Stable Oxygen Isotope Records (δ18O) of a High-Andean Cushion Peatland in NW Argentina (24° S) Imply South American Summer Monsoon Related Moisture Changes During the Late Holocene

plants muticus ( extend the knowledge base on the modern ecology of these peatlands, we investigated the stable isotope composition of bulk material and cellulose ( δ 18 O, δ 13 C, δ N) of the dominant cushion-forming species andina (Juncaceae) and Z. muticus (Cyperaceae) as well as water samples ( δ 18 O, δ 2 H) of several pools interspersed within the peatland. We further applied a multiproxy approach for a peat core from CTP spanning the last 2900 years with XRF scanning, bulk geochemistry and stable isotope analyses on bulk peat and cellulose size fractions. ) due to local variable evaporative enrichment. Within the peat core, we observe considerable multi-centennial variations in δ 18 O composition of cellulose conﬁrmed by all size fractions. Based on the regional relation between decreasing δ 18 O prec values with increasing precipitation amounts and 18 O enrichment in the peatland waters due to evaporation, we suggest an interpretation of our δ 18 O cellulose record as moisture proxy for CTP. This interpretation is corroborated by a high correlation between oxygen isotopes, peat growth and geochemical data. Accordingly, CTP indicates dryer conditions between 2190 and 2120, 1750 and 1590, 1200 and 1080 and since 130 cal. yr BP, whereas periods with increased humidity prevailed from 2750 to 2250 and from 600 to 130 cal. yr BP. Temporal changes in the match to South American Summer Monsoon (SASM) reconstructions suggest impacts of other large-scale atmospheric variability modes or a different SASM expression at our southerly location.

High-elevation cushion peatlands are promising archives for paleoenvironmental studies in their extreme habitat of the Central Andean highlands between ∼4000 and 4800 m a.s.l. The Cerro Tuzgle cushion peatland (CTP, 24 • 09 S, 66 • 24 W), located in the NW Argentine Andes, is formed by the vascular cushion plants Oxychloe andina (O. andina) and Zameioscirpus muticus (Z. muticus). To extend the knowledge base on the modern ecology of these peatlands, we investigated the stable isotope composition of bulk material and cellulose (δ 18 O, δ 13 C, δ 15 N) of the dominant cushion-forming species O. andina (Juncaceae) and Z. muticus (Cyperaceae) as well as water samples (δ 18 O, δ 2 H) of several pools interspersed within the peatland. We further applied a multiproxy approach for a peat core from CTP spanning the last 2900 years with XRF scanning, bulk geochemistry and stable isotope analyses on bulk peat and cellulose size fractions. Modern samples of O. andina and Z. muticus expose significant differences in cellulose δ 18 O, e.g., between leaves and rhizomes of O. andina ( δ 18 O l−r = 4.11 ) and between leaves of O. andina and Z. muticus ( δ 18 O l−l = 2.8 ). Modern water samples exhibit strong isotopic differences between single water pools (max. δ 18 O = 13.09 ) due to local variable evaporative enrichment. Within the peat core, we observe considerable multi-centennial variations in δ 18 O composition of cellulose confirmed by all size fractions. Based on the regional relation between decreasing δ 18 O prec values with increasing precipitation amounts and 18 O enrichment in the peatland waters due to evaporation, we suggest an interpretation of our δ 18 O cellulose record as moisture proxy for CTP. This interpretation is corroborated by a high correlation between

INTRODUCTION
The Andes represent a prominent longitudinal climatic barrier for the South American continent. Two major atmospheric circulation systems influence the climate of the Andes, the South American summer monsoon (SASM) and the Southern Hemisphere westerly winds (SHW) (Garreaud et al., 2003). Interactions between these systems affect climate conditions of the entire Southern Hemisphere. Hence, interest in paleoclimate archives from the Central Andes has grown during the recent decades (e.g., Flantua et al., 2016). Between these systems, the Arid Diagonal is developed, characterized by low annual precipitation amounts and a limited number of paleoclimate archives. Due to this lack of continuous and high resolution records, knowledge on the climatic evolution of the Andean region during the Holocene is still limited and insufficient compared to the hemispheric and global significance of this region.
Suitable archives to fill the gap of knowledge are high-Andean cushion peatlands, as several studies could provide in the past (Schittek et al., 2016;Kock et al., 2018Kock et al., , 2019. Continuous, high accumulation rates ensure highly resolved records for paleoenvironmental investigations with a multiplicity of proxy data (Schittek, 2014). While stable isotope analyses are a well-established tool in paleoecological investigations of Northern Hemisphere peatlands (e.g., Ménot-Combes et al., 2002;Moschen et al., 2009), isotope analyses on cushion peatlands reside in an initial stage. Different to the Northern Hemisphere dominant peat-forming species Sphagnum, a nonvascular plant, the cushion-forming species are vascular plants. Vascular plants are able to regulate their gas exchange by stomata, affecting the isotope fractionation processes of both oxygen (δ 18 O) and carbon isotopes (δ 13 C) during metabolic processes (Ménot and Burns, 2001).
The stable isotope values of peatland plants underlie a strong dependency on the respective source values of atmospheric CO 2 and plant available H 2 O. Overall, carbon isotope fractionation is dependent on the atmospheric source value, on the difference between intercellular and atmospheric partial pressure of CO 2 , which is controlled by the stomatal conductance in vascular plants, and on the biochemical fractionation (Farquhar et al., 1989). During more humid phases, an increasing stomatal conductance is leading to decreased δ 13 C values in plant organic matter. During more arid phases, an increased leaf to air vapor pressure gradient is leading to a closure of the stomata to prevent passive water loss as a result of transpiration (Amesbury et al., 2015a). One of the first responses to water deficit in vascular plants is the inhibition of growth, which allows plants to reduce their transpiration rates and to avoid metabolic and cellular damage caused by low water potentials (Cominelli et al., 2008). Besides carbon isotopes, oxygen isotopes are very often applied within paleoecological investigations (e.g., Ménot-Combes et al., 2002). The δ 18 O value of plant cellulose is directly related to the respective leaf water (δ 18 O LW ) and enriched by about 27.0 due to biochemical fraction during formation of primary photosynthates (glucose) (DeNiro and Epstein, 1981;Sternberg, 2009). The aperture of plant stomata has a strong influence on δ 18 O LW . Stomatal closure, caused, e.g., by drought, lead to a decrease of stomatal conductance and increased enrichment of δ 18 O LW values during transpiration (Farquhar et al., 2007). Due to the dependence on the respective water source value (δ 18 O SW ) and the leaf water enrichment, reported δ 18 O values for extracted cellulose of vascular plants in peatlands cover a considerable range (Brenninkmeijer et al., 1982;Hong et al., 2009;Amesbury et al., 2015b).
Stable isotope investigations on cushion peatlands in South America prior to this study have been executed in Peru (Distichia peatlands; Skrzypek et al., 2011;Engel et al., 2014) and Bolivia (Gouze et al., 1987). Further studies on vascular plants in peatlands only have been conducted in Switzerland (Ménot and Burns, 2001), China (e.g., Hong et al., 2000Hong et al., , 2009Liu et al., 2018) and New Zealand (Amesbury et al., 2015a,b). For carbon isotopes, a relationship between δ 13 C and temperature (Ménot and Burns, 2001;Skrzypek et al., 2011;Engel et al., 2014;Amesbury et al., 2015a) was described. Ménot and Burns (2001) also reported that on an altitudinal gradient the atmospheric CO 2 concentration is a stronger determinant for δ 13 C values of peat plants than temperature. Hong et al. (2000Hong et al. ( , 2009 interpret δ 18 O as a proxy for surface air temperature changes, while Amesbury et al. (2015b) used δ 18 O for tracking precipitation moisture sources.
Here we investigated modern water and plant (leaves, rhizomes, fruits) samples of the Cerro Tuzgle cushion peatland (CTP) as well as a 2.6 m composite peat profile. In the peat profile, we analyzed the major and trace elements, ash content, humification tendency, organic carbon and total nitrogen isotope composition of bulk peat and carbon and oxygen isotope composition of cellulose. Prior to cellulose extraction, samples were separated into three different size fractions (<200 µm, 200-1000 µm, >1000 µm). Additional macrofossil analyses were used to identify changes between different plant species within the record. As a general hypothesis for the study we postulate that changes in moisture conditions at CTP are mirrored in δ 18 O cell values of the peat profile and that disturbances in such a relation, e.g., because of decomposition, can be detected within the multi-proxy approach.
Overall, the study strives (1) to set-up a modern dataset as base for interpretation of the isotope records, (2) to establish the first continuous cellulose stable carbon and oxygen isotope record of cushion peatlands from the Central Andes, (3) to discuss moisture fluctuations at CTP in the context of available SASM reconstructions.

