries later, maximum Poaceae and a decrease of Nothofagus pollen amounts indicate moister and less windy conditions linked to distinctly lower Fe/Mn from AD 1550–1850 (Figure 4), a period related to the LIA. During the second half of the 20th century the highest carbonate values occur (TIC10 = CWP) coinciding with a decrease of lake levels (Figure 3). The final part of the record is characterized by the onset of human activities since AD 1830 corroborating Mayr et al. (2005).

Mechanisms of climate change

Despite the Southern Hemispheric summer minimum of solar insolation at 10,000 cal. BP (Berger and Loutre, 1991), the Early Holocene was the warmest period in southern South America. At Laguna Azul, this is evidenced by strongest zonal winds with maxima at 11,600 and 9000 cal. BP. Early-Holocene weakening of the thermohaline circulation in the Northern Hemisphere reduced heat transfer from the South Atlantic to the North Atlantic. Thus, thermal energy increased in the Southern Oceans causing an air-temperature rise in the Southern Hemisphere (Ganopolski et al., 1998; Hodell et al., 2001).

A similar mechanism explains the start of the Neoglacial in South America. Although the Southern Hemisphere experienced increased insolation, the Neoglacial cooling is observed since at least 5000 cal. BP (Glasser et al., 2004). The reason is cooling of the Northern Hemisphere since 9000 cal. BP (Berger and Loutre, 1991). In turn, increased thermohaline circulation resulted in heat export from the South Atlantic to the North Atlantic, triggering sea-ice growth around Antarctica and glacial advances on the continents. This caused global feedbacks at the end of the Middle Holocene, such as the termination of the African Humid Period (Ganopolski et al., 1998) or the establishment of modern El Ni ñ o-Southern Oscillation (ENSO) frequencies (Clement et al., 1999; Moy et al., 2002; Sandweiss et al., 1996).

In contrast to the Late-Glacial–Holocene transition, temperature and zonal wind intensity are decoupled since 7000 cal. BP with strengthened and poleward shifted SHW (Varma et al., 2012). This agrees with more pronounced ENSO frequencies in response to increased insolation (Clement et al., 2000) and reached the highest intensity during the past 3500 years (Moy et al., 2002; Villa-Martínez et al., 2003).

In synthesis, Fletcher and Moreno (2012) suggest zonal changes of SHW on millennial timescales between 14,000 and 5000 cal. BP that switched to a centennial pattern after 5000 cal. BP, controlled by intensified ENSO (Moreno et al., 2014). Markgraf (1998) developed a comparable conclusion, linking modern seasonal shifts of SHW to a pattern that established 4000 cal. BP when seasonality of insolation increased in the Southern Hemisphere. These forcing mechanisms operate at different strengths and timescales (Bentley et al., 2009) and their causal relationships are not yet completely understood (Garreaud et al., 2013). Additional centennial-scale variability is observed in our record and well represented by geochemistry (TIC maxima) characterizing lake level low stands, that is, warm/dry and more windy conditions with higher evaporative stress such as the RCA (TIC8), the MCA (TIC9) and the CWP (TIC10).

Regional variability of Southern Hemispheric Westerlies

The search for a better knowledge about SHW variability stimulated research in Patagonia, the only continental landmass that intersects with the SHW wind belt. To learn how climatic changes observed at Laguna Azul contribute to the understanding of SHW variability, we compare our findings with other records from South America (Figure 7). Overall, studies agree that the core region of SHW migrated latitudinally between equatorward positions during glacial and poleward positions during interglacial conditions (Toggweiler et al., 2006). Currently, SHW vary seasonally between 40–60 °S (Kohfeld et al., 2013), while an expanded band (30–60 °S) is assumed for the Holocene (Fletcher and Moreno, 2011).

OPEN IN VIEWER

Figure 7.

Holocene millennial-scale SHW reconstructions for sites east of the Andes from south (left) to north. (a) Total annual flux (TAF) of silt grains at Isla de los Estados (Björck et al., 2012). (b) Long-distance and wind-delivered pollen amounts of Nothofagus for Laguna Azul (this study) and (c) for Laguna Potrok Aike (Wille et al., 2007); (d) Chironomid-based mean annual temperature (MAT) from Laguna Potrok Aike (Massaferro and Larocque-Tobler, 2013). (e) Hypothesized latitudinal migration of SHW for Lago Cardiel based on lake-level variations (Quade and Kaplan, 2017). (f) Mean sortable silt (MSS) from Lago Castor (Van Daele et al., 2016). All data are smoothed with a 5-point running average (except e) and shown on their own independent timescales.

