Forcing of the LGM Mesoamerican Monsoon.

The observation of similar modern and LGM Mesoamerican monsoon strength is a surprising result because of the lower specific humidity associated with cooler temperatures, such as is indicated by 3–4 °C sea-surface temperature (SST) depression relative to modern in the Cariaco Basin (17) and lower land-surface temperatures (LST) on the Yucatan Peninsula (20). Thus, the low LGM δ¹⁸O values may be explained by a similar degree of air-mass rainout as today. The stalagmite δ¹⁸O values also contain a temperature imprint: a decrease in LST of ∼3–4 °C is responsible for making the stalagmite δ¹⁸O values higher by ∼0.6–0.8% due to the water–carbonate precipitation temperature (13), leading to apparently slightly weaker monsoons during cold periods. However, this influence on the stalagmite δ¹⁸O profile (+0.6 to +0.8%) is small relative to the observed magnitude of δ¹⁸O changes related to monsoon strength (ca. 3–4%), and it would have been offset by the decrease in δ¹⁸O values of precipitation during cold periods due to the atmospheric temperature effect. Because our observations contradict a simple thermodynamic forcing of monsoon strength via atmospheric water vapor content, an alternative mechanism for maintaining an active LGM monsoon must be considered.

A key similarity between today and the LGM is that local summer insolation at 20°N reached precessional-scale minima, and we suggest that the similar monsoon strength was forced by nearly equal summer insolation despite the high ice volume, lower temperature, and lower CO₂ (21) boundary conditions at the LGM (Fig. 1). Today, monsoon strength is positively correlated both with the magnitude of the ocean-to-land temperature gradient (ΔT_SST-LST) and proximity to the ITCZ, and we hypothesize that similar processes affected the Mesoamerican monsoon during the LGM in response to similar orbital forcing. However, insolation forcing is unlikely to be the only driver on monsoon strength during deglaciation, because weak monsoon intervals during the YD and HS1 also occurred at times of higher summer insolation and low temperatures.

Atlantic Ocean Circulation and ITCZ Forcing of the Mesoamerican Monsoon.

An alternative hypothesis to explain the similar LGM and modern monsoon strength, potentially acting in concert with insolation forcing, is via the latitudinal position of the ITCZ. Today, the ITCZ migrates seasonally in response to the warmest SSTs, reaching its northern most extent ∼10–12°N in late boreal summer over the eastern Pacific warm pool (22) off southwestern Mexico. On millennial time scales, the ITCZ latitudinal position has been shown to be forced by the strength of the ocean circulation (AMOC) (23), so we compared our record to the ocean circulation ²³¹Pa/²³⁰Th proxy in the North Atlantic ocean. Our Mesoamerican monsoon reconstruction shares several features with the ²³¹Pa/²³⁰Th AMOC proxy, including intermediate AMOC during the LGM, a pronounced weakening during HS1, and an Early Holocene increase. Because our record displays key similarities to the ²³¹Pa/²³⁰Th record, we suggest that Mesoamerican monsoon strength is also related to the vigor of the AMOC and its consequent influence on the latitude and strength of the ITCZ. Active AMOC maintains high SST in the North Atlantic as a result of cross-equatorial northward transport of warm surface waters, which in turn favors a northerly location of the ITCZ, and vice versa.

