Assessing The Impact Of Uncertainties In Seismic Risk Models On Portfolio Loss Estimates

Evaluation of the potential socio-economic losses stemming from natural hazards, such as earthquakes, is crucial for catastrophe risk management. Nowadays, the portfolio loss estimation process is mainly carried out by means of complex probabilistic seismic risk assessment models, whose development involves the estimation of various parameters essential to characterize the seismic hazard of the region, the exposed inventory, as well as its vulnerability to ground shaking. Hence, it is of interest for modellers and users to fully understand the relationships between input and output variables, and of the conditions/assumptions under which they hold. This is critical when the estimates of several model parameters are uncertain and their inadequate treatment can potentially lead to biased loss estimates. In this regard, we dedicated the first part of this work to evaluate the impact that variations in different input parameters may cause to the final risk results through a comprehensive sensitivity analysis conducted on the Iranian region of Isfahan. By systematically modifying the values of a key set of parameters of the probabilistic seismic hazard model and subsequently by calculating their impact on resulting losses, we generated curves linking input-output normalized variations of some common risk metrics. In the second part, the focus turns on the vulnerability model, where we probabilistically assess the impact that the uncertainty in the vulnerability curves selection has on the loss estimates dispersion in our specific exercise. The outcomes of this study can be useful to practitioners and risk managers in the construction and usage of these models, and in supporting the communication about how uncertainties in the hazard and vulnerability components can affect seismic risk estimates for given portfolios.