Cultural implications of late Holocene climate change in the Cuenca ,

Tripti Bhattacharyaa,1, Roger Byrnea, Harald Böhnelb, Kurt Wogaub, Ulrike Kienelc, B. Lynn Ingrama,d, and Susan Zimmermane

Departments of aGeography and dEarth and Planetary Science, University of California, Berkeley, CA 94720; bCentro de Geociencias, Universidad Nacional Autónoma de México, Querétaro 76230, Mexico; cClimate Dynamics and Landscape Evolution, Helmholtz Centre Potsdam, German Research Centre for Geosciences, 14473 Potsdam, Germany; and eCenter for Accelerator Mass Spectrometry, Lawrence Livermore National Laboratory, Livermore, CA 94550

Edited by Karl W. Butzer, The University of Texas at Austin, Austin, Texas, and approved January 1, 2015 (received for review March 26, 2014) There is currently no consensus on the importance of climate agriculture (9). Moreover, this area was controlled by powerful change in Mesoamerican prehistory. Some invoke drought as a pre-Columbian city-states in the Classic and Post-Classic periods, causal factor in major cultural transitions, including the abandon- and served as an important site of contact between highland ment of many sites at 900 CE, while others conclude that cultural Mexico and the cultures of the Atlantic Gulf Coast (10). factors were more important. This lack of agreement reflects The site of Cantona, in the northern Cuenca Oriental, is the fact that the history of climate change in many regions of particularly intriguing because of its size and mysterious aban- Mesoamerica is poorly understood. We present paleolimnological donment. The city drew its economic importance from the ex- evidence suggesting that climate change was important in the ploitation of obsidian at the site of Oyameles-Zaragoza, and was abandonment of Cantona between 900 CE and 1050 CE. At its a key supplier of obsidian to settlements along the Gulf of peak, Cantona was one of the largest cities in pre-Columbian Mexico (11, 12). The site covered over 12 km2 and housed an Mesoamerica, with a population of 90,000 inhabitants. The site is estimated population of 90,000 at its peak, but was abandoned located in the Cuenca Oriental, a semiarid basin east of . between 900 CE and 1050 CE (Fig. 1). Several authors have We developed a subcentennial reconstruction of regional climate speculated that drought was responsible for this abandonment from a nearby maar lake, Aljojuca. The modern climatology of the (11), but before the present study, no detailed paleoclimate region suggests that sediments record changes in summer mon- records have been available for the Cuenca Oriental. The present EARTH, ATMOSPHERIC,

