Obsidian Sourcing

Obsidian is a volcanic glass that is formed when volcanoes erupt. They are most commonly found within rhyolite flows. Because the eruption process occurs as flesig lava mixes with continental crust, each eruption has a unique elemental signature that can indicate the obsidian's source. Because obsidian was a valuable cutting tool - still valuable enough to be used in hospitals for neurosurgery today - it was widely used by prehistoric societies as both a prestige time and a highly useful tool.

The variation in trace elements, typically rubidium (Rb), strontium (Sr), yttrium (Y), zirconium (Zr), and niobium (Nb) but can also include thorium (Th), iron (Fe), barium (Ba) and manganese (Mn). While it is variation at the part-per-million (ppm) level which distinguishes one from the other, the calibrations of portable x-ray fluorescence devices (XRF) are able to distinguish these even at the sub source level. (Rudamaker et al. 2013). For the instruments to be capable of this level of performance, they need to be properly calibrated with reference standards that ensure that the results are replicable (Speakman and Shackley et al. 2013). The use of quantitative units is critical as there is a vast literature documenting the chemical concentrations of obsidian sources worldwide - the best way to identify provenance is to use the same quantitive units as these historical data.

We've compiled a database of obsidian for the world from the existing literature, which includes over 15,000 individual sources compiled from the measurement of 28,000 source samples. From this, we've built an application that can use a dual-layer fingerprinting algorithm to identify most likely sources.
Stacks Image 5
The program can function in one of two ways - as a Bayesian estimation which uses the lat/long of the archaeological site to inform a prior probability based on proximity or as a direct hypothesis-test using scaled Z-scores to evaluate best fits. Because we built the obsidian database from the literature, we also compiled spatial information about each source as well as descriptions of its local geology and ethnographic history when applicable.
Stacks Image 7
Stacks Image 13
We also integrated traditional chenometrics used in obsidian sourcing, such as principle components analysis (PCA) and elemental ratio plots. In both cases the artifacts will be displayed in a scatterplot form while the data we have for each source can be shown as a ellipse.

We know that a fingerprinting algorithm can only get so far - while it is true that every obsidian source has a unique trace elemental signature, that does not mean that geochemical d
öppelgangers do not exist. We provide all the significance testing and posterior probabilities for all viable sources to allow archaeologists to evaluate the data to make use of their prior knowledge of their sites.
Stacks Image 17
Users can adjust the search to their specifications - using political maps, lat/long boundaries, etc. The model sensitivity can also be adjusted. The model sensitivity is the % allowed variation from all trace elements used in sourcing. This can be as low or high as the user prefers. Typically, a lower model sensitivity prioritizes exact geochemical matches while a higher model sensitivity prioritizes the overall pattern of the variation.

If you want to source obsidian, you can send your samples to PRI where we can do it in house,
or you can email to inquire about using the app and database yourself.