Reliable assessment of seismic site class using stochastic approaches

Seismic site classification plays a vital role in quantification of earthquake hazard based on weighted-average shear wave velocity of overlying 30 m soil deposit (V-s30). National Earthquake Hazards Reduction Program states six different seismic site classes, and corresponding limits of V-s30 are pre-assigned. While acquiring shear wave velocity (V-s) through different geotechnical engineering test practices, varying degrees of uncertainty mainly linked to inherent, measurement and transformation uncertainty are naturally adhered into the V-s30 estimation process. These types of uncertainties induce wide variability in the concluding estimate and proliferate the risk of safe and sound decision-making process. This present study aims to address the aforementioned hurdle by implementing stochastic approaches in order to reliably assess seismic site class and hence seismic site response. An approximate method, i.e., First-Order Reliability Method (FORM), and two simulation approaches, i.e., Monte Carlo Simulation and Importance Sampling, have been deployed in this study to obtain the reliable seismic site class which has not been attempted earlier. Field procured measurements from high strain soil test, i.e., Standard Penetration Test, have been utilized to estimate the V-s using multiple well-established transformation models ultimately yielding 324 V-s30 outcomes. This study highlights the data scatter at every step of routine geotechnical investigations commencing from field measurement to data transformation and bestows a novel methodology to reliably assess the seismic site class. Further, statistical variation of the mean V-s30 value has also been performed to depict the trend of stochastic outcomes in response to the extent of accounted uncertainties. This novel study of stochastically quantifying a seismic site class aids to cater the implications of uncertainties and helps to obtain a reliable seismic site response.

Automated summary

This study addresses the uncertainty problem in seismic site classification by implementing stochastic approaches. An approximate method and two simulation approaches are utilized to obtain reliable seismic site classes. Field measurements from high strain soil tests are used to estimate the necessary parameters, resulting in 324 seismic site class outcomes. The study highlights the data scatter in routine geotechnical investigations and proposes a novel methodology for reliable seismic site classification.