To further test the hypothesis of an ITCZ position influence on the Mesoamerican monsoon, we compared our record to other hydrologic proxy records in the global monsoon regions temporally (Fig. 2) and spatially (Fig. S3) to AMOC variations for the Early Holocene, YD, HS1, and LGM time slices. The records exhibit a strong similarity in character, with pronounced hydrologic anomalies during the LGM, deglacial, and Holocene time periods that are antiphased between the hemispheres, as would be predicted by an ITCZ control on monsoon strength. Prominently, active monsoon conditions during the LGM are apparent in the Mexico speleothem reconstruction and in high magnetic susceptibility measurements in Lake Petén Itzá in Guatemala (19, 24), and in similar Early Holocene and LGM (ca. 24 ka) Central American monsoon strength (25), demonstrating wetness on both the Atlantic and Pacific coasts of Mesoamerica and Central America. Based on these observations, we conclude that the active LGM monsoon was aided by a proximal location of the ITCZ to the Mexican landmass, which facilitated moisture delivery into the monsoon system. This mechanism is consistent with modern climate dynamics, such as cooling similar to a modern La Niña event (Fig. S4), which results in increased ΔT_SST-LST, a more northerly position of the ITCZ (1), and low δ¹⁸O values (18) associated with the monsoon.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Hydrological proxies for the North and South American monsoon regions show antiphased behavior. (A) Mexico stalagmite δ¹⁸O reconstruction of the Mesoamerican monsoon. (B) Summer insolation average (JJA) for 20°N. (C) Lake Petén Itzá magnetic susceptibility record on the original chronology (light blue) for the Holocene (19) and an updated (blue) chronology for the Late Glacial Period (24). (D) Reflectance in Cariaco Basin sediment cores (26). (E) Ti concentrations in Cariaco Basin sediment (32). (F) Asian Monsoon proxy reconstruction from Dongge and Hulu Caves (3). (F) Ti concentrations in Cariaco basin sediments, positively correlated with greater continental runoff (32). (G) Ti/Ca ratio in northeastern Brazil margin sediments, a proxy for continental runoff (28). (H) Western Amazon stalagmite monsoon reconstructions from Santiago (orange) and Tigre Perdido Caves (light green) (8, 27). (I) Stalagmite monsoon reconstruction from Botuverá Cave, southern Brazil (4). (J) Lake level reconstructions for the Bolivian altiplano (31). Stalagmite δ¹⁸O records have been adjusted for the change in δ¹⁸O values of the ocean, denoted with the subscript “swc.” The y axis for the South American monsoon records are reversed for better comparison with northern Hemisphere records.

Continental river runoff in the Cariaco Basin of northern South America (26) shows intermediate levels during the LGM, indicating that the monsoon was somewhat weaker than today, although interpretation of the LGM-to-modern change in this record is complicated by the influence of deglacial sea-level rise on ocean circulation. The Mesoamerican monsoon data are similar to observations of the EAM (3), which shows similar sea-water–corrected δ¹⁸O values during the LGM as today, a likely response to the common forcing of summer insolation. Similarly, speleothem δ¹⁸O records from South America [Botuverá (4) and Santiago and Tigre Perdido (8, 27) Caves] indicate similar to somewhat stronger monsoon strength at the LGM relative to today, but continental runoff in northeastern Brazil appears to have been low (28).

Following an active LGM monsoon, the Mesoamerican monsoon collapsed abruptly at the onset of HS1 (18 ka) and remained weak for at least 700 y. General Circulation Model (GCM) results indicate an AMOC slowdown during freshwater release into the North Atlantic during Heinrich events, which provokes a southward displacement of the ITCZ (29). A southward displaced ITCZ during HS1 is supported by observations of minimum Ti concentrations in the Cariaco Basin (26), increased wetness in northeast Brazil (28), expansion of paleolakes and glaciers on the Bolivian altiplano (30, 31), and a weak EAM (3) (Fig. S3). The Mesoamerican monsoon was also weak during the YD (6) in southwestern Mexico, Lake Petén Itzá (low magnetic susceptibility), and in the Cariaco Basin (low Ti concentrations) (32) when SST was depressed by 3–4 °C (17). AMOC was weakened during the YD relative to interstadial conditions (15). In contrast, locally wet conditions mark the YD in South America (Fig. 2), but to a lesser extent than during HS1. The ITCZ was thus apparently weaker and displaced to the south during the YD (Fig. S3).

