Hydroclimate and ENSO Variability Recorded by Oxygen Isotopes From Tree Rings in the South American Altiplano

Hydroclimate variability in tropical South America is strongly regulated by the South American Summer Monsoon (SASM). However, past precipitation changes are poorly constrained due to limited observations and high‐resolution paleoproxies. We found that summer precipitation and the El Niño‐Southern Oscillation (ENSO) variability are well registered in tree‐ring stable oxygen isotopes (δ18OTR) of Polylepis tarapacana in the Chilean and Bolivian Altiplano in the Central Andes (18–22°S, ∼4,500 m a.s.l.) with the northern forests having the strongest climate signal. More enriched δ18OTR values were found at the southern sites likely due to the increasing aridity toward the southwest of the Altiplano. The climate signal of P. tarapacana δ18OTR is the combined result of moisture transported from the Amazon Basin, modulated by the SASM, ENSO, and local evaporation, and emerges as a novel tree‐ring climate proxy for the southern tropical Andes.

The South American Summer Monsoon (SASM) system dominates most of the South American seasonal hydroclimate (Marengo et al., 2012;Vuille et al., 2012). The buildup of the Bolivian High, an anticyclonic system situated over Bolivia, during the austral summer is a key atmospheric feature associated with the SASM. The Bolivian High blocks the atmospheric circulation from the west, allowing the mid-to upper-level easterly winds to entrain near-surface upslope flow and moisture transport from the Amazon basin toward the Altiplano (Garreaud et al., 2003). This moisture originates in the Atlantic Ocean and is transported toward the Amazon Basin by easterly trade winds during the austral summer (Lenters & Cook, 1997). Moisture is recycled over the Amazon basin before being transported toward the Altiplano (Garreaud, 1999;Segura et al., 2020). Thus, most precipitation (i.e., 75%-90%) in the Altiplano falls during the austral summer (December to March; Garreaud et al., 2003;Vuille & Keimig, 2004), coincident with the mature phase of SASM from December to February (Raia & Cavalcanti, 2008), although occasional winter snowfall sourced from the Pacific is possible (Vuille & Ammann, 1997). At interannual scales, this moisture transport to the Altiplano from the Amazon Basin is greatly influenced by sea surface temperatures (SST) in the tropical Pacific (Sulca et al., 2018) with below (above) normal precipitation during El Niño (La Niña) years (Vuille, 1999).
There are few instrumental records of climate in tropical South America prior to 1950, especially in the high Andes including the Altiplano. This is further compounded by the high interannual and spatial precipitation variability in the region due to its complex topography and the multitude of climate drivers (Vuille & Keimig, 2004). Long-term high-resolution climate proxies are therefore of great importance to extend climate information into the past and to evaluate the present climate dynamics in multicentennial time scales. Tree rings are high-resolution (annual to subannual) climate proxies with proven value for climate reconstructions. For more than a decade, a great effort has been made to expand the South American tree-ring network to the tropical Andes with special emphasis on the semiarid western Altiplano where Polylepis tarapacana grows (16°-23°S; Argollo et al., 2004;Christie et al., 2009;Morales et al., 2004Morales et al., , 2012Morales et al., , 2020Moya & Lara, 2011;Soliz et al., 2009). This species forms the world's highest elevation forests (4,000-5,100 m a.s.l.; Braun, 1997). Given the quality of its ring-width climate signal and its longevity of up to 750 years, it represents the most valuable paleoclimate archive to develop tree ring-based reconstructions in the tropical Andes. However, very little research has been done to evaluate the climate signal in the stable oxygen isotopes contained in tree-ring cellulose of P. tarapacana (Ballantyne et al., 2011;Rodriguez-Caton et al., 2021).
Previous studies have shown that the degree of moisture recycling and rainout upstream over the Amazon Basin largely controls δ 18 O in precipitation over the tropical and Central Andes and the Bolivian lowlands (Baker et al., 2015;Hoffmann et al., 2003;Kanner et al., 2013;Vimeux et al., 2005Vimeux et al., , 2011Vuille & Werner, 2005;Vuille et al., 2012). These studies were based on models and natural archives mostly located north of 19°S; thus, less is known about the driest and most distal part of the SASM domain located at the southwestern margin of the Altiplano adjacent to the Atacama Desert (Galewsky & Samuels-Crow, 2015).
