R EPORTS 33. R. C. Finkel, K. Nishiizumi, J. Geophys. Res. 102 (C12), 42. A. Schiller, U. Mikolajewicz, R. Voss, Clim. Dyn. 13, California, Lawrence Livermore National Laboratory 26699 (1997). 325 (1997). under contract no. W-7405-Eng-48. NSF Paleoenvi- 34. M. Stuiver et al., Radiocarbon 40, 1041 (1998). 43. We thank D. Shindell, H. Wright, T. Johnson, D. ronmental Arctic Sciences (PARCS) contribution 35. F. Yiou et al., J. Geophys. Res. 102, 26783 (1997). Peteet, and two anonymous reviewers for comments; number 212. 36. P. Foukal, Geophys. Res. Lett. 29, 2089 (2002). Y. Axford, J. Briner, and A. Werner for field assistance; Supporting Online Material 37. J. L. Lean, Y. M. Wang, N. R. Sheeley, Geophys. Res. and J. Bright and S. McMillan for laboratory assist- www.sciencemag.org/cgi/content/full/301/5641/1890/ Lett. 29, 2224 (2002). ance. Supported by a Packard Fellowship in Science DC1 38. G. Bond et al., Science 278, 1257 (1997). and Engineering (F.S.H.); NSF grant nos. Materials and Methods 39. J. D. Haigh, Science 272, 981 (1996). ATM-9996064 and ATM-0318404 (F.S.H.), Table S1 40. D. T. Shindell, D. Rind, N. Balachandran, J. Lean, P. EAR-9808593 (D.K.), and ATM-0081478 ( Y.S.H.); and References and Notes Lonergan, Science 284, 305 (1999). an Israeli Science Foundation grant (A.S.). Radiocar- 41. D. T. Shindell, G. A. Schmidt, M. E. Mann, D. Rind, A. bon dating was performed, in part, under the auspices Waple, Science 294, 2149 (2001). of the U.S. Department of Energy by the University of 30 June 2003; accepted 28 August 2003 (SnO2). To assess the history of smelting, we Intensive Pre-Incan Metallurgy measured the concentrations in lake sedi- ments of five metals (Ag, Bi, Pb, Sb, and Sn) Recorded by Lake Sediments associated with ore composition (9). Of these metals, Pb serves as the cornerstone of our interpretations for two reasons. First, the In- from the Bolivian Andes can smelting technology relied on argentifer- Mark B. Abbott1*† and Alexander P. Wolfe2† ous galena [soroche (Pb, Ag)S] as a flux during smelting, which was conducted in The history of pre-Columbian metallurgy in South America is incomplete be- charcoal-fired, wind-drafted furnaces lined cause looting of metal artifacts has been pervasive. Here, we reconstruct a with clay (huayras)(12, 13). The use of millennium of metallurgical activity in southern Bolivia using the stratigraphy soroche led to excessive Pb volatilization, of metals associated with smelting (Pb, Sb, Bi, Ag, Sn) from lake sediments resulting in lake sediment concentrations that deposited near the major silver deposit of Cerro Rico de Potosı´. Pronounced are orders of magnitude higher than those of metal enrichment events coincide with the terminal stages of Tiwanaku culture the other analyzed metals. Second, Pb is (1000 to 1200 A.D.) and Inca through early Colonial times (1400 to 1650 A.D.). largely immobile once deposited in lake sed- The earliest of these events suggests that Cerro Rico ores were actively smelted iments (14, 15). Although molecular diffu- at a large scale in the Late Intermediate Period, providing evidence for a major sion rates for Ag are higher than those for Pb pre-Incan silver industry. (16, 17), in both cases they are insignificant in comparison to average sediment accumu- New World metallurgy emerged in the Andean highly sought by royalty for symbolic and lation rates in LL (ϳ1 mm yearϪ1). Thus, region of South America between the Initial ritual purposes (6), the geographic distribu- postdepositional mobility is not a confound- Period (1800 to 900 B.C.) and the Early Hori- tion, intensity, and timing of Late Intermedi- ing factor in the interpretation of the record, a zon (900 to 200 B.C.) (1). The oldest well-dated ate Period silver mining in the Andes remains conclusion supported by the largely parallel archaeological site containing metal artifacts is unclear. Here, we infer a regional history of trends observed for each of the metals. Mina Perdida (Lurı´n Valley) in coastal Peru, metallurgy from lake sediments retrieved ad- Before 1000 A.D., concentrations of all five where hammered foils and gilded copper are jacent to the largest silver deposit of the metals in the sediments of LL were low and preserved in contexts dating to 1400 to 1100 Bolivian tin belt. stable, representing natural background levels B.C. (2). The tradition of sheet-metal working Laguna Lobato (hereafter LL) (7) is small of metal accumulation (Fig. 2). An additional (hammering, gilding, annealing, and repousse´) (0.2 km2), relatively deep (11 m), and occu- 15 samples spanning the earlier period from remained pervasive in the Andes throughout the pies a nonglacial catchment of 3.9 km2 (Fig. 2000 B.C. to 600 A.D. have similarly low Pb Early Intermediate Period (200 B.C. to 600 1). The lake overflows only during the wet A.D.) and the Middle Horizon (600 