STUDY AREA
NW Argentine, High-Andean "Puna" The CTP (24 • 09 S, 66 • 24 W, 4,350 m a.s.l., Figure 1a) is located in the northwest Argentine Andes close to the Cerro Tuzgle (24 • 09 S, 66 • 24 W) in the province of Jujuy. The area is characterized by recent tectonic processes and geothermal activity, including several seasonal and permanent springs in the so called Cerro Tuzgle -Tocomar geothermal volcanic area (Giordano et al., 2013). The occurrence of block glaciers or highelevation cryogenic soils is not reported for the study area. The region is part of the internally drained Puna comprising the southern part of the Altiplano-Puna plateau (Allmendinger et al., 1997;Norini et al., 2014). Due to differences in precipitation amount, the Argentine Puna can be subdivided into a semihumid part in the east, where the CTP is situated and an arid area in the west (Cabrera, 1968;Troll, 1968;Kuentz, 2009).
Its location is in the north to the Arid Diagonal at the crossroads between tropical and extratropical precipitation regimes (Vuille and Ammann, 1997;Garreaud et al., 2009). The weather station at the Concordia mine (4,144 m a.s.l.) shows the main climatic characteristics of the study area (Figure 1b).
Annual temperatures range between 2 and 10 • C (mean annual temp.: 6.9 • C) with temperature maxima prevailing in the austral summer months. Annual precipitation sums are 109 mm on average while more than 90% occur between November and March, indicating the main source for meteoric water reaching the CTP is the SASM. Cold front or cuts off events that lead to precipitation in the austral winter months are rare (Figure 1b).

Cushion Peatlands as Environmental Archives
In the recent decade, high-elevation cushion peatlands turned out to be high-potential archive for paleoenvironmental investigations (Squeo et al., 2006;Engel et al., 2014;Schittek et al., 2015). These sensitive ecosystems are comparably less vulnerable to the availability of water due to their special textural adaptations like cushion-growth that increase the resistance to water fluctuations (Schittek, 2014). Thus, they have the potential for continuous records from the harsh environment in the (semi-) arid region of the Central Andes. High-Andean cushion peatlands can accumulate several meters of thickness, which are composed of Juncaceae like Oxychloe, Distichia and Patosia and are frequently supplemented by Cyperaceae of the genera Zameioscirpus and Phylloscirpus (Ruthsatz, 2008). Furthermore, cushion peatlands are very sensitive to climatic changes due to their occurrence close to the limits of vegetation growth and contingent upon temperature, water economy and growing period (Ruthsatz, 1977).
A number of high-elevation peatlands occur in the study region (Werner, 1974). In contrast to most of the cushion peatlands in the area, CTP is situated between comparatively flat slopes of outcropping Ordovician volcanic rocks and sedimentary units as well as Cretaceous conglomerates and sands. Thus, it is protected against increased allochthonous sediment input  Mina Concordia (1950-19901966 and1970 are missing) data based on Bianchi et al., 2005]. Map data source is GLCF World Data and DGM-GTOPO30. from lateral stream channels or valleys during extreme rainfall events (Schittek et al., 2016). A presently inactive alluvial fan at the southernmost section of the peatland developed a natural dam and led to continuous accumulation of peat in the past. Modern vegetation of the investigated section of the peatland is characterized by stable mats formed by Oxychloe andina (O. andina), partially supplemented by Zameioscirpus muticus (Z. muticus). O. andina is forming dense cushions by keep growing at their tops and form rhizomes at the bottom (Figure 2b). The leaves are lignified and have a prickle at their tops, potentially as protection against grazing of hoofed animals (Figure 2c) (Rauh, 1988;Ruthsatz, 2008;Schittek et al., 2016). Z. muticus grows in small gaps and fissures of the O. andina cushions. Furthermore, Cyperaceous species like Z. muticus can overgrow open peat surfaces that were damaged by mechanical, frost or drought harms, and thus can protect the peatland against erosion (Ruthsatz, 2008). In contrast to O. andina, Z. muticus has a higher physiological and ecological tolerance, showing, e.g., greater resistance against frost or droughts. However, the species cannot compete with the growth of the cushion plant Oxychloe. The structure of their rhizomes is too insecure, the leaves are neither lignified nor possess a prickle and the growth does not always continue at their tops (Ruthsatz, 2008).
As the CTP is situated in a valley of an ephemeral stream, the Juncaceous cushion plants are interspersed by shallow pools, which can be connected by superficial water flows, originating from springs and during rainfall events from run-off from the surrounding slopes, but also occur isolated without superficial connection (Figure 2a). Underground flow can additionally lead to hydrologic connectivity within cushion peatlands (Schittek et al., 2012). However, for CTP, consisting of homogeneous peat with interspersed minerogenic matter of smaller grain sizes, the water flow velocity is likely to be low.