At Laguna Azul, we observed four intervals with varying SHW modes during the Holocene. Low lake levels because of increased SHW intensity with high evaporation rates occurred from 10,100 to 8300 cal. BP with the lowest lake levels and development of ectogenic meromixis and since 3000 cal. BP creating seasonal anoxia in combination with eutrophication. Low zonal winds with increased precipitation caused higher lake levels from 11,200 to 10,100 and from 8300 to 3000 cal. BP

A less arid period with weaker SHW marks the Early Holocene (11,200–10,100 cal. BP) at Laguna Azul. Simultaneously, wetter conditions were reported from the lee-sites Rio Rubens (52 °S: −0.5; Markgraf and Huber, 2010), Laguna Potrok Aike (52 °S: −0.6; Hahn et al., 2013) and Lago Cardiel (49 °S: −0.8; Quade and Kaplan, 2017). The second value provided in brackets is the correlation coefficient between precipitation and zonal wind speed according to Moreno et al. (2014). Whether the SHW moved poleward or equatorward during the Early Holocene can be checked. Further to the north, Lago El Salto (42 °S: +0.8) turned drier in a luv position indicative of a southward shift of SHW (Moreno and Videla, 2016). There is additional evidence from Lake Tamar (53 °S: +0.8) of wet conditions from a luv position (Lamy et al., 2010), while Isla de los Estados in a lee position (55 °S: −0.8) was dry (Björck et al., 2012), both indicative of stronger SHW. For the marine record Palmer Deep (65 °S, luv of Antarctic Peninsula) no signs of SHW presence were recognized (Domack et al., 2001). We conclude that the core of SHW was located between 53 and 60 °S in the Early Holocene in agreement with the SHW position (57 °S) suggested by Quade and Kaplan (2017) (Figure 7).

For the early Middle Holocene, Laguna Azul recorded the lowest lake levels and ectogenic meromixis as a consequence of the highest zonal winds (10,100–8300 cal. BP). A lake-level drop of 30 m is also documented for Laguna Potrok Aike (52 °S: −0.6) from 9300 to 6900 cal. BP (Anselmetti et al., 2009; Gebhardt et al., 2012; Kliem et al., 2013; Zolitschka et al., 2013) coincident with most intense SHW reconstructed from long-distance transport of AFT pollen peaking at 8600 and 7500 cal. BP (Figure 7). Further north, Lago El Salto (42 °S: + 0.8) and Lago Condorito (41 °S: + 0.6) were drier from 11,600 to 7600 and 10,000 to 7600 cal. BP, respectively (Moreno et al., 2010; Moreno and Videla, 2016) indicating weaker SHW. At 46 °S Lago Castor experienced a minor increase in SHW intensity (Van Daele et al., 2016), while Lago Cardiel (49 °S: −0.8) is characterized by decreasing lake levels from 10,000 to 7800 cal. BP (Quade and Kaplan, 2017). South of Laguna Azul, Isla de los Estados (55 °S: −0.8) experienced no SHW influence until 7200 cal. BP (Björck et al., 2012) (Figure 7). In sum, between 10,000 and 7500 cal. BP the core region of SHW migrated northward to 50–52 °S, in agreement with the SHW position of 51 °S reconstructed for Lago Cardiel (Figure 7). More northern sites document little (Lago Castor) or no (Lagos El Salto and Condorito) influence of an SHW intensity increase.

For the late Middle-Holocene, a distinct poleward shift of SHW to 65 °S is documented until 3360 cal. BP for Palmer Deep (Domack et al., 2001). Furthermore, an increase in SHW intensity was noted for Isla de los Estados (55 °S: −0.8) reaching wind-intensity maxima between 4600 and 3300 cal. BP (Björck et al., 2012) (Figure 7). To the north, Lago Condorito (41 °S: +0.6) was wetter after 7000 cal. BP (Moreno et al., 2010), while Lago Cardiel (49 °S: −0.8) recorded a higher lake level between 7500 and 5000 cal. BP (weaker SHW) and a lake level lowering (stronger SHW) from 5000 to 3300 cal. BP (Quade and Kaplan, 2017). Lago Castor (46 °S) experienced strengthened SHW (7500–3800 cal. BP; Figure 7) with a maximum from 5000 to 3800 cal. BP (Van Daele et al., 2016). On the other hand, Laguna Potrok Aike (52 °S: −0.6) documents a lake level rise and lower Nothofagus pollen amounts (6900–3300 cal. BP) coincident with decreasing SHW intensity (Mayr et al., 2007b), just like the record from Laguna Azul (this study) between 7500 and 3000 cal. BP. In sum, strong SHW intensities were recorded from 65 to 41 °S. This is at the same time when Lagunas Azul and Potrok Aike experienced less windy conditions than before. One possible explanation is an expansion of the SHW belt with the highest zonal winds at the northern and southern margins while the central part at 52 °S weakened.

