摘要:Plenty of analytical methods were introduced to clarify provenance problems of white marble artefacts and
building materials. The goal of all these techniques is to determine individual features of different marble
quarries or at least to cluster them into reasonable groups with localised geographic origin.
The study presented here is a part of a multi-method investigation technique, developed to clarify Western
Anatolian white marble provenance questions. Beside of the instrumental analytical investigations the analysis
of fabric became a key technique in the last years. The quantitative texture analysis (QTA), the combination of
the quantitative fabric analysis and extraction of fractal properties of the calcite grain boundaries was applied
on Western Anatolian white marbles. The advantage of the QTA is that it can be performed on the same thin
section as used for cathodoluminescence microscopy and therefore requires no extra material.
Thin sections of marble samples from different Western Anatolian occurrences were prepared and digitally
enhanced images of these thin sections were processed. Various parameters from the images themselves and
from vectorised contours of grain boundaries were calculated. Based on the distribution of the derived
parameters the rock samples were grouped into distinct categories. These clusters represent different tectonic
and geological units. Having defined the grouping criteria for the rock samples, this categorisation was then
applied to the thin sections of archaeological artefacts determining the supposed provenance. The results were
compared to categorisations based on other methods like isotope geochemistry, trace element analysis, and
cathodoluminescence investigation. These techniques mutually support each other resulting in clearly defined
provenance groups and provide an opportunity to organise them into a decision tree scheme. The decision tree
paves the way towards a logically set of analytical techniques avoiding unnecessary analytical steps: the digital
imaging and feature extraction methods provide quantitative values, therefore parameter intervals can be
defined for different provenance groups.