The development of a network of P. tarapacana oxygen isotope chronologies emerges as a valuable tool to complement the present collection of stable oxygen isotope records from the tropical Andes (Vuille et al., 2012), mainly from ice cores, lake sediments, and speleothems, to infer the local-and large-scale climate variability and changes in the region during the last millennium. Stable oxygen isotopes in tree rings (δ 18 O TR ) reflect the δ 18 O signal of the water absorbed by the roots but also other processes, such as soil evaporation and leaf transpiration. While water absorption by roots is a non-fractionating process (e.g., Bariac et al., 1990;Dawson & Ehleringer, 1993), soil evaporation and leaf transpiration (the process of plant water evaporation through leaf stomata) can result in enriched δ 18 O TR signatures due to preferential evaporation of the lighter isotope ( 16 O). Finally, during cellulose synthesis, oxygen atoms can be exchanged between phloem carbohydrates and xylem water, contributing to a higher imprint of source water (Cheesman & Cernusak, 2017;Gessler et al., 2009;Sternberg et al., 1986).
Here, we used four δ 18 O TR site chronologies of P. tarapacana spanning from 1950 to 2015 along a latitudinal-aridity gradient from 18°S to 22°S in the South American Altiplano to investigate the potential of this network to develop a precipitation reconstruction for the southern tropical Andes. Since, δ 18 O TR carries a strong signal of the water used by the trees, we hypothesize that our δ 18 O TR records reflect hydroclimate and thus register the SASM and ENSO variability in this region.

Stable Oxygen Isotope Tree-Ring Records
Our annually resolved P. tarapacana δ 18 O TR site chronologies, Guallatire (GUA), Frente Sabaya (FSA), Irruputuncu (IRR), and Uturunco (UTU), span from 1950 to 2007-2015 and are located at ∼4,500 m a.s.l. along a latitudinal-aridity gradient ( Figure 1a, Table S1 in Supporting Information S1). Site characteristics and methods to develop the δ 18 O TR site chronologies can be found in Supporting Information S1 (Notes S1 and Table S1 in Supporting Information S1) and Rodriguez-Caton et al. (2021). High correlations between the four sites ( Figure S1 in Supporting Information S1) supported merging the site δ 18 O TR chronologies and obtaining a composite P. tarapacana δ 18 O TR chronology by computing a robust mean of the four site chronologies. This  (Table S2 in Supporting Information S1) is shown in all panels: mean JFM δ 18 O (‰) in precipitation for GNIP stations (circles) for the periods depicted in Table S3 in Supporting Information S1, and annual δ 18 O values for tree rings (triangles) and ice cores (squares) for the period 1950-1984. The symbols are colored according to the δ 18 O ratio (‰), following the color bar in (b). composite chronology is almost identical to the first principal component of the four chronologies ( Figure S2b in Supporting Information S1), which explains around 79% of the total variance.
We analyzed other δ 18 O paleoclimate proxy records based on the Altiplano and surrounding areas, as well as observed and reanalysis-based precipitation amounts and precipitation δ 18 O ( Figure 1, Table S2 in Supporting Information S1). The records include ice cores (Hoffmann et al., 2003;Thompson et al., 1984Thompson et al., , 1998Vimeux et al., 2009), δ 18 O TR from an individual P. tarapacana tree in Argentina (Ballantyne et al., 2011), and Cedrela odorata δ 18 O TR from tropical lowlands (Baker et al., 2015;Brienen et al., 2012). We calculated average δ 18 O for 1954-1984, the common period between these proxies. Observed δ 18 O in precipitation was obtained from the International Atomic Energy Agency-Global Network of Isotopes in Precipitation (IAEA-GNIP) from five stations closest to the study area ( Figure 1, Table S3 in Supporting Information S1). We also used 2.5° grid resolution IsoGSM precipitation and precipitation δ 18 O for the period 1979-2018 (Yoshimura et al., 2008). IsoGSM is an isotope-enabled reanalysis product that realistically portrays the δ 18 O seasonal cycle and interannual variability in this region (Hurley et al., 2019).

Climate Data
We used Climatic Research Unit 0.5° grid resolution precipitation data set version 4.04 (CRU TS v.4.04; Harris et al., 2020) and CHIRPS precipitation data (Funk et al., 2015) based on remote sensing and meteorological stations for the period 1981-2020 with a spatial resolution of 0.05° × 0.05°. Monthly regional precipitation anomalies for 1950-2008 were also developed by averaging precipitation data from 32 meteorological stations located in the study region (Table S4 in Supporting Information S1; Rodriguez-Caton et al., 2021). We used 1° grid resolution ERA5 reanalysis temperature data (Dee et al., 2011) with a time span of 1979-2018 and global SST from NOAA NCDC Extended Reconstructed SST (ERSST version 3b) with a grid resolution of 2.0° × 2.0° from 1854 to present (Smith et al., 2008).

