Weatherill G. 2022. Impact of directivity on probabilistic seismic hazard calculations in New Zealand. Lower Hutt (NZ): GNS Science. 78 p. (GNS Science report; 2022/01). doi:10.21420/RETZ-D556.
Abstract
Directivity of ground motion in the near-field region of large earthquakes is a well-observed phenomenon whose impacts can be particularly detrimental to structures that sit close to an earthquake rupture. Predictive models of this phenomenon and its amplifying effect on ground motion have been available for more than 20 years, with recent projects such as the Next Generation Attenuation West 2 (NGA-West2) yielding several advanced models for application in probabilistic seismic hazard analysis (PSHA). Despite the availability of such models, their application in national- or regional-scale PSHA has still been limited. Among the reasons for this are the divergence in amplitude and spatial pattern of amplification, the sometimes complex geometrical calculations that are required to implement them and the increased computational demand in the seismic hazard calculation to integrate over the additional uncertainties that these models explicitly capture. In the present study, we incorporate explicit directivity modelling into the current seismic hazard model for New Zealand. This exploratory work is undertaken prior to the finalisation of the forthcoming National Seismic Hazard Model (NSHM) for New Zealand and is intended to understand the feasibility and implications of incorporating explicit models of directivity into the seismic hazard calculations. This report begins with an overview of the theoretical framework for incorporating directivity into PSHA and the considerations for selecting and applying directivity models to seismic hazard analysis. To execute the calculations, the OpenQuake seismic hazard and risk calculation software has been adapted to support the implementation of directivity using a hypocentre randomisation approach, in which uncertainty in the hypocentre location in future ruptures is modelled explicitly within the hazard integral. Using the directivity amplification model of Bayless et al. (2020), we run full PSHA calculations for New Zealand based on the existing seismic hazard model of Stirling et al. (2012). The sensitivity of the resulting seismic hazard for the 475- and 2475-year return periods is shown via difference maps for the whole country and uniform hazard spectra for selected cities. The seismic hazard results show that, when the uncertainty in hypocentre location and the influence of multiple sources close to a site are accounted for, the impacts on seismic hazard at intermediate to long periods are more moderate than the near-fault amplification factor found in the current New Zealand Standard for structural design. The largest increases in hazard can be seen in the western and central South Island. Here, design accelerations increase by as much as 20–30% at spectral periods of 3 s, while, for much of the rest of the country, the changes are less than 5–10% for the return periods of interest in ordinary building design. The explanations for these observations are broken down in detail, and their implications for the next generation NSHM of New Zealand and for future seismic design codes are discussed. Changes in estimated seismic hazard resulting from the incorporation of directivity into the PSHA framework are compared against the deterministically calibrated ‘Near-Fault’ factor incorporated into the 2005 New Zealand seismic design code (NZS1170.5). These suggest that the geographic regions affected most by directivity in PSHA and the regions where the Near-Fault factor should be applied remain broadly consistent. Current models of directivity predict higher amplification over greater distances than that implied by the Near-Fault factor, but the incorporation of uncertainty in hypocentre location and the resultant effects of directivity amplification and de-amplification from different faults coincide to moderate these higher values. Further analysis is necessary before recommending modifications to the existing Near-Fault factors in NZS1170.5, however. (The author)