Modern Water and Plant Samples
In October 2013, two crossing transects (from northwest to southeast and from southeast to northwest) were set to enable a systematic sampling of the peatlands surface waters. Furthermore, various modern plant samples of different species (O. andina and Z. muticus) were taken in a close radius around the coring location in 2012. In November 2014, further water samples were taken from selected positions within the peatlands surface. The peat water sample was extruded by hand from a respective peat sample close to the surface of the intact cushion.
Modern plant samples were separated into leaves, rhizomes, and fruits of O. andina and Z. muticus in the field. In the laboratory, the separated plant parts were washed several times with deionized water to remove all visible sedimentary remains. After freeze-drying, the plants were milled with a high-speed mill grinder (Retsch, MM 400) to generate homogeneous samples. Since dispersion was not necessary for these samples, cellulose extraction after Wissel et al. (2008) was conducted without a treatment of NaOH.
For δ 18 O cell analyses, about 275 µg of freeze-dried cellulose was weighed in silver capsules, crimped and stored for at least 24 h in a vacuum drier at 100 • C before analyses. Samples were then pyrolysed at 1450 • C by using a high temperature pyrolysis analyzer (HT-O, HEKAtech) and measured on-line with a coupled IRMS (Isoprime, GV Instruments) to determine the oxygen stable isotope ratio. Peak integration of m/z 28, 29, and 30 was used to determine the oxygen content (O). The δ 13 C-values were determined using 200-300 µg of dried cellulose weighed into tin foil cups. Samples were combusted at 1080 • using an element analyzer (EuroEA, Eurovector) interfaced on-line to an isotope IRMS (Isoprime, GV Instruments). Peak integration of m/z 44, 45 and 46 was used to determine the carbon content (C). The element contents of C and O respectively the C/O ratio can be used as a quality indicator for the extracted cellulose in comparison to the theoretical C/O ratio in cellulose of 0.90 (Wissel et al., 2008).
The overall precision of replicate analyses is estimated to be better than 5% (rel.) for carbon and oxygen content, <0.1 for δ 15 N and δ 13 C and <0.25 for δ 18 O.
Stable isotope analyses of the water samples were carried out using a Cavity Ringdown Spectrometer (L2130-I, Picarro).
Results are reported as δ values relative to the VSMOW (Gonfiantini, 1978). Internal standards, calibrated against VSMOW, Standard Light Antarctic Precipitation, and Greenland Ice Sheet Precipitation were used to ensure long-term stability of analyses. The precision of the analytical system was ≤ 0.1 for δ 18 O and ≤1.0 for δ 2 H.

Core Sampling and Geochemical Measurements
In December 2012, the core Tuz-694 which reached a total depth of 8 m was extracted in 1 m segments by percussion coring. In this study, we present investigations on the upper 3 m of the core Tuz-694. The core was opened lengthwise with a modified core saw, cut into two halves, photographed and described lithologically.
For major and trace elements analyses, the first core half was measured with a resolution of 2 mm using an ITRAX X-ray fluorescence (XRF) core scanner (Cox Analytical Systems; Croudace et al., 2006). A molybdenum (Mo) tube at 30 kV and 10 mA was used with an exposure time of 10 s for every measurement respectively. The same core half was cut in one centimeter slices, from which boundary areas were removed to exclude possible contamination, which were used for cellulose extraction. The second half was accordingly prepared in 1 cm slices for stable isotope (δ 13 C OM , δ 15 N bulk ) and element content analyses as well as loss on ignition (LOI 550 ) and humification (T 535 ) measurements on bulk material. From these samples, subsamples at 8 cm intervals were selected for macrofossil analyses.
Prior to the cellulose extraction after Wissel et al. (2008), bulk peat samples were treated with NaOH (5%) to ensure dispersion and enable separation into three size fractions by sieving at 200 and 1000 µm (established fractions: <200 µm, 200-1000 µm, >1000 µm). Oxygen and carbon isotope measurements of the extracted cellulose followed methods described above.
For stable isotope and element content analyses as well as for LOI 550 and T 535 , bulk peat samples were freeze-dried and milled (Retsch, MM 400). For the analyses of the stable isotope ratios of carbon (δ 13 C OM ) and total carbon content (TC) as well as total nitrogen isotopes (δ 15 N bulk ) and nitrogen content (TN) one measurement was conducted using tin capsules and combusted at 1080 • C in an elemental analyzer (EuroEA, Eurovector) with an automated sample supply linked to an isotope ratio mass spectrometer (Isoprime, Micromass). TN and TC contents were calculated according to the amounts of N 2 and CO 2 released after sample combustion (peak integration) and calibrated against element standards. Various tests with 5% HCl proved that decalcification was not necessary for the determination of organic carbon content and for δ 13 C OM measurements, thus, the TC content is an equivalent of the total organic carbon content (TOC). Loss on ignition (LOI 550 ) determination followed standard procedures with an initial milled material weight of 25-30 mg (see Heiri et al., 2001). Accordingly, the ash content was calculated by subtracting the LOI 550 from 100%. Humification measurements followed the procedures recommended by Blackford and Chambers (1993) and Chambers et al. (2011), standardized to a target value of 20 mg organic matter in the sample.

