Probabilistic seismic hazard analysis for spatially distributed infrastructure

Abstract

SUMMARY Two key issues distinguish probabilistic seismic risk analysis of a lifeline or portfolio of structures from that of a single structure. Regional analysis must consider the correlation among lifeline components or structures in the portfolio, and the larger scope makes it much more computationally demanding. In this paper, we systematically identify and compare alternative methods for regional hazard analysis that can be used as the first part of a computationally efficient regional probabilistic seismic risk analysis that properly considers spatial correlation. Specifically, each method results in a set of probabilistic ground motion maps with associated hazard‐consistent annual occurrence probabilities that together represent the regional hazard. The methods are compared according to how replicable and computationally tractable they are and the extent to which the resulting maps are physically realistic, consistent with the regional hazard and regional spatial correlation, and few in number. On the basis of a conceptual comparison and an empirical comparison for Los Angeles, we recommend a combination of simulation and optimization approaches: (i) Monte Carlo simulation with importance sampling of the earthquake magnitudes to generate a set of probabilistic earthquake scenarios (defined by source and magnitude); (ii) the optimization‐based probabilistic scenario method, a mixed‐integer linear program, to reduce the size of that set; (iii) Monte Carlo simulation to generate a set of probabilistic ground motion maps, varying the number of maps sampled from each earthquake scenario so as to minimize the sampling variance; and (iv) the optimization‐based probabilistic scenario again to reduce the set of probabilistic ground motion maps. Copyright © 2012 John Wiley & Sons, Ltd.

Seismic Hazard and Loss Analysis for Spatially Distributed Infrastructure in Christchurch, New Zealand

“…The approach was recommended in Han and Davidson (2012) based on a comprehensive conceptual and empirical (for Los Angeles, California) comparison of available approaches. Han and Davidson (2012) offers for additional detail. We summarize the method, which we call the Extended OPS method, here.…”

“…In Han and Davidson (2012), this step was used instead of the OPS method ( In Han and Davidson (2012), this step was used instead of the OPS method (…”

“…The sensitivity analysis was done to provide insight into how to set those user-specified parameters to optimize the method across steps. (Note that it is straightforward to include spatial correlation among the residuals, as was done in Han and Davidson (2012), but to be consistent with the NZ NSHM, that was not done here.) 1).…”

Integrating disciplinary contributions to achieve community resilience to natural disasters

“…Earthquake, storm surge flooding, and wind communities share analogous challenges of how to represent the regional hazard in a way that is probabilistic and computationally efficient, while still accounting for spatial correlation that is so important when considering spatially distributed infrastructure. Nevertheless, the three literatures and professional communities are largely distinct and have sought solutions in parallel, using slightly different terminology and different approaches, offering great opportunities to transfer solutions from one hazard domain to the others (Apivatanagul et al 2011;Han and Davidson 2012;Toro et al 2010). More multidisciplinary efforts across hazards could facilitate awareness of when such opportunities exist.…”

Protecting Electricity Networks from Natural Hazards

“…Holistic multi-risk analyses also requires the indirect tangible consequences to be integrated (see e.g., Garcia-Aristizabal et al 2015), but network analysis under such conditions is complex. Optimization-based approaches (e.g., Han and Davidson 2012;Miller and Baker 2015) can be a valid strategy to make such a multi-risk analysis a computationally feasible task. Such approaches often use proxy measures as indirect indicators for network performance, which are usually more tractable measurements (Miller and Baker 2015).…”