Relationships Between Climate and δ 18 O Tree-Ring Records
We computed Pearson correlations between P. tarapacana δ 18 O TR chronologies and monthly precipitation anomalies for the common period 1950-2007 from August of the previous tree growing season to May of the current growing season using the R package treeclim (R Core Team, 2020; Zang & Biondi, 2015). The significance of the correlation coefficients was assessed by stationary bootstrapped confidence intervals (Biondi & Waikul, 2004).
Based on the results from monthly correlations, we used January, February, and March (JFM) precipitation for both CHIRPS (PP CHIRPS ) and CRU (PP CRU ) to develop spatial correlations with individual-site and composite P. tarapacana δ 18 O TR chronologies using Spearman rank. Field correlations were also computed between JFM precipitation (PP IsoGSM ) and precipitation δ 18 O (PP δ 18 O IsoGSM ) from the isotope-enabled IsoGSM reanalysis data set, averaged over the location of our four study sites (three grid cells, since two locations shared a grid point).
To assess the influence of ENSO on δ 18 O TR , we calculated spatial correlations between the δ 18 O TR chronologies and JFM global SST. Singular Spectral Analysis (Vautard, 1995;Vautard & Ghil, 1989) was performed to determine the dominant, nonstationary modes of variability of the frequency domains of the composite δ 18 O TR chronology and the JFM SSTs for the Niño 3.4 region (5°N-5°S, 170°W-120°W). We grouped the dominant modes of variability when they oscillated in similar frequencies (Notes S2 in Supporting Information S1). We also calculated probability density functions for δ 18 O TR and precipitation at each study site during El Niño/La Niña years and tested for significant differences according to the Mann-Whitney test ( Figure S3 and Table S5 in Supporting Information S1).

Precipitation δ 18 O and Proxy Records in the Altiplano and Surrounding Areas
GNIP observations and ice-core proxy data show more negative δ 18 O values as elevation increases (Figure 1a). This altitude effect likely reflects the first-order isotopic Rayleigh distillation with decreasing temperature (Figure 1c) and orographic precipitation as air masses approach the Andes and are lifted along the eastern slopes. This depletion in heavy isotopes as air masses move from the lowlands in the Amazon basin toward the higher altitude environments in the Altiplano is consistent with the distribution of precipitation δ 18 O seen in IsoGSM ( Figure 1b). However, Polylepis δ 18 O TR from the Altiplano has more positive δ 18 O signatures than δ 18 O TR from Cedrela trees growing in lowland areas. This may be the result of increasing aridity and hence enhanced evaporation toward the southwest of the Andes (Figure 1d). Within our P. tarapacana network, an enrichment of δ 18 O TR is also evident toward the southern and more arid sites (Table S1 in Supporting Information S1).

Precipitation and ENSO Signal in Polylepis tarapacana δ 18 O TR
At all sites, the most significant correlations between δ 18 O TR and monthly precipitation occur during the concurrent growing season (Figure 2a). The four P. tarapacana δ 18 O TR chronologies show a significant negative correlation with precipitation during the peak of the wet season (JFM), which coincides with the mature phase of SASM (Figures 1d and 2a). The highest correlations between δ 18 O TR and monthly precipitation occur in January, the rainiest month, with correlations about −0.6 (alpha <0.05) at the northernmost site GUA. Years with extremely low (high) precipitation are characterized by significantly more positive (negative) δ 18 O TR values (Figure 2b). The spatial extent of the CRU precipitation signal embedded in the δ 18 O TR records covers much of the Altiplano, reaching correlation coefficients close to −0.8 (alpha <0.05) at the northern sites of the study region (Figure 2c). The significant negative correlation coeffcients between precipitation and the four δ 18 O TR site chronologies extend into the adjacent lowlands eastward from the Altiplano, indicating linkages to upstream rainout.
