GEOPHYSICAL RESEARCH LETTERS, VOL. 30, NO. 14, 1772, doi:10.1029/2003GL017191, 2003

Improving climatic signal representation in tropical ice cores: A case study from the Quelccaya , Anton Seimon Department of Geography/CIRES, University of Colorado, Boulder, USA Received 24 February 2003; revised 24 February 2003; accepted 19 May 2003; published 30 July 2003.

[1] The published records for Quelccaya, Peru are patterns found at Quelccaya, snow pit studies suggest that shown to contain numerous errors in annual layer annual layer discrimination can still be difficult [Thompson determination, introduced by the subjective methodology et al., 1984]. Accordingly, a subjective methodology was of the original analysis. A new methodology is presented that utilized by the original investigators prioritizing visible aligns and then merges d18O sample measurement profiles stratigraphy to determine the separation of annual layering, from the two Quelccaya cores. Absolute dating of the with other parameters used for supporting reference, stratigraphy is established by comparison of prominent [Thompson et al., 1986]. Simple layer counting and a single dust anomalies with documented reference- reference marker event, strong anomalies in large partic- marker events. The results demonstrate that the original ulates ascribed to the A.D. 1600 eruption of sample measurements can register a comprehensive [Thompson and Mosley-Thompson, 1989], were used to millennial history at sub-annual resolution, thus date the ice core strata. also offering great potential for clearer interpretations of [4] This study presents new results obtained by careful ENSO signals in tropical Andean ice cores. INDEX TERMS: re-analysis of selected segments of the two cores and offers 3344 Meteorology and Atmospheric Dynamics: Paleoclimatology; a new approach to their dating. Climatic representation is 1827 Hydrology: Glaciology (1863); 1863 Hydrology: Snow and brought into sharper focus by merging the d18O data series ice (1827); 3394 Meteorology and Atmospheric Dynamics: from the cores and establishing the precise time line of their Instruments and techniques. Citation: Seimon, A., Improving strata. climatic signal representation in tropical ice cores: A case study from the Quelccaya Ice Cap, Peru, Geophys. Res. Lett., 30(14), 2. Methodology 1772, doi:10.1029/2003GL017191, 2003. 2.1. Data [5] Annual layer data are from WDC-Glaciology files at 1. Introduction the NOAA National Geophysical Data Center [Thompson, 18 [2] Two deep ice cores from the Quelccaya Ice Cap, Peru 1992]. Individual d O sample values are extracted from (14S, 71 W; 5,670 m) in 1983 were the first obtained in figures in publications by the original analysis team as tropical latitudes and have yielded a 1500-year annually- specified in the text, since access to the original sample resolved climate record [Thompson et al., 1985]. Climatic measurements could not be obtained. interpretations of the Quelccaya ice stratigraphy continue to 2.2. Procedures receive widespread application, yet statistical correlation between the two Quelccaya ice cores has never been pre- [6] The present investigation aims to improve the reso- sented to evaluate the internal consistency of the climate lution of annual climatic signals registered in stratigraphy of signals between the two records (150 m apart). The original the Quelccaya cores by reanalyzing original sample mea- isotopic analyses were based on water samples as the ice surements with a new methodological approach that is could not be retrieved frozen at that time. The subjectively guided by the following logic. Since the ice cores were determined annual-layer parameter averages from both cores drilled only 150 m apart atop a broad, low-relief plateau, are available to the research community, but the individual they must have sampled nearly identical stratigraphic pro- files. It follows that the individual sample measurements of sample measurements used to create these averages have 18 never been released, eliminating opportunities for indepen- d O, obtained at 5-cm intervals along each core must dent assessment and validation of results. Although several therefore represent independent data points of that stratig- other long ice core