A 2,300-year-long annually resolved record of the South American summer monsoon from the Peruvian Andes

Hydrology and Hydrology and .

Pumacocha is the uppermost lake in a topographically isolated catchment with limited groundwater reserves and no fluvial input. As a result, the lake’s hydrologic mass balance is dominated by precipitation. Data from the closest weather station in Cerro de Pasco (10.68 °S, 76.25 °W; 4330 m asl) show that regional precipitation averages 1,000 mm yr^-1 with approximately 60% occurring between December and March during the monsoon season. From these data we estimate that precipitation over the Pumacocha watershed averages approximately 1.67 × 10⁶ m³ yr^-1. This is approximately 23% greater than the total lake volume (1.36 × 10⁶ m³), thereby maintaining hydrologically open conditions and a surface water residence time of less than a year. As a result, the influence of evaporation on the lake’s hydrologic and isotopic mass balances is minimal, a point illustrated by δ¹⁸O and δD measurements on 27 monthly lake water samples collected at Pumacocha between August 2006 and July 2008 (Fig. 3 and Fig. S3). These samples constrain a local meteoric water line (LMWL) that is offset from, but parallel to, the global meteoric water line (GMWL; Table S3). This indicates that the isotopic composition of the lake water reflects seasonal variation in free from the effects of evaporation. If Pumacocha surface waters were influenced by evaporation, they would plot along the local evaporation line (LEL), which is constrained by water samples from closed basin lakes in the region that are sensitive to evaporation (Fig. 3). This LEL is oblique to the GMWL and LMWL and intersects the LMWL at a δ¹⁸O value of -13.6‰. Additional support for the limited role of evaporation is provided by the close agreement between annual mean (-13.0‰) and modeled annual mean (-12.9‰ ± 1.5) at Pumacocha (Table S4; refs. 5, 6). These data support the contention that Pumacocha reflects and that other influences on are negligible. The monthly water samples also record a strong seasonal cycle, indicating that responds rapidly to changes in and is sensitive to variations in (Fig. S3). No relationship was observed between and air, or lake-surface temperature, which is consistent with work that shows the amount of precipitation and rainout along the trajectory of an air mass are the primary factors that determine (7).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

(A) Surface water samples from Pumacocha (black circles) and regional hydrologically closed basin lakes (gray triangles). Pumacocha surface water samples plot along a LMWL (black line) that parallels the GMWL (dashed line). Regional lakes with closed basin hydrologies plot along a LEL (gray line), which is oblique to both the LMWL and the GMWL and intersects the LMWL at -13.6‰ δ¹⁸O (VSMOW; gray square). (B) Same as in A but expanded to illustrate seasonal isotopic variability in the Pumacocha surface water.

Pumacocha and .

Modern limnological data from Pumacocha demonstrate a clear relationship among , , and . Authigenic calcite was collected from littoral vegetation at Pumacocha in August 2006 and May 2008, and from deepwater sediment traps deployed between August 2006 and May 2007. When converted to Vienna standard mean ocean water (VSMOW) for comparison with , the annual and three-year-average values are identical within error to measured mean annual (Table S4). This indicates that the calcite is precipitated in isotopic equilibrium with Pumacocha lake water and that measurements can be used to reconstruct past variability. Considering the evidence that tracks , the correspondence between and supports using down-core measurements as a proxy for reconstructing . Although we recognize that the links among , , and are empirical, the modern observational data support this working model as a reasonable first-order approximation.

Temperature Effect on Pumacocha .

Modern limnological and paleoclimate data suggest that temperature has a minor influence on the integrated signal at Pumacocha. Sediment trap averaged -12.8‰ and varied by 0.8‰, which is about twice that attributable to temperature-controlled fractionation during calcite formation. On longer time scales, varied by up to 2‰ between the LIA and CWP (Fig. 4A). Accounting for temperature-controlled oxygen isotope fractionation during the formation of precipitation from water vapor (+0.6‰ °C^-1; ref. 8) and calcite from lake water (-0.22‰ °C^-1; ref. 9), a 2‰ shift requires a 5.5 °C temperature change. This far exceeds estimates of South American LIA cooling inferred from glacial equilibrium line altitude changes in Bolivia (1.1 to 1.2 °C; ref. 10) and lacustrine temperature reconstructions based on changes in lake productivity in central Chile (0.7 to 0.9 °C; ref. 11). Although a 0.7 to 1.2 °C cooling could account for 0.3 to 0.5‰ (15 to 25%) of the Pumacocha shift between the LIA and the CWP, the remaining variance is unexplained by temperature. These modern limnological and paleoclimate data suggest that temperature is, and has been, a minor influence on the integrated signal at Pumacocha and that variability is attributable to changes in the isotopic composition of precipitation.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

(A) The Pumacocha record of SASM intensity. In B and C, the Pumacocha data are shown with a gray line for comparison with (B) from Cascayunga Cave, Peru, a record of lowland precipitation (22) and (C) from Quelccaya, Peru (23) smoothed with a 5-point moving average.

