Previous studies have documented a systematic discrepancy between probabilistic seismic hazard maps and historical records of shaking (Stein et al., 2015; Brooks et al., 2016, 2017, 2018, 2019; Salditch et al., 2020; Salditch, 2021). In almost every case, shaking predicted by the hazard map is higher than that historically observed. The consistency of this overprediction for different regions suggests a fundamental issue with the methodology of the maps. It is important to understand this discrepancy, and the contributors to it, because these maps are used worldwide to develop building codes for earthquake resilient structures. Billions of dollars-worth of property - and countless lives - rely on the maps’ accuracy. This dissertation explores four potential contributors to the discrepancy between California seismic hazard maps and the California Historical Intensity Mapping Project (CHIMP): exclusion of site effects in the map, inconsistent minimum magnitudes between map and data, ground-motion intensity conversion equations (GMICE) used to compare the maps (measured in peak ground acceleration or velocity) to data (measured in seismic intensity), and the use of the conservative mean instead of the best-estimate median hazard. It concludes by examining the total reduction of the discrepancy when all effects are considered together. To account for site effects, I incorporated the velocity of seismic waves based on the rock type at the site in models that predict shaking amplification or deamplification which differs from the current maps which use a regional average. Considering site effects increases the maps’ predicted ground motion at most sites and, when considered alone, slightly increases the overprediction. Inconsistent minimum magnitudes between maps and data can arise because most hazard maps assume the effects of earthquakes as small as magnitude 5 would contribute slightly to maximum shaking at a site. However, CHIMP only includes data for earthquakes greater than roughly magnitude 6, which contribute larger shaking. Accordingly, the difference in minimum magnitudes between maps and data could contribute to map overprediction. I modeled the fractional contribution of earthquakes with magnitudes between 5 and 6 in the hazard maps and removed the shaking from the smaller earthquakes. I find that the scaled maps still overpredict, although by a lesser degree than the original maps, and thus some of the overprediction stems from inconsistent minimum magnitudes between map and data. I next considered the effect from CHIMP data and hazard maps being measured in different units, one directly characterizing shaking (intensity) and the other reflecting ground acceleration, and whether the GMICE used to convert between the two contribute to the discrepancy. I found a previously undocumented bias in the equation stemming from an incorrect assumption. New GMICE reduce this bias and substantially improve the discrepancy. Finally, the mean hazard has long been used as a conservative estimate over the median, which may better reflect that observed. I approximated the decrease in hazard from switching from the mean to the median. This switch reduces the overprediction. By incorporating the examined adjustments together, I find that the discrepancy is reduced by about half for a 475-yr return period map and can be fully accounted for in a 2475-yr return period map. Although this thesis focuses geographically on California, the results are expected to be applicable elsewhere, with necessary adjustments (e.g., minimum magnitudes will vary based on the available data in a study region).

Creator

• Gallahue, Molly Margaret

DOI

• https://doi.org/10.21985/n2-vmsq-wc38

Subject

• Geophysics

Language

• en

Alternate Identifier

• etdadmin_upload_979215
• http://dissertations.umi.com/northwestern:16479

Date created

• 2023-01-01

Resource type

• Dissertation

Rights statement

• In Copyright