← Back to Sessions

Why Seismic Hazard Models Appear to Overpredict Historical Shaking Observations: An Intensely Simple Answer

Description: 

To understand how well seismic hazard maps predict actual shaking, we use performance metrics to compare them via hindcasting. Our hindcasting approach uses seismic intensities derived from documented shaking effects, which provide the only source of information for shaking caused by historical earthquakes, and often the most abundant information for instrumentally recorded events. Using intensities allows longer observation periods for map performance evaluation, but has limitations in uncertainty and spatial coverage. Comparison of modern hazard maps and historical intensities also requires conversion between forecasted metrics such as peak ground acceleration (PGA) and intensity. Recent maps for California, Japan, Italy, Nepal, and France appear to consistently overpredict historically observed intensities. However, numerical simulations show that observed shaking is equally likely to be above or below predictions. The consistency of results from independently developed models and datasets in different countries and tectonic settings suggests a possible systematic bias in the hazard models, the observations, or both. Analysis of possible causes shows that much of the discrepancy is due to a subtle issue: the Ground Motion Intensity Conversion Equations (GMICE) equations used to compare the maps with historically observed intensities. Gallahue and Abrahamson (2023) showed that currently used GMICEs introduce a bias when used to convert hazard maps, overestimating the PGA-equivalent intensity value by as much as 1 intensity unit for above average ground motions that often control seismic hazard. Improved GMICE will be important for seismic hazard performance evaluation and ground motion modelling efforts. For mitigating earthquake risk, it is encouraging that much of the apparent overprediction of earthquake hazards results from the conversion equations rather than a systematic effect in the hazard modeling approach. Thus although any given hazard map may overpredict or underpredict shaking due to chance or parameter choice, we find no evidence for underlying systematic problems with hazard mapping.

Session: The 2023 USGS National Seismic Hazard Model and Beyond [Poster Session]

Type: Poster

Date: 5/1/2024

Presentation Time: 08:00 AM (local time)

Presenting Author: Leah

Student Presenter: No

Invited Presentation: 

Authors