soonal precipitation. Elemental geochemistry (X-ray fluorescence) climate is strongly seasonal, with rainfall primarily occurring from AND PLANETARY SCIENCES and δ18O from authigenic calcite indicate a centennial-scale arid May through October (Fig. S1). This summertime increase in interval between 500 CE and 1150 CE, overlaid on a long-term convective activity is driven by the seasonal heating of the land drying trend. Comparison of this record to Cantona’schronology surface, and is linked to the broader circulation regime known as suggests that both the city’s peak population and its abandonment the North American, or Mexican, Monsoon (13). High-resolution occurred during this arid period. The human response to climate paleoclimatic records can help us understand the long-term vari- change most likely resulted from the interplay of environmental ability of this monsoonal system. and political factors. During earlier periods of Cantona’s history, Here we report on a subcentennially resolved paleoclimate ANTHROPOLOGY increasing aridity and political unrest may have actually increased record from a maar lake, Aljojuca, in the eastern Cuenca Ori- the city’s importance. However, by 1050 CE, this extended arid ental, 30 km south of Cantona (Fig. 1). We discuss the impli- period, possibly combined with regional political change, contrib- cations of the record for our understanding of late Holocene uted to the city’s abandonment. Significance Mesoamerica | paleoclimate | late Holocene | Cantona | paleolimnology Researchers have long invoked drought to explain the demise key uncertainty in the history of pre-Columbian Meso- of many pre-Columbian Mesoamerican sites. However, the cli- Aamerica is the precise relationship between climatic change matic history of many regions of Mesoamerica remains poorly and major cultural transitions. While some scholars argue for understood. This includes the region around Cantona, a large a direct causal link between aridity and the abandonment of fortified city in highland Mexico that was abandoned between many sites (1–4), others argue that anthropogenic landscape 900 CE and 1050 CE. We used stable isotopes and elemental transformations were more important in facilitating cultural concentrations from lake sediments to reconstruct past climate, change (5, 6). A focal point of this debate is the dramatic cultural and found evidence of regional aridity between 500 CE and change that occurred during the Terminal Classic, from ∼850 1150 CE. In the initial phase of drought, Cantona’s population CE through 1000 CE. Much of this debate has focused on the grew, possibly as a result of regional political instability. How- Maya lowlands (3, 7), although researchers have also explored ever, by 1050 CE, long-term environmental stress likely con- linkages between climate and cultural change in Terminal Classic tributed to the city’s abandonment. Our work highlights the highland Mexico (1, 2, 4). However, these comparisons have been interplay of environmental and political factors in past human hindered by a lack of high-resolution paleoclimate records: While responses to climate change. paleoclimate records from across Mesoamerica suggest arid con- ditions at some point during this interval (1–3, 6–8), patterns of Author contributions: T.B., R.B., H.B., and U.K. designed research; T.B., R.B., H.B., K.W., U.K., B.L.I., and S.Z. performed research; B.L.I. contributed new reagents/analytic tools; climatic change may have been regionally heterogeneous. T.B., R.B., H.B., K.W., and S.Z. analyzed data; and T.B. and R.B. wrote the paper. The Cuenca Oriental, or Oriental Basin, offers an important set- The authors declare no conflict of interest. ting in which to assess hypotheses about past human−environment This article is a PNAS Direct Submission. relationships. Located east of the Basin of Mexico, in the eastern Data deposition: Radiocarbon dates and raw proxy data have been deposited in the Trans-Mexican Volcanic Belt (TMVB), the region is marginal National Oceanic and Atmospheric Administration/National Climatic Data Center Paleo- for maize agriculture today. Paleoclimate records from this area climatology database (www.ncdc.noaa.gov/paleo/study/17735). therefore have broad implications for our understanding of the 1To whom correspondence should be addressed. Email: [email protected]. “ importance of past climate change in altering the northern This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. frontier of Mesoamerica,” defined as the northern limit of maize 1073/pnas.1405653112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1405653112 PNAS Early Edition | 1of6 Downloaded by guest on September 27, 2021 North Most core sections featured laminated sediments (Fig. S2). Our N Central Bayesian radiocarbon-based age model indicates that the cores 24˚N cover the past 6,200 y and feature a high sedimentation rate, with South a mean sedimentation rate of 3.5 mm/y, although the sedimen- tation rate varies over time (Fig. S3 and SI Materials and Meth- 0 5 km ods). In this paper, we focus on the last 3,800 y of the record, 22˚N when multiple proxies provide robust evidence of a dry phase between ∼500 CE and 1150 CE (Fig. 2). While the paleoclimate patterns recorded at Aljojuca may have implications for several pre-Columbian sites, we focus on Cantona, since it is the closest Tula 20˚N major pre-Columbian site to the lake. Teotihuacan Cantona In closed basin lakes, the oxygen isotopic ratio of carbonates Mexico City Aljojuca reflects changes in the ratio of evaporation to precipitation (E/P), Cholula Maar as a high ratio of E/P preferentially concentrates 18Oinlakewater 18˚N (14). In samples of authigenic carbonates from Aljojuca, we found that oxygen and carbon isotopes covary strongly, indicating closed basin conditions (Fig. S4) (14). We therefore interpret higher values of δ18O[‰ relative to Vienna Pee Dee Belemnite (VPDB)] 16˚N as indicative of an increase in the E/P ratio and a reduction in 0 200 km summer monsoon strength. We also interpret changes in sediment geochemistry as an 106˚W 104˚W 102˚W 100˚W 98˚W 96˚W indicator of changes in the strength of the summer monsoon. Increases in the concentration of aluminum, expressed as mass % Fig. 1. Locations of Cantona (red square) and other major sites in central Al2O3, reflect terrigenous material (i.e., feldspars and secondary highland Mexico (black circles), along with location of Aljojuca maar. To- clay minerals) washed into the lake during periods of high pre- pography from 1,500 m above sea level (masl) to 4,500 masl, contoured by 500 m. Gray line shows outline of the Cuenca Oriental. (Inset) Map shows cipitation (Fig. S5). The full suite of elemental geochemistry data Google Earth satellite imagery of region around Cantona, with outline of supports this interpretation, as fluctuations in concentrations of the city shown in gray. Labels designate the southern, central, and northern aluminum covary positively with fluctuations in immobile, ter- architectural clusters within the city. Modified from ref. 11. rigenous elements like titanium (Fig. S5). However, this terrige- nous flux may also increase during periods of anthropogenically induced erosion. Our moving average-filtered proxy data show climate variability in Mesoamerica. We also explore the cultural that starting at 0 BCE/CE, values of δ18O become progressively implications of these paleoclimate changes for Cantona and more enriched, with the most enriched values occurring between other important pre-Columbian sites in highland Mexico (Fig. 1). 500 CE and 1100 CE. Sedimentary aluminum concentrations begin to decline at ∼200 BCE, with minimum values between 550 Results CE and 1200 CE (Fig. 2). In 2007, we recovered cores from the maar lake Aljojuca that The oxygen isotope and aluminum proxy records reflect slightly extend from the sediment−water interface to a depth of 12 m. different trends. Most notably, aluminum concentrations show

1500 1000 500 0 500 1000 1500 CE/ BCE

Peak population of 90,000 inhabitants } A IIIII I Cultural Chronology IV at Cantona } } Population reduction Evidence of coup, transition to more B and abandonment militaristic state 14

12 (mass %)

3 10 C O 2

Al 8 −10

−5

D O (‰ VPDB) 30 18 0 δ 20 Asteraceae High−Spine 10 (% total pollen) E absence) Zea Pollen (presence/ 0 500 1000 1500 2000 2500 3000 3500 Age (cal. yr. B.P.)

