Uncertainty range in probabilistic seismic risk metrics resulting from multiple hazard models

Quite frequently, catastrophes impact densely populated areas of the world, and hence mitigation and proper management requires risk evaluations. Because of the uncertain nature of extreme natural hazards and the lack of data, it becomes necessary to forecast the potential damage and losses before the event happens. Therefore, CAT models build on scenarios that represent all possible realizations of the hazard in terms of frequency and intensity. Probabilistic risk models require the characterization of such hazards, the exposure on infrastructure in the evaluated region, and the vulnerability of these infrastructures. The main objective of this research is to compare Loss Exceedance Curves (LECs), Probable Maximum Loss curves (PMLs), and Average Annual Losses (AAL) using four different available seismic hazard models for Chile, denoted hereafter as GAR15, BID, CIGIDEN, and ASLAC. To isolate the effect of changing the hazard model in the risk results, the exposure and vulnerability information is fixed to the one available from the Global Assessment Report, GAR 15 and GAR ATLAS 2017. Results show differences due to the variability on the information used in each model i.e., historical data, estimation of model parameters, the mathematical structure of the model, the value of model inputs and scale of transformation. Imprecise probability theory, logic trees, and frequency and severity blends used by CAT modelers are the approaches applied and compared herein to propose either model blending, or an interval of possible realizations. While uncertainty in risk estimations can be considerable, they need to be understood only as a representation of reality, and hence, uncertainty is just a characteristic needed to deal with correctly in communications and decision making. Moreover, results from models set a risk benchmark used in decisions consistent with that risk.