A strengthening of the Mesoamerican monsoon coincided with the YD/Early Holocene transition in Greenland (6), and was associated with a peak in summer insolation associated with the precession cycle at ca. 9.5 ka. Subsequently the monsoon gradually weakened during the Holocene, and was punctuated by a weak monsoon interval between 9.3 and 7.4 ka spanning the 8.2-ka event (14). The Holocene monsoon weakening supports the hypothesis of an orbital pacing (3, 33) of the Mesoamerican monsoon, whereby decreasing summer insolation at 20°N resulted in a southward displacement of the ITCZ. Similar trends are also evident in decreasing Holocene wetness in the Cariaco Basin, the EAM, and a monsoon strengthening in South America.

Regional Records of the Mesoamerican Monsoon.

A synthesis of Mesoamerican climate records (Fig. 3), including a well-dated central Mexico glacial chronology (34) and lacustrine records from the Yucatan Peninsula (19, 35) and Central Mexico (36⇓–38), shows that paleoclimate varied coherently since the LGM (all records are shown on a calendar age scale; see SI Text for methods). Our observation of an active LGM monsoon is supported by wet conditions inferred from high magnetic susceptibility in Lake Petén Itzá sediments (19), during which time pollen-inferred LSTs were lowered by ∼4 °C (35). Maximum glacial extent in highland Mexico between ca. 21 and 17.5 ka (34) (Hueyatlaco-1 advance) coincided with a high Lake Chalco level (37) and a high abundances of isoetes spores (an aquatic fern of oligotrophic lakes) in Lake Pátzcuaro in the western Mexican highlands (36). High abundances of open water (38) and high neutral diatoms (pH ∼7; medium to low electrical conductivity) (39) indicate fresh lakes in the Basin of Mexico. The available evidence supports a wet and cold LGM in Mesoamerica, although the radiometric control of the central Mexico lake chronologies is poor (SI Text). The wet conditions inferred from the lake records suggest that the LGM monsoon was similar in effective strength relative to today, despite the lower specific humidity associated with decreased atmospheric temperatures during cold periods.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Comparison of Mesoamerica paleoclimate records shows a coherent monsoon response and a wet LGM. (A) Mexico stalagmite δ¹⁸O monsoon record. (B) JJA mean insolation for 20°N latitude. (C) Interval of sedimentary hiatus on west shore of L. Texcoco. (D) Sum of open-water diatoms in main basin of L. Texcoco (38). (E) Percentage of high neutral diatoms in L. Chalco, southern Basin of Mexico (39). (F) Percentage of isoetes spores in L. Pátzcuaro sediments. (G) Percent Ca, a proxy for marl precipitation during dry conditions (36). (H) Lake Petén Itzá magnetic susceptibility record (19). (I) Glaciation altitudinal limits on Iztaccíhautl (34). (J) Pollen-inferred land surface temperatures on the Yucatan Peninsula (35); black star is the age of the earliest securely dated human remains in Mexico (49), dating to the YD.

Widespread evidence for dry conditions in Mesoamerica documents a coherent Mesoamerican monsoon collapse at the LGM to HS1 transition and during the YD (Fig. 3). Gypsum deposition in Lake Petén Itzá was attributed to evaporative conditions and low lake levels (19) during Heinrich stadials and the YD, when LSTs were similar to the LGM (35). In the Mexican highlands, a decrease in circumneutral (pH ∼7) diatom abundance in Lake Chalco sediments (37) is evident during HS1, and glaciers retreated upslope. These observations suggest that moisture was a limiting factor in glacier mass balance during HS1, a contention supported by low levels of Lake Chalco between 18.5 and 14.5 ka (37) and a sedimentary hiatus on the margins of ancient Lake Texcoco (38). Post-YD monsoon strengthening is indicated by low stalagmite δ¹⁸O values, lake filling in southwestern Mexico at 10.85 ka (40), and establishment of a mesic Yucatan forest by 11.25 ka (19). An abrupt Early Holocene monsoon weakening between 7.5 and 9.0 ka was associated with the only major Holocene glaciation on Iztaccíhuatl, (Milpulco-2) (34). A Late Holocene drying trend is evident in higher stalagmite δ¹⁸O values by ca. 3.5 ka and increased marl precipitation in Lago Pátzcuaro in western Mexico (36).

