J. Geomag. Geoelectr., 38, 1339-1348, 1986 Paleointensities of the Geomagnetic Field Obtained from Pre-Inca Potsherds near Cajamarca, Northern Peru (Received December 28, 1985; Revised May 12, 1986) The changes in geomagnetic field intensity in Peru in the last 1500 years was investigated by the Thelliers' method applied to ceramics excavated near Cajamarca, northern Peru. The age interval spanned by these samples is from 1500 B. C. to 1300 A. D., which is subdivided into seven different cultural periods. Ages of the individual periods were determined by combining C-14 data, stratigraphy, and the stylistic evolution of the ceramics in the pre-inca cultures of northern Peru. Thirteen good and 7 acceptable results were obtained among the 24 samples subjected to the experiments. The results are in broad agreement with the ones already reported for the period 1500 B. C. -500 A. D., and a maximum at about 700 B. C. and a minimum at about 0 A. D. are well defined by the present data. The large scatter in the previous results are concluded to be due to the errors in the assignment of ages to the ceramics. The importance is stressed of using materials with well defined ages for archeomagnetic experiments. 1. Introduction and Description of Samples Variation of the geomagnetic paleointensity in the last few thousand years in Peru has already been studied by NAGATA et al. (1965), KITAZAWA and KOBAYASHI (1968), GAMES (1977), and GUNN and MURRAY (1980), but there appears to be a large scatter among the reported data, which is not very typical of the paleointensity data from well studied regions such as Europe or Japan. This discrepancy seems to be caused, in large part, by inappropriate age assignments for the materials used in experiments. In archeologically well studied areas, the ages of the materials are assigned based on well established time scales obtained from the combination of independent evidences such as old documents, cultural style, and radiometric ages, and there is little possibility that quoted ages are seriously in error if they are younger than a few thousand years. In Peru and other South American countries, the situation is somewhat different. Archeomagnetic samples (potsherds, bricks, etc.) from Peru were usually obtained from experts working in archeology or anthrolopology, and ages (mostly carbon-14 based data) for the samples were often quoted from these sources. C-14 ages are obtained on materials like carbonized wood, grains, bones, or shell found in the same site as were the potsherds. However, it is not uncommon that one excavation
M. KONO et al. locality contains several different cultural levels, and mixtures of potsherds belonging to different cultures or ages are often found even in a single horizon. Even if C-14 ages were obtained for the materials from the same horizon, they may not be sufficient by themselves because the age of potsherds in a horizon may differ from that of the organic material from which a data was obtained. It is also a common situation that C-14 ages contain much larger errors than the statistical errors quoted as the uncertainties, because of non-ideal effects such as leaching of organic components, weathering, and so on. Except GUNN and MURRAY (1980) who determined the ages of the ceramics by the thermoluminescence method, the former authors quoted dates of such nature for their samples. Therefore, some of the assigned C-14 ages may be erroneous and so may not be good time measures for the samples. For the problems in calibration of C-14 ages to calendar years, the readers are referred to KLEIN et al. (1982). To circumvent these difficulties, we used samples of potsherds which have reasonably well determined ages. They are from pre-inca ruins of Cajamarca, northern Peru, where the Tokyo University Team has been continuing excavations for many years (TERADA and ONUKI, 1982, 1985). About 100,000 artifacts were unearthed at Huacaloma excavation site, the southeastern suburb of Cajamarca. The location of the site and the vertical section of the layers in the excavated trench are shown in Fig, 1. This site contains several layers of different cultures in succession, for which ages were assigned by the Tokyo University Team in the following manner (TERADA and ONUKI, 1982, 1985). The stratigraphic section, which is only slightly confused by slumps, clearly tells us the order of cultural evolution at this site. The total period was therefore divided into several cultural stages based on stratigraphy and comparison with other archeological sites in northern Peru. Fig. 1. Simplified chlonological section of the southern half of the Trench Eat Huacaloma excavation site, near Cajamarca, northern Peru. No vertical exaggeration. Inset shows the location of the site. (after TERADA and ONUKI, 1982).
Paleointensities of the Geomagnetic Field Obtained from Pre-Inca Potsherds A number of C-14 age data were obtained for these cultural stages (Table 1). In most cases, C-14 age data from the same layer are consistent (e. g., Initial Cajamarca or Late Huacaloma in Table 1), but quite scattered ages were obtained in others (e. g., Layzon). In some cases, the scatter of ages for the same layer is real and reflects the fact that one cultural period spans a finite length of time. However, there must be errors when the ages in adjacent periods overlap. It is noteworthy that the statistical error attached to a single C-14 age does not give a good measure of the actual error in the assigned age. Besides C-14 ages and the stratigraphic positions, the styles of ceramics and other materials were extensively studied based on the comparison of cultural development in northern Peru. Best estimate ages (Table 1) for individual cultural stages were determined by combining these three methods: stratigraphy, C-14 dates, and evolution in the style of ceramics. The last of these methods is quite useful and also Table 1. Carbon 14 ages for Cajamarca excavation site. * These three dates are not considered representative, because of the disturbance of the layer containing carbonized material. ** Dates from Layzon site, about 3km west of Huacaloma site.
