Quaternary Geochronology 10 (2012) 359e366
Contents lists available at SciVerse ScienceDirect
Quaternary Geochronology journal homepage: www.elsevier.com/locate/quageo
Research paper
OSL dating in multi-strata Tel: Megiddo (Israel) as a case study Naomi Porat a, *, Geoff A.T. Duller b, Helen M. Roberts b, Eli Piasetzky c, Israel Finkelstein c a
Geological Survey of Israel, 30 Malkhe Israel St., Jerusalem 95501, Israel Institute of Geography and Earth Sciences, Aberystwyth University, Aberystwyth SY23 3DB, UK c Tel Aviv University, Tel Aviv 69978, Israel b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 13 October 2011 Received in revised form 24 November 2011 Accepted 28 November 2011 Available online 3 December 2011
Megiddo, one of the most important mounds (Tel) in the Levant, was inhabited almost continuously from the 7th millennium to the 4th C. BC and archaeological remains have accumulated to a height of ca. 20 m. Megiddo features a significant number of destruction levels, some of which can be correlated to wellknown historical events. Other destruction levels are less well dated, and in order to improve the chronological control, an OSL dating campaign was designed, particularly for those periods where the radiocarbon calibration curve incorporates large errors on radiocarbon dates. Twenty-six samples were collected from a range of archaeological periods, excavation areas and sediment types. In-situ gamma and cosmic dose rates were obtained either with Al2O3:C dosimeters that were buried at the site for 2 months or with a calibrated gamma scintillator. Very-fine-sand quartz was extracted and measured using conventional SAR to obtain the equivalent dose (De). The OSL age of many samples is older than the expected archaeological age and their De values are usually scattered. This suggests that sediments were continuously reused and recycled at Tel Megiddo without exposure to sunlight and very little fresh sediment was added directly from dust to the archaeological accumulation, challenging the basic requirement for OSL dating. Using combined criteria of sequential stratigraphic order of the samples and the over-dispersion of the measured De values helped to reject the samples that yield ages which fail to represent the age of their archeological context. Twelve of the 26 OSL ages had to be rejected, but the 14 ages which did pass the criteria agree very well with the expected archaeological ages. Thus analysis of a single sample is ineffective for determining an archeological age for a given context. Sediments from in-between building stones are more suitable than those taken from floors, streets and ash layers; samples from destruction layers should be avoided. Megiddo provides an example of the difficulties in OSL dating in a multi-period, complex archaeological site. Ó 2011 Elsevier B.V. All rights reserved.
Keywords: OSL Dating Megiddo Israel Archeology
1. Introduction The ancient Tel (archaeological mound) of Megiddo overlooks the fertile Jezreel Valley in northern Israel. Furnished with abundant water supply, Megiddo is strategically located on the main highway of the Ancient Near East, which led from Egypt to Mesopotamia. It is mentioned in the Bible and in texts of all great civilizations of the neighboring lands e Egypt, Assyria and Hatti. The mound, which covers an area of 11 ha, was excavated by three expeditions in the past and by a Tel Aviv University-led team in recent years (http://megiddo.tau.ac.il/index.html). It features remains of 30 settlements, starting in Neolithic times in the 7th
* Corresponding author. Tel.: þ972 2 5314298; fax: þ972 2 5380688. E-mail address:
[email protected] (N. Porat). 1871-1014/$ e see front matter Ó 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.quageo.2011.11.011
millennium BC and ending in the Persian period in the 4th century BC, that created a ca. 20 m accumulation of debris. Megiddo provides evidence for the first urbanization process in the Levant in the late 4th millennium BC. In the Bronze Age it was the hub of a city-state which dominated the fertile valley and in the Iron Age it served as an administrative center for the Northern Kingdom of Israel. Megiddo features many elaborate monuments e palaces, temples, fortifications and a sophisticated water-system. We chose Tel Megiddo for this case study for three reasons. Firstly, the absolute chronology of Megiddo, based on historical records, stratigraphy, ceramic typology, and radiocarbon dates, is well-established; that is, one can compare the results determined by optically stimulated luminescence (OSL) to secure historical and radiometric dates. Secondly, Megiddo features an unparalleled number of layers with well-defined relative chronology, i.e. one can retrieve samples for OSL dating from secure contexts, over a range
360
N. Porat et al. / Quaternary Geochronology 10 (2012) 359e366