records have since been obtained from the raphy; they are considered to be reliable in accuracy since tropical , the Quelccaya cores remain highly relevant in analysis was performed independently on each series by presenting annually-resolvable stratigraphy much further different laboratories [Thompson et al., 1989], and the two back in time than any of their counterparts. data series exhibit excellent correspondence. Any differ- [3] The reconstruction of annual climate and diagnosis of ences in published annual layer sample averages between climatic variability is dependent upon properly establishing the two ice cores must therefore result from the combination the representation of climate in measured parameters in of subjective errors in the assignment of annual layers well-dated ice stratigraphy. Despite the clear seasonality in [Thompson et al., 1985], and the limitations of objective precipitation, oxygen isotope (d18O) and dust horizon interval sampling. 2.3. Reanalysis Strategy 18 Copyright 2003 by the American Geophysical Union. [7] This study focuses on d O as a key signifier of 0094-8276/03/2003GL017191$05.00 annual layering since this parameter exhibits both a distinct

HLS 4 -- 1 HLS 4 - 2 SEIMON: QUELCCAYA REANALYSIS annual sinusoid and marked interannual variability in response to climatic processes [Dansgaard, 1964; Vuille et al., 2003]. Annual layer sequences that abruptly decorrelate are studied to identify where subjective errors have resulted in improper chronological alignment of the two cores. After the suggested adjustments are made to reestablish synchro- nicity, absolute dating is performed on the now-linked strata by matching anomalous dust signatures to major events listed in a multi-centennial history of regional . These events are known to trigger major earth movements and cause injection of dust into the atmosphere [e.g. Henderson et al., 1999].

3. Analysis 3.1. Reestablishing Synchronous Profiles [8] Statistical correlation was first performed on annual layer means of d18O from the two Quelccaya cores for the 500 year period 1483–1983 to establish degrees of correspondence. Correlation patterns as a function of time reveal offsets in alignment in several multiyear sequences, confirming that errors must also exist for these years in the dating of the ice core profiles of at least one of the cores (Figure 1). Numerous significant decorrelation and recorrelation points are apparent, each shift indicative of inconsistent annual layer assignment between the cores. Strongest correlation is found for multidecadal periods of the 17th and 19th centuries, rather than in the more recent strata of the 20th century as might be expected. A partic- ularly significant loss of correlation is evident between the Figure 2. (a) Depth profiles of 5-cm interval d18O annual layers ascribed to the years 1764–1805. An oppor- measurements from Quelccaya Summit (thick) and Core 1 tunity to investigate this case further is presented by (thin line) ice cores, for the period ascribed to 1775–1825 examining individual d18O sample measurements ascribed by Thompson et al. [1986; redrawn from their Figure 1]. to 1775–1825, displayed for both ice cores in Thompson et (b) Published annual d18O means for Summit (black) and al. [1986; their Figure 1] to validate their dating methodol- Core 1 (gray) for 1775–1825 [Thompson, 1992]. (c) As in ogy, and redrawn here in Figure 2a. Figure 2b for reanalyzed and time-corrected annual d18O 18 [9] Reanalysis of the d O sample measurements reveals means; (d) 5-yr moving window correlation scores for errors in the original annual layer determination. Correlation Summit and Core 1 annual d18O means shown in Figures 2b between the unadjusted sample series is relatively poor (dashed) and 2c (solid). (R = 0.51), likely owing to slight differences in ice flow characteristics at each site causing the series to drift in and out of phase, which is to be expected (Figure 2a). The offset sustained prior to 1806 (Figure 2b). Strong corre- published annual mean d18O values derived from this data spondence between sample profiles can be reestablished similarly correlate at only R = 0.54, however, due to a 1-yr once slight adjustments are made to bring the data into alignment by