Solar Forcing of Drought Frequency in the Maya Lowlands

See allHide authors and affiliations

Science  18 May 2001:
Vol. 292, Issue 5520, pp. 1367-1370
DOI: 10.1126/science.1057759


We analyzed lake-sediment cores from the Yucatan Peninsula, Mexico, to reconstruct the climate history of the region over the past 2600 years. Time series analysis of sediment proxies, which are sensitive to the changing ratio of evaporation to precipitation (oxygen isotopes and gypsum precipitation), reveal a recurrent pattern of drought with a dominant periodicity of 208 years. This cycle is similar to the documented 206-year period in records of cosmogenic nuclide production (carbon-14 and beryllium-10) that is thought to reflect variations in solar activity. We conclude that a significant component of century-scale variability in Yucatan droughts is explained by solar forcing. Furthermore, some of the maxima in the 208-year drought cycle correspond with discontinuities in Maya cultural evolution, suggesting that the Maya were affected by these bicentennial oscillations in precipitation.

Lake Chichancanab, on the north-central Yucatan Peninsula, Mexico (Fig. 1), possesses sediments that preserve a highly sensitive record of past climate changes (1, 2). In 1993, we retrieved a 4.9-m sediment core from Lake Chichancanab. The 9000-year-long record of oxygen isotopes and gypsum concentrations from the core was used to infer relative changes in the ratio of evaporation to precipitation (E/P). The record demonstrated that the period 800 to 1000 A.D. was the driest of the middle to late Holocene and that it coincided with the collapse of Classic Maya civilization in the 9th century A.D. The cause of this drought remained elusive, however, although it apparently was widespread in the Neotropics (3–6) and Sahel (7). Here we present analyses from new Lake Chichancanab sediment cores and demonstrate that the terminal Classic drought was only one episode in a recurrent pattern of dry events that occurred during the past 2.6 millennia.

Figure 1

Location of Lakes Chichancanab (CH) and Punta Laguna (PL) relative to precipitation (mm/year) on the Yucatan Peninsula. Mean annual temperature and rainfall near Lake Chichancanab average 25.4°C and ∼1300 mm, respectively, and most of the precipitation occurs during the rainy season between May and September. Mean annual temperature and rainfall near Lake Punta Laguna average 25.7°C and ∼1520 mm, respectively. Annual evaporation losses are considerable (1521 mm/year measured at Santa Rosa, 12 km from Lake Chichancanab) at both locations and, consequently, the lake volumes and dissolved solute concentrations are highly sensitive to changes in E/P.

In May 2000, we returned to Lake Chichancanab and retrieved a high-resolution sediment record that enabled us to refine paleoclimatic interpretations for the region over the last 2600 years. Two cores were taken side-by-side in 11 m of water near the deepest area of Chichancanab's central basin. Because uppermost sediments were unconsolidated, the upper 70 cm of each core were extruded vertically in the field and sectioned at 1-cm intervals. Consolidated deposits below 70-cm depth were transported to the laboratory in polycarbonate tubes, extruded, and sampled at 1-cm intervals for element and stable-isotope analysis. A 1.9-m-long composite section was constructed with the two cores and yielded a basal age of 2600 years before the present (yr B.P.).

Core chronology for the past 2.6 millennia is based on 12 accelerator mass spectrometry (AMS)–14C dates on identifiable terrestrial material (Table 1). This represents a substantial improvement in dating resolution compared with the 1993 core that was studied previously, in which terrestrial material for AMS-14C dating was encountered at only two sediment depths (2). In addition, sedimentation rates in the 2000 core are 40 to 80% greater than those in the 1993 core, averaging 0.7 mm/year from 190- to 85-cm depth and increasing to 0.9 mm/year from 85 cm to the top of the core. Sediment ages above and below 85 cm were calculated with two depth-age linear-regression equations (Table 1). On average, each 1-cm sample represents 11 to 14 years of sediment accumulation.