Fig. 2. (A) Cultural chronology at Cantona. Numerals designate phases of occupation, while gray bars show other major events in relation to multiproxy

paleoenvironmental evidence. Heavy black lines represent a four-point moving average filter. (B) Mass % Al2O3, indicative of rain-induced slopewash; (C) δ18O, representative of E/P ratios; (D) high-spine Asteraceae pollen, which may be an indicator of anthropogenic landscape disturbance; and (E) maize pollen presence, with blue bars indicating presence of maize pollen at a particular level. See Materials and Methods for more details. Gray bar highlights multiproxy evidence of an extended dry period. Top axis notes age in calendar years AD/BC, with 2-σ range of calibrated radiocarbon dates within the sequence in red. Details on dates are available in Fig. S3.

2of6 | www.pnas.org/cgi/doi/10.1073/pnas.1405653112 Bhattacharya et al. Downloaded by guest on September 27, 2021 a return to higher values between 400 CE and 500 CE, while city, and walled streets that restricted mobility suggest a milita- δ18O values remain enriched. Differences between these proxy ristic government that both repelled outside attack and exercised archives may reflect different timescales of response to climate strong control over the city’s population (18). Other hypotheses, changes, or sensitivity to other processes. For instance, fluctua- like internal revolt or elite takeover, may also be consistent with tions in aluminum may in part be sensitive to changes in an- the evidence of destruction and cultural change in the archaeo- thropogenic land use. However, we suggest that aluminum logical record. We refer to this period as a coup to maintain fluctuations primarily reflect changes in climate rather than an- consistency with the dominant explanation in the existing archaeo- thropogenic land use, at least in the pre-Columbian period. Our logical research on Cantona (18). We also note that recent re- age model does show increases in sedimentation rate between search suggests that Teotihuacan was abandoned at ∼550 CE, 250 BCE and 50 CE and after 1570 CE, which might suggest an possibly as a result of internal revolt (19). anthropogenic influence (Fig. S3). However, there is a lower Despite intensifying aridity between 500 CE and 1150 CE, sedimentation rate in the drought interval, between 500 CE and population rose again in the period defined as Cantona III 1150 CE, consistent with reduced terrigenous flux into the lake (600−900 CE), reaching an estimated peak of 90,000 at 700 CE. system as a result of climate drying. If sediments strongly reflected Archaeologists working at the site have suggested that part of this anthropogenic erosion, we would expect increased proportions of population increase likely resulted from migration from other terrigenous indicators (i.e., Ti or Al) in this time period. Instead, areas of Mexico (18). Migrants from the north may have been we observe the opposite, with elevated proportions of calcium, driven by increasing aridity (20). This time period was likely indicative of a greater proportion of authigenic minerals, like one of turmoil elsewhere in highland Mexico, as a result of calcite (Fig. S5). Teotihuacan’s earlier decline, the decline of Cholula between Pollen records can provide important indicators of anthropo- 650 CE and 850 CE, and eruptions of the volcano genic activity, with weedy taxa (e.g., high-spine Asteraceae, (20, 21). It is possible that this turmoil may have created a flux of Amaranthaceae, or Poaceae) and crop pollen often indicating migrants to other rising centers of regional power (18, 20). Can- agriculture (15, 16). The pollen record at Aljojuca does not re- tona is similar to other regional powers that emerged in the late flect land use practices associated with Cantona, since the city is Classic, because it is strategically located, on a basaltic flow with 30 km from the lake, but it can be used to evaluate whether the a view of the surrounding landscape (2, 10, 18). This defensible area around the lake itself was strongly influenced by human position may have contributed to the site’s persistence, and might activities. Increasing percentages of high-spine Asteraceae pol- have helped the city attract migrants fleeing political turmoil. len at ∼1700 CE may result from Spanish conquest and sub- The growth of Cantona III was short-lived. Between ∼900 CE EARTH, ATMOSPHERIC,

sequent land transformation (Fig. 2). Before at least 1300 CE, we and 1050 CE, designated as Cantona IV, the city declined. This AND PLANETARY SCIENCES do not see a comparable increase in Asteraceae pollen that may interval was marked by the increased construction of defensive indicate pre-Columbian settlement around the lake. The full infrastructure and emigration. By 1050 CE, the city and its sur- suite of pollen data may represent the combined influence of roundings contained a remnant population of an estimated 5,000 climatic drying and anthropogenic land use (Table S1). occupants. This population reduction occurs within the period of Most importantly, despite any pollen evidence for anthro- peak aridity at Aljojuca (Fig. 2). Analysis of the age uncertainty pogenic activity around Aljojuca, it is important to note that associated with the δ18O proxy values suggests that despite age