of thousands of years. Finally, this study is part of an extensive interdisciplinary, comparative dating project that was initiated at Megiddo (http://www-nuclear.tau.ac.il/megiddo-2010/). Shortlived samples from secure contexts in 12 strata have so far been radiocarbon-dated and several layers have been sampled for archaeo-magnetism (e.g. Shaar et al., 2011), rehydroxilation (Wilson et al., 2009) and OSL dating. The project is intended to promote the new dating methods and decrease uncertainties, as well as to facilitate as-accurate-as-possible dating of layers that cannot be dated by 14C because of lack of organic material, or the accuracy of 14C is hampered because of the nature of its calibration curve, which produces large errors. The latter is true, for instance, for the Iron Age IIB-C, in the 8th to 6th centuries BC e a period of time of great historical importance in the Levant. Megiddo was built on a chalky hill near a spring, and is surrounded on most sides by alluvial plains. Quartz necessary for OSL dating is found in minor quantities in all types of sediments at the site. The source of this quartz is aeolian (Yaalon and Ganor, 1979; Dayan et al., 2008) and it was deposited directly on the site as primary air fall. At the time of deposition this quartz would be expected to have been well-bleached and is therefore suitable for OSL dating. Other means by which quartz could have been brought to the site is as construction material from nearby. For example, sun-baked bricks were made from the alluvial sediments surrounding Megiddo, which contain aeolian quartz that has been washed from the nearby hills. These bricks form a substantial component of the Tel sediments, particularly of destruction layers. The assumption that will be tested is that either the fires that ravaged the site at the time of destruction, or exposure to enough sunlight before new constructions took place, were enough to fully heat or bleach the quartz and reset the clock for OSL dating. For the OSL ages of these samples to match the age of the archeological context, resetting needed to have taken place at each given destruction layer or sediment deposit. If this did not happen, or only partially happened, then the OSL age will be incorrect, normally yielding ages much older than the archaeological context.
2. Methods Several archaeological contexts were targeted for OSL dating (Table 1): a) Destruction layers with ashes and mud-brick deposits which originated from collapsed walls. b) Well bedded sediments in streets, courtyards and house floors, found near ovens (within ash debris layers) or deposited by water in small ponds. These probably settled as dust directly on the site and were incorporated into the sediments. c) Sediment that accumulated directly from dust in the interstices between and under building stones soon after construction. A total of 26 samples were collected by drilling into the excavated sections with a 100 diameter drill bit and placing the sediment in light-tight bags under cover, to prevent any exposure to sunlight. A complementary sediment sample for dosimetry was collected from each point. Only very-fine-sand (75e125 mm) quartz was found in the sediment samples and this was extracted for OSL dating using routine laboratory procedures (Porat, 2007). Briefly, after sieving for the selected grain size, carbonates were dissolved by soaking the samples overnight in 10% HCl solution followed by concentrated H2O2 to dissolve organic matter. The rinsed and dried samples were then passed through a Frantz magnetic separator to remove heavy minerals, insoluble carbonates and most feldspars. Etching in 42% HF for 40 min removed any remaining feldspars and the outer 15 mm of the quartz grains, followed by soaking in 16% HCl overnight to dissolve any fluorides which may have precipitated.
2.1. Dose rates The concentrations of U, Th and K, measured by ICP-MS (for U, Th) or ICP-AES (for K), were used to calculate the alpha and beta dose rates. Each sediment sample was measured twice, and the averaged concentrations were used. Uncertainties of 3%, 5% and 10% for K, U and Th, respectively, were derived from comparison to international standards and repeated measurements. For about half of the samples the gamma and cosmic dose rates were obtained with Al2O3:C dosimeters (Burbidge and Duller, 2003) that were buried for 2 months at the holes drilled for sampling. Prior to deployment the dosimeters were exposed to sunlight behind a glass window. They were then wrapped in black plastic to prevent inadvertent exposure to daylight, and then sealed in 1.5 mm thick copper tube to shield them from beta particles. For the other half of the samples, the gamma and cosmic dose rates were measured in the field using a calibrated 200 diameter portable gamma scintillator, after the sampling holes were widened. 2.1.1. Cosmic dose Tel Megiddo was abandoned in the 4th C BC, when it attained its final height. Excavations since the beginning of the 20th century have lowered the surface of the Tel and created large pits and baulks. Therefore the cosmic dose fraction of the measured dose rate was separated from the gamma component by subtracting the expected cosmic dose calculated from current burial depth, using values from Prescott and Hutton (1994). The cosmic dose was subsequently adjusted to the historic burial depth prior to the beginning of excavations, as assessed from excavation reports. Using this historic burial depth and the archaeological age of each sample, the cosmic dose was further modeled for a continuous sediment accumulation prior to the 4th century BC and a constant burial depth since then. The largest correction, for the oldest and deepest sample, increased the total dose rate by 2%. This adjusted and modeled cosmic dose rate was used for age calculations, with estimated uncertainty of 8%. 