When moving from north to south along the aridity gradient in the Altiplano, the precipitation signal in δ 18 O TR weakens at both local (Figure 2a) and regional (Figure 2c) scales, but at the same time, the signal starts to propagate to the southeast toward the La Plata Basin (Figure 2c site UTU). However, overall, there is high consistency in the pattern of spatial correlation among the four study sites. This is further supported by 46%, 43%, and 42% of explained variance when using the composite δ 18 O TR chronology as a predictor for instrumental, CRU, and CHIRPS precipitation, respectively ( Figure S4 in Supporting Information S1).
Across the southern tropical Andes, the composite P. tarapacana δ 18 O TR chronology shows a clear and consistent spatial correlation pattern with CRU and CHIRPS precipitation data (Figures 3a and 3b). These results are consistent with correlations between IsoGSM reanalysis precipitation (JFM PP IsoGSM ) and reanalysis precipitation δ 18 O averaged across the four study sites (PP δ 18 O IsoGSM ; Figure 3c). However, while the sign of the correlation is the same, the location of the most significant correlation between PP IsoGSM and PP δ 18 O IsoGSM is displaced upstream over the southwestern Amazon basin (Figure 3c). Our results also show that precipitation amount is lower during El Niño than La Niña events along the aridity gradient with significant differences at the wetter sites GUA and FSA ( Figure S3b and S5 in Supporting Information S1). Accordingly, the northernmost and wettest site GUA shows significantly more enriched δ 18 O TR during El Niño versus La Niña major events ( Figure S3c in Supporting Information S1), as well as stronger correlations with central and eastern equatorial SSTs compared with the rest of the sites ( Figure S3d in Supporting Information S1).

Discussion and Conclusions
Our findings reveal that P. tarapacana δ 18 O TR can serve as a valuable proxy for recording precipitation variability over the South American Altiplano. All the study sites show negative correlations between δ 18 O TR and precipitation (i.e., years with more rainfall correspond to more negative δ 18 O TR ) during austral summer (JFM) of the tree's current growing season. The sensitivity of our four δ 18 O TR records to summer precipitation agrees with the fact that most precipitation across this region falls during austral summer as a consequence of the SASM (Garreaud et al., 2003;Vuille et al., 2012). Indeed, the highest correlation occurs in January, which corresponds to the rainiest month in the study region. Although precipitation is also important during December, correlations were not significant for this month. This might be related to the fact that δ 18 O TR was measured from cellulose, which depending on the species and area can accumulate late in the growing season (Rathgeber et al., 2016).
The IsoGSM data indicate that precipitation δ 18 O over the Altiplano is related to precipitation upstream over the Bolivian lowlands and southwestern Brazil (Figure 3c). This is consistent with previous studies, that show that precipitation δ 18 O in the tropical Andes is dominated by Rayleigh-type fractionation during rainout over the Amazon basin (i.e., more negative precipitation δ 18 O results from enhanced convective activity upstream, resulting in the depletion of the heavy isotopes in the remaining water vapor) (Risi et al., 2012;Vuille & Werner, 2005). The influence of upstream processes over the Amazon basin on the precipitation δ 18 O has been observed in tree rings (Baker et al., 2016;Brienen et al., 2012), as well as in models and observations (Vimeux et al., 2005(Vimeux et al., , 2011 in the Bolivian lowlands and other proxies in Peru (Hurley et al., 2019;Kanner et al., 2013;Vuille et al., 2012). While our δ 18 O TR records also seem to capture this signal of upstream rainout, they show the highest correlations with precipitation over the Altiplano and surrounding areas (Figures 3a and 3b). This increased signal of precipitation over the Altiplano can be the result of several processes. First, a local amount effect, where more precipitation over this region results in more depleted precipitation δ 18 O (the heavier isotopologues, H 2 18 O, tend to collect in the precipitation, leaving the water vapor and subsequent precipitation isotopically lighter). Second, local evaporation covaries strongly with precipitation over the Altiplano (Vuille & Keimig, 2004). Cloudy and rainy conditions reduce solar radiation and temperature; hence, evaporation is reduced and soil water is less enriched in heavy isotopes. This effect may contribute to the enhanced local precipitation signal seen in our δ 18 O TR records. Third, the δ 18 O TR signature not only depends on source water δ 18 O, but can also reflect evaporative conditions at leaf level during plant transpiration, where drier conditions can result in enriched plant water δ 18 O due to the preferential evaporation of the lighter isotope ( 16 O) (Barbour et al., 2004;Roden & Ehleringer, 1999). Under increased aridity, however, Pex (the fraction of carbonyl oxygen atoms that exchange at the cambium during biosynthesis) increases, inducing decreased δ 18 O TR (Cheesman & Cernusak, 2017;Szejner et al., 2020), which may partly counterbalance the effect of leaf transpiration. Overall, we interpret the signal in our δ 18 O TR records to be associated with (a) upstream rainout over the Amazon basin, (b) the local precipitation amount effect, (c) an effect of local evaporation on soil water in the Altiplano, and (d) potential physiological processes at the tree level (Rodriguez-Caton et al., 2021). This would result in δ 18 O TR being a valuable proxy for hydroclimate over the Altiplano, sensitive to local precipitation, local evaporation, and upstream rainout.