Table 1

Age analysis of sediment core CH-23-V-00 from Lake Chichancanab. Radiocarbon analyses were performed solely on terrestrial organic matter (seed, charcoal, carbonized wood fragments) because aquatic materials are subject to significant hard-water lake error associated with the dissolution of limestone bedrock (devoid of14C) in the karst watershed of Chichancanab. Radiocarbon dates were converted to calendar years with the program OxCal version 3.5 (33) and atmospheric data from Stuiver et al.(34). Sediment depths were converted to age with two equations derived by linear regression of depth-age points above and below 85 cm. 0 to 85 cm: age (cal yr B.P.) = −10.276 + 10.563 × depth (cm), sedimentation rate = 0.09 cm/year,r 2 = 0.999; 85 cm to bottom: age (cal yr B.P.) = −388.18 + 15.15 × depth (cm), sedimentation rate = 0.07 cm/year, r 2 = 0.993.

View this table:

The Chichancanab core displays considerable stratigraphic variation in the proportions of organic matter, calcium carbonate, and gypsum (Fig. 2, A and B). Lake Chichancanab water is near saturation for gypsum (CaSO4) today (sulfate, 2545 mg liter−1; calcium, 693 mg liter−1), but the mineral precipitates only in shallow water near the lakeshore. During periods of drier climate, lake volume is reduced and CaSO4 saturation is exceeded. Under such dry conditions, gypsum precipitates throughout the entire lake and is preserved in the sediment record, thereby providing a lithologic proxy of past increases in E/P.

Figure 2

Element and isotope results from Chichancanab core CH-23-V-00 versus calendar year (B.C./A.D.). (A) Weight % carbonate determined by coulometric titration (35). (B) Weight % total S (solid line) measured with a Carlo Erba CNS analyzer and gamma-ray attenuation bulk density (dashed line) measured with a GEOTEK Multi-Sensor Core Logger. (C) δ18O of the gastropod Pyrgophorus coronatus(36). Dashed line and crosses correspond to the raw data and bold line represents a five-point running mean through the data. Vertical line at 2.5‰ is the mean value of all data.

Gypsum concentrations in the core were determined by measuring weight percent total S. Gypsum dominates the sediment during three dry periods: 475 to 250 B.C., 125 to 210 A.D., and 750 to 1025 A.D. (Fig. 2B). Yucatan droughts were stronger and more frequent before 1100 A.D., similar to observations in the United States northern Great Plains (8). The oldest arid period occurred between ∼475 and 250 B.C. (2500 to 2200 14C yr B.P.), and previous results from the 1993 Chichancanab core indicated that this event was preceded by a protracted period (∼5000 years) of minimal gypsum precipitation and low 18O isotope (δ18O) values, indicating relatively low E/P (2). Evidence for late Holocene climatic drying at ∼2500 14C years B.P., after moist conditions in the early to middle Holocene, is documented throughout the circum-Caribbean. For example, in a core from Lake Miragoane, Haiti, increased aridity at ∼2400 14C yr B.P. was inferred from an increase in δ18O, indicating higher E/P and lower lake levels (9). At the same time, pollen assemblages show a shift from mature mesic forests dominated by Moraceae to dry forest taxa with abundant weeds (10). A similar late Holocene drying trend is detected in sediment cores from Lake Valencia (11), Venezuela, and the offshore Cariaco Basin (6).

The S peak between 125 and 210 A.D. is consistent with pedological and archaeological evidence for a severe dry period at El Mirador, northern Guatemala, that may have contributed to the site's abandonment around 150 A.D. (12). Aridity between 130 and 180 A.D. has also been inferred on the basis of shell δ18O from sediment cores in Lake Salpeten, Guatemala (13). This dry period coincides with pervasive Preclassic site abandonment in the Maya Lowlands (14) that may have been a response to widespread drought on the Yucatan Peninsula (15).

The most recent episode of gypsum deposition (750 to 1025 A.D.) supports our previous finding of prolonged drought between ∼800 and 1000 A.D. (2, 16), which coincided with the Classic Maya collapse (15, 17). The density record, which reflects the presence of gypsum, suggests two distinct events centered on 800 A.D. (terminal Classic) and 1020 A.D. (Fig. 2B), similar to the δ18O record from nearby Lake Punta Laguna (16). Evidence for drier climate from other areas in Mexico, Central America, and northern South America indicates that the terminal Classic drought was widespread (3–6). Two distinct rainfall minima between ∼800 and 1000 A.D. have also been documented in the Sahel and coincided with the terminal Classic drought in Yucatan (7,18, 19). The circum-Caribbean and the Sahel are directly linked through the seasonal migration of the Atlantic intertropical convergence zone, and this correspondence is not surprising (20, 21).