SASM Variability and .

The principal control on over tropical South America is Raleigh-type fractionation during rainout (7). This is reflected by observations that show is more negative during the austral summer wet season, despite warmer temperatures, and more positive during the dry austral winter, when temperatures are generally cooler (12). There is, however, a slight offset of about 1–2 mo between the most intense precipitation and the most depleted at Pumacocha. Similar observations have been made in the Andes of Bolivia and attributed to rainout processes upstream over regions with slightly different precipitation seasonality (13). Hence is not simply a measure of precipitation amount at the site, but rather of the intensity of monsoon-related convective activity along the trajectory of an air mass (12–14). Year-to-year variability documented by observations, proxy data, and stable isotope-enabled global climate models show that is lower during intense monsoon seasons and vice versa (7). This isotopic response to monsoon variability is preserved and enhanced along an air mass trajectory, even if the precipitation signal at the site where stable isotopes are measured is not exactly the same as over the core region of convective activity, in this case the Amazon Basin (7). Hence the isotopic monsoon signal tends to be much more spatially coherent than monsoon precipitation, which is more strongly affected by microclimatic effects (14). This “isotopic memory” of an air mass and its influence on downstream δ¹⁸O records like the one from Pumacocha means that local records reflect large-scale changes in the intensity of the SASM (7). We therefore interpret from Pumacocha as a proxy for the intensity of monsoon precipitation upstream.

Modern interannual variations in SASM intensity are caused by a number of factors. Pacific SST anomalies associated with El Niño–Southern Oscillation (ENSO) play a primary role in modulating SASM variability through changes in the Hadley and Walker circulation over tropical South America and the Pacific (15, 16). Changes in tropical Atlantic SSTs have also been linked to interannual variations in SASM activity and drought conditions over the Amazon Basin (15, 17). Soil moisture provides an important land surface feedback that has been shown to affect SASM precipitation patterns (18). All these factors lead to distinct variations in SASM activity, only loosely coupled to ITCZ variations over the tropical oceans on interannual time scales. Over the tropical Andes, summer precipitation is predominantly modulated by eastern equatorial Pacific SST (15). Cooling in the eastern equatorial Pacific (e.g., La Niña events) typically leads to reduced meridional baroclinicity and anomalous upper-level easterly flow over the Andes, which enhances near-surface moisture influx from the Amazon Basin, and vice versa. Therefore, although the SASM moisture supply lies almost exclusively in the tropical Atlantic, perturbations in atmospheric circulation caused by SST anomalies in the tropical Pacific play an important role mediating the intensity of the SASM in general and over the tropical Andes in particular (15).

SASM variability is less well understood on decadal and longer time scales because limited instrumental data in the tropics complicates investigations of long-term trends. Nonetheless, observations show that Atlantic SST variations play a significant role modulating SASM variability on multidecadal time scales (19). Paleoclimate studies similarly suggest that the tropical Atlantic may have been an important factor driving tropical South American rainfall variability on longer time scales and during past periods with different climatic boundary conditions (20, 21).

SASM Variability and Mean State Changes.

Using the empirical relationships among Pumacocha , , and , and the documented close relationship between SASM intensity and (7, 12), we interpret down-core variations in Pumacocha as representative of the intensity of the SASM. Accordingly, high is taken to reflect a less intense SASM, whereas lower represents a more intense SASM.

The Pumacocha record is characterized by considerable decadal-scale variability with three pronounced century-scale mean state changes during the MCA, LIA, and CWP, indicating major changes in the strength of the SASM (Fig. 4A). During the MCA, increased between A.D. 900 and 1100 with two distinct decadal-scale peaks from A.D. 935 to 950 and A.D. 1000 to 1040, indicating considerable reductions in SASM intensity. Medieval aridity was followed by a prolonged decrease in during the LIA starting as early as A.D. 1300, with minimum values between A.D. 1400 and 1820. This trend signifies the most dramatic increase in the strength of the SASM identified in the 2,300-year-long Pumacocha record. After A.D. 1820, generally increased to the present, especially after A.D. 1900. Between A.D. 1908 and 2007, increased from -15.0‰ to -12.9‰ at an average rate of 0.02‰ per year, which is the steepest and most rapid rise in the 2,300-year-long Pumacocha record. This increase reflects a long-term reduction in the intensity of the SASM that is similar in magnitude to the MCA aridity, but with a more abrupt onset.

