Four short cores of Laguna El Calvario (LCA) were retrieved in 2017 using a UWITEC^© gravity corer. All the cores were lithologically characterized through X-radiographs and visual descriptions. The LCA SHC-4 core was selected for performing the multiproxy analysis given it was the longest (33 cm long; Figure 1 Supplementary Material ). The sedimentological analysis included a loss on ignition, X-ray fluorescence analysis (XRF), and inductively coupled plasma mass spectrometry (ICP-MS). Loss-on ignition was assessed at a contiguous 1 cm interval to estimate organic, inorganic, and carbonate contents (Heiri et al., 2001). The chronology of the Laguna El Calvario record was established through ²¹⁰Pb and ¹⁴C dating techniques. The ²¹⁰Pb activities (dpm g⁻¹) were estimated for the first 15 cm through its daughter radionuclide ²¹⁰Po which is in secular equilibrium ( Table 1 Supplementary Material ). The chemical procedure included the acid digestion of sediment samples with the addition of ²⁰⁹Po as a yield tracer and the deposition of the ²¹⁰Po onto ultrapure silver discs (Flynn, 1968). Activities were quantified in a Canberra Quad Alpha Spectrometer until a 1 σ error was achieved. Additionally, five bulk sediment samples were analyzed for AMS radiocarbon dating in the Direct AMS laboratory, USA. Levels 0–1 cm and 6–7 cm were analyzed for the ¹⁴C measurements in order to check any ¹⁴C reservoir effect, a common issue in high Andean lakes. According to the ²¹⁰Pb model these levels correspond to the second half of the 20^th century, −60 and −9 cal yrs BP, respectively. To estimate the reservoir effect in the Laguna El Calvario, the ¹⁴C ages from levels 0–1 cm and 6–7 cm were averaged resulting in a reservoir age of 1041 ¹⁴C yrs BP (950 cal yrs BP), which was subtracted from the remaining three older radiocarbon ages ( Table 1 ). Then, the radiocarbon ages were calibrated with the SHCal20 curve (Hogg et al., 2020) and the age-depth model was computed using Plum (Aquino-López et al., 2018) with the rPlum R package (Blaauw et al., 2020). rPlum is a recently developed Bayesian approach that permits computing integrated chronologies without pre-modeling the ²¹⁰Pb ages but combining other chronostratigraphic markers such as the ¹⁴C ages.

The XRF scanning was employed to measure the variability of geochemical elements on unprocessed sediments using an ITRAX core scanner (Cox Analytical Systems, Croudace et al., 2006) at the GEOPOLAR laboratory of the University of Bremen, Germany. Measures of the XRF series were established at each 2 mm interval and then a molybdenum tube at 40 kV and 10 mA was applied with an exposure time of 5 s for every measurement. The concentration of the different minerals was expressed in total counts (cnts) and elements over 100 cnts were selected. To support XRF data, continuous and discrete sediment samples were analyzed using an ICP-MS at UC-Davis facilities, from total digested sediment samples. The chemical procedure briefly consists of digesting ~250 mg of sediment with a mix of strong high-purity acids (HNO₃, HCl, HCLO₄, HF; Suprapure^®Merck) in several steps until total dissolution, using screw-top PFA-Teflon™ vials and a hotplate Teflon™ PFA coating.

Water content, organic matter, and grain size influence the scanning densities on XRF analysis (Zhang et al., 2020; Mondal et al., 2021), therefore ICP-MS was used to verify the trends of the measured elements included in this work. The elements Ti, K, Sr, and Rb can be associated with allochthonous detrital input into the lake (Guyard et al., 2007; Davies et al., 2015). The Ti/coh ratio was used as a proxy of clastic input (Ohlendorf et al., 2014; Schittek et al., 2016). The Zr/Ti ratio was used to infer the grain size given that Zr is highly abundant in the coarse silt fraction whereas Ti can be found in the clay to fine-silt fraction (Oldfield et al., 2003). Therefore, high Zr/Ti ratio values indicate increased coarse silt influx (Shala et al., 2014). The Si, P, Cd, U, and Cu elements were standardized by Titanium deposition.