The width of the rings of P. tarapacana is also a valuable proxy for precipitation and moisture changes in this area. However, tree-ring width records show opposite relationships with climate between consecutive growing seasons (e.g., Soliz et al., 2009) with the signal of the previous year being most useful for reconstructing precipitation (Morales et al., 2012). The climate signal of P. tarapacana δ 18 O TR is consistent across seasons and strongest for the current growing season (for a detailed analysis, see Rodriguez-Caton et al., 2021). Although measuring δ 18 O TR can be expensive and time consuming, δ 18 O TR provides complementary information to tree-ring width and has lower tree-to-tree variability (i.e., five trees are sufficient to capture the common variability in our P. tarapacana stands).
The depletion of heavy isotopes in precipitation δ 18 O and ice-core δ 18 O with elevation ( Figure 1) is consistent with prior observations on the isotopic composition of precipitation in the Bolivian and Chilean Altiplano (Aravena et al., 1999;Fiorella et al., 2015;Hardy et al., 2003;Valdivielso et al., 2020). In contrast, δ 18 O TR seems to reflect the northeast-southwest aridity gradient across the region with more negative values for Cedrela odorata east of the Andes in the wetter tropical lowlands of Peru and Bolivia (Baker et al., 2015;Brienen et al., 2012) and more positive values for semiarid high-elevation P. tarapacana (Ballentyne et al., 2011;Rodriguez-Caton et al., 2021). Similarly, the more enriched δ 18 O TR signatures toward the southern and more arid sites of our network (Table S1 in Supporting Information S1) are likely due to reduced precipitation and therefore reduced cloudiness and higher solar radiation, driving higher soil-water evaporation and leaf transpiration.
Toward the southern portion of our study area, the δ 18 O TR variability seems to be less influenced by SST changes in the tropical Pacific, and the JFM precipitation signal is slightly weakened. The weaker strength of El Niño signal in the δ 18 O TR chronologies from the southern drier sites in comparison with the northernmost wetter sites reflects the weaker influence of El Niño in precipitation over the southwestern Altiplano ( Figure S5 in Supporting Information S1) and agrees with the pattern of ENSO sensitivity reported for a network of P. tarapanca chronologies across the Altiplano (Crispín-DelaCruz et al., 2022). In addition, the significant negative correlation between δ 18 O TR and precipitation to the southeast of the Andes (i.e., north-central Argentina), most evident in the southern sites, suggests a possible influence of interannual variations in the intensity of the South American Low-Level Jet and extratropical cold air incursions (Garreaud, 2000;Vuille & Keimig, 2004). The southern sites (20-22°S) are located at the southern latitudinal limit of the SASM influence in the transition zone between the tropics and the subtropics, where complex interactions between atmospheric processes in low latitudes and midlatitudes may affect the δ 18 O TR precipitation signal. How these multiple factors may influence the δ 18 O TR climate signal of the southern sites deserves an in-depth analysis and will be assessed in future work. Despite the differences along the gradient, the composite δ 18 O TR chronology, explaining almost 80% of the δ 18 O TR total variability, is strongly related to both austral summer precipitation and tropical Pacific SSTs.
Stable oxygen isotopes in the tree rings of the long-lived species P. tarapacana hold immense promise for reconstructing tropical South American hydroclimate of the past millennium. The strong and highly consistent precipitation signal embedded along the new network of δ 18 O TR chronologies in the Altiplano represents a new tool for evaluating year-to-year and multi-centennial precipitation variability in this region. Our findings are relevant for generating more reliable regional and local hydroclimate reconstructions in the Central Andes, as well as potentially contributing to reconstruct past ENSO variability. This will be useful for improving global circulation models and providing information for a better management of water resources in this region.