← Back to Sessions

Analytical Approximations for Propagating Epistemic Uncertainty and Modeling Virtual Faults for Areal Sources in Seismic Hazard Analysis

Description: 

The computational requirements in Probabilistic Seismic Hazard Analysis (PSHA) have significantly increased with recent advances in ground-motion models (GMMs) and source characterization. The epistemic uncertainty is usually modeled using logic trees that discretize the distribution of the uncertain input parameter. The total number of end branches is the product of the number of branches on each node which grows rapidly as the complexity of the logic trees grows. Similarly, the aleatory variability for the source characterization for areal source zones can be modeled by filling the zone with virtual faults that sample a distribution of strikes, dips, and locations. The combination of the epistemic branches from the logic tree with a large number of virtual faults leads to long run times. To limit the run times, each node of the logic trees is often limited to just 3 branches, but this does not work well for non-ergodic GMMs. For virtual faults, the discrete samples of the strikes, dips, and locations is often crude to limit the number of cases in the hazard calculation. We propose an alternative approach to compute the hazard using analytical approximations for the distribution of the hazard or distribution of the input parameters for the GMMs (e.g., distance metrics) which can be combined with the full analytical distribution for a given node of the logic tree and the full distribution for the range of geometries for virtural faults. Rather than using exact hazard calcuations with crude sampling of the inputs, we use analytical approximations for the hazard calculation with the full distribution of the inputs. The calculation times can be reduced by one to two orders of magnitude with a small loss of accuracy. For the tails of the epistemic range of the hazard, the loss of accuracy from the analytical approximation is partly offset by using the full distributions rather than the crude discrete samples for each node of the logic trees and the aleatory range of the fault geometries for virtual faults. The analytical approaches make PSHA using non-ergodic GMMs practical on desktop computers.

Session: New Directions in Environmental, Seismic Hazard and Mineral Resource Exploration Studies - II

Type: Oral

Date: 4/17/2025

Presentation Time: 05:15 PM (local time)

Presenting Author: Maxime

Student Presenter: No

Invited Presentation: 

Poster Number:

Authors