2.1.2. Moisture contents Reliable estimates of the time-averaged moisture contents in the sampled sediments is crucial for age calculations, as an increase of 5% in moisture contents increases the age by roughly 4%. In Megiddo some excavation baulks had been exposed for more than a year prior to sampling, and the sediment was dry, whereas others baulks were sampled during the excavation season, with moisture contents approaching natural conditions. Moisture measurements of these relatively fresh samples ranged from 15% to 25%. The latter maximum value represents deeply buried floors in early summer. As most samples were buried under more than 2 m of sediments, a depth at which there is little seasonal variation in moisture or temperature, a 20 5% moisture content was used for these; 15 4% and 10 3% moisture contents were used for samples buried at 1e2 m and 20% (see Table 2 for all O-D values). Error bars were removed for clarity. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
800 700 600 500 400 300 300
400
500
600
700
800
900 1000
In-situ gamma+cosmic (µGy/a) Fig. 2. Comparison between the combined gamma and cosmic dose rate measured in the field and that calculated from the radioactive elements and burial depth. Red triangles are samples measured using a portable gamma scintillator and blue squares were measured using Al2O3:C dosimeters. Sample MGD-11 is circled. The 1:1 line is solid, and the dashed lines show the 10% range. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
holes (10% of the grains, as suggested by Rodnight et al. (2006), does provide the expected De for MGD-7 (in this case the second component), but for MGD-12 the De is underestimated. Thus micro-aliquot measurements using the single grain system and the ensuing statistical analyses do not clearly resolve such thoroughly mixed young and older grains, and cannot detect archaeologicallymeaningful ages. 3.3. Performance of OSL ages When the OSL ages are compared with the expected archaeological ages and their uncertainties, which are based on stratigraphy, typology of pottery and radiocarbon dates, only 14 of the 26 OSL ages fall within 10% of the expected archaeological ages (Fig. 3a). The remaining ages are all overestimated, some by several thousand years, and for many samples not even one aliquot provides the expected De. Ages calculated for the highly scattered
samples using the minimum age model do not capture the correct archaeological age (Table 2). The fact that the OSL ages are biased toward being older than the known archeological ages and that some of the ages are overestimated by many centuries, cannot be explained by imprecise field gamma measurements or other uncertainties concerning dose rate estimates. That such one-sided bias in the ages exists suggests that perhaps the major cause of overestimation must be associated with the presence in the sediment of quartz grains which had not undergone resetting at the time they were deposited in the sampled layer. This conclusion is also implied by the single grain (micro-aliquot) results. Samples collected from living floors and courtyards perform no better than samples from destruction layers; however the two samples collected from gaps within stone walls (MGD-6 & MGD-11) are in good agreement with the expected age (Table 1). One of these samples (MGD-11) is the oldest, and its OSL age agrees remarkably well with the expected archaeological age (Table 1). This type of setting e dust that had accumulated between stone building blocks soon after construction e is apparently highly suitable for OSL dating (e.g. Porat et al., 2006), however this is not common at Megiddo.
3.4. Unbleached quartz and sediment reuse A major source for unbleached grains could be sun-dried mud bricks. These were made from the alluvial plain sediments or from older destruction layers, and could contain old, unbleached grains. As they were not fired, only dried in the sun, only a very small fraction of the quartz grains was fully exposed to the sun and bleached. Fallen mud bricks from walls form a significant component of the destruction layers and cannot always be discerned from ordinary rubble. If bioturbation by burrowing rodents had occurred it would have translocated bleached grains from the surface into buried, older sections and would bias the ages toward lower De values. Except for sample MGD-6, which is situated exceptionally close to the surface (Table 2), this is not observed at Tel Megiddo. In fact, several samples that were collected from well-bedded and wellpreserved sediments, which do not show any evidence for layer mixing or disturbances, are overestimated. It appears that some periods at Tel Megiddo are more prone to containing older sediments than others. The Iron Age II samples (Table 1) deviate from the expected age on average by 62% and the Late Bronze samples by 34%, whereas the Iron Age I and the Early Bronze Age samples deviate by only 17 and 5%, respectively. This could indicate that in certain periods (more than in others)
Fig. 4. Radial plots showing the micro-aliquot (single grain holes) results and data modeling for samples MGD-7 and MGD-12. The upper shaded region in each panel indicates the De range 1s expected from the archaeological age; and the lower shaded region indicates the MAM De.
