Back analysis for estimating the larger seismic actions applied to an ancient Greek dry stone monument in the last millennium

This paper presents an investigation of the response to past earthquakes of the facade colonnade of the classical temple of Apollo Epikourios, Peloponnese, Greece, which is still standing, damaged but in full length and in most of its height. The investigation was based on field studies and sophisticated FE simulations and was carried out in three stages. Initially, it was determined which major damage of the colonnade must be attributed solely to earthquake actions. Afterwards, the efficiency of the selected for use numerical tool (Abaqus) was examined by numerically reproducing a seismic experiment on a colonnade consisted of three columns capped with four epistyle (architrave) blocks. The results were quite satisfactory, although it appeared that the numerical predictions underestimate the magnitude of the dynamic behaviour of the colonnade; this finding was taken into account during the evaluation of the results of the last part of the work. Within the third stage, a detailed numerical model was created, reproducing the façade colonnade of Apollo temple in the state which (according to field study results and reasonable assumptions) it was preserved in the period of the 13th–14th century AD when several blocks collapsed (possibly simultaneously) from its architrave. The model was excited with records of six earthquakes with different characteristics, scaled with gradually increasing factors. The numerical results showed that a simultaneous toppling of most of the architrave blocks in question could have happened as the result of an earthquake with a PGA around 0.60 g and predominate period around 0.6 s, or the result of few (3–4) earthquakes of similar frequency content and PGA around 0.40–0.55 g. Therefore, the relative conclusions of the field studies were strengthened, the maximum level of shaking occurred at the region of the monument at least during the last millennium was estimated, and the selected numerical tool and the adopted modelling parameters were further verified.