changes in mass % Al2O3 are not synchronous with the appearance uncertainty, the entire interval of Cantona IV likely coincided ANTHROPOLOGY of maize pollen or weedy taxa, as we may expect if the record with dry conditions (Figs. S3 and S6). The precise social changes primarily reflected changes in local, anthropogenically induced leading to abandonment remain unknown, but it may have been erosion instead of climatic processes (Fig. 2) (17). In addition, both internal unrest generated by heightened resource scarcity, or the geochemistry and stable isotope data show evidence of a drying change in regional power dynamics associated with the rise and trend between 100 BCE and 1150 CE, culminating in peak aridity decline of other sites (18, 20, 22). After the close of this arid between 500 CE and 1150 CE. The overall similarity of the alu- interval, there was a return to much wetter conditions, as well as minum and oxygen isotope proxy records suggests a climatic driver significant social changes in the Cuenca Oriental and adjacent for the observed changes in both proxies, especially because δ18O areas, including the resurgence of Cholula (20). Cantona itself, values of calcite are less sensitive than other proxies to anthropo- however, was permanently abandoned. genic effects. Pre-Columbian Vulnerability to Climate Change in the Cuenca Oriental. Comparison of Paleoclimate and Cultural History. In this section, we A comparison of climate and cultural history suggests some cor- compare the cultural history of Cantona with the paleoclimatic relation between changes in climate and cultural events: The hy- record developed from Aljojuca. We also discuss the implica- pothesized coup that resulted in the transition to Cantona III tions of this paleoclimatic record for other important sites in occurred during a period of drying climate, and the site’sfinal highland Mexico. Recent archaeological research at Cantona has abandonment may have occurred during an extended arid period produced a radiocarbon chronology that indicates occupation between 500 CE and 1150 CE. However, during Cantona III, the from 600 BCE to ∼1050 CE (Fig. 2). Detail on Cantona’s ex- site’s population actually grew during the early part of this dry cavation and interpretation is available in SI Background. phase. Identifying correlations between cultural history and cli- The first phase of occupation, Cantona I (600 BCE to 50 CE), mate changes can be difficult: Radiometric dating uncertainty may marked the expansion of the city from two smaller preexisting hinder attempts to identify synchronicity in climatic and cultural villages. Isotope and geochemical evidence indicates a relatively histories (23, 24). Some of these concerns can be ameliorated by wet climate for the majority of this interval. The second phase of quantifying the age uncertainty associated with proxy records, which occupation, Cantona II (50–600 CE), included a period of strong wehaveattemptedtodousingBayesian age modeling techniques expansion, by the end of which the city covered 1000 ha and had (SI Materials and Methods and Figs. S3 and S6). Furthermore, more than 60,000 inhabitants. Climate exhibited a slow drying correlations do not necessarily reveal the social processes that de- trend over this interval. The archaeologists working at the site termine vulnerability to climate change (23, 24). Other lines of have suggested that the end of Cantona II was likely a period of evidence can reveal the complexity of human responses to drought: major turmoil: They have hypothesized that the destruction of For colonial Mexico, archival sources have provided important many structures in the civic-religious center suggests a military evidence on the ways existing administrative practices and social coup between 550 CE and 600 CE (11, 18). Subsequently, conflicts preconditioned responses to drought (25). Land use his- decreases in religious iconography, increased fortification of the tories inferred from detailed geomorphic evidence have shed light