simple pattern matching of the annual sinus- oids in d18O. Reassignment of samples to create synchro- nous annual layers of matched d18O sinusoids brings the climatic representation of the data into greatly improved focus (Figure 2c). Correlation improves to R = 0.96 once adjustments are applied (Figure 2d). Since the two d18O profiles sustain consistent behavior, despite high degrees of interannual variation throughout this 50 year segment, they affirm the premise that the cores must display strong correlation since they represent the same stratigraphy. This indicates that the variations exhibited are true climatic signals, rather than glaciological noise. [10] To serve the primary objective, the rendering of clear climate signals, this concept is developed further here by merging the two data series, once alignment has been Figure 1. Correlation scores (5-yr moving window) for properly established, to create a combined ‘‘best’’ data Quelccaya Core 1 and Summit ice cores of published d18O series (Figure 3). This profile can be viewed as one that annual layer means [Thompson, 1992] from A.D 1483– might be found on the Quelccaya Ice Cap in a column 1983. Time advances from right to left. midway between the actual boreholes; it utilizes equal SEIMON: QUELCCAYA REANALYSIS HLS 4 - 3

drift away. This scenario raises the possibility that regional might preserve a history of paleoseismicity by the introduction of distinct dust horizons characterized by an anomalous ratio of large to small particulates relative to that found in dust layers of non-seismic origin. Since sampling in the Quelccaya case was performed at 5-cm intervals only, and was thus not continuous, not all potential earthquake- related events are expected to be represented in both ice core data series. In addition, during the austral summer prevailing easterly flow should tend to advect dust plumes from the most active seismic zone along the Pacific coast away from Quelccaya, diminishing the likelihood for anom- alous dust deposition. Figure 3. Combined time-corrected sample series incor- [14] Despite these limitations, particulate anomalies porating all d18O sample measurements from Quelccaya strongly suggestive of earthquake occurrence are amply Summit core and Core 1 (line), and annual layer means evident in all the individual sample profiles presented in pub- (triangle), derived from all samples between annual d18O licationsof Quelccaya ice core data [Figure 1 in Thompson et maxima labeled for January of hydrological year. Doc- al., 1986; Figures 3–5 in Thompson and Mosley-Thompson, umentary reports identify strong El Nin˜o events in 1803-04 1987; Figure 5 in Thompson and Mosley-Thompson, 1989]. and 1791, and regional in 1783 [Ortlieb, 2000]. These collectively cover 223 of the possible 452 years between the start of historical documentation (beginning with the Spanish conquest in 1532) and ice core drilling in contributions from each core, and thus yields a profile with 1983. At least 26 major earthquake events can be identified twice as many sample points as each of its constituent parts. that appear to fit the stratigraphic signatures of prominent The resulting gains in resolution effectively double the particulate anomalies, and also indicate when time correc- available inference available on climate at annual scales, tions should be applied (Table 1). while disparities and ambiguities are minimized. [15] Interestingly, the A.D. 1570–1645 sequence shown 3.2. Time Calibration by Thompson and Mosley-Thompson [1989] in order to highlight the Huaynaputina eruption actually presents a [11] The second phase of reanalysis concerns fixing the compelling case for shifting their time line forward by annual layering in time. This is critical for studies compar- 4 years. The years 1604-09 featured three large magnitude ing ice core data to that of other high temporal resolution earthquakes in southern Peru, beginning with one of the climate proxies such as tree-ring studies and coral growth strongest events in the historical record on 24 November patterns. Poor correspondence has been found between the 1604 [Dorbath et al., 1990; IGP, 2001]. It follows that the Quelccaya time series and proxies for El Nin˜o events anomalies in large particulates