Composite Profile, Radiocarbon Dating and Data Analyses
The upper sections of the second and third Tuz-694 core segment were determined as caving material because of loose bedding and deviations in XRF results and were thus discarded. A small hiatus occurs between 98 and 100 cm due to loss of material owing to the coring procedure. Moreover the retrieved core segments were compacted unequally during the coring due to the high water content. Within the first 1 m section, the peat material was compacted to 80 cm (20% compaction), the peat material of the second segment to 49 cm (51% compaction) and the third segment was compacted to 67 cm (33% compaction). To build a composite profile, the core segments were linearly decompacted by using a correction factor (core length/compacted length). The developed composite profile is the basis for the age-depth model. All depths mentioned in the text are decompacted composite core depths except for the depth given in the stratigraphical description below.
In an earlier contribution (Schittek et al., 2016), the calibration of radiocarbon ages for the Tuzgle composite record was based on the IntCal13 calibration data set . Meanwhile, Marsh et al. (2018) proposed a mixed-curve model Calibrated ages (2σ ) were calculated using CALIB 7.0.4 (Stuiver et al., 2013) and the SHCal13 data set (Hogg et al., 2013). Given are the radiocarbon ages and their measurement error, the mean calibrated ages (2σ error in brackets) and the modeled median ages (2σ error in brackets).
between IntCal13 and SHCal13 for the regions affected by the SASM and receiving unknown amounts of carbon dioxide input from both hemispheres. Although atmospheric CO 2 from the Northern Hemisphere is brought to the eastern Andes during the austral summer, its potential influence declines with increasing southern latitude. At the latitude of CTP, the impact of northern hemispheric CO 2 should be negligible. Accordingly, radiocarbon ages for the Tuzgle composite record were recalibrated here with the SHCal data set, the age-depth model was recalculated (Tuz.v2) and geochemical proxies from Schittek et al. (2016) are also shown on this updated age-depth scale. The modeled ages for the decompacted core depths are the medians of 800 Monte Carlo simulations also used to generate confidence intervals integrating the probabilistic nature of calibrated 14 C-dates (Higuera et al., 2009; MCAgeDepth software package). All discussed ages are calibrated ages BP (cal. yr BP), if not mentioned otherwise. Pearson correlation coefficients (p > 0.05) were determined by using the R software (version 3.2.2) and the package "rioja" (version 0.9-15; Juggins, 2017) to describe parameter relations.

Stratigraphy and Chronology
The stratigraphy of core Tuz-694 is dominated by homogeneous, faintly layered peat, partially interspersed with clastic material (Figure 3). Between 14 and 8 cm depth, sand is supplementing the highly organic peat. A short hiatus, occurring between 100 and 98 cm, is caused by core loss due to the coring procedure. At 134 cm, a fine gravel dominated layer occurs. Between 245 and 230 cm, an increased content of silt is observed. The chronology is based on nine AMS radiocarbon dates and the drilling date as anchor point for the core top ( Table 1).
The age-depth model (Figure 3) shows a continuous peat accumulation of the CTP for the last ∼2900 cal. years. The average time-resolution of our CTP record is about 11 years per centimeter with a range between 7 to 18 years.

Modern Data
The δ 18 O values of modern water scatter between −11.1 and 2.0 , while the respective δ 2 H values vary between −78.2 and −13.8 (Supplementary Table S1). Overall, the data reveal a very good correlation between δ 18 O and δ 2 H (r = 0.99). The extruded water sample shows relatively low δ 18 O (−8.7 ) and δ 2 H (−65.1 ) values, however, only a small offset to isotopic of an adjacent pool (−9.5 δ 18 O and −71.1 δ 2 H) could be revealed.
The analysis of plant samples from the CTP reveal distinct differences between O. andina and Z. muticus ( Table 2). With regard to the cellulose yield, a large difference between leaves of O. andina (19.3%) and Z. muticus (10.9%) becomes apparent. The difference between the rhizomes is much smaller (O. andina 14.1% and Z. muticus 16.6%) and almost indistinguishable with respect to the small sample number. The isotope measurements of the extracted cellulose indicate that leaves of Z. muticus have lower δ 18 O cell values (28.9 ) than leaves of O. andina (31.68 ), but that Z. muticus has slightly higher δ 18 O cell values in rhizomes compared to O. andina (28.2 to 27.6 ).
The δ 13 C cell dataset indicate values as expected for C3 plants (Pate, 2001). δ 13 C cell values are consistently enriched in leaves and rhizomes of Z. muticus compared to leaves and rhizomes of O. andina (leave δ 13 C cell: Zam−Oxy = 0.6 , rhizome δ 13 C cell: Zam−Oxy = 0.8). Whereas the intraplant difference between leaves and rhizomes in the carbon isotope composition (δ 13 C l−r ) is very low, major differences occur in the oxygen isotope composition of the investigated species ( δ 18 O l−r = 0.73 for Z. muticus; δ 18 O l−r = 4.11 for O. andina). Comparing distinctive leaverhizomes pairs within the same cushion of O. andina results in an even higher spread ( δ 18 O l−r = 4.95 ).

Core Data
Bulk Isotope Composition, Element Contents, LOI 550 and Humification Values of δ 13 C OM reveal low variability between −26 and −25 during most parts of the record (Figure 4). Highest values occur between 1700 and 1270 cal. yr BP (up to −24.5 ), before a decreasing trend toward lower δ 13 C OM values prevail until 230 cal. yr BP. The δ 15 N bulk values shift considerably between 4.0 (∼2060 cal. yr BP) and 1.1 (∼390 cal. yr BP) with phases of higher (1730 to 1340, 700 to 610, and around 110 cal. yr. BP) and lower (2910 to 2300, 1240 to 1010, 500 to 350 and 50 to −30 cal. yr BP) values. The TOC content ranges between 42.8 FIGURE 3 | Core images, stratigraphy and age depth model of the core Tuz-694. 14 C-ages with their respective uncertainties are shown on the left side of the core images. The only hiatus in the stratigraphy is present between 98 and 100 cm. The age depth model is based on nine 14 C-dates. The red dots indicate the mean calibrated ages with error bars (2 s). Gray band represents the modeled uncertainty and the black line the 50th percentiles.  Table S2). Similarly, only a weak correlation occurs between δ 13 C OM and δ 15 N bulk (r = 0.16). T 535 values reveal a negative correlation with δ 13 C OM (r = −0.4) and with TOC (r = −0.39).