Pollen, diatom, and chironomid records were analyzed at 2 cm discrete intervals so each biological proxy record consisted of 17 samples. For the pollen analysis, 1 cm³ of these sediment samples were processed following the standardized methods outlined by Faegri and Iversen (1989), including KOH treatment, sieving, acids (HCl and HF for carbonate and silicate removal), and acetolysis. In levels with low pollen concentration, an additional 2 cm³ of sediment was processed. For the estimation of pollen concentration (grains cm⁻³), tablets of Lycopodium clavatum were added to each sediment sample (Stockmarr, 1971). The palynomorphs were determined at the most detailed taxonomic level under a microscope (400–1000×) aided by pollen taxonomic keys (Heusser, 1971; Markgraf and D'Antoni, 1978) and reference samples from the Laboratorio de Paleoecología y Paleoclimatología of the Centro de Estudios Avanzados en Zonas Áridas (CEAZA). The basic pollen sum includes a minimum of 300 terrestrial pollen grains per sample while paludal, aquatic, and/or non-pollen taxa such as zygospores (Zygnemataceae) and microalgae were incorporated in a separate sum. Relative abundances were calculated for each taxon. Pollen accumulation rates (grains cm⁻² yr⁻¹) were calculated employing the pollen concentration values and sedimentation rate derived from the age-depth model.

Two (2) grams of sediment were processed for diatom analysis following the methodology outlined by Battarbee (1986). Each sample was dried, oxidized with H₂O_2, and heated for 2 minutes in a microwave. Finally, permanent preparations were mounted using Naphrax^®. A minimum of 600 valves were counted to determine the relative abundances. For absolute abundances, the aliquot method (Battarbee, 1986) was used, following random transects. Results are expressed in valves per gram of dry sediment. The taxonomic literature on diatom determination included the monographic works of Metzeltin and Lange-Bertalot (1998); Metzeltin and Lange-Bertalot (2007), Rumrich et al. (2000), and Lange-Bertalot et al. (2017), among others.

Five (5) grams of wet sediment were processed for chironomid analysis. The sediment samples were deflocculated using 10% KOH solution at 50–70°C for 30 minutes and sieved through 100 and 200 um mesh sizes. Larval head capsules (HC) were hand-sorted from the residual sediment and mounted on permanent slides with Hidromatrix® mounting media. Taxonomic identification was performed under a Nikon Phase optic microscope at a magnification of 100–1000× with reference to available taxonomic literature (Massaferro and Brooks, 2002; Massaferro et al., 2013; Cranston, 2019).

Given the nature of each of the analyzed proxies in the Laguna El Calvario record, their sensitivity and response time to a given environmental/climatic change may differ from one to another. So we decided to analyze each of them separately and therefore integrate their signal in the discussion section taking into account the different spatial/temporal scales represented by each other.

The retrieved core of Laguna El Calvario spans the last 2400 cal yrs BP ( Figure 2 ). The part of the age–depth model based on the three radiocarbon dates is the most accurate possible concerning the given ages ( Table 1 ). The ²¹⁰Pb radioactive activities showed good exponential decay and the unsupported activities occurred in the first 15 cm reaching the age of 1833 AD ( Figure 2 , Table 1 Supplementary Material ), determining a supported activity of 0.85 ± 0.17 dpm g⁻¹. The mean sedimentation rate ranges around 0.13 ± 0.01 cm year⁻¹ in the first 5 cm.

The lithological description and organic content of the sedimentological record of Laguna El Calvario (LCA-SHC4; 33 cm) are summarized in Figure 3 . Between 33 and 28 cm the sediment is composed of a porous dark brownish silty sand with gravel-sized clasts. The organic matter and carbonate contents are around 25% and 3%, respectively, while the inorganic density exhibits its highest values (0.21–0.33 g cm⁻³). The sediment composition shifts to clayey silt from 28 cm to the top of the core. Intercalated light–dark brown and green laminated layers are present around 20–28 cm and 2–8 cm whereas a brown homogeneous layer is present between 20 and 8 cm. The content of organic matter and clasts fluctuates around 22–42%, and 2.6–3.9%, respectively. Inorganic density ranged between 0.1 and 0.2 g cm⁻³.