M. KONO et al. important, since ceramics of a certain age are often found in layers belonging to later stages because of slumping, reuse of old materials by later people, and other causes. The time covered in Huacaloma excavation site is 1500 B. C. to 1300 A. D. and is divided into six (or seven) cultural stages or periods (Table 1); Early Huacaloma (assigned period of 1500 B. C. to 1000 B. C.), Late Huacaloma (1000 B. C. to 500 B. C.), Layzon (500 B. C. to 200 B. C.), Initial Cajamarca (200 B. C. to 200 A. D.), Early Cajamarca (200 A. D. to 600 A. D.), and Middle Cajamarca (600 A. D. to 1300 A. D.). Initial Cajamarca period is subdivided into Phase A (older) and B (younger) based on stratigraphy (TERADA and ONUKI, 1982, p. 255), though C-14 ages were not discriminated. 2. Paleointensity Experiments and Results Paleointensity experiments were carried out by the Thelliers' method (THELLIER and THELLIER, 1959) modified by COE (1967) and by Shaw's method (SHAW, 1974). We shall only report the results of the Thelliers' method in the present paper. Experimental procedure for the Thelliers' method is as follows. After the measurement of the natural remanent magnetization (NRM), samples were heated the Curie temperature was reached. Heatings were made twice to the same temperature, first in a nonmagnetic space (residual field less than 200nT) and then in Paleointensity estimates were obtained from the linear relation between the NRM and TRM (thermoremanent magnetization) components on the Arai diagrams, using least squares method where errors in both coordinates are taken into account (KONO and TANAKA, 1984). The magnetic minerals in all the samples were Ti-poor titanomagnetite with reversible heating and cooling curves (Fig. 2 (a)) in thermomagnetic analysis, and even when irreversibility was observed, the change due to heating was not severe (Fig. 2 (b)). In the course of heat treatments, the directions of remanence did not change directional changes in remanence are shown in Fig. 3. From these results, these samples seem to be well suited for paleointensity experiments. Examples of successful paleointensity experiments are given in Fig. 4. Only the Arai diagrams of samples with standard errors (in the slope of the fitted line) smaller included or excluded from the regression analysis. The latter points are affected by viscous remanence (VRM) at low temperatures or alteration due to high temperature oxidation near the Curie point. In one group of samples heated together as one set, the from the linear trend (e. g., HL1342-A in Fig. 4). It was found by examining the recorder chart that the assigned temperature was not reached in the second run due perhaps to the setting error. These points were accordingly excluded from linear regression analysis. In calculating the slope (and therefore the paleointensity), it was made sure that six or more points belonged to the linear portion, and that this portion covered at least
Paleointensities of the Geomagnetic Field Obtained from Pre-Inca Potsherds Fig. 2. Thermomagnetic curves of typical samples. The applied field was 0.55T and the experiments were level for (a) is indicated by a horizontal line. (a) Nearly reversible case. (b) Slightly irreversible case. a third of the total length of the fitted straight line between the intercepts at the NRM and TRM axes. Linear regression was done following the method of KONO and TANAKA (1984). Data with large standard errors were not rejected at this stage, but that does not affect the result very much because multiple data were available for each period and data with large errors were given a small weighting (see below). Thus the slope of the NRM-TRM relation was determined for 21 out of 24 samples subjected to the experiment (Table 2). The quality of the calculated paleointensities differ greatly in Table 2. The standard error (SF in Table 2) was calculated for each datum to give the reliability of the paleointensity estimate. This value combines the effects of both the variance in linear regression and that in quadratic regression to represent the non-ideal, nonlinear behavior (KONO and TANAKA, 1984). The mean paleointensity for an individual cultural period was calculated using the inverse square of the standard error (1/S2F) as the weight of individual data (Table 3). In doing so, all the data in Table 2 were included regardless of their standard errors. Because small weights were attached to data with large standard errors, there was no significant change if they
M. KONO et al. Fig. 3. Examples of changes in the direction of the NRM component in the course of the paleointensity experiments. N indicates the NRM direction, while the small numerals show the temperature in degrees Celsius. Remanence directions are arbitrarily placed for convenience, and only the relative changes are significant. Equal area projection. 3. Discussion and Conclusions The Thelliers' method was applied to 24 potsherd samples from pre-inca ruins of were obtained. The high rate of success is perhaps related to good stability in both the magnetic and chemical sense. As the ferromagnetic minerals in these samples are Ti-poor titanomagnetites (see Fig. 2), the original firing of these ceramics should have Samples used in the present study came from different fragments of broken pottery bearing different patterns, even when sample names are quite similar, such as HL521-A, B, C, and E. Nevertheless, most of the data in one period agree with each other quite well. (Data such as HL1341-B are not exceptions if we consider the large
Paleointensities of the Geomagnetic Field Obtained from Pre-Inca Potsherds Fig. 4. Examples of Arai diagrams for Thelliers' method. Intensities of both the NRM and TRM components are normalized by the NRM magnitudes of individual samples. Points are arbitrarily displaced in the vertical direction. Zero level for the NRM component of each sample is indicated by horizontal line to the right. standard errors they have). This is probably because the age assignments were appropriate. Only the samples of Middle Cajamarca period show a large scatter among the three values obtained. Although the number of data is not very large, at least two of them are quite well defined (Table 2). It seems, therefore, that the difference among them is real. There are two possibilities to account for this large scatter of data. One is errors in age assignments for some of these samples and the other is that a large amplitude fluctuation of geomagnetic intensity really occurred in this period. In this respect, it is interesting that GUNN and MURRAY (1980) reported a fluctuation Figure 5 compares the present results with previous data reported as "good" by the original authors. Our data are broadly consistent with data already reported for the time interval 1500 B. C. to 400 A. D. From 1500 B. C., the intensity increased to a minimum at about 0 A. D. seems well defined by the present data. As the standard
M. KONO et al. Table 2. Results of the Thelliers' method. NRM: Intensity of NRM; T1, T2: Temperature interval in which NRM-TRM relation is linear; N: number of points in the linear interval; b: gradient of NRM-TRM relation; F: paleointensity; SF: standard error calculated by the method of KoNo and TANAKA (1984). Table 3. Summary of the paleointensity results. n0: Number of samples subjected to Thelliers' method; n: Number of paleointensity data; F: Mean paleointensity for the period; s. e.: Standard error of the mean paleointensity.