Cellulose Oxygen and Carbon Isotopes
The C/O ratios of cellulose from all three investigated size fractions fall into a small range around 0.9 and, thus, indicate that the extraction provided cellulose free of inorganic and organic contaminants (Supplementary Figure S1, see Wissel et al., 2008 for details).
In general, the δ 18 O cell values of the different size fractions fluctuate between 28.0 and 31.0 and show comparable trends (Figure 5). Conspicuous peaks (e.g. ∼1320 cal. yr BP) or phases with increased (e.g., 2210 to 2120, 1800 to 1600 or 1160 to 1080 cal. yr BP) or reduced values (e.g., 2530 to 2230 or 610 to 130 cal. yr BP) occur in all three time series within the same depths. The high correlation between the fractions < 200 µm and 200-1000 µm (r = 0.93) confirms a strong similarity between these size fractions. δ 18 O cell values of the fraction > 1000 µm are comparable to those of the two smaller fractions, however, with a reduced variability and slight differences in the temporal development. This is confirmed by weaker, but still remarkably high correlations (r ≥ 0.59).
Comparing all size fractions, the δ 13 C cell values vary on average between −22.0 and −23.0 . However, a transition from lower to higher variable δ 13 C cell values after 1080 cal. yr BP is prominent. Further, a slight trend toward decreasing values after 1080 cal. yr BP can be distinguished that is mostly visible in the fraction > 1000 µm. The two smallest fractions reveal a high correlation (r = 0.61). Correlations between the smallest and the largest (r = 0.35), as well as between the middle and largest fraction (r = 0.52) are lower. Within fractions, correlations between δ 18 O cell and δ 13 C cell reach r values between 0.17 and 0.48.
In comparison to δ 13 C OM values, δ 13 C cell values are enriched by about ∼3 on average. This difference is comparable with the results of Ménot and Burns (2001) for vascular plants and of Moschen et al. (2009) for Sphagnum peatlands respectively. The trends of δ 13 C cell and δ 13 C OM show similarities that are confirmed by correlation coefficients between r = 0.34 and r = 0.42. However, it is conspicuous that δ 13 C OM values are not decreasing within the upper 50 cm. No correlation occurs for δ 13 C cell and TOC, while slight negative correlations between δ 18 O cell and TOC values exist (r ≤ −0.26). A slight negative correlation can be found between T 535 and δ 13 C cell (r ≤ −0.4) while no correlation exists between T535 and δ 18 O cell (r ≥ 0.1).
Visual inspection of size fractions prepared for cellulose extraction revealed further details of sample compositions. According to nine inspected samples, the largest size fraction (>1000 µm) is mainly comprised of rhizomes with contributions reaching almost 100% and minor admixtures of leaves. The medium size fraction (200-1000 µm) in comparison contains larger amounts of leave fragments resulting roughly in a 40/60% mixture of rhizomes fragments and leaves. The content of the smallest size fraction (<200 µm) was unidentifiable.

Geochemical Composition From XRF Measurements
Sulfur (S) and Calcium (Ca) values reveal a close similarity (r = 0.99) with very low intensities (total counts = cnts) between 2900 and 2070 cal. yr BP and again from 1060 to −60 cal. yr BP (Figure 6)

Implications From Modern Hydrology
The isotopic composition of modern water samples gives an insight into the present hydrological system of CTP. In the headwater as well as the main accumulation section of the peatland (Figure 7), two springs occur (Schittek et al., 2016). These springs are presumably seasonal cold-water springs, located in the Ordovician basement rocks that are characterized by secondary permeability (Giordano et al., 2013). Giordano et al. (2013) describe this part of the Cerro Tuzgle-Tocomar geothermal region as recharge area, thus, the CTP is affected by groundwater of recent precipitation periods and not by deeper geothermal groundwater sources. Following this concept, the isotopic source values of the peatlands' water are directly linked to the isotope composition of local precipitation, though local evaporation processes during infiltration may have additional influence. However, a multiyear smoothing of the input isotope signal has to be expected due to soil/groundwater residence time and the smoothing effect of the peat water reservoir.
The isotopic composition all CTP water samples are compared with the Global Meteoric Water Line [GMWL, δ 2 H = 8.13δ 18 O+10.8, Rozanski et al. (1993)], as the existing database (e.g., Global Network of Isotopes in Precipitation -GNIP) is not sufficient to develop a Local Meteoric Water Line. Irrespective of the observed spatial heterogeneity, the samples reveal regularity as they fall on a single line in δ space, clearly describing the modern evaporation line (LEL) for CTP (Figure 7). This is also true for the extruded plant water sample. The slope of the observed relation (δ 2 H = 4.85 δ 18 O -25.05) is typical for evaporating surface waters (e.g., Mook, 2001) that have to be expected due to the local environmental conditions. We interpret this behavior as another evidence for a single, isotopic homogeneous water source for CTP, i.e., that meteoric water (precipitation, young groundwater) recharges the peatland. Thus, older and potentially isotopically different groundwater is not involved in the hydrology of the peatland which in turn is closely connected to the meteoric water cycle. This is supported by low electrical conductivity values of the water (Schittek et al., 2016).
The surface water data set from October 2013 gives further insights into the spatial heterogeneity of water isotopes. Samples close to the two springs (Tuz-1, Tuz-2,  are characterized by a low degree of 18 O enrichment compared to spring values, thus low evaporative enrichment (Figure 8). All other samples of the CTP suggest a higher degree of evaporative enrichment, e.g., in isolated pools (Tuz-5) that are surrounded by dense cushion vegetation without interspersed channels. The water samples of the lake section (Tuz-9 and Tuz-10) exhibit the strongest degree of evaporation due to their large open water surface prone to evaporation. On the one hand, these spatial heterogeneities indicate the strong effect of microtopographical differences for evaporative isotope enrichment of surface water in cushion peatlands. Similar effects have already been described for Sphagnum peatlands (Ménot-Combes et al., 2002). On the other hand, they point toward a limited movement respectively exchange of water within the peatland despite superficial connections through FIGURE 7 | Isotopic composition of CTP modern water samples, sampled in October 2013 (diamonds) and November 2014 (circles). The black filled circle represents the peat water sample (see Supplementary Table S1). The dashed circles represent different degrees of evaporation; (1) low evaporation, (2) slight to medium evaporation, (3) strong evaporation. Also shown are the GMWL (δ 2 H = 8.13δ 18 O+10.8, Rozanski et al., 1993)  visible rivulets (Figure 2). It seems likely that with increasing dryness or temporal drought, the number of isolated pools in the peatland will grow causing a respective response toward higher δ 18 O and δ 2 H values in these pools. The implication for a microtopographical habitat, i.e., a potentially temporal stable single cushion in the peat land is that variations in water availability through changes in e.g. evaporation, precipitation, connectivity or growth are systematically translated into a respective response of the local source water isotopic signature.