Regarding the geochemical results, changes in XRF-Molybdenum Incoherent/Coherent ratio (inc/coh) follow the LOI organic matter (%) and water content (%) ( Figure 3 ). The elements Ti, K and Sr show similar trends throughout the core according both, XRF and ICP-MS data ( Figure 4 ; Tables 2 , 3 Supplementary Material ), which validates the data obtained with XRF. Indeed, the strong coherence among these curves suggests the same origin of variability ( Table 2 Supplementary Material ). The sparse plant cover in the surroundings and the hydrological changes associated with the precipitation regime dynamics can modulate the terrigenous input into the lake therefore, the Ti/coh ratio was applied as an indirect indicator of precipitation. Thus, high values of Ti/coh ratio along with Ti, K, and Sr in the ICP-MS were recorded between 32 and 28 cm ( Figure 4 ). Between 28 and 20 cm, Ti/coh ratio shows highly variable values with maxima around 27 and 21 cm concomitant with the occurrence of a dark layer ( Figure 4 ). Between 20 and 8 cm, these ratio values are more stable but show a rising trend. Ti, K, and Sr and the Ti/coh ratio values display a declining trend in the last 8 cm. Additionally, XRF data including Rb showed similar trends ( Figure 4 ).

The Zr/Ti ratio, a proxy of grain size changes, shows peak values and high variability that are concomitant with laminated sediments between 28 and 20 cm and then, in the last 8 cm ( Figure 4 ). The silica data based on the XRF analysis were not included in the results due to the low counts. In this regard, the Si/Ti ratio in the ICP_MS analysis shows different trends of the detrital-related elements since increases are observed in the topmost 6 cm of the record ( Figure 4 ). So, the variation of silica in the record could be attributed to intra-lake processes such as productivity. The P/Ti ratio from the ICP-MS data also shows the highest values in the topmost 6 cm supporting the silica interpretation. The ICP-MS Cd/Ti and U/Ti ratios were considered as proxies of sulfidic conditions and organic sedimentation given that the U content is normally well correlated with organic fluxes reaching the bottom of the lake. In fact, Cd/Ti and U/Ti ratios present their highest values in the topmost 8 cm. Finally, the ICP-MS Cu/Ti is considered an indicator of mining activity in the region, which displays the highest concentrations in the uppermost 6 cm.

The palynological record shows a high plant diversity with pollen taxa defining the three main vegetation belts in the western semiarid Andean region: Poaceae, Adesmia-type, and Nassauvia-type (high and lower Andean); Ephedra (sub-Andean), and Chenopodiaceae and arboreal taxa (pre-Andean) ( Figure 5A ).

Between 32 and 22 cm (2400–1000 cal yrs BP; Figure 5A ), the pollen record is dominated by Poaceae (up to 22%), Chaethantera/Oriastrum (6–17.3%), Senecio-type (6–12%), Oxalis (3–11%), Ephedra (<10%), Chenopodiaceae (9–14%) and Arenaria (4–13%). The cold-tolerant pollen types such as Nassauvia-type (up to 3.6%), Adesmia-type (up to 6.2%), Laretia-type (up to 5%) reach their highest values at the top of the zone (1500–1000 cal yrs BP) whereas Poaceae percentages gradually decline (8.5%). Arenaria values increase from 4 to 12% since 1300 cal yrs BP. Cyperaceae and Myriophyllum, paludal and aquatic taxa, display low frequencies (<2%) while the non-pollen palynomorph Pediastrum presents maximum values of ~28 cm. Pollen accumulation rates (PAR) for local terrestrial taxa are relatively low ranging between 9 and 25 pollen grains cm⁻² yr⁻¹.

Between 22 and 12 cm (1000–75 cal yrs B; Figure 5A ), the pollen assemblages are characterized by the slight recovery of Poaceae values up to 14.5% along with the decline of Chaethantera/Oriastrum, Nassauvia-type and Adesmia-type. Ephedra (up to 21.6%), Montiaceae (4–17%) and Verbenaceae (2–7%) values increase during this period. Cyperaceae percentages remain under 2% while Pediastrum values strongly decline to almost zero. Spirogyra spores display a mild increment and the Zygnema spores percentages increase at 16 cm (100 cal yrs BP). PAR values are the lowest for the whole record (total terrestrial pollen ca. 2 grains cm⁻² yr⁻¹) increasing towards the end of this period (up to 40 grains cm⁻² yr⁻¹).

