Investigation of the Effect of Debris-Induced Damage for Constructing Tsunami Fragility Curves for Buildings

Abstract

Catastrophe models quantify potential losses from disasters, and are used in the insurance, disaster-risk management, and engineering industries. Tsunami fragility and vulnerability curves are key components of catastrophe models, providing probabilistic links between Tsunami Intensity Measures (TIMs), damage and loss. Building damage due to tsunamis can occur due to fluid forces or debris impact; two effects which have different implications for building damage levels and failure mechanisms. However, existing fragility functions are generally derived using all available damage data for a location, regardless of whether damage was caused by fluid or debris effects. It is therefore not clear whether the inclusion of debris-induced damage introduces bias in existing functions. Furthermore, when modelling areas likely to be affected by debris (e.g., adjacent to ports), it is not possible to account for this increased likelihood of debris-induced damage using existing functions. This paper proposes a methodology to quantify the effect that debris-induced damage has on fragility and vulnerability function derivation, and subsequent loss estimates. A building-by-building damage dataset from the 2011 Great East Japan Earthquake and Tsunami is used, together with several statistical techniques advanced in the field of fragility analysis. First, buildings are identified which are most likely to have been affected by debris from nearby ‘washed away’ buildings. Fragility functions are then derived incorporating this debris indicator parameter. The debris parameter is shown to be significant for all but the lowest damage state (“minor damage”), and functions which incorporate the debris parameter are shown to have a statistically significant better fit to the observed damage data than models which omit debris information. Finally, for a case study scenario simulated economic loss is compared for estimates from vulnerability functions which do and do not incorporate a debris term. This comparison suggests that biases in loss estimation may be introduced if not explicitly modelling debris. The proposed methodology provides a step towards allowing catastrophe models to more reliably predict the expected damage and losses in areas with increased likelihood of debris, which is of relevance for the engineering, disaster risk-reduction and insurance sectors.