Implications From Modern Vegetation
The modern plant samples reveal distinct isotopic differences between O. andina and Z. muticus ( Table 2). In contrast to Z. muticus ( δ 18 O l−r = 0.7 ) a huge offset in δ 18 O between rhizomes and leaves of O. andina ( δ 18 O l−r = 4.1 ) is evident. Even larger differences between leaves and roots were reported for the vascular plant Empodisma (∼ δ 18 O l−r = 8.4 to 11.8 ) (Amesbury et al., 2015b). We suggest that this offset arises from dissimilar leaf structures. Leaves of O. andina are described as lignified (Ruthsatz, 2008), thus, might possess a higher protection against transpiration, leading to increased 18 O/ 16 O ratios of leaf water and, finally, of leaf cellulose relative to Z. muticus. Considering these intraspecies δ 18 O cell ranges, vegetation changes should have a minor influence on the δ 18 O signal in the size fractions, i.e., 200-1000 µm, since values of Z. muticus leaves and rhizomes are within the limits of O. andina. Given the small species difference for rhizomes ( Table 2), such effects should be negligible for the largest size fraction (>1000 µm).
In comparison with other investigations on vascular (Amesbury et al., 2015a) and non-vascular peatlands (e.g., Loisel et al., 2009), the ranges of δ 13 C cellulose values are small for both species (Table 2). However, previous studies reveal that interspecies variability can have significant influences on the δ 13 C cell signal (Ménot and Burns, 2001). In the present study, mean δ 13 C cell values of Z. muticus are enriched for leaves (δ 13 C Zam−Oxy = 0.61 ) and rhizomes (δ 13 C Zam−Oxy = 0.84 ) respectively in contrast to O. andina. Thus, the potential δ 13 C cell change to be expected for a complete (but unlikely) vegetation change from one species to the other would amount to 0.6-0.8 . Further investigations on the range of δ 18 O cell and δ 13 C cell variations within a single cushion in the same microenvironment could help identify causes for the observed isotopic differences.

Peat Forming Vegetation and Decomposition
To use cushion peatlands as continuous high-resolution archives aspects like changes in plant composition need to be considered. Peatlands consisting of Oxychloe are poor in species diversity, in contrast to e.g., Distichia dominated peatlands, are especially suitable for paleoenvironmental studies (Ruthsatz, 2008) because species-specific effects are limited. This notion applies with regard to the CTP record, where only O. andina and Z. muticus dominate the peatland. Our measurements on modern plant samples indicate that changes in the occurrence between the species could cause a smoothing of δ 18 O variations recorded in the peat core and might lead to a slight shift in δ 13 C cell values (see above). However, macrofossil analyses confirm that O. andina is the dominant peat forming species in CTP within the whole record (Schittek et al., 2016;Schittek, unpublished data). Furthermore, macrofossil analyses indicate an increase of Z. muticus between 80 cm and 130 cm depth (Schittek et al., 2016;Schittek, unpublished data), without any visible response in our isotope proxies excluding any vegetational bias in the records.
Another question is, whether different size or compositional fractions of the peat possess different (climate-) signals. Correlation analyses reveal a strong respectively good accordance between our two smaller size fractions (r = 0.93 for δ 18 O, r = 0.61 for δ 13 C), indicating that the fractions < 200 µm and 200-1000 µm comprise the same signal. Weaker correlations of δ 18 O of these fractions with the fraction > 1000 µm can be linked to the composition of the size fractions since the largest fraction mainly consist of rhizomes while the smaller fractions have varying contributions of rhizomes and leaves. According to our modern plant data (see above), O. andina dominated phases should possess an offset δ 18 O between the smaller (enriched) fractions and the largest fraction dominated by (depleted) rhizomes. However, this expectation is not met by our data, even if some parts of the record exhibit more depleted δ 18 O cell > 1000 values. Most likely a direct comparison between the isotopic compositions of living leaves with already decomposed leaves or other plant parts in the peat is not feasible. A way to overcome this problem could be to use already died off leaves age groups of preceding years of still living plants. Modern δ 13 C cell values of O. andina exhibit a small deviation between leaves and rhizomes ( δ 13 C l−r = 0.2 ), thus, the development in every fraction is expected to be very similar. This assumption is mainly confirmed despite some deviations regarding the δ 13 C cell > 1000 record. Reasons for these deviations could be a change in the contribution of the remaining plant parts (leaves, rhizomes) to the composition of the smaller fractions. Overall, the general isotopic trends in all fractions reveal a good match, wherefore we argue that the primordial environmental signal is imprinted and inherited in all three size fractions.
Decomposition can have an influence on the stable isotope composition of bulk peat or cellulose. Burial and humification of peat material removes at first the more labile organic fractions, thus may significantly change the δ 13 C values of bulk organic matter (Ménot and Burns, 2001) toward enriched δ 13 C values in the residual peat (e.g., Novak et al., 2010). A raised degree of humification can potentially be identified by low T 535 values (e.g., Chambers et al., 2011). In the CTP record, the variation of T 535 is very limited indicating that the influence of decomposition on isotope records of bulk peat is negligible, after the step of initial removal of mobile and soluble phases. Starting from fresh plant litter, cellulose is also decomposed to a comparable or even larger degree as lignin after initial removal of soluble components (e.g., Sjöberg et al., 2004;Fioretto et al., 2005). However, not much is known on isotopic changes through cellulose decomposition especially with respect to the oxygen isotopic composition of cellulose. During an artificial aging experiment, Schleser et al. (1999) observed a maximum deviation of 0.3 from the initial δ 13 C value of wood cellulose, however, with increasing degree of cellulose decomposition the δ 13 C values of the remaining cellulose again reached their initial values. Thus, we expect isotope effects of decomposition on cellulose, despite strong reduction of cellulose content, to be small. Identical (within analytical uncertainty) mean cellulose oxygen and carbon stable isotope values of all size fractions further support this concept.