The ratio of 18O to 16O in a tropical closed-basin lake is controlled mainly by E/P because of the preferential loss of 16O from lake water during evaporation. The δ18O of the water is recorded in the shells of ostracods and gastropods, which precipitate their shells at a constant offset from oxygen isotopic equilibrium that is species-specific (22). The δ18O record from Chichancanab generally supports our E/P inferences based on the S record. Oxygen isotope values are greater than average between 800 and 1090 A.D. (Fig. 2C) when S content is high. Maximum δ18O values in this interval are not substantially greater than in other parts of the record, but low δ18O values (<2.5‰) are conspicuously absent. Similarly, high S content between ∼475 and 250 B.C. is associated with elevated mean δ18O values that decrease rather abruptly at about 150 B.C. Curiously, the S peaks between 125 and 210 A.D. are not associated with high average δ18O values, and the corresponding density peaks in this period are not as high as those recorded in some of the other drought events (Fig. 2B).

Firm sediments below 70 cm were analyzed in their polycarbonate core tubes with a GEOTEK MultiSensor Core Logger (MSCL). Gamma-ray attenuation (GRA) was measured at 0.5-cm intervals and converted to sediment bulk density with a water-aluminum calibration standard. The GRA bulk density stratigraphy is nearly identical to the weight % S profile (Fig. 2B), because authigenic gypsum is denser than the shell-bearing organic matter that composes most of the sediment core. The double peak in density (gypsum) between 750 and 1000 AD may have blurred when the core was extruded and sampled because of the difficulty in sectioning the dense nodular gypsum in this interval. Consequently, the weight % S measured on discrete samples shows a single, broad high rather than two distinct peaks between 750 and 1000 A.D.

Because the density record is at higher resolution (0.5-cm intervals) than either the weight % S or δ18O record and was measured on whole cores before extrusion and sampling, we used it for time series analysis. Spectral analysis of the density signal revealed power concentrated at periods of ∼810, 208, 100, 50, and 39 years (Fig. 3A). The peak at 208 years dominates the spectrum and is significant at the 95% confidence level, whereas the peaks at 50 and 39 years are significant at the 80% level. The periods near ∼100 and 50 years may be harmonics of the ∼208 year cycle. Time series analysis of the δ18O record from Lake Punta Laguna (16), located ∼150 km northeast of Chichancanab, also reveals a 208-year cycle over the last 2000 years (Fig. 3B) that is coherent and in phase with the density record from Chichancanab (Fig. 4B). A 200-year cycle has also been noted in foraminifera abundance records from the Cariaco Basin off Venezuela and is attributed to solar forcing of century-scale variability in regional upwelling and trade-wind intensity during the Holocene (23).

Figure 3

(A) Power spectra of Chichancanab GRA bulk density record (black line) and Δ14C for the last 2500 years (dashed gray line) (34). (B) Power spectra of the δ18O signal measured on the ostracodCytheridella ilosvayi in a sediment core from Lake Punta Laguna (black line) (16) and Δ14C (dashed gray line) (34). Power spectra were calculated with the program Analyseries (37) by interpolating the signals at uniform 5-year intervals and linear detrending. All estimates were made with a Bartlett window, 1/3 lag, and no prewhitening constant was applied. (**) Denotes peaks in the GRA bulk density and δ18O signals that are significant at the 95% confidence level; (*) denotes significance at the 80% level. The peak at a period of ∼208 years dominates the spectra of E/P proxies in Yucatan and Δ14C (34), suggesting that the 208-year drought cycle on the Yucatan Peninsula is related to variations in solar activity. (C) Coherency and phase from cross-spectral analysis of the Punta Laguna δ18O and Δ14C signals. The signals are coherent (∼0.7) and ∼180o out of phase at the 208-year period.