N. Porat et al. / Quaternary Geochronology 10 (2012) 359e366
and poorly bleached grains are mixed. Those samples with intermediate O-D (10e20%) but with matching ages usually contain a few aliquots with very high De values; once these are removed the calculated age conforms to expected archaeological age (Fig. 1e). It is more difficult to explain why samples with low over-dispersion do not agree with the expected archeological age. Possibly these sediments are derived from older bricks that when reused were not exposed at all to sunlight. Thus the O-D value is not sufficient for correct quality identification and further selection is needed.
6 5 12 13
4 28 3
Samples in ran k stratigraphic order
14 2 23 24
22 21 1 29 10 9 31
8 7 30 15 27
26 25 11
2000
3000
4000
5000
6000
7000
365
8000
OSL age (years before 2010) Fig. 5. All samples arranged in stratigraphic order, from the oldest to the youngest archaeological levels (bottom to top on vertical axis), with the OSL ages on the horizontal axis. Full, red circles indicate samples that, within errors, do not agree with the stratigraphic order. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)
older material from underlying rubble was extensively reused for construction.
3.5.2. Sequential order enforced according to stratigraphy The 26 samples can be organized in the expected chronological order (Table 1), from the oldest (MGD-11) to the youngest (MGD-6). This order is based on archeological considerations, that is, on stratigraphy and relative ceramic chronology. In the same excavation area (same baulks) it is just superposition in the local sections. Between areas the standard archeological tools of stratigraphy and pottery typology are used. It should be emphasized that for this kind of ordering of the samples, the known absolute ages of their archeological context are not needed and were not used. The ordering of the samples according to relative chronology allows rejecting results that do not follow the sequence. Fig. 5 demonstrates the procedure and the results: The samples are arranged and plotted in sequential rank order according to their relative stratigraphic position, from the oldest to the youngest (from bottom to top on the vertical axis); at the same time the OSL ages are given on the horizontal axis. Each sample is stratigraphically younger than the one below it and its OSL age should be equal or younger. Samples which are older than the previous ones by more than the 1s errors (of both samples) are rejected, until the next lower or the same OSL age is encountered. The results show that the OSL ages of twelve samples out of the measured 26 samples are not in stratigraphic order. Out of the five samples which have large O-D values (above 20%) four are not in stratigraphic order (Table 2). Possibly each parameter cannot be used on its own for screening samples but their combination increases our confidence in the ages when no age control is available. When comparing Fig. 3a to b, which show the OSL ages against the expected archaeological ages with their archaeological, historical and/or 14C uncertainties, it appears that the sequential order forcing captures exceptionally well all the samples with correct OSL ages, whereas criteria based on the O-D values are less successful. The straight dotted line in Fig. 3a is a linear fit to the data with c2 per degree of freedom of 0.98 and a slope of 1.11 0.1, which support the agreement claim.
3.5. Identifying reliable samples 4. Conclusions To separate the OSL ages which correspond to the archeological age of their context from samples which yielded overestimated OSL ages, two approaches were used, one based on laboratorymeasured parameters and the other on field archeology. 3.5.1. Over-dispersion (O-D) This parameter is an indication of the scatter within the sample beyond that which would be expected from experimental uncertainties. It defines intra-sample scatter, and high values indicate a high scatter among the individual De measurements. Fig. 3b distinguishes between samples with low (20%, seven samples shown as blue circles, two not shown) O-D values. Many samples with concordant ages have low O-D values, whereas most overestimated samples are also over-dispersed (Table 2 and Fig. 3b), suggesting that the two parameters are associated and that the major cause of age overestimation is scatter within the samples, whereby well bleached
The OSL results suggest that sediments were continuously reused and recycled at Tel Megiddo. There was not enough time for certain sediments to be exposed to sunlight, as they were rapidly incorporated into other building materials or as fills for constructions. Very little fresh sediment was added directly from dust to the archaeological accumulation, but apparently at some periods this is more prevalent than in others. The constant recycling and reuse of old or collapsed material challenges the basic requirement for OSL dating and therefore makes analysis of a single sample ineffective for determining an archeological age for a given context. The fine grain size of the quartz meant that between 3 and 5 grains were measured together in micro-holes; using a single grain holder with holes of w150 mm diameter may help to resolve the few wellbleached grains in these samples. Samples collected from in-between building stones perform well, whereas those taken from floors, streets and ash layers are less suitable; samples from destruction layers should be avoided.