Between 12 and 0 cm (75 cal yrs BP to present; Figure 5A ), shrubs such as Ephedra, Chenopodiaceae, and Senecio-type percentages declined at the expenses of Poaceae (10–37%) and herbs values such as Arenaria (12–28%) and Montiaceae (13–21.6%). Other taxa percentages such as Verbenaceae (5–10%) and Plantago (1–2.5%) increase during this period similar to Cyperaceae percentages that show an increment towards the top of the record (2.6–12.8%). The non-pollen palynomorphs (NPPs), such as the algae Pediastrum and the Zygnema spores, values also show a significant increase whereas Spyrogyra spores are absent in this zone. PAR values are the highest for the whole sequence reaching up to 126 grains cm^–2 yr^–1.

The diatom record of the Laguna El Calvario includes 15 taxa that have relative abundances over 5% ( Figure 5B ). Between 32 and 19 cm (2400–700 cal yrs BP), diatom assemblages are characterized by a high frequency of Cymbella spp. (20–42%), Halamphora veneta (15–25%) and Gomphonema spp. (20–50%). Associated taxa include Planothidium aff. biporomum (0–5%), Pinnularia aff. brebissonii (1–6%), Craticula pampeana (1–11.6% at 25–27 cm), Craticula halophila (0–7%), Caloneis silicuta (0–3%) and Nitzchia spp. (0–4%). Small fragilariods, including Pseudostaurosira pseudoconstruens, display a short-lived peak at the top of the period (ca. 1000 cal yrs BP). Diatoms abundance fluctuates between 60 and 130 × 10⁷ valves per gram of dry sediment ( Figure 5B ). Between 19 and 5 cm (700 cal yrs BP to 1970 AD), Gomphonema spp. (up to 45%) and the small fragilarioids (with more than 45% at the top of the period) dominate while Cymbella spp. and Halamphora veneta values decline. Other secondary taxa such as Fragilaria capucina var., Amphora copulata, Ulnaria ulna, and Encyonema silesiacum showed minor increments. The diatom record presents a major diatom assemblage change in the last 5 cm (1970 AD to present) when most of the taxa values decline at expenses of the small fragilarioids (82–92%; Figure 5B ). Ulnaria ulna exhibited a minor increase in the most surficial sample (5.6%).

The chironomid record is characterized by the presence of 3 morphotypes ( Figure 5C ). The dominant morphotype is Cricotopus sp1 with ca. 80% of the total relative abundance associated with Cricotopus sp 2 (ca.15%) and Parochlus < 5%. From 32 to 14 cm (2400–120 cal yrs BP), the chironomid assemblages are dominated by the morphotypes Cricotopus sp1 and Cricotopus sp2. The Head Capsules (HC) counts are highly variable ranging between 2 and 33 ( Figure 5C ). Between 14 and 0 cm (from 120 cal yrs BP to 2014 AD), the chironomid record shows a clear alternation of Cricotopus and Parochlus abundance whereas the HC counts ranged from 31 and 2. Cricotopus sp1 dominates the whole record except for a peak of Parochlus at 2 cm. Cricotopus sp 2 co-dominates the records but disappears in the topmost 2 cm ( Figure 5C ). Given the low counts of HC, we interpreted with caution the chironomid record which was supported with the other proxies.

The multiproxy analysis of Laguna El Calvario allows us to have a broad perspective of the past environmental changes during the last 2400 years in the western semiarid Andes (29°S). Following the lithological and biological changes in the sedimentary record ( Figures 3 – 5 ) four main phases can be distinguished.

In order to compare the long-term environmental changes of the Laguna El Calvario (LCA) record with other paleorecords of western South America, (1) the Ti/coh ratio was selected as a proxy of runoff-induced basin erosion ( Figure 6A ); (2) the high Andean grassland taxa Poaceae and lowland xerophytic Ephedra as a moisture availability proxy while Chaetanthera/Oriastrum, Adesmia type, Azorella type, Nassauvia type as a proxy of cold conditions ( Figure 6B ) and; (3) the P/Ti ratio (ICP-MS) and diatom concentration as a lake productivity proxies ( Figure 6C ).