Not only had the southern location of SHW reached the Drake Passage during the late-Middle Holocene, also gradually colder conditions established in Patagonia since the onset of the Neoglacial between 7000 and 5000 cal. BP. There is evidence from glaciers of the South Patagonian Icefield (50 °S) expanding since 5800 cal. BP (Porter, 2000), moraines at Lago Argentino (51 °S) document glacial advances after 6000 cal. BP (Kaplan et al., 2016) and a sediment record obtained from Almirantazgo Fiord (54 °S) shows first glacial re-advances at 5700 cal. BP (Bertrand et al., 2017). Furthermore and related to increased precipitation, a stalagmite developed in the luv of the superhumid Andes in Marcelo Arevalo Cave (53 °S) after 5700 cal. BP (Schimpf et al., 2011). Although the onset of the Neoglacial evidences colder temperatures and thus suggests a northward migration of SHW since 7000 cal. BP, the record of Palmer Deep at 65 °S still documents high SHW intensity until 3360 cal. BP (Domack et al., 2001) – a contradiction that needs to be resolved.

During the Late Holocene, marine productivity decreased at Palmer Deep (65 °S) after 3360 cal. BP. At Isla de los Estados (55 °S: −0.8) weaker SHW occurred after 3300 cal. BP (Björck et al., 2012) as well as at Lake Tamar (53 °S: + 0.8) after 5500 cal. BP (Lamy et al., 2010) and at Lago Fagnano (54 °S: -0.6) after 6000 cal. BP (Waldmann et al., 2010). The response at Punta Yartou (54 °S: +0.6) shows wetter and colder conditions after 5060 cal. BP (Mansilla et al., 2016) and at Laguna Azul (52 °S: −0.8) more Nothofagus pollen, both indicative of higher SHW intensity. Lago Castor (46 °S: transitional) shows a decrease in sortable silt (Van Daele et al., 2016) and Lago Cardiel (49 °S: −0.8) higher lake levels, that is, less zonal winds for both sites since 3300 cal. BP. In the north, Lagos El Salto (42 °S: +0.8) and Condorito (41 °S: +0.6) continued to have wetter conditions (Moreno et al., 2010; Moreno and Videla, 2016) as does marine site GeoB3313-1 (41 °S: +0.8) after 4000 cal. BP (Lamy et al., 1999, 2010). In conclusion, weaker SHW south of 54 °S is associated with increased SHW intensity at Laguna Azul, decreased SHW intensity at Lagos Cardiel and Castor (Figure 7) and continuously high zonal winds to 41 °S. Whether this is in response to the SHW belt with its windier margins being shifted northward following the onset of Neoglaciation with a delay or the result of other processes remains an open question.

Overprinting the millennial SHW pattern, there are century-long cold/wet and warm/dry periods during the Late Holocene. Their onset coincides with a distinct MAT decline (3800–3200 cal. BP). Almost simultaneously, the proxy for wind intensity (Nothofagus) indicates the onset of higher zonal winds for Lagunas Azul and Potrok Aike, while independent SHW proxies decrease for Islas de los Estados and Lago Castor (Figure 7). Although MAT displays no major variability for the Late Holocene, sediments evidence three distinct periods with lower lake levels centred at 2200 and 1000 cal. BP and during the last century (Figure 6) coeval to Northern Hemispheric RCA, MCA and CWP. Furthermore, we recognize the DACP and the LIA as pronounced moist periods with high lake levels and the maximum expansion of Poaceae

Best represented is the cool/wet LIA detected at many sites from Palmer Deep near Antarctica (Domack et al., 2001), via Lago Fagnano on Tierra del Fuego (Waldmann et al., 2010) towards the marine record GeoB-3313 off Chile (Lamy et al., 2010) with many Patagonian sites in between: Lagunas Potrok Aike (Haberzettl et al., 2005), Azul (this study), Las Vizcachas (Fey et al., 2009), Cháltel (Ohlendorf et al., 2014), Cari-Laufquén (Cartwright et al., 2011) and Lagos Guanaco (Moreno et al., 2010), del Desierto (Kastner et al., 2010), Cardiel (Ariztegui et al., 2010), El Salto (Moreno and Videla, 2016) and Puyehue (Boës and Fagel, 2008).