detailed in Thompson et al. [Ortlieb and Machare´, 1993], though can be improved if [1986] might have been largely comprised of ash from the time scale is adjusted for possible missing years Huaynaputina deposited regionwide [da Silva and Zielinski, [Michaelsen and Thompson, 1992]. In this study new 1998], but actually only deposited on the ice cap in associ- techniques are developed that greatly improve the time- ation with the intense seismic activity that commenced four fixing capabilities for the Quelccaya cores, and might also years later. Elsewhere in this sample profile prominent find application in ice core studies elsewhere. anomalies can be tied to major earthquakes in 1582, 1590, [12] The original Quelccaya analysis utilized the A.D. 1600 (attending the Huaynaputina eruption) and 1630; 1600 eruption alone as reference marker to date the ice this fortifies the argument that the original time calibration cores. In contrast, the present investigation draws upon a of the ice cores is offset by several years, even for its multitude of time-reference indicators from a catalog of defining event. This would explain inconsistencies noted significant earthquakes in Peru since the Spanish conquest by other investigators between Quelccaya and other annually in the 16th century [IGP, 2001]. Large earthquakes repre- resolved paleoclimate proxies, and affirms that reanalyzed sent an occasional source for potentially identifiable strati- sample measurements can be absolutely dated for much of graphic markers in ice core data [Henderson et al., 1999]; the first 450 years of the ice core strata. this is especially so in southern Peru, where one of the most [16] In the data series for 1775–1825, discussed above, seismically active zones in the world underlies the Atacama major earthquakes can be fitted precisely to particulate Desert and adjacent arid Andean occidental crest. The earthquake signals in ice core data will be detailed else- where, but its application to this study can be briefly Table 1. Reference Marker Events in Sample Segments described. Number of Events [13] Major earthquakes occur at roughly decadal intervals Change in the central Andean region and generate large debris Core and Years This study Previous Required? clouds from landslides and other earth movements. Under 1. Summit 1928–1947 7 0 No a favorable wind regime these particulates will drift across 2. Summit 1864–1905 5 0 Yes 3. Summit 1775–1825 3 0 Yes the Andes and deposit dust particles upon the surface. The Core 1 1775–1825 4 0 Yes recent origin of disturbance suggests a size separation 4. Summit 1570–1645 7 1 Yes process, whereby more massive particulates will sediment 5. Summit 1452–1550 0a 0 Unknowna to the surface relatively quickly while smaller particulates aCannot yet be determined with data available. HLS 4 - 4 SEIMON: QUELCCAYA REANALYSIS anomalies in the stratigraphic sequence if both time series [19] Acknowledgments. This work is supported by NASA-Earth Sys- are shifted back by one year, in which case earthquakes can tem Science fellowship NGT5-30399. Thanks go to Roger Barry for extensive comments and feedback. Tom Dickinson assisted with preparation be tied to 1821 (Core 1), 1813 (both cores), 1812 (Core 1), of graphics. 1799 (Core 1), 1784 (Summit) and 1777 (Summit). The likelihood that these anomalies can find comparable corre- References spondence with other phenomena seems remote. As a result, Bradley, R. S., M. Vuille, D. R. Hardy, and L. G. Thompson, Low latitude ice cores record Pacific sea surface temperatures, Geophys. Res. Lett., confidence is high that the paleoearthquake reference 30(4), 1174, doi:10.1029/2002GL016546, 2003. 18 markers can establish that the reanalyzed d O sample series Dansgaard, W., Stable isotopes in precipitation, Tellus, 16, 436–468, 1964. presented in Figure 3 is also precisely fixed in time for all of de Silva, S. L., and G. A. Zielinski, Global Influence of the AD 1600 eruption of Huaynaputina, Peru, Nature, 393, 455–458, 1998. the years shown, with inferences on subannual climatic Dorbath, L., A. Cisternas, and Y. Dorbath, Assessment of the size of large variation now feasible for each of these annual layers. and great historical earthquakes in Peru, Bulletin of the Seismological Soc. of America, 80, 551–576, 1990. Henderson, K. A., L. G. Thompson, and P.