Comment on "Seismic Hazard Analysis for the UK: Sensitivity to Spatial Seismicity Modelling and Ground Motion Prediction Equations" by K. Goda, W. P. Aspinall, and C. A. Taylor

Abstract

In the course of a recent paper on seismic‐hazard analysis issues for the United Kingdom by Goda et al. (2013), the authors apply model tests derived from Musson and Winter (2012) to the seismic source‐zone model for the United Kingdom used to inform the United Kingdom national annex to the Eurocode 8 building code (Musson and Sargeant, 2007; hereinafter: BGS2007), and also to two spatially smoothed approaches, and report better results from the latter. They do not mention that similar tests on the BGS2007 model are actually reported in BGS2007. Given that I have been advocating testing of seismic‐hazard source models for over twelve years, it would be a strange thing if the BGS2007 model had been released without such testing. BGS2007 describes how the initial model performed poorly when tested, and was modified in the course of the project to give acceptable results, an example of how testing can be used as part of an iterative process to improve model performance. The tests made by BGS2007 were made on the basis of comparing simulated earthquake catalogs derived from the model with the historical catalog. The principle (Musson, 2000, 2004) is that it should be credible that the historical catalog is a member of the set of possible catalogs derived from the model, which is something that can be checked statistically. The tests in BGS2007 were made against the number of M w 4.5 events observed in 300 years, a dataset believed to be complete. The value of M w 4.5 was used in the test in BGS2007 because this was the lower‐bound magnitude used in the hazard calculations, and therefore defined the hazard‐relevant seismicity. Goda et al. (2013) tested on events above M w 4.0 in 260 years, also a complete dataset according to BGS2007. It would be …