Comparison with other highly resolved paleoclimate records from tropical South America shows that SASM mean state changes captured by the Pumacocha record were regional in scale with similar timing, direction, and magnitude. Specifically, the Pumacocha record tracks the 900-year-long Cascayunga Cave record (6.09 °S, 77.23 °W, 930 m asl), which is interpreted as a record of South American rainfall (Fig. 4B; ref. 22). The correspondence between these δ¹⁸O records during their common period from A.D. 1050 to 2006 is notable because Cascayunga Cave is 3,400 m lower than Pumacocha and over 600 km to the north. Interestingly, the timing of mean state changes in the two records is nearly identical, although both the degree and magnitude of δ¹⁸O variability are more pronounced in the Pumacocha record. Their similarities, however, support our contention that the Pumacocha record reflects regional scale precipitation variability and provides supporting evidence that strengthening of the SASM during the LIA was not strictly an Andean response.

The Pumacocha record also shares many features with the annually resolved Quelccaya Ice Cap δ¹⁸O () record, which is approximately 500 km southwest of Pumacocha (13.93 °S, 70.83 °W; 5670 m asl; Fig. 4C; ref. 23). The timing and direction of mean state changes in these records are similar during the LIA and CWP, although the Quelccaya ice core shows greater variability and slightly later minimum values during the LIA. The MCA is less pronounced in the Quelccaya record than in the Pumacocha record, although increased between A.D. 980 and 1100 with two peaks centered on A.D. 990 and 1080. It is not clear why the MCA is less evident at Quelccaya because, although considerably lower resolution (100 years per sample), tropical Andean ice cores from Nevado Huascaran (24) and Sajama (25) have broadly elevated from A.D. 900 to 1300 and A.D. 900 to 1500, respectively. Notably, the Quelccaya record has historically been interpreted as a paleo-temperature record, although a number of authors have noted that precipitation amount likely plays a major role driving variability (e.g., ref. 26). Because the Pumacocha and Cascayunga Cave records reflect SASM precipitation variations, their similarities with Quelccaya are consistent with the notion that Rayleigh distillation during rainout in the SASM region strongly influences isotopic records of tropical precipitation.

The close agreement in the timing, direction, and magnitude of mean state changes in δ¹⁸O during the MCA, LIA, and CWP from lake sediment, speleothem, and ice core records supports the idea that a common large-scale mechanism influenced δ¹⁸O reaching these central Andean sites spanning 11° latitude and 4,740 m of elevation. Based on the work presented here, the most likely cause of these documented shifts in is a change in SASM intensity as all three sites receive the majority of their annual precipitation during the monsoon season.

We compared the Pumacocha record to volcanic and solar radiative forcing (Fig. 5; e.g., 27, 28), reconstructed NH temperatures (29), variations in the latitude of the ITCZ (30, 31), and North Atlantic SST’s (32, 33) to help identify the relevant factors that contributed to decadal variability and mean state changes in the SASM during the MCA, LIA, and CWP (Fig. 5). North Atlantic SST’s are represented by the Mann et al. (32) reconstruction of SST from 60 °N to the equator and 75.5° to 7.5 °W and referred to generally as the North Atlantic. These variables were selected because numerous lines of evidence suggest that ocean–atmosphere processes, which are important drivers of modern SASM variability, contributed to past SASM changes (e.g., ref. 34). Volcanic and solar forcing are included because changes in radiative forcing from these sources have been identified as leading influences on NH temperature variability during the MCA and LIA (27, 35), which model simulations suggest impact the tropical ocean–atmosphere systems, particularly in the North Atlantic (34, 36).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

(A) Reconstructed radiative forcing (blue) for the last 1,000 years from ref. 27 and residual ¹⁴C for the last 2,300 years (Δ¹⁴C; a proxy for solar variability; red; ref. 28), (B) reconstructed Northern Hemisphere temperatures (29), (C) % Ti from the Cariaco Basin, Venezuela (30), interpreted as a proxy for the location of the ITCZ, and (D) reconstructed SST anomalies in the North Atlantic (32). The superimposed gray line in each of the graphs shows the Pumacocha record for comparison.