Bhattacharya et al. PNAS Early Edition | 3of6 Downloaded by guest on September 27, 2021 on the ways the ancient Maya adapted to climate change (26, 27). Centennially resolved lacustrine records from central Mexico However, detailed correlative studies, like the one presented here, suggest a dry phase between ∼700 CE and 1000 CE (8). In ad- can identify the extent to which cultural change and environmental dition, researchers have reported evidence of late Classic droughts stressors track each other, and therefore can serve as the first step in the trace metal record from the Cariaco Basin (29). However, in more detailed analyses of human−environment interactions. there are substantial differences between the Cariaco Basin and Our results suggest that the abandonment of Cantona co- the Aljojuca records, especially after 1200 CE, suggesting that the incided with one of the driest intervals in the past 3,800 y. Cariaco Basin may not be a good analog for the climate of the However, the strong growth of Cantona III also occurred during Cuenca Oriental (29). this arid interval. This emphasizes the important interplay of There are also some similarities between the Aljojuca record environmental and political stresses: A drying climate no doubt and a speleothem record from the Sierra Madre del Sur, in stressed agricultural production and water resources, and may western Mexico. Both paleoclimate records show a late Holo- have been a contributing factor to the coup that resulted in the cene dry period between ∼500 CE and 950 CE (1). However, in militaristic government of Cantona III. The strong, centralized the Aljojuca record, this dry interval extends to 1150 CE, and control exercised by this new ruling structure might have regu- both records show diverging trends between 400 BCE and 100 lated access to resources to adapt to resource scarcity. It is also CE, possibly as a result of regionally different climatic patterns possible that Cantona’s strategic location may have attracted (Fig. 3). These apparent differences may also be due to differing migrants, fleeing environmental stress or political upheavals processes controlling how climate is recorded in lakes and spe- elsewhere in central Mexico, although further research is nec- leothems. For instance, lake calcite δ18O often reflects overall essary to explore this hypothesis. Some researchers have linked lake water balance (e.g., E/P changes), while speleothem calcite these late Classic political upheavals with climatic drying (1, 2). It δ18O reflects the degree of prior rainout of air masses, or the is important to note, however, that while sites like Cholula un- characteristics of local precipitation. derwent significant periods of decline between 650 CE and 850 CE, other factors apart from climate, like invasion or volcanic Climatic Interpretation. We suggest that the Aljojuca paleoclimate eruptions, may have played a significant role in these cultural record reflects multidecadal to centennial scale changes in sum- transitions. In addition, other sites like Cacaxtla reached their mertimeconvectiverainfall.Theprolongeddryphaserecorded peak in this time interval (20). The broader instability of this from 500 CE to 1150 CE corresponds reasonably well with late time period, which may have been in part due to climatic drying, Holocene dry phases reported from the Yucatan, despite different may have briefly increased the importance of Cantona and other regional climatologies. While a long-term drying trend over the sites in defensible locations (20). Holocene may be attributable to precessional forcing (30), the Future research should focus on clarifying the factors that may termination of the dry phase at 1150 CE implies some other cause. have increased or decreased vulnerability to drought at Cantona We suggest that these changes may result from teleconnections and other pre-Columbian sites. Existing archaeological research connected to changes in the Atlantic or Pacific basins. does suggest some vulnerability to long-term rainfall reductions Evidence from the modern instrumental record and modeling at Cantona. Total annual precipitation near Aljojuca is ∼800 experiments suggest that changes in regional sea surface tem- mm/y. This is considered marginal for maize agriculture (28). peratures (SSTs) can force large-scale, coherent patterns of Some researchers have hypothesized that Cantona imported a rainfall change over Mesoamerica. Analysis of instrumental data portion of its food from regions to the south (18), near Aljojuca, suggests that developing warm conditions in the eastern equa- or from the Gulf lowlands (20). Food storage facilities were torial Pacific are tied to dry summer conditions over large likely an important component of long-distance trade networks, regions of Mexico and Central America (31–34). An anoma- and an important adaptation to the unpredictability of local lously warm tropical North Atlantic is linked to increased rainfall maize yields. Archaeological surveys have located 83 silo-like over the Yucatan and southern Mexico, and decreased rainfall structures for grain storage (18). Many are clustered within the over northern Mexico (35, 36). Tropical Pacific and Atlantic SST civic-religious center, or are located at key strategic locations changes may also induce meridional shifts in the mean position throughout the city, suggesting that ruling elites controlled re- of the Intertropical Convergence Zone (37). Proxy records of El source access (18). These structures may have been additionally Niño−Southern Oscillation conditions at ∼1000 CE are equiv- important during periods of rainfall reduction, when lowered ocal: some records show enhanced El Niño-like conditions (38– agricultural production placed strain on the population. 40), while others suggest a La Niña-like mean state (41). A re- Water is also scarce in the modern landscape around Cantona. cent 2000-y SST record identifies cooler SSTs in the tropical The closest surface water, a presently saline lake, is over 5 km Atlantic and reduced Atlantic Meridional Overturning Circula- south of the city (11). Archaeological surveys reveal evidence of tion strength between 650 CE and 800 CE (42), which could be cisterns excavated into stone floors throughout the city, which linked to reduced precipitation over regions of Mesoamerica in may have stored rainwater. Given the paucity of surface water in the early part of the dry interval we identify. Further modeling the region, these structures may have been the primary mecha- studies and efforts to synthesize regional SST and terrestrial nism for storing water in the city (18). This reliance on rainwater precipitation records may clarify the spatial patterns and mech- collection suggests that the city would have been especially vul- anisms behind late Classic drought in Mesoamerica. nerable to drought. Further work to clarify land use patterns and develop a well-dated history of resource management practices Materials and Methods will play an important role in clarifying human−environment Core Recovery. In March 2007, a joint team from Universidad Nacional de linkages at Cantona. Autónoma de México (UNAM) and the German Research Centre for Geo- sciences recovered one long core and two shorter cores, with lengths of 12 m, Comparison with Other Paleoclimatic Records. The Aljojuca record 6 m, and 8 m respectively, from Aljojuca maar using an Usinger hydraulic displays several similarities to paleoclimatic records from other piston coring system. The cores are largely laminated, with occasional turbi- ∼ ditesorsectionsofhomogeneousclay.Weidentifiedfourtephralayersat regions of Mesoamerica. A long-term drying trend, from 100 depths of 353 cm, 551 cm, 885 cm, and 1,041 cm. Core sections were correlated BCE to 1100 CE, is also present in the Punta Laguna ostracod using magnetic susceptibility, high-resolution photographs, and tephra layers. δ18O record in the northern Yucatan (7) (Fig. 3). The period of maximum aridity at Aljojuca roughly coincides with drought Chronology. Core chronology was established by radiocarbon dating of 15 intervals at Punta Laguna (7), as well as a period of increased samples (SI Materials and Methods and Figs. S2 and S3). Three dates from drought frequency inferred at Lake Chichancanab (Fig. 3) (3). 383 cm represent replicate dates, one from a charcoal fragment, one from

4of6 | www.pnas.org/cgi/doi/10.1073/pnas.1405653112 Bhattacharya et al. Downloaded by guest on September 27, 2021 1500 1000 500 0 500 1000 1500 CE/ BCE A −10

