Rethinking epistemic and aleatory uncertainties for seismic hazard scenarios: A case study of the Lembang and Cimandiri faults in Indonesia

Probabilistic Seismic Hazard Assessment (PSHA) is a widely used tools to evaluate the threat of seismic events in earthquake-prone regions and is particularly useful for engineering decision-making and setting construction design standards. However, outside of these communities the results of PSHA analysis are non-intuitive, particularly for disaster risk managers. In these cases, specific hazard scenarios are often used to demonstrate the potential scale of the hazard challenge. For scenario-based seismic hazard calculations the aleatory uncertainties are traditionally accounted for by calculating multiple realisations of the ground shaking intensity measure for a given ground motion prediction equation (GMPE). Epistemic uncertainties are usually estimated in earthquake scenarios by considering a weighted statistic - usually the mean or median - of two to four GMPEs. In this study we show that this approach usually overestimates the ground shaking for any particular region.

We propose an updated approach where we calculate ground motions using all available GMPEs instead of a subset of equations.  Our GMPE set for the test area in West Java, Indonesia, includes 26 equations relevant for Active Shallow Crust environments. Using the Global Earthquake Model OpenQuake-engine we calculate 1000 realisations of each GMPE, merge the histograms of all realisations for all GMPEs into a single ground motion prediction set for each site location. We show that this histogram approximates a lognormal distribution. We show that the mean or median both overestimate the likely ground motions by over 71% and 37% respectively compared to the maximum of the kernel density estimator, which better represents the peak of the distribution. We apply this new method to investigate the shaking distribution from a number of earthquake rupture scenarios on the Lembang Fault and the Cimandiri Fault and test the impacts of a potential joint rupture across both faults, a situation often deemed to be the worst-case scenario for the region.