Open AccessCase study of geochemical clustering as a tool for tracing sources of clays for archaeological and modern bricks
Author(s) -
Ričardas Taraškevičius,
Vaidotas Kazakauskas,
Saulius Sarcevičius,
Rimantė Zinkutė,
Sergej Suzdalev
Publication year - 2019
Publication title -
baltica
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.242
H-Index - 17
eISSN - 1648-858X
pISSN - 0067-3064
DOI - 10.5200/baltica.2019.2.2
Subject(s) - dendrogram , geology , cluster analysis , lithology , mineralogy , hierarchical clustering , sampling (signal processing) , geochemistry , archaeology , computer science , geography , artificial intelligence , population , demography , filter (signal processing) , sociology , genetic diversity , computer vision
The study was conducted using 14 hierarchical clustering ways and combining them with 4 inter-related sets of elements, i.e. the contents of Al, Ba, Ca, Fe, Ga, K, Mg, Mn, Na, Nb, P, Rb, Si, Sr, Th and Ti determined by EDXRF in 44 splits of 10 archaeological and 4 modern bricks, as well as in 38 clay samples from two quarries. Empirical scoring of tree dendrograms of archaeological samples helped to identify Complete Linkage, Weighted Pair Group Average and Ward’s methods as the most suitable for sourcing. Successful identification of geochemical clustering methods for fingerprinting sources of bricks is determined by the intrinsic features of the geochemical composition of stonework or quarries: their similarity, determined by the geochemical peculiarities of clay indicators, such as Al, Rb, Ga, K, Th, Fe, Ti and Nb, and differences, expressed by the elements that are more abundant in sands and silts (Si, Na), carbonates (Ca, Sr, Mg), organic matter (P) and other lithological-mineralogical tracers. It has been found that the mean values of the geochemical composition of the allied objects have much more useful fingerprinting properties. It is strongly recommended for source fingerprinting to select not only typical lithological-mineralogical samples, but also homogeneous sampling sets excluding possible outliers. It has been shown that each raw clay material has its own specific geochemical features. This is an essential useful feature for source fingerprinting using clustering of the objects of interest.