Analysis of Shear Wave Velocity Measurements for Prediction Uncertainties in Southern California
AdvisorLouie, John N.
Geological Sciences and Engineering
StatisticsView Usage Statistics
With the relatively sparse number of direct VS30 measurements in Southern California, researchers have created various proxies to estimate this predictor of ground motion. Utilizing 400 publicly available VS30 site characterization measurements, the goal of this study is to evaluate the data and spatial uncertainty of VS30 predictions. We calculated VSZ for depths of 10, 30, 50 and 100 m as well as Z500, Z1000, Z1500 for the 82 new site measurements presented in this study. For all 400 publicly available site characterizations, we made predictions based on the Southern California Earthquake Center Community Velocity Model Version 4.0 (Magistrale et al., 2000) and a topographic-slope proxy developed by Wald & Allen (2007). The subsets of our data set, "rock" (44 sites), "soil" (118) and "basin" (238) are based on the site class map of Wills et al. (2000). The mean values of basin and soil subset predictions based on the SCEC CVM are within 5% and 15% respectively when compared to measurement means. Less than half of the soil site predictions are within ±20% of measured values. Less than half of the predictions based on topographic slope are within ±20% of the measured values. Through statistical analysis, a log normal distribution best fits the complete dataset. VS30 and spatial standard deviation maps created from natural-log transformed data show a spatial standard deviation or uncertainty of >0.30 natural-log units in the Los Angeles Basin and >0.38 ln units outside the basin in areas where predictions are made. The spatial uncertainty does not reflect the epistemic uncertainty associated with the measured values. Unlike proxies based on geologic correlation of units, and topographic slope proxies, hazard maps created from indicator kriging show NEHRP hazard class E is likely in the Los Angeles Basin. More dense measurements are feasible and can readily decrease the uncertainty of predictions.