−5 O (‰ VPDB) 18

δ 0 B

1 ) 3 1.2

1.4 C −1.5 1.6 Density (g/cm

−0.5 O (‰ VPDB)

18 1000 δ 0.5 D 800

600

400 Precipitation (mm/yr)

0 500 1000 1500 2000 2500 3000 3500 EARTH, ATMOSPHERIC,

Age (cal yr. B.P.) AND PLANETARY SCIENCES

Fig. 3. Comparison between Aljojuca authigenic calcite δ18O record with a four-point moving average (A) and regional paleoclimate records: (B) Lake Chichan- canab sediment density record (3); (C) Punta Laguna ostracod oxygen isotope record (5), smoothed with a five-point moving average as in the original text; and (D) the Juxtlahuaca precipitation record, reconstructed from the δ18O of speleothem calcite (1). All data downloaded from National Oceanic and Atmospheric Ad- ministration/National Climatic Data Center Paleoclimatology database. Vertical axes are flipped so that “down” indicates drier conditions. See Comparison with Other Paleoclimatic Records for discussion. ANTHROPOLOGY a wood fragment, and one from a pine scale from the same depth. To avoid significant morphological overlap between the two pollen types makes it the influence of hardwater error on radiocarbon dates, we focused on difficult to distinguish domesticated maize from teosinte (46). Our identi- dating terrestrial macrofossils, or, in their absence, concentrations of pine fications of maize pollen must therefore be regarded as tentative. pollen obtained via standard palynological techniques and then separated via cell sorter (Fig. S3). Two dates were obtained from charcoal recovered Stable Isotopes. We measured the oxygen and carbon stable isotope ratios from tephra layers at 353 cm and 551 cm. of calcite on bulk sediment samples, containing thin laminations of authi- We use the Bayesian-based age modeling software, Bacon, to create the genic calcite. We inspected samples to verify the absence of shells or large age model, which estimates sediment accumulation rates using a Markov grains of calcite. Isotope ratios are reported in δ notation in ‰ relative to Chain Monte Carlo technique (43). This technique allows for the explicit VPDB. Samples were dried at 50 °C for 24 h and measured on an Isoprime modeling of age uncertainties associated with each depth in the core (SI Micromass mass spectrometer. Analytical precision for δ18O was ± 0.037‰, Materials and Methods and Figs. S3 and S6). and ±0.027‰ for δ13C, and is reported relative to the standard NBS-19 (n = 12). Samples are spaced by ∼20 y. Elemental Geochemistry. Initial qualitative X-ray diffraction revealed that core We interpret changes in δ18O as indicative of changes in the ratio of lithology was dominated by plagioclase, quartz, secondary clay minerals, evaporation and precipitation (E/P). An increase in the evaporation relative 18 18 and authigenic calcite. To quantify fluctuations in elemental geochemistry, to precipitation increases the O content of lakewater: H2 O tends to 16 16 we combusted sediments at 550 °C and analyzed powder pellets using a evaporate more slowly than H2 O, and the higher O content of atmo- ThermoFisher Scientific QUANT’X EDXRF Analyzer. Major element concen- spheric water vapor means that 16O is preferentially returned to the lake in trations are expressed in mass % oxide, while minor element concentrations precipitation (47). A higher (lower) ratio of evaporation to precipitation are expressed in parts per million. The precision for major and trace elements would therefore result in higher (lower) δ18O. The strong covariance of reported here ranged between 0.5% and 3.0%. Samples are spaced by ∼40 y. oxygen and carbon isotope ratios in authigenic calcite (R2 = 0.82) suggests

We interpret the mass % of Al2O3 as an indicator of input of plagioclase a hydrologically closed lake system that loses water primarily to evaporation and secondary clay minerals from the maar slope walls into the lake system. (14, 47) (Fig. S4). Groundwater provides a major source of water into the The flux of Al increases during wet conditions, when enhanced slopewash maar lakes of the Cuenca Oriental. However, previous studies suggest that transports terrigenous material into the lake system (Fig. S5). the phreatic zone aquifers around fall on the local meteoric water line, and largely reflect the isotopic ratio of precipitation (48). Pollen. Samples for pollen analysis (1.25 cm3) were taken at ∼50-y intervals and processed using standard palynological methods (44). We counted 300 grains ACKNOWLEDGMENTS. We thank Christiane Hassel (Indiana University persampleat400× magnification, except for three samples where concen- Bloomington Flow Cytometry Core Facility), Tim Teague and Wenbo Yang trations were low, and, in total, identified 108 different pollen and spore taxa (University of California, Berkeley), Tom Guilderson and Paula Zermeño (Table S1). We also scanned three slides from each level at 100× to identify (Lawrence Livermore National Laboratory), Gabriela Castañeda (UNAM Juriquilla), and several undergraduate assistants. We also thank John Chiang, maize (Zea mays subsp. mays) pollen, monoporate grains with long axis ≥ 90 μ ≥ μ David Wahl, Liam Reidy, and Cindy Looy for helpful discussions. This material m and pore annulus 12 m, as an indicator of local agriculture. We also used is based upon work supported by the National Science Foundation (NSF) Nomarski interference contrast microscopy to eliminate the possibility that Graduate Research Fellowship under Grant DGE 1106400. Additional funding these grains represent Tripsacum (45, 46). Unfortunately, teosinte, the wild comes from NSF DDRIG (BCS-13333370), a University of California (UC) Mu- relative of maize, is native to this area of the Cuenca Oriental, and the seum of Paleontology Graduate Research Award (Summer 2013), grants