Paleointensities of the Geomagnetic Field Obtained from Pre-Inca Potsherds Fig. 5. "Good" paleointensity data reported by previous authors (NAGATA et al., 1965; KITAZAWA and KOBAYASHI, 1968; GAMES, 1977; GUNN and MURRAY, 1980), and the present data (open circles with error bars). Our data are shown as the mean and standard errors for each period. For the present data and for those by Gunn and Murray, the age scale is in calender years. error for each period is small, small amplitude fluctuation can also be distinguished. We may therefore conclude that the use of well dated materials for paleointensity determination will clarify the real changes in the geomagnetic field. In concluding, we must emphasize the importance of proper age assignments in archeomagnetic studies. The large scatter in the previous "good" data between 0 and 1600 A. D. (Fig. 5) may, in part, be attributed to large errors in the ages. The success we had in the present study owes much to the good age control for our samples. It should always be kept in mind that the standard deviations attached to C-14 ages are only statistical errors, and not a true measure of the actual errors in the age data (cf. Table 1). This paper is a contribution from "Geophysical Studies of the Central Andes" carried out in 1980-1982 with grants from Ministry of Education, Science and Culture (Grant Nos. 504204, 56041015, 57043017). We thank Ryozo Matsumoto for the assignment of ages and for discussion. We are also grateful to Martin Aitken for helpful comments on the manuscript. Paleomagnetic measurements were done at the paleomagnetic laboratory of the University of Tokyo, for which we are indebted to Minoru Ozima and Toshiyuki Tosha. REFERENCES COE, R. S., Paleo-intensities of the earth's magnetic field determined from Tertiary and Quaternary rocks, J. Geophys. Res., 72, 3247-3269, 1967. GAMES, K. P., The magnitude of the paleomagnetic field: a new non-thermal, non-detrial method using sun-dried bricks, Geophys. J. Roy. Astron. Soc., 48, 315-329, 1977. GUNN, N. M. and A. S. MURRAY, Geomagnetic field magnitude variations in Peru derived from archaeological ceramics dated by thermoluminescence, Geophys. J. Roy. Astron. Soc., 62, 345-366, 1980.
M. KONO et al. KITAZAWA, K. and K. KOBAYASHI, Intensity variations of the geomagnetic field during the past 4,000 years in South America, J. Geomag. Geoelectr., 20, 7-19, 1968. KLEIN, J., J. C. LERMAN, P. E. DAMON, and E. K. RALPH, Calibration of radiocarbon dates, Radiocarbon, 24, 103-150, 1982. KONO, M. and H. TANAKA, Analysis of the Thelliers' method of paleointensity determination, 1. Estimation of statistical errors, J. Geomag. Geoelectr., 36, 267-284, 1984. NAGATA, T., K. KOBAYASHI, and E. J. SCHWARZ, Archaeomagnetic intensity studies of South and Central America, J. Geomag. Geoelectr., 17, 399-405, 1965. SHAW, J., A new method of determining the magnitude of the paleomagnetic field, application to five historic lavas and five archaeological samples, Geophys. J. Roy Astron. Soc., 39, 133-141, 1974. TERADA, K, and Y. ONUKI (editors), Excavations at Huacaloma in the Cajamarca Valley, Peru, 1979, 351pp., Univ. of Tokyo Press, Tokyo, 1982. TERADA, K. and Y. ONUKI (editors), The Formative Period in the Cajamarca Basin, Peru: Excavations at Huacaloma and Layzon, 1982, 345pp., Univ. of Tokyo Press, Tokyo, 1985. THELLIER, E. and O. THELLIER, Sur l'intensite du champ magnetique terrestre dans le passe historique et geologique, Ann. Geophys., 15, 285-378, 1959.