Immediately apparent from these comparisons is a remarkable correspondence between the Pumacocha record of SASM rainfall and reconstructed NH temperatures during the last 2,000 years (Fig. 5B). This relationship is particularly evident during the MCA, LIA, and CWP mean state changes. For example, the two greatest reductions in SASM intensity in the Pumacocha record were coincident with NH temperature maxima during the MCA and CWP. Conversely, the SASM was stronger than at any other point in the last 2,300 years when NH temperatures were at a 2,000-year low during the LIA. A cooling of the Northern Hemisphere leads to a southward displacement of the ITCZ, as documented in modeling and paleoclimate studies because of adjustments in the Hadley circulation balancing the need for enhanced northward heat transport (29, 30, 33). Although the Atlantic ITCZ serves as the main moisture conduit for SASM precipitation over the continent (37), the SASM and the marine ITCZ are separate systems, only loosely connected on interannual time scales (38). The remarkable similarities between the Pumacocha record and the reconstructions of both tropical SST’s and the latitude of the ITCZ (29), however, suggest that the coupled tropical ocean–atmosphere system plays an important role in modulating SASM rainfall on decadal and longer time scales. This relationship is particularly apparent during the MCA, LIA, and CWP.

Weakening of the SASM observed in the Pumacocha record during the MCA is contemporaneous with generally elevated radiative forcing and warm NH temperatures (Fig. 5 A and B; ref. 29). At the same time, SST reconstructions from the North Atlantic show warmer, although variable, conditions (32, 33), whereas those from the equatorial Pacific indicate La Niña-like cooling (32). Although there is some debate regarding the tropical Pacific’s response to radiative forcing, the available data support the occurrence of prolonged La Niña-like conditions during the MCA. Warm North Atlantic and cool eastern equatorial Pacific SSTs, as suggested by the proxy data, would shift the latitude of the ITCZ northward over both ocean basins (39, 40). In line with this model, elevated titanium concentrations (% Ti) in marine sediments from the Cariaco Basin off the coast of Venezuela occurred when North Atlantic SSTs peaked during the MCA, suggesting that the Atlantic branch of the ITCZ was persistently northward when the North Atlantic was warm (Fig. 5; ref. 30). Consistent with an intensification of the SASM when the ITCZ is shifted southward, the Pumacocha record shows that the SASM was considerably reduced during the MCA when peak % Ti in the Cariaco Basin record indicates that the ITCZ was persistently northward. The similarities between these records suggests that the northward shift in the ITCZ in response to tropical North Atlantic warming and eastern equatorial Pacific cooling during the MCA contributed to a reduction in the intensity of the SASM.

During the LIA, the Pumacocha record shows a strengthening of the SASM from A.D. 1400 to 1820 that was stronger than during any other period in the last 2,300 years. This strengthening occurred at the same time that diminished solar irradiance and elevated volcanism acted in concert to reduce radiative forcing and cool the NH (29, 35). Also during this interval, SST reconstructions show that the North Atlantic cooled (31) while the equatorial Pacific exhibited El Niño-like conditions (32). Proxy records of ITCZ variability from lakes in the western Pacific (31) and the Cariaco Basin (30) in the Atlantic indicate that the ITCZ was displaced considerably southward over both basins during the LIA (Fig. 5C). In the Atlantic, the southward ITCZ displacement is consistent with climate model simulations that show that NH cooling during the LIA would lower North Atlantic SSTs and strengthen the northern branch of Hadley Cell circulation, shifting the Atlantic ITCZ southward (34). The southerly position of the ITCZ over the Pacific and Atlantic during the LIA is coincident with the prolonged maximum in SASM rainfall documented by the Pumacocha record. This suggests that NH cooling and adjustments in the position of the ITCZ were more important in determining the mean state of the SASM during the LIA than enhanced El Niño-like conditions in the Pacific, which on interannual time scales tend to reduce rainfall over much of the SASM domain.

Increasing in the Pumacocha record after A.D. 1820 indicates a long-term weakening of the SASM, particularly after A.D. 1900. This reduction in SASM rainfall follows increases in radiative forcing (27), instrumental and reconstructed NH temperatures (29), and SSTs in the central equatorial Pacific and North Atlantic (32). These conditions are similar to those associated with weakening of the SASM during the MCA with the exception of apparent El Niño-like warming in the equatorial Pacific (32). The relationship between the Pumacocha record and reconstructed North Atlantic SSTs, evident even in the modern record, suggests that the weakened mean state of the SASM since the end of the LIA may be related to a northward shift in the Atlantic ITCZ in response to North Atlantic and NH warming.