Provenance Determination of Vinica Terra Cotta Icons Using Self-Organising Maps

Institut za hemija, PMF, Univerzitet Sv. Kiril i Metodij, Arhimedova 5, 1001 Skopje, Republic of Macedonia.
Annali di Chimica (Impact Factor: 0.99). 06/2007; 97(7):541-52. DOI: 10.1002/adic.200790036
Source: PubMed


In the Vinica Fortress, Republic of Macedonia, 50 undamaged terra cotta icons and 100 fragments, all dated 6th-7th century, were found. In order to determine the provenance of these unique terra cotta icons, the mass fractions of 19 different chemical elements were previously determined in ten fragments of the terra cotta icons and thirty three samples of clays from eight different sites from the region. Due to the dimensionality and complexity of the experimental data, the archaeometric results were treated with self-organising maps (SOM). The results obtained using SOM were compared with the ones obtained using principal component analysis. Both chemometric methods revealed that Vinica terra cotta icons were made from clay from Grncarka, 2.5 km South-East from the Vinica Fortress.

Download full-text


Available from: Orhideja Grupce
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: In the last decades several analytical techniques have been used in archaeological ceramics studies. However, instrumental neutron activation analysis, INAA, employing gamma-ray spectrometry seems to be the most suitable technique because it is a simple analytical method in its purely instrumental form. The purpose of this work was to determine the concentration of Ce, Co, Cr, Cs, Eu, Fe, Hf, K, La, Lu, Na, Nd, Rb, Sb, Sc, Sm, Ta, Tb, Th, U, Yb, and Zn in 160 original marajoara ceramic fragments by INAA. Marajoara ceramics culture was sophisticated and well developed. This culture reached its peak during the V and XIV centuries in Marajó Island located on the Amazon River delta area in Brazil. The purpose of the quantitative data was to identify compositionally homogeneous groups within the database. Having this in mind, the data set was first converted to base-10 logarithms to compensate for the differences in magnitude between major elements and trace elements, and also to yield a closer to normal distribution for several trace elements. After that, the data were analyzed using the Mahalanobis distance and using the lambda Wilks as critical value to identify the outliers. The similarities among the samples were studied by means of cluster analysis, principal components analysis and discriminant analysis. Additional confirmation of these groups was made by using elemental concentration bivariate plots. The results showed that there were two very well defined groups in the data set. In addition, the database was studied using artificial neural network with unsupervised learning strategy known as self-organizing maps to classify the marajoara ceramics. The experiments carried out showed that self-organizing maps artificial neural network is capable of discriminating ceramic fragments like multivariate statistical methods, and, again the results showed that the database was formed by two groups.
    Full-text · Article ·
  • Source

    Full-text · Book · Jan 2011
  • Source

    Full-text · Book · Jan 2011
Show more