Paleoclimatic Implications From the CTP Proxy Dataset
The δ 18 O values recorded in plant cellulose are controlled by several (local) influences like evaporation degree of the surface water or changes in peat composition (see above). However, a major factor is the isotopic composition of the source water. Climate modeling studies (e.g., Vuille et al., 2003;Vuille and Werner, 2005) Table S3). The connection between precipitation amount and δ 18 O is supported by Ballantyne et al. (2011)  Based on this fundamental relation, we argue that increased precipitation amounts at the study site are directly connected with more depleted source water for CTP cushion plants and vice versa. Furthermore, we argue that recent precipitation forms the primary source of water for the cushion plants. Since water influx into the peatland from the north or from older groundwater can be excluded (see above), water solely reaches the CTP in form of (1) direct precipitation on the peatland and superficial run-off from the surrounding slopes as well as from the ephemeral stream and (2) recharge from two springs from shallow groundwater (see also Schittek et al., 2016).
Accordingly, during phases of increased humidity, depleted precipitation reaches CTP. Due to a high water volume (but nearly constant surface area), evaporation has potentially lower influence on the δ 18 O values of the water within the peatland. With sufficient water supply (considering site conditions), we expect the vascular plants O. andina and Z. muticus to have comparably open stomata, leading to comparably less 18 O enrichment in the leaf water (e.g., Roden et al., 2000). Overall, these factors operate additive and lead to relatively depleted δ 18 O cell values within the cushion plants. Vice versa, periods with increased aridity are characterized by low but enriched precipitation reaching the study area, stronger influence of local evaporation effects, reduced stomatal conductance and increasing fractionation with enriched δ 18 O LW values, and finally enriched δ 18 O cell values. Thus, we interpret δ 18 O cell values to be a proxy for moisture conditions at CTP (Figure 9b).
Beside oxygen isotopes, the stable carbon isotope compositions of bulk organic matter or cellulose also provide insights into local environmental and climatic conditions (Figure 9a). Through the stomatal aperture transpiration and photosynthesis in leaves are coupled processes, which in turn have strong impacts on the stable carbon and oxygen isotope composition of new photosynthates. Due to this basic dependency of both variables on stomatal aperture, i.e., because gas exchange through the stomata influences both carbon and water cycle of the plants, a certain relation can be expected (Smedley et al., 1991;Barbour et al., 2002;Chaves et al., 2004;Gessler et al., 2009). Since photosynthetic carbon fixation solely influences the plant carbon cycle, such a relation might be very variable in strength. Nevertheless, further evidence for our interpretation of δ 18 O cell as moisture indicator for CTP may be derived from a correlation between δ 13 C cell and δ 18 O cell . Indeed, positive significant correlations appear in our cellulose stable isotope records (Supplementary Table S2). Correlation coefficients between δ 13 C cell and δ 18 O cell reach values of about r = 0.5 for the medium size fractions. Overall, these positive correlations between δ 13 C cell and δ 18 O cell support our interpretation that more enriched δ 18 O cell values are indicative for more severe moisture stress at CTP.
The ratio of Mn/Fe is used as an indicator of redox conditions in CTP (Schittek et al., 2015(Schittek et al., , 2016. Accordingly, anoxic conditions in the peat water column and presumably stable water conditions are indicated by high Mn/Fe values (Figure 9d). Principally, the place of signal generation differs between these proxies (leave, "catotelm") which might already cause temporal deviations. Also, Mn and Fe can be of lithogenic origin and transported to the CTP while redeposition processes can further hinder an unambiguous interpretation. δ 18 O cell values can be affected in detail by several factors like precipitation amount, moisture source, moisture transport history or the degree of  (Morales et al., 2012). Mark that all values are annualized based on the original data (presented in transparent colors) and averaged by 51 years. rainout upstream (Insel et al., 2013). Further investigations are needed to explain the differences between these two proxies connected to the peatland hydrology in more detail.
The striking parallel curves of S and Ca counts (r = 0.99) suggests the hypothesis that these element counts are linked to gypsum precipitates, developed under dry conditions without strong water-table fluctuations (e.g., Litaor et al., 2004). Although the values might indicate secondary gypsum precipitates after core opening, we argue that the ion concentration did not change in the plant pore water, and thus, reflect climatic conditions during synthesis. Interestingly, the phase of strongly increased counts of S and Ca appear together with highest values of δ 18 O cell (Figure 9c), which again strengthen our interpretation of δ 18 O cell as proxy for moisture conditions at CTP.