Mesoamerican and Global Paleomonsoons.

Based on a comparison of our monsoon reconstruction to orbital forcing, SST, and AMOC variations, we conclude that the Mesoamerican monsoon responded similarly to the shared global forcings evident in the paleomonsoon responses in Asia and South America (3, 33) over the last 22 ka (Fig. S3). Based on correlations to paleoclimate proxy records, we conclude that the primary controls on Mesoamerican monsoon strength were (i) local summer insolation values driven by the precession cycle, and (ii) the AMOC-controlled position of the ITCZ. The conclusion that AMOC was active during the LGM is supported by the low ²³¹Pa/²³⁰Th ratios in the North Atlantic Ocean, and detailed observations of ²³¹Pa/²³⁰Th ratio in Atlantic Ocean sediments that document a shallower deepwater overturning cell, but with a rate of AMOC similar or greater in strength than today (41). The paleoceanographic data thus document an active AMOC during the LGM, favorable for forcing an active LGM Mesoamerican monsoon via a northerly position of the ITCZ. GCM results also support our conclusions of an active LGM monsoon. In a combined modern and LGM run of the community climate model (CCM3) (42), wet summer conditions in southwestern Mexico southward to 20°N are simulated for the LGM, and we show that a strong monsoon extended to at least 17.4°N. Conversely, freshwater “hosing” model experiments, where a large volume of freshwater is added to the North Atlantic Ocean to simulate meltwater discharge during Heinrich stadials, result in a southward displacement of the ITCZ (23, 29). The model experiments are consistent with increased ²³¹Pa/²³⁰Th ratios in the North Atlantic Ocean during HS1 and the YD (15), recording a weakening of the AMOC and a southward displacement of the ITCZ. The tropical hydrologic response appears to be strongest and most coherent during HS1, with clear evidence for a southward displacement of the ITCZ, with a weaker response in the YD and Early Holocene (Fig. S3).

Our monsoon reconstruction differs in important characteristics from the East Asian and South American monsoons (Fig. 2). Most prominently, monsoon collapses during HS1 and the YD were of apparently equal magnitude in Mesoamerica, whereas the EAM displays a weaker YD response relative to HS1 which is more similar to the AMOC forcing. There is a less consistent hydrological response in the SASM during the YD, with the expansion of altiplano lakes, increased runoff to the Brazil margin, and lower δ¹⁸O in the Botuverá stalagmite recording a stronger monsoon (28, 31, 43, 44), yet stalagmites in the western Amazon and Amazon river runoff indicating a weaker monsoon (8, 45). Part of this discrepancy is likely due to the presence of a precipitation dipole over the western Amazon and northeast Brazil, where there is an opposite rainfall response to climate events during the Late Glacial (44, 46). For example, the YD was somewhat wetter than today in northeast and southern Brazil, but in the western Amazon was marked by a transition from wet to dry conditions (8, 27, 44). We also observe enhanced Holocene δ¹⁸O anomalies and hence rainfall variability in Mesoamerica relative to Asia and South America. These observations suggest that the abrupt transitions and amplified isotopic variability in the Mesoamerican monsoon may be a result of its proximal location to the AMOC center of action in the North Atlantic Ocean, possibly enhanced by a high sensitivity or a threshold response of monsoon strength to ITCZ position.

Implications for Mesoamerican Monsoon Forcing and Emergence of Agriculture.

