Validation of ground motion simulations using precarious rocks

Abstract

Southern California Earthquake Center (SCEC) have developed the Broadband Platform (BBP) that computes 0-100 Hz seismograms for historical and scenario earthquakes using several ground motion simulation methods. Simulating an earthquake involves the development of a source, whereby values are assigned to a model and simulations are run utilising physics-based ground motion simulation methods. Ground motion simulations on the SCEC BBP are developed for a major reverse fault in central Otago, New Zealand, using ancient precariously-balanced rocks (PBRs; unstably balanced rocks on top of pedestals) to validate the simulations, rather than the standard approach of using instrumental strong motion records for validation. The Dunstan Fault, a 60 km long reverse fault, is responsible for the uplift of the Dunstan Mountains (1500-1600 m). PBRs are abundant within a few km of the southwestern end of the fault, and are therefore conveniently located for validating simulated peak ground acceleration (PGA) from M >7 near-field Dunstan Fault earthquakes. The fragility age (age since the PBR reached the present unstable morphology), and fragility (the PGA required to topple the PBR, based on field-based estimates), are compared to the recurrence interval and simulated ground motions of Dunstan Fault earthquakes. Earlier studies show cosmogenic Beryllium 10 (Be10) exposure dates for two PBRs are in the range of 40,400 to 55,300 years B.P., and the Dunstan Fault to show a recurrence interval of about 8000 years. Therefore, the PBRs have likely experienced repeated large earthquakes where ground-motions did not exceed their fragilities (i.e. PGAs no greater than c. 700 cm/s/s). The PBR fragilities fall within the range of PGAs produced by the simulations (160-1330 cm/s/s), with about 16% of them exceeding the highest fragility. Decreasing kappa from the default value of 0.04 to a value more representative of Otago (0.016) results in an overall increase of simulated PGAs by about 30%, resulting in 66% of the simulations plotting above the highest PBR fragility (700 cm/s/s). This suggests there are parameters that lead to predicted PGAs that exceed the PBR fragilities. A combination of effects can explain the over-prediction. Factors such as hanging wall effects, fault dip, directivity, kappa and site amplification are concluded to have large influences on ground motions. The over-prediction in rock PGA is concluded to be due to one or more of these factors that are simulating unrealistic conditions for a real Dunstan Fault earthquake. This research represents the first effort at using PBRs to validate ground-motion simulations in New Zealand, and has been jointly supported by University of Otago, QuakeCoRE, and GNS Science.