Expert elicitation of model parameters for the 2018 National Seismic Hazard Assessment - Summary of workshop, methodology and outcomes

The 2018 National Seismic Hazard Assessment (NSHA18) aims to provide the most up-to-date and comprehensive understanding of seismic hazard in Australia. As such, NSHA18 includes a range of alternative models for characterising seismic sources and ground motions proposed by members of the Australia earthquake hazard community. The final hazard assessment is a weighted combination of alternative models. This report describes the use of a structured expert elicitation methodology (the ‘Classical Model’) to weight the alternative models and presents the complete results of this process. Seismic hazard assessments are inherently uncertain due to the long return periods of damaging earthquakes relative to the time period of human observation. This is especially the case for low-seismicity regions such as Australia. Despite this uncertainty, there is a demand for estimates of seismic hazard to underpin a range of decision making aimed at reducing the impacts of earthquakes to society. In the face of uncertainty, experts will propose alternative models for the distribution of earthquake occurrence in space, time and magnitude (i.e. seismic source characterisation), and how ground shaking is propagated through the crust (i.e. ground motion characterisation). In most cases, there is insufficient data to independently and quantitatively determine a ‘best’ model. Therefore it is unreasonable to expect, or force, experts to agree on a single consensus model. Instead, seismic hazard assessments should capture the variability in expert opinion, while allowing that not all experts are equally adept. Logic trees, with branches representing mutually exclusive models weighted by expert opinion, can be used to model this uncertainty in seismic hazard assessment. The resulting hazard assessment thereby captures the range of plausible uncertainty given current knowledge of earthquake occurrence in Australia. For the NSHA18, experts were invited to contribute peer-reviewed seismic source models for consideration, resulting in 16 seismic source models being proposed. Each of these models requires values to be assigned to uncertain parameters such as the maximum magnitude earthquake expected. Similarly, up to 20 published ground motion models were identified as being appropriate for characterising ground motions for different tectonic regions in Australia. To weight these models, 17 experts in seismic hazard assessment, representative of the collective expertise of the Australian earthquake hazard community, were invited to two workshops held at Geoscience Australia in March 2017. At these workshops, the experts each assigned weights to alternative models representing their degree of belief that a particular model is the ‘true’ model. The experts were calibrated through a series of questions that tested their knowledge of the subject and ability to assess the limits to their knowledge. These workshops resulted in calibrated weights used to parameterise the final seismic source model and ground motion model logic trees for NSHA18. Through use of a structured expert elicitation methodology these weights have been determined in a transparent and reproducible manner drawing on the full depth of expertise and experience within the Australia earthquake hazard community. Such methodologies have application to a range of uncertain problems beyond the case of seismic hazard assessment presented here.