to 1100 season (December to March), and it has no 65°W60°W 55°W A.D.) (3). By 1000 A.D., large-scale copper hydrological connection with surface waters ° smelting and bronze production is evident at draining Cerro Rico, 6 km west of the lake. 10 S BRAZIL sites such as Bata´n Grande on the northern Because westerly winds prevail for 8 months PERU Peruvian coast (4). Beginning in the Late of the year (April to November) (8), LL is Intermediate Period (1100 to 1450 A.D.), strategically located to record atmospheric intensive copper working became widespread deposition of metals volatilized during smelt- Cuzco Quelccaya BOLIVIA on the Bolivian altiplano, with the production ing or transported as fine-grain particulates. Lake Tiwanaku of materials of copper-tin alloy (i.e., bronze), This study is based on a 74.5-cm core recov- Titicaca LaPaz in contrast to the copper-arsenic artifacts ered from the deepest portion of the lake and Oruro Study site found in Peru (5). By this time, silver and dated by 210Pb, 137Cs, (table S1), and 14C gold were well-established as precious metals analyses (9) (table S2). 20°S Potosí among Andean cultures. Although silver was Cerro Rico lies within a zone of xenother- Pacific mal mineralization related to Middle Tertiary Ocean PARAGUAY intrusions (10). In addition to native silver, CHILE ARGENTINA 0km 400 1Department of Geology and Planetary Science, Uni- the richest ores contain combinations of acan- 2 versity of Pittsburgh, Pittsburgh, PA 15260, USA. De- thite (Ag S), andorite (PbAgSb S ), chlorar- Fig. 1. Location map of the study site in rela- partment of Earth and Atmospheric Sciences, Univer- 2 3 6 gyrite (AgCl), matildite (AgBiS ), miargyrite tion to the Tiwanaku capital, Lake Titicaca, sity of Alberta, Edmonton, AB T6G 2E3, Canada. 2 Potosı´, and the Quelccaya ice core. The shaded (AgSbS ), pyrargyrite (Ag SbS ), and tetra- *To whom correspondence should be addressed. E- 2 3 3 area indicates the central Bolivian tin belt (27), mail: [email protected] hedrite [(Ag, Cu, Fe, Zn)12Sb4S13](11). Tin which broadly corresponds to the crest of the †Both authors contributed equally to this work. is associated primarily with cassiterite Andean cordillera. www.sciencemag.org SCIENCE VOL 301 26 SEPTEMBER 2003 1893 R EPORTS concentrations (ϳ20 ␮ggϪ1), despite stable Thus, Holocene climate variability exerted little centrations did not return to levels recorded isotopic evidence for pronounced hydrological influence on nonpollution metal fluxes to the during the earlier smelting peak, suggesting changes in this time interval (18) (fig. S1). lake’s sediments. Metal concentrations initially technological advances by Incan metallurgists rose well above background shortly after 1000 aimed at minimizing volatile losses of silver. A.D., reaching a first peak around 1130 to 1150 This notion is supported by distinctive spikes of A Year A.D. A.D. (Fig. 2). Concentrations of Pb exceed 100 these three metals immediately after the Span- ␮ggϪ1 in this interval, approximately one-third ish arrival at Potos´ı (1545 A.D.). Early Colonial 2000 1900 1800 1700 1600 1500 1400 1300 1200 1100 1000 900 800 700 600 0 of the peak Pb burden reached subsequently in smelting used bellowed Castilian stone furnaces 10 }CRS 210Pb early Colonial times. Such enrichment trends that had proven successful elsewhere in the 20 are directly comparable to those reported from Andes, but repeatedly failed at Cerro Rico (12). 30 lakes proximal to major Medieval mining sites These furnaces overheated the ore, thus volatil- 40 50 in Europe (19). Because we know of no natural izing metals, including the silver targeted for 60 processes capable of inducing this degree of Pb extraction. Accordingly, the maximum sedi- Depth in core (cm) 70 enrichment, and given the proximity of LL to a ment Ag enrichment of the last millennium AMS 14C 80 major source of metal pollution, we associate occurs in conjunction with colonial experimen- Tiwanaku B Archaeology { Altiplano the magnitude of metal enrichment in the lake’s tation. Due to the failure of Spanish extractive Inca • Arrival of Pizarro at Potosí (1545) sediments with the intensity of ore smelting. techniques, the indigenous huayra technology • Exhaustion of surface ore (1572) This initial rise in metal concentration coin- was retained at Cerro Rico, leaving the smelting Lake Titicaca lowstand C Quelccaya precipitation decline cides with the late stages of the Tiwanaku Em- process largely in the hands of Incan metallur- 300 pire that controlled the Lake Titicaca basin and gists under colonial rule. In the 27 years that Pb extended beyond Oruro in the southern alti- followed the Spanish arrival, thousands of ac- 200 plano (Fig. 2). Very few Tiwanaku silver arti- tive huayras adorned the mountain at any given 100 facts appear to have survived, except for early time (11, 12).