Our conclusion of an active LGM summer monsoon similar to today differs from prior interpretations of wetness in central and southwestern Mexico. Previous workers suggested that such wetness was derived from winter extratropical air masses guided by a southward-displaced jet stream (20, 36) and/or polar outbreaks (nortes) (19). If the extratropical moisture hypothesis is correct, then winter precipitation would have been maximal during periods associated with the most extreme southward displacement of the ITCZ during the YD and HS1 when AMOC was slow. To test this prediction, we analyzed isotopic data for Chihuahua in northern Mexico (28.63°N, 106.07°W), which receives both summer and winter rainfall. Because winter precipitation δ¹⁸O values [−9.2 ± 2.9% Vienna standard mean ocean water (VSMOW)] are about 3% lower than summer (−6.2 ± 3.3% VSMOW), the winter precipitation hypothesis would predict that stalagmite δ¹⁸O values would have decreased by several permil during the YD and HS1. In fact, we see the opposite: δ¹⁸O values increased during these times. Further, the winter rainfall hypothesis suffers from other weaknesses. Today, rainfall is dominated by the summer monsoon, with winter rainfall providing less than 4% to annual totals to southwestern Mexico because of low specific humidity in winter air masses (Figs. S5–S7). Further, neither enhanced winter westerlies nor polar outbreaks alone can account for observed regional wetness in Mesoamerica (Fig. 2), because orographic effects today limit winter precipitation to windward slopes (Fig. S8). This contrasts with widespread monsoon wetness observed both today and during the LGM. A drastic shift in the latitude of the westerlies deep into the tropics is also difficult to reconcile with modern and Late Glacial climate dynamics and requires a no-modern-analog climate state. GCM output indicates the zone of maximum winter westerly precipitation over North America at the LGM was restricted to ca. 30 and 40°N (42), and it is difficult to envision the presence of pluvial lakes at these latitudes (47) if the winter rainfall band was displaced to 17°N latitude.

Our monsoon reconstruction has implications for Mesoamerican plant domestication, because agricultural emergence after human colonization was hypothesized to have been impossible during the Late Glacial Period because of low CO₂, low temperatures, and aridity (48). Although our data support an active monsoon during the LGM (18–22 ka), the earliest securely dated human remains in Mexico significantly postdate the LGM to 12,650 ± 70 calendar y B.P. (10,755 ± 75 ¹⁴C y B.P.) (49), coinciding with the YD weak monsoon interval (Fig. 1). However, the presence of the earliest known pre-Clovis artifacts in North America, dated to as early as 15.5 ka, and their arrival in South America (Monte Verde site) by ca. 14.5 ka (50, 51), suggests that humans may have also been present in Mesoamerica during the B/A interstadial, when the Mesoamerican monsoon was strong (19), and CO₂ concentrations of ca. 240 ppm approached Early Holocene concentrations (ca. 260 ppm). In contrast, had humans traversed Mesoamerica during HS1 between 17.0 and 14.5 ka, they would have encountered a weak monsoon when CO₂ was <220 ppm. Neither HS1 nor the YD appears to present suitable monsoon conditions to support agriculture, and no solid evidence of human occupation of Mesoamerica during the wet LGM is available. Further, the absence of evidence for agricultural emergence during the B/A may be related to the relatively short duration of suitable climatic conditions, absence of a significant human population, or other factors. Thus, our data most strongly support the hypothesis that agriculture in Mesoamerica was first possible when a strong monsoon coincided with climate warming and high atmospheric CO₂ concentrations ca. 11 ka.

Our data can further constrain the climatic conditions associated with maize domestication, likely from a wild teosinte (Zea Mays ssp. parviglumis; hereafter “teosinte”) ancestor in the Balsas River Basin by ∼9.0 ka (10, 52). Maize was likely cultivated with squash at lake margins sometime during the earlier half of the interval between 10.0 and 5.0 ka (40), and archeological evidence documents the presence of maize starch on grinding tools by ca. 8.7 ka in the Balsas River drainage (52). Our data, including the Holocene section from Cueva del Diablo, located near the town of Teloloapan where a modern stand of teosinte is present (40), suggests that the Early Holocene was marked by highly variable rainfall, from very wet at 9.6 ka, to the most pronounced Holocene dry period between 9.0 and 7.2 ka, and a return to wetter climate between 7.0 and 4.0 ka. The occurrence of maize and squash phytoliths suggests that lake margin agriculture may have been used to exploit the high water tables in this environment as an adaptation to the vagaries of an unstable Holocene climate.