Objective assessment of source models for seismic hazard studies: with a worked example from UK data

Abstract

Up to now, the search for increased reliability in probabilistic seismic hazard analysis (PSHA) has concentrated on ways of assessing expert opinion and subjective judgement. Although in some areas of PSHA subjective opinion is unavoidable, there is a danger that assessment procedures and review methods contribute further subjective judgements on top of those already elicited. It is helpful to find techniques for objectively assessing seismic source models that show what the interpretations physically mean in terms of seismicity. Experience shows that well-meaning but flawed design decisions can lead to source models that are incompatible with the seismic history that was used as input. In this paper a method is demonstrated in which large numbers of synthetic earthquake catalogues, that match the completeness thresholds of the historical catalogue, are generated. The study area can be divided into a grid of uniform cells, and the number of earthquakes in each cell in both the historical catalogue and each simulated catalogue are then counted. Comparison of the historical pattern and a set of 1,000 simulated patterns, using a X² test, shows if the historical pattern is credibly a member of the set of outcomes obtainable from the seismic source model. A second method is to chart the distribution of a large sample of simulated catalogues in terms of magnitude frequency, and observe whether the historical catalogue is comfortably within this distribution, or an outlier. If it proves impossible to replicate the historical catalogue using the model, it casts doubt on whether the model is a valid depiction of the seismicity rates that will govern the future hazard. At the very least, the disparity needs careful investigation to ensure the model is error-free. A worked example is presented here for the UK, using a source model that was used in Global Seismic Hazard Map (GSHAP), compared to one that was artificially constructed to be credible but flawed. Two tests find the GSHAP model to be an acceptable representation of the pattern of seismicity in the UK, while the artificial model is conclusively rejected.