Figure 4

(A) Comparison of Δ14C (red line) (30), δ18O record from Lake Punta Laguna (black line) (16), and GRA bulk density from Lake Chichancanab (blue line). Peaks in δ18O indicate increased E/P (drought) and coincide with minima in 14C production that are caused, in part, by increased solar activity (38). M, S, and W denote the Maunder, Sporer, and Wolf sunspot minima, respectively. (B) Bandpass filter centered at 208-years of the Δ14C record (red line) (34), δ18O signal from Lake Punta Laguna (black line) (16), and the GRA bulk density signal from Lake Chichancanab (blue line). The 208-year filtered component of δ18O is in phase with inferred solar variability (or 180° out of phase with Δ14C production) for the last 1500 years. The 208-year component of the Punta Laguna δ18O and Chichancanab density signals are in phase between 0 and 1200 A.D.

A 206-year cycle is prevalent in records of cosmogenic nuclide production including both 14C and 10Be (Fig. 3) (24, 25), and is believed to reflect solar variability or a combination of solar forcing and oceanic response. Periods of higher solar activity correspond to times of lower cosmonuclide production. Comparison of the Punta Laguna δ18O and the 14C production (Δ14C) records shows an antiphase relation for the past 2000 years (Fig. 4) (25). Higher E/P (δ18O) coincides with lower 14C production, implying that drought occurred during times of increased solar activity (Fig. 4). A similar relation between drought and high solar radiation has been demonstrated in lake sediment records from equatorial east Africa that span the past 1100 years (26).

The mechanism by which changes in solar activity cause E/P shifts in Yucatan is not certain. An amplifying mechanism is required to obtain a significant climate response from rather small variations in solar output (27). Hypothesized mechanisms include changes in the ultraviolet part of the solar spectrum, which affects ozone production and stratospheric temperature structure (28), and the effect of cosmic ray intensity on cloud formation and precipitation (29). Sensitivity experiments conducted with atmospheric general circulation models imply that changes in solar output may affect global mean temperature, humidity, convection, and intensity of Hadley circulation in the tropics (30–32). Mean annual rainfall on the Yucatan Peninsula varies by a factor of 5 over a distance of only ∼500 km between the semiarid northwest coast (500 mm/year) and the southern lowlands of northern Guatemala and Belize (2500 mm/year) (Fig. 1). Consequently, any solar-forced change in the strength or position of Hadley circulation or tropical convective activity would be expected to affect rainfall in the region.

A comparison of the Chichancanab GRA bulk density record with bandpass filters of the signal, centered at periods of 50 and 208 years, shows that the large bicentennial-scale drought events were often composed of multiple ∼50-year oscillations. The record demonstrates that the arid events centered at 485 B.C. and 285 B.C. were part of the 208-year cycle (Fig. 5). Similarly, the droughts between 125 and 210 A.D. (associated with Preclassic Abandonment), at ∼800 A.D. (associated with terminal Classic Collapse), and ∼1020 A.D. also fit the pattern of 208-year drought recurrence. The Maya were highly dependent on rainfall and surface reservoirs as their principal water supply (15). Consequently, these multidecadal- to multicentury-scale oscillations in precipitation probably had a detrimental impact on Maya food production and culture.

Figure 5

(A) GRA bulk density record from Lake Chichancanab (black line) shown relative to a bandpass filter (dashed gray line) of the signal centered at ∼50 years (frequency = 0.02 ± 0.005). (B) GRA bulk density record from Lake Chichancanab (black line) shown relative to a bandpass filter (dashed gray line) of the signal centered at ∼208 years (frequency = 0.0048 ± 0.0008). Centennial-scale Yucatan droughts recurred at a period of ∼208 years and were often composed of multidecadal oscillations with a period near 50 years. Shown to the right are the subdivisions of Maya cultural evolution. Some of the drought events coincide with important Maya cultural discontinuities such as the Preclassic Abandonment (PA) between 150 and 200 A.D. and the Terminal Classic Collapse (TC) between 750 and 900 A.D.

  • * To whom correspondence should be addressed. E-mail: dhodell{at}


View Abstract

Navigate This Article