-N. Lin, Recording of El Nin˜o in 4. Applications ice core d18O records from Nevado Huascara´n, Peru, J. Geophys. Res., 104(D24), 31,053–31,065, 1999. [17] This investigation indicates that individual sample IGP, Instituto Geofı´sico del Peru´, Cata´logo Sı´smico del Peru´ 1471–1982, measurements would more effectively serve as the basis for Centro Nacional de Datos Geofı´sicos-Sismologı´a, Lima, 2001. paleoclimate analysis with this data series, rather than Michaelsen, J., and L. G. Thompson, A comparison of proxy records of El Nin˜o/Southern Oscillation: Monograph, in El Nin˜o: Historical And subjectively determined layer averages made widely avail- Paleoclimatic Aspects Of The Southern Oscillation, edited by H. F. Diaz, able by the original investigators. The results of the align- pp. 323–348, 1992. ment and absolute dating correction exercises for selected Ortlieb, L., The documented historical record of El Nin˜o in Peru, in El Nin˜o and the Southern Oscillation, edited by H. F. Diaz and V. Markgraf, New sample series of the Quelccaya ice cores demonstrate York, Cambridge Univ. Press, 207–295, 2000. considerable gains in accuracy, and therefore in climatic Ortlieb, L., and J. Machare´, Former El Nin˜o Events - Records From Western representation too. The annual sinusoids shown in Figure 3 South-America, Global And Planetary Change, 7, 181–202, 1993. suggest that the reanalyzed data series exhibits strong Thompson, L. G., E. Mosley-Thompson, P. Grootes, and M. Pourchet, Tropical glaciers: potential for ice core paleoclimatic reconstructions, representation of significant El Nin˜o events attended J. Geophys. Res., 89(D3), 4,638–4,646, 1984. by regional , with marked positive anomalies in Thompson, L. G., E. Mosley-Thompson, J. F. Bolzan, and B. R. Koci, A 1500 annual mean d18O similar to that reported from Nevados year record of tropical precipitation in ice cores from the Quelccaya Ice Cap, Peru, Science, 229, 971–973, 1985. Huascara´n at 9S[Henderson et al., 1999] and Sajama at Thompson, L. G., E. Mosley-Thompson, W. Dansgaard, and P. M. Grootes, 18S[Bradley et al., 2003]. There is undoubtedly much The as recorded in the stratigraphy of the tropical Quelccaya potential for further improvement by systematic applica- Ice Cap, Science, 234, 361–364, 1986. Thompson, L. G., and E. Mosley-Thompson, Evidence of abrupt climate tion of these methods to the complete set of Quelccaya change during the last 1500 years recorded in ice cores from the tropical sample measurements. In particular, since the Quelccaya Quelccaya Ice Cap, Peru, in Abrupt , W. H. Berger and stratigraphy is annually resolvable for 1500 years, far L. D. Labeyrie, Reidel, Dordrecht, pp. 99–110, 1987. longer than that of all other low-latitude ice cores, this Thompson, L. G., and E. Mosley-Thompson, One-half millennia of tropical climate variability as recorded in the stratigraphy of the Quelccaya Ice indicates the potential for a thorough reanalysis to reveal a Cap, Peru, in Aspects of Climatic Variability in the Pacific and the western complete millennial history of ENSO. Such a record would Americas, edited by D. H. Peterson, AGU, Washington D.C., pp. 15–31, be suitable for inter-comparison with other paleoclimate 1989. Thompson, L., Quelccaya Ice Core Database, WDC-A Paleoclimatology proxies from within a region that shows strong sensitivity Contribution #92-008. NOAA/NGDC Paleoclimatology Program, to ENSO-related climatic perturbations. Boulder, CO, 1992. [18] In summary, this study finds that discrepancies in the Vuille, M., R. S. Bradley, M. Werner, R. Healy, and F. Keimig, Modeling 18O in precipitation over the tropical Americas, Part I: Interannual two original ice core profiles for Quelccaya were introduced variability and climatic controls, J. Geophys. Res., 108(D6), 4174, by subjective errors, and that the original data available in doi:10.1029/2001JD002038, 2003. individual sample measurements offer great potential to ÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀÀ register a comprehensive millennial climate history at sub- A. Seimon, Dept. of Geography, University of Colorado, Boulder, CO annual resolution. 80302-0260, USA.