Source of the Stone Used to Build a French Cathedral

Mediterraneum Meeting, Multivariate Analysis and Chemometry Applied to Environment and Cultural Heritage, 2nd ed., Ventotene Island, Italy, Europe, 1-4 June 2008

Feature Selection Applied to Study the Source of the Stone Used to Build a French Cathedral

C. Pizarro, C. Sáenz-González, S. Rodríguez-Tecedor, J. M. González-Sáiz
Department of Chemistry, University of La Rioja, C/ Madre de Dios 51, 26006 Logroño, Spain

The source of the stone used to build the “Nôtre-Dame de Senlis” cathedral in Senlis (Northern France) has been determined by using compositional data taken from the Limestone Sculpture Provenance Project database.
The application of stepwise linear discriminant analysis (SLDA) has enabled to select, with the classification algorithm, the most discriminant variables between five nearby limestone quarries around Senlis.
Once the quarries has been discriminated, SIMCA has been applied in order to obtain a class model for each quarry in an attempt to assign the samples from “Nôtre-Dame de Senlis” to one of the five categories considered.

Knowing the source of the stone used to build a monument can help identify the different building campaigns and show patterns of exchange in stone [1]. Moreover, it could permit to discriminate the original construction from any restorations or additions.
Differences in the microscopic structure of stones from the same stone formation are often undetectable. Therefore, petrography is not capable of discriminating stone sources within the same sedimentary facie and new analytical methods are needed [2]. On the other hand, the composition of the stone is not constant in the whole stone formation and discrimination among different sources is possible by means of compositional fingerprinting.
The Limestone Sculpture Provenance Project [3] has built a database that includes compositional analysis of samples taken from sculptures in museum collections, quarries and monuments, carried out by Neutron Activation Analysis (NAA). Since applications based on NAA coupled with multivariate analysis techniques seem to show great promise for provenance studies[1] and given the great improvement in the classification results obtained with the application of a variable selection algorithm[4], we decided to perform a more indepth study to test the feature selection to compositional data from limestone.
However, taking into account the compositional variation of both the “object” under study and the quarries as well as the errors in the analytical procedure, results only can be shown as statistical probabilities [1].

Materials & Methods
The Limestone Sculpture Provenance Project database was used as source of the compositional data used in this study. The compositional data, obtained by NAA, comprises the concentrations of twenty five elements (Na, K, Rb, Cs, Ba, Sc, La, Ce, Eu, Lu, Hf, Th, Ta,

Author: Ghg_admin
Date: 10/01/17