Regional Paleoclimate Comparison
Based on our δ 18 O cell record, we observe phases with enhanced and reduced moisture supply at the CTP (24 • S, Figure 10a). Wetter conditions prevailed from 2750 to 2250 and from 600 to 130 cal. yr BP, while drier conditions can be identified between 2190 and 2120, 1750 and 1590, 1200 and 1080 and since 130 cal. yr BP. In order to interpret our findings in the context of changes in the large scale atmospheric circulation, we compare the CTP record with local and supraregional paleoclimate records of the high Andes between 10 and 24 • S (Figure 10).
Regional atmospheric circulation is mainly driven by (1) the strength of the Atlantic Meridional Overturning Circulation, (2) latitudinal changes of the Intertropical Convergence Zone (ITCZ), (3) the dipole-like pattern of the South Atlantic Convergence Zone and the South American Low Level Jet (SALLJ) and (4) the position of the Bolivian High (Liebmann and Mechoso, 2011;Marengo et al., 2012;Vuille et al., 2012;Rojas et al., 2016;Novello et al., 2018). During austral spring, moisture transport from the North Atlantic to the Andean highlands starts with a weakening of the meridional temperature gradient between October and November, leading to a southward displacement of the SALLJ. Simultaneously, connected to this shift of the ITCZ over the tropical Atlantic and Pacific, deep convection over the central part of the continent starts to develop (Hastenrath, 1997;Marengo et al., 2012). Finally, this leads to the development of the SASM during the summer months, thus, to N-S moisture transport reaching the NW Argentine Andes (Garreaud, 2000;Vera et al., 2006;Vuille et al., 2012). In the northern Central Andes, numerous archives are describe to have a close δ 18 O dependence to the SASM development (e.g., lake sediments: Bird et al., 2011a,b;speleothems: Kanner et al., 2013; ice caps: Thompson et al., 2013;tree rings: Ballantyne et al., 2011). Comparing these sites (see Figure 10d) with the CTP δ 18 O cell record (Figure 10a) several phases with strong similarities and differences can be detected.
In If we assume that no shift of the seasonal precipitation regimes took place, thus, rainfall occurs only between November and March, other reasons linked to the overturning circulation systems need to be considered. Vuille and Keimig (2004) stated that due to easterly upper-and midtropospheric wind anomalies, three out of phase SASM modes can be differentiated. A northern mode is leading to enhanced convective activity and tropical precipitation in the Andes between ∼10 and ∼20 • S, while a northeastern mode is affecting the eastern Cordillera of the high Andes between ∼15 and ∼22 • S by increased precipitation amounts. Contrary, a southern mode leads to increased precipitation in the southern Central Andes between ∼20 and ∼30 • S, due to a southern displacement of the Bolivian High and a Rossby wave dispersion of the Southern Hemisphere extratropics (Vuille and Keimig, 2004). A more frequent out-of-phase SASM in the past could be a possible explanation for temporal deviations between the proxy records, although extratropical precipitation events as cold air cut offand cold frontal snowfall are typical for the austral winter (Vuille and Ammann, 1997;Barett et al., 2009). To disentangle the paleoclimatic features in the southern Central Andes, more isotope studies on cushion peatlands are needed.
Temperature reconstruction of the Amazon Basin as the source region for the Ammonium concentration in the Nevado Illimani (Kellerhals et al., 2010; Figure 10e) as well as local temperature reconstruction from the high-Andean Laguna Chepical (32 • S, de Jong et al., 2013) excludes the influence of local temperatures on the δ 18 O cell values of the CTP, which could possibly extend/reduce a vegetation-active period. Previous works already showed that the amount effect and not temperature explains a significant fraction of isotopic variations on seasonal and interannual time scales (e.g., Bird et al., 2011b;Vuille et al., 2012).
As archives covering the last 2900 years are scarce, a local paleoclimate comparison is difficult. Nevertheless, taking small age depth model uncertainties into account, precipitation reconstruction for the Altiplano of Bolivia and Northern Chile based on tree-ring width series of Polylepis tarapacana also reveals a good progression with the δ 18 O cell values of the CTP (Morales et al., 2012 ; Figure 10f).

CONCLUSION
For the understanding of cushion peatlands as highly suitable paleoclimate archives our study could reveal several important findings. Modern plant samples expose significant differences in the isotope composition of the dominant plant species and clearly indicate the need for analyses of the modern vegetation prior to the interpretation of any core sample results. Modern water samples reveal strong isotopic differences for single water pools of up to δ 18 O = 13.09 due to evaporation, implying limited movement respectively exchange of water within the peatland despite superficial rivulets. Overall, modern water samples describe a local evaporation line for CTP.
Macrofossil analyses of bulk samples showed that the dominating peat forming species of CTP is O. andina. Despite temporal changes in composition of the deposited peat between O. andina and Z. muticus no detectable influence on the isotopic composition of peat cellulose did occur. Microscopic inspection of peat size fractions revealed distinct differences in their composition (leaves, rhizomes). Irrespective of that, the respective isotope records of all three size fractions are overall consistent.
Increasing precipitation in the study area is connected with decreasing δ 18 O prec values while less precipitation is accompanied by enriched δ 18 O prec values (IAEA/ WMO, 2004). Also, during more arid phases the effects of local evaporation on the isotopic composition of peat water is stronger than during humid phases. Additionally, the stomata of the cushion plants are constricted or closed under water stress, leading to respectively enriched δ 18 O LW values, and finally enriched δ 18 O cell values. Considering these factors together we argue that δ 18 O cell values are a suitable proxy for moisture conditions at the CTP reflecting local humidity changes during the last 2900 cal. yr BP that are closely connected to the SASM. Discrepancies in the temporal match with other SASM reconstructions suggest impacts of other largescale atmospheric variability or a different SASM expression at the southerly location of the CTP. According to this finding, predominant dryer conditions dominated at the CTP from 2190 to 2120, 1750 to 1590 and are dominating since 130 cal. yr BP, while more humid conditions prevailed between 2750 and 2250 and between 600 and 130 cal. yr BP. The investigated size fractions show similar developments in their δ 18 O cell records, however, smaller deviations occur most likely indicating either different compositions or decomposition histories. The obvious differences between the CTP record and other existing climate records reveal the need for further investigations regarding the causes for those discrepancies and provide potential to reveal the history of the spatial (latitudinal) evolution of the SASM during the late Holocene.

DATA AVAILABILITY
All datasets generated for this study are included in the manuscript and/or the Supplementary Files. Raw data can be found in the Supplementary Table S1 and will additionally be made available via the National Oceanic and Atmospheric Administration (NOAA) Paleoclimate database.

AUTHOR CONTRIBUTIONS
All authors have made substantial contributions to the manuscript in an equal way and approved the submitted version. SK, KS, and AL contributed conception and design of the study. KS and FS provided approval for publication of the content. HV, CO, and HW performed the statistical analysis of the data and helped by the interpretation. SK, AL, and KS wrote sections of the manuscript. LL and JK revised the manuscript critically for important intellectual content.

FUNDING
This work was funded by the German Academic Exchange Service (DAAD) PROMOS -scholarship for Thesis preparation 2015 to SK, German Science Foundation to FS (SCHA 14-1/2), the German Federal Ministry for Education and Research to KS (ARG 06/009), and the National University of Jujuy to LL (SeCTER) which is greatly appreciated.