Bhattacharya et al. PNAS Early Edition | 5of6 Downloaded by guest on September 27, 2021 from Berkeley’s Stahl Foundation (Summer 2012), UC Berkeley’s Larsen Fund, Grant HA 2756/8-1. This article is LLNL-JRNL-657962. We thank three and German Research Foundation International Ocean Discovery Program reviewers and the editor for their helpful comments.

1. Lachniet M, Bernal JP, Asmerom T, Polyak V, Piperno D (2012) A 2400 Mesoamerican 23. Butzer KW (2012) Collapse, environment, and society. Proc Natl Acad Sci USA 109(10): rainfall reconstruction links climate and cultural change. Geology 40(3):259–262. 3632–3639. 2. O’Hara SL, Metcalfe SE, Street-Perrott FA (1994) On the arid margin: The relationship 24. Butzer KW, Endfield GH (2012) Critical perspectives on historical collapse. Proc Natl between climate, humans and the environment. A review of evidence from the Acad Sci USA 109(10):3628–3631. highlands of central Mexico. Chemosphere 29(5):965–981. 25. Endfield GH (2012) The resilience and adaptive capacity of social-environmental sys- 3. Hodell DA, Brenner M, Curtis JH (2005) Terminal Classic drought in the northern Maya tems in colonial Mexico. Proc Natl Acad Sci USA 109(10):3676–3681. lowlands inferred from multiple sediment cores in Lake Chichancanab (Mexico). Quat 26. Luzzadder-Beach S, Beach TP, Dunning NP (2012) Wetland fields as mirrors of drought Sci Rev 24(12-13):1413–1427. and the Maya abandonment. Proc Natl Acad Sci USA 109(10):3646–3651. 4. Stahle DW, et al. (2011) Major Mesoamerican droughts of the past millennium. Ge- 27. Dunning NP, Beach TP, Luzzadder-Beach S (2012) Kax and kol: Collapse and resilience ophys Res Lett 38(5):L05703. in lowland Maya civilization. Proc Natl Acad Sci USA 109(10):3652–3657. 5. Heine K (2003) Paleopedological evidence of human-induced environmental change 28. Heisey PW, Edmeades GO (1999) 1997/98 World Maize Facts and Trends – Maize in the - area (Mexico) during the last 3500 years. Rev Mex Cienc Geol Production in Drought Stressed Environments: Technical Options and Research Re- 20(3):235–244. source Allocations (Int Maize Wheat Improve Cent, Mexico City). 6. Sanders WT (2003) Collapse and abandonment in Middle America. The Archaeology 29. Haug GH, et al. (2003) Climate and the collapse of Maya civilization. Science – of Settlement Abandonment in Middle America, eds Inomata T, Webb RW (Univ Utah 299(5613):1731 1735. Press, Salt Lake City), pp 193–207. 30. Haug GH, Hughen KA, Sigman DM, Peterson LC, Röhl U (2001) Southward migration 7. Curtis JH, Hodell DA, Brenner M (1996) Climate variability on the Yucatan Peninsula of the intertropical convergence zone through the Holocene. Science 293(5533): – (Mexico) during the past 3500 years, and implications for Maya cultural evolution. 1304 1308. Quat Res 46(1):37–47. 31. Bhattacharya T, Chiang JCH (2014) Spatial variability and mechanisms associated with – 8. Metcalfe S, Davies S (2007) Deciphering recent climate change in central Mexican lake El Niño-induced drought in Mexico. Clim Dyn 43(12):3309 3326. records. Clim Change 83:169–186. 32. Giannini A, Kushnir Y, Cane MA (2000) Interannual variability of Caribbean rainfall, – 9. Armillas P (1969) The arid frontier of Mexican civilization. Trans N Y Acad Sci 31: ENSO, and the Atlantic Ocean. J Clim 13(2):297 311. 33. Seager R, et al. (2009) Mexican drought: An observational modeling and tree ring 697–705. study of variability and climate change. Atmosfera 22(1):1–31. 10. Plunket P, Uruñuela G (2005) Recent research in Puebla prehistory. J Archaeol Res 34. Magaña V, Vásquez JL, Pérez JL, Pérez JB (2003) Impact of El Niño on precipitation in 13(2):89–127. Mexico. Geofis Int 42(3):313–330. 11. García-Cook AG, Merino Carrion BL (1998) Cantona: Urbe Prehispánica en el Altiplano 35. Kushnir Y, Seager R, Ting M, Naik N, Nakamura J (2010) Mechanisms of tropical At- Central de México. Lat Am Antiq 9(3):191–216. lantic SST influence on North American precipitation variability. J Clim 23:5610–5628. 12. Knight CLF, Glascock MD (2009) The Terminal Formative to Classic Period obsidian 36. Wang C, Lee S, Enfield DB (2008) Climate response to anomalously large and small assemblage at Palo Errado, , Mexico. Lat Am Antiq 20(4):507–524. Atlantic Warm Pools during the summer. J Clim 21:2437–2450. 