A number of records available for the last millennium also confirms a three partition in MCA, LIA and CWP: Lagos Puyehue (Boës and Fagel, 2008), El Salto (Moreno and Videla, 2016), del Desierto (Kastner et al., 2010), Guanaco (Moreno et al., 2009), Cipreses (Moreno et al., 2014), and Lagunas Potrok Aike (Haberzettl et al., 2005), Azul (this study) and Las Vizcachas (Fey et al., 2009). This picture is supported by regional palaeohydric balance integrations from south-eastern Patagonia documenting a positive water balance for the LIA (500–100 cal. BP) and negative water balances for the CWP (since AD 1919) and the MCA (1400–500 cal. BP; Echeverria et al., 2017). Moreover, temperature reconstructions on a continental-scale for South America reveal colder conditions from AD 1380 to 1920 (equivalent to the LIA), whereas two warmer phases were recognized for AD 1150–1320 (equivalent to the MWP) and for the CWP (Ahmed et al., 2013). In addition, the pronounced warm/dry period of the RCA at Laguna Azul was recognized at other Patagonian lakes, such as Lagos El Salto, Cipreses and Guanaco (Moreno et al., 2009, 2014; Moreno and Videla, 2016) pointing to overarching regional climatic control disregarding their luv/lee location west or east of the Andes.

The loss in zonal symmetry of SHW was suggested to be responsible for the onset of centennial climate fluctuations related to ENSO variability with teleconnections across South America (Fletcher and Moreno, 2012). This increase in ENSO frequency accelerated during the past 3800 years and peaks during the MCA (Moy et al., 2002). For the same time interval stronger linkages between SHW, insolation (solar activity, astronomical forcing) and ENSO were postulated by Kilian and Lamy (2012). Changes in solar activity seem to be capable of climate implications unrelated to zonal symmetry and follow a global character. A well-known example was the Maunder solar minimum, which together with large volcanic eruptions was one of the triggers responsible for the LIA (Owens et al., 2017).

Conclusion

Position and strength of SHW are documented by the multiproxy study of Holocene sediments from Laguna Azul in semi-arid south-eastern Patagonia highlighting millennial- to centennial-scale hydroclimatic variability. The lacustrine history of Laguna Azul started with higher lake levels in the Early Holocene (11,200–10,100 cal. BP), which are linked to weaker SHW reflected by generally increased moisture availability east of the Andes (cf. Mancini et al., 2008). During the Early Holocene, the core of SHW was located between 53 and 60 °S. From 10,100 to 8300 cal. BP low lake levels and high salinity document the driest conditions. Together with anoxia and methanogenesis, evidence for ectogenic meromixis relate to the highest SHW intensities centred around 50 °S with increased evaporation indicated by drier conditions east of the Andes (cf. Mancini et al., 2008). Between 8300 and 3000 cal. BP, an again higher lake level was linked to warmer and moister conditions at Laguna Azul, indicative of less intense SHW. However, regional comparison documents an expansion of stronger SHW to 65–41 °S, a discrepancy that needs to be investigated by further studies. After 4000 cal. BP, MAT indicates the Neoglacial temperature drop. Lacustrine conditions became more variable with century-long warm/dry (RCA, MCA, CWP) and cold/moist (DACP, LIA) periods under generally increased SHW intensity. These short climatic anomalies are also known from sites west of the Andes and from the Northern Hemisphere. At 3300 cal. BP SHW pulled back from latitudes south of 54 °S and contracted to 54–41 °S.

Despite many environmental and climate investigations from south-eastern Patagonia, substantial knowledge gaps still exist. To improve our understanding of SHW variability, an extended network of high-resolution terrestrial and marine records with quantitative reconstructions as well as reliable time control are mandatory to develop consistent interpretations. This includes the studied site, where high-resolution diatom and chironomid records would contribute additional quantitative reconstructions. Moreover, integration of critical sites by climate modelling would be beneficial to determine the different modes of forcing that control the climate of South America on a regional scale with influences on the entire Southern Hemisphere and beyond.

Supplemental Material