The relative high persistence and dominance of humid taxa throughout the pollen record of the Laguna El Calvario are consistent with the establishment of a wet Late Holocene inferred by several paleorecords in central Chile (30–35°S). Increased winter precipitation has been recorded from 4000 cal yrs BP on in the subtropical Andes (Veit, 1996; Grosjean et al., 1997; Espizua, 2005; Martel-Cea et al., 2016; Tiner et al., 2018; Frugone-Álvarez et al., 2020; Mayta and Maldonado, 2022), the lowlands (Jenny et al., 2002b; Jenny et al., 2003; Villa-Martínez et al., 2003 and Frugone-Álvarez et al., 2017) and the coastal areas (Maldonado and Villagrán, 2002 and Maldonado and Villagrán, 2006). Intense glacial activity in El Encierro valley (29°S; Figure 1A ) located northwards of Laguna El Calvario was recorded before 2600 cal yrs BP (Grosjean et al., 1998; Zech et al., 2006). So after this period of glacial advance (3000–2600 yrs cal BP; Grosjean et al., 1998), the Laguna El Calvario basin could have been formed when ice retreated around 2600 cal yrs BP. High lake levels in the Laguna El Negro Francisco in the Andes at 27°S (4125 m a.s.l., Figure 6D) (Grosjean et al., 1997) and a smooth trend to wet conditions in small lakes at 30°S (3900–3800 m a.s.l.) (Mayta and Maldonado, 2022) were recorded after 2700 cal yr BP. This chronology is almost consistent with the high erosion around the lake basin ( Figures 6A, B ) and the downward distribution of the vegetation belts in Laguna El Calvario before 1500 cal yrs BP ( Figures 4 , 6C ). On the other hand, grain size and geochemical analysis from two lakes located further south (Laguna El Cepo and Laguna Cerritos Blancos, 30°S, 2900–3800 m a.s.l.; Figure 1A ) suggest an increment of storm frequency since 2200 cal yrs BP ( Figure 6F ) (Tiner et al., 2018) whereas the pollen assemblages of Laguna Quebrada Parada and Laguna Corralito ( Figure 1A ) pinpoint the establishment of wetter conditions from 1900 cal yrs BP (Mayta and Maldonado, 2022). Synchronous wet conditions inferred Laguna El Calvario record around 2400–1600 cal yrs BP were recorded in Laguna Chepical (32°S; Figures 1B , 6G ) that shows wetter than present conditions between 2700–1300 cal yrs BP (Martel-Cea et al., 2016). Coastal and lowland records reflect wetter conditions after 3000 cal yrs BP peaking after 1800 cal yrs BP that were associated with an equatorward position/migration of the northern border of the SWW (Jenny et al., 2002a; Maldonado and Villagrán, 2002; Jenny et al., 2003).

Between 1500 and 800 cal yrs BP, the Laguna El Calvario record reflects a high variability of precipitation at the centennial scale along with the persistence of cold conditions as recorded by the dominance of high-altitude pollen types. The other paleorecords from the Elqui valley (30°S; Figure 1A ) also show wetter conditions coupled with an increase in storm frequency (Tiner et al., 2018; Mayta and Maldonado, 2022). However, the Laguna Chepical record (33°S) reflects a decrease of moisture by this time but an extended ice cover season ( Figure 5G ) (Martel-Cea et al., 2016). In the lowlands, the Laguna Aculeo (33°S; Figure 1A ) record pinpoints wet conditions between 2500–700 cal yrs BP (Villa-Martínez et al., 2003) while Palo Colorado (32°S; Figure 1A ), a coastal record, displays a retraction of wet indicators around 700 cal yrs BP (Maldonado and Villagrán, 2006).