13. Higgins RW, Yao Y, Wang XL (1997) Influence of the North American monsoon system 37. Chiang JCH, Kushnir Y, Giannini A (2002) Deconstructing Atlantic intertropical con- on the U.S. summer precipitation regime. J Clim 10(10):2600–2622. vergence zone variability: Influence of the local cross-equatorial sea surface tem- 14. Leng MJ, Marshall JD (2004) Palaeoclimate interpretation of stable isotope data from perature gradient and remote forcing from the eastern equatorial Pacific. J Geophys lake sediment archives. Quat Sci Rev 23(7-8):811–831. Res 107(D1):10.1029/2000JD000307. 15. Piperno DR (2006) Quaternary environmental history and agricultural impact on 38. Conroy JL, Overpeck JT, Cole JE, Shanahan TM, Steinitz-Kannan M (2008) Holocene vegetation in Central America. Ann Mo Bot Gard 93:274–296. changes in eastern tropical Pacific climate inferred from a Galapagos lake sediment 16. Park J, Byrne R, Böhnel H, Molina Garza R, Conserva M (2010) Holocene climate record. Quat Sci Rev 27(11-12):1166–1180. change and human impact, central Mexico: A record based on maar lake pollen and 39. Moy CM, Seltzer GO, Rodbell DT, Anderson DM (2002) Variability of El Niño/Southern sediment chemistry. Quat Sci Rev 29(5-6):618–632. Oscillation activity at millennial timescales during the Holocene epoch. Nature 17. Piperno DR, et al. (2007) Late Pleistocene and Holocene environmental history of the 420(6912):162–165. Iguala Valley, Central Balsas Watershed of Mexico. Proc Natl Acad Sci USA 104(29): 40. Yan H, Sun L, Wang Y, Huang W, Qiu S (2011) A record of the Southern Oscillation – 11874 11881. Index for the past 2000 years from precipitation proxies. Nat Geosci 4:611–614. 18. García-Cook A, Martínez Calleja Y (2012) Sistemas de almacenamiento en Cantona, 41. Cobb KM, Charles CD, Cheng H, Edwards RL (2003) El Niño/Southern Oscillation and Puebla [Storage systems in Cantona, Puebla]. Almacenamiento Prehispanico del norte tropical Pacific climate during the last millennium. Nature 424(6946):271–276. de Mexico al Altiplano Central [Prehispanic Storage from Northern to Central High- 42. Wurtzel JB, et al. (2013) Mechanisms of southern Caribbean SST variability over the land Mexico], eds Bortot S, Michelet D, Darras V (Centro de Estudios Mexicanos y last two millennia. Geophys Res Lett 40(22):5954–5958. Centroamericanos, Mexico City), pp 91−107. Spanish. 43. Blaauw M, Christen JA (2011) Flexible paleoclimate age-depth models using an au- 19. Beramendi-Orosco LE, et al. (2009) High-resolution chronology for the Mesoamerican toregressive gamma process. Bayesian Anal 6(3):457–474. urban center of Teotihuacan derived from Bayesian statistics of radiocarbon and 44. Faegri K, Iversen J (1989) Textbook of Pollen Analysis (John Wiley, New York). archaeological data. Quat Res 71(2):99–107. 45. Whitehead DR, Langham EJ (1965) Measurement as a means of identifying fossil 20. Evans ST (2008) Late Classic, Classic collapse, and Epiclassic. Ancient Mexico and maize pollen. Bull Torrey Bot Club 92(1):7–20. Central America: Archaeology and Culture History (Thames and Hudson, London), 46. Holst I, Moreno JE, Piperno DR (2007) Identification of teosinte, maize, and Tripsacum pp 315–376. in Mesoamerica by using pollen, starch grains, and phytoliths. Proc Natl Acad Sci USA 21. Siebe C, Abrams M, Macias JL, Obenholzner J (1996) Repeated volcanic disasters in 104(45):17608–17613. Prehispanic time at Popocatépetl, central Mexico: Past key to the future? Geology 47. Criss RE (1999) Nonequilibrium fractionation and isotopic transport. Principles of 24(5):399–402. Stable Isotope Distribution (Oxford Univ Press, Oxford), pp 139–183. 22. Conserva ME, Byrne R (2002) Late Holocene vegetation change in the Sierra Madre 48. Issar A, Quijano JL, Gat JR, Castro M (1984) The isotope hydrology of the ground- Oriental of Central Mexico. Quat Res 58(2):122–129. waters of central Mexico. J Hydrol 71(3-4):201–224.

6of6 | www.pnas.org/cgi/doi/10.1073/pnas.1405653112 Bhattacharya et al. Downloaded by guest on September 27, 2021