Search

Interdisciplinary investigations of damaged cultural heritage objects have nowadays become standard practise. Numerous techniques in various fields may generate large amounts of data, difficult to interpret. Machine learning was applied to data collected from samples of a painting to build a predictive model of potential further biodeterioration and consequent damage to the paintings. We used this strategy for a deteriorated 17th century Celje Ceiling, tempera painting in a wooden frame, covering 143 m2 of the ceiling. An interdisciplinary approach, from monitoring microclimatic conditions and wood moisture content, to examinations using UV and VIS photography, analyses of paint layers and decay products by Raman and FTIR spectroscopy, SEM-EDS, and protein binders by ELISA, was used to assess current condition of the painting. Data from 535 painting's samples were subjected to analysis by machine learning methods. Moulding, the main cause of biodeterioration, was found to be strongly influenced by the position of the painting in the room and related microclimatic conditions. The presence of a coating layer, protein binders and pigments, such as goethite, ultramarine and kaolinite, had important role in mould development as well. The painting was mainly damaged by currently non-active, presumably xerophilic Aspergillus species, as determined by microscopy, cultivation and amplicon sequencing. [Display omitted] • Four century-old tempera painting, Celje Cieling, covering 143 m2 of ceiling was analysed using a multidisciplinary approach. • Microclimatic conditions in the chamber and wood were monitored over three years. • Data obtained by painting analyses were subjected to machine learning to create a predictive model of further deterioration. • Microscopy, culture-dependent and -independent methods were used to assess microbial contamination in selected parts of the painting. [ABSTRACT FROM AUTHOR]