Magnetic Fingerprinting of Central Mediterranean Obsidian Source Groups
Abstract
The Central Mediterranean islands of Lipari, Palmarola, Pantelleria, and Sardinia are known to be the only sources of obsidian in that region (Tykot, 2002). The obsidian on each source island has a unique chemical composition (i.e., a fingerprint) that has allowed Tykot (2002) to provenance Neolithic obsidian artifacts from archaeological sites throughout the Central Mediterranean to these four sources. Thus, Tykot (2002) revealed material distribution patterns otherwise inaccessible to archaeologists. An exploratory study by McDougall et al. (1983) demonstrated the potential use of magnetic properties to distinguish Central Mediterranean obsidians. If each obsidian source group has a unique magnetic fingerprint, magnetic provenancing might serve as a quick, inexpensive, and non-destructive alternative to chemical provenancing. Our research continues the work started by McDougall et al.; we set out to characterize the Central Mediterranean obsidians and search for magnetic fingerprints that distinguish them. Our preliminary study of the magnetic properties of 30 geologic obsidian specimens from Sardinia A, B, and C subgroups and the three other Central Mediterranean island sources suggests that each source has a unique combination of magnetic properties that could allow magnetic provenancing of archaeological obsidians. For example, the combination of natural remanent magnetization (NRM; median [med] = 2.46 x 10-4 Am2/kg, interquartile range [IQR] = 9.77 x 10-4 Am2/kg) and magnetic susceptibility (low field, low frequency magnetic susceptibility [χlflf]: med = 27.0 x 10-6 m3/kg, IQR = 75.9 x 10-6 m3/kg) seems to have good potential to discriminate the groups. However, statistical analysis shows that there is significant overlap in the confidence intervals of these variables, a limitation which appears to be partially imposed at this point by the small sample size. Discriminant analysis of several magnetic variables also shows potential to assist in classification (i.e., provenancing) of Central Mediterranean obsidians. We have obtained discriminant functions that correctly classify over 80% of the obsidian specimens in leave-one-out cross validations using as few as three log- transformed magnetic parameters: NRM, χlflf, and saturation isothermal remanent magnetization (SIRM; med = 7.59 x 10-3 Am2/kg, IQR = 2.85 x 10-2 Am2/kg). Backfield coercivity of remanence, median destructive field, percent frequency effect, and saturation coercivity seem to contribute the least to group discrimination, yet their effects are statistically significant. Even though initial tests of discriminant analysis look promising, it should be noted that the small sample sizes and relatively large number of variables used in the discriminant analyses violate the test's assumptions and may make the resulting discriminant functions unusable; this will be addressed in future research. The results from our preliminary study suggest that the Central Mediterranean sources of obsidian might have characteristic and distinguishable magnetic fingerprints. However, many more geologic obsidian specimens must be measured to define the ranges of the source groups' magnetic properties and to further test this hypothesis.
- Publication:
-
AGU Spring Meeting Abstracts
- Pub Date:
- May 2009
- Bibcode:
- 2009AGUSMGP11I..04W
- Keywords:
-
- 1540 Rock and mineral magnetism;
- 1594 Instruments and techniques