366
N. Porat et al. / Quaternary Geochronology 10 (2012) 359e366
Using combined criteria of stratigraphic order of the samples and the over-dispersion of the measured De values can help reject the samples that yield ages which fail to represent the age of their archeological context. At Megiddo, 12 of the 26 OSL ages have to be rejected on the basis of these criteria, but the 14 ages which did pass the criteria agree very well with the expected archaeological ages. Acknowledgements This project was partially funded by the Megiddo expedition, Institute of Archaeology, Tel Aviv University. We thank R. Madmon for assistance in field work, Z. Dolgin for sample preparation, and D. Shtober, O. Yoffe and S. Ehrlich for chemical analyses. Editorial handling by: R. Grun References Bailey, R.M., Arnold, L.J., 2006. Statistical modelling of single grain quartz De distributions and an assessment of procedures for estimating burial dose. Quaternary Science Reviews 25, 2475e2502. Brennan, B.J., Schwarcz, H.P., Rink, W.J., 1997. Simulation of the gamma radiation field in lumpy environments. Radiation Measurements 27, 299e305. Burbidge, C.I., Duller, G.A.T., 2003. Combined gamma and beta dosimetry, using Al2O3:C, for in situ measurements on a sequence of archaeological deposits. Radiation Measurements 37, 285e291. Dayan, U., Ziv, B., Shoob, T., Enzel, Y., 2008. Suspended dust over south-eastern Mediterranean and its relation to atmospheric circulations. International Journal of Climatology 28, 915e924. Duller, G.A.T., 2008. Single-grain optical dating of Quaternary sediments: why aliquot size matters in luminescence dating. Boreas 37, 589e612. Galbraith, R.F., Roberts, R.G., Laslett, G.M., Yoshida, H., Olley, J.M., 1999. Optical dating of single and multiple grains of quartz from Jinmium rock shelter,
northern Australia. Part I: experimental design and statistical models. Archaeometry 41, 339e364. Lepper, K., Agersnap Larsen, N., McKeever, S.W.S., 2000. Equivalent dose distribution analysis of Holocene eolian and fluvial quartz sands from central Oklahoma. Radiation Measurements 32, 603e608. Mayya, Y.S., Morthekai, P., Murari, M.K., Singhvi, A.K., 2006. Towards quantifying beta microdosimetric effects in single-grain quartz dose distribution. Radiation Measurements 41, 1032e1039. Olley, J., Caitcheon, G., Murray, A.S., 1998. The distribution of apparent dose as determined by optically stimulated luminescence in small aliquots of fluvial quartz: implications for dating young sediments. Quaternary Geochronology 17, 1033e1040. Olley, J.M., Pietsch, T., Roberts, R.G., 2004. Optical dating of Holocene sediments from a variety of geomorphic settings using single grains of quartz. Geomorphology 60, 337e358. Porat, N., 2007. Analytical procedures in the luminescence dating laboratory (in Hebrew), Technical Report TR-GSI/2/2002, Geol. Survey of Israel, Jerusalem, 33 pp. Porat, N., Rosen, S.A., Avni, Y., Boaretto, E., 2006. Dating the Ramat Saharonim Late Neolithic Desert Cult site. Journal of Archaeological Sciences 33, 1341e1355. Prescott, J.R., Hutton, J.T., 1994. Cosmic ray contributions to dose rates for luminescence and ESR dating: large depths and long-term time variations. Radiation Measurements 23, 497e500. Rodnight, H., Duller, G.A.T., Wintle, A.G., Tooth, S., 2006. Assessing the reproducibility and accuracy of optical dating of fluvial deposits. Quaternary Geochronology 1, 109e120. Shaar, R., Ben-Yosef, E., Ron, H., Tauxe, L., Agnon, A., Kessel, R., 2011. Geomagnetic field intensity: how high can it get? How fast can it change? Constraints from Iron Age copper slag. Earth and Planetary Science Letters 301, 297e306. Wilson, M.A., Carter, M.A., Hall, C., Hoff, W.D., Ince, C., Savage, S.D., Mckay, B., Betts, I.M., 2009. Dating fired-clay ceramics using long-term power law rehydroxylation kinetics. Proceedings of the Royal Society A 465, 2407e2415. Wintle, A.G., Murray, A.S., 2006. A review of quartz optically stimulated luminescence characteristics and their relevance in single aliquot regeneration dating protocols. Radiation Measurements 41, 369e391. Yaalon, D.H., Ganor, E., 1979. East Mediterranean trajectories of dust-carrying storms from the Sahara and Sinai. In: Morales, C. (Ed.), Saharan Dust. John Wiley and Sons, pp. 187e193.