After 800–750 cal yrs BP, a marine core record located at 41°S inferred a less humid interval as a result of the southward displacement of SWW (Lamy et al., 2001) in concordance with an increase of pollen types indicating drier conditions in Laguna El Calvario, Laguna Quebrada Parada and Laguna Corralito (Mayta and Maldonado, 2022). Superimposed on these long-term conditions, a high frequency of ENOS-modulated storms was recorded by different proxies in Laguna Aculeo (33°S) and Laguna del Maule (36°S; Figure 1A ) (Jenny et al., 2002a; Frugone-Álvarez et al., 2020, respectively). So, dry general climatic conditions under a high variability associated with the ENSO may have prevailed north of 32°S but more marked southwards of 33°S.

After 1800 AD, the higher temporal resolution of the Laguna El Calvario record, increasing from 0.02 to 0.13 cm yr⁻¹, provides quite detailed data for the last two centuries ( Figure 2 Supplementary Material ). Drier than present conditions were replaced by colder ones as reflected by the increase in upper Andean belt elements (cold-tolerant taxa) around 1850 AD. This is synchronous to the last part of the Little Ice Age, recorded in central Chile lowlands as lower spring-summer temperatures (von Gunten et al., 2009b), glacial expansion in the high Andes at 35°S (Espizua and Pitte, 2009) and increased precipitation in central Chile (33–34.5°S; LeQuesne et al., 2006). Between 1900 and 1950 AD, wet pollen indicators (Poaceae) abundance increased at the expense of cold tolerant taxa pollen. Besides, a synchronous higher variability of runoff indicators related to an upward shift of the zero-isotherm altitude might suggest an increase in liquid precipitation. This increasing trend of precipitation was also recorded by the low-frequency signal of three ring chronologies of central Chile (33°–34.5°; LeQuesne et al., 2006).

During the last part of the 20^th century, similar to modern dry conditions established in Laguna El Calvario synchronously to the mediterranean Andes (Martel-Cea et al., 2016) and the Altiplano (Morales et al., 2012). The recent decline of Poaceae and the increase of lowland shrubs (Ephedra and Chenopodiaceae) may represent an ongoing upward displacement of plant communities in the subtropical Andes. This process can be directly linked to the ongoing climate change characterized by the establishment of extreme drought conditions (also called Mega Drought; Garreaud et al., 2019) that broadly affects the Andean lake basins in central Chile (Fuentealba et al., 2021). Regarding regional warming, warmer autumns and springs (Burger et al., 2018) may have influenced the nutrient enrichment (high productivity, OM, and P) in Laguna El Calvario ( Figure 6D ; Figure 2 Supplementary Material ) but may have led in a drastic decline of primary productivity in the plant communities at the same time (Rudloff et al., 2021).

Climate variability in the subtropical Andes during the late Holocene, mainly after ~2000 cal yrs BP has been attributed to El Niño Southern Oscillation, whose positive phase (El Niño) originates rainy winters in central Chile (Jenny et al., 2002a; Maldonado and Villagrán, 2006; Martel-Cea et al., 2016; Frugone-Álvarez et al., 2020). Wetter than present conditions around 2400 and 1500 cal yrs BP as well as colder conditions until 800 cal yrs BP in Laguna El Calvario is near concomitant with the increased El Niño activity inferred by several tropical records from the Eastern Pacific with a range of ca. 2000 to ca. 1000 cal yrs BP ( Figures 6G, I ) (Moy et al., 2002: 2000–1000 cal yrs BP; Rein et al., 2005: 2000–1300 cal yrs BP; Conroy et al., 2008: 2000–1500 cal yrs BP). Therefore, ENSO may have played a key role in the hydrological dynamics in the Laguna El Calvario at least between 2000 and 1500 cal yrs BP. Most ENSO paleorecords show a more weakened activity or La Niña-like phase in the past 800 years that may explain the stable dry conditions displayed by the Laguna El Calvario record (Moy et al., 2002; Rein et al., 2005; Conroy et al., 2008). Some discrepancies among records of central Chile and the subtropical Andes during the last millennium (i.e., Laguna Aculeo, Laguna del Maule, and neoglacial advances) could be attributed to other large atmospheric anomalies such as Southern Annular Mode (SAM) that also could have played a key role in the past at the interannual scale (Vuille and Milana, 2007; Dätwyler et al., 2020). However, further studies are needed to elucidate the implications of the coupling SAM and ENSO modes of variability in the past by evaluating the sensibility of paleorecords of the semiarid Andes of central Chile.