Development of tests for long-term earthquake ground motion forecasts in New Zealand

Technical Abstract

Earthquake hazards analysis uses past behaviour to forecast future behaviour. The methodology of probabilistic earthquake hazard analysis, which estimates the strength of ground shaking that might occur in a future time period, is complex. Therefore, we need to find ways of verifying the results of an analysis so we can be confident in using the results for practical applications such as building design. In our study we develop a method to test the output of the New Zealand probabilistic seismic hazard (PSH) model. The general testing method uses the historical earthquake record to count up the number of times strong shaking has impacted 26 New Zealand cities and towns (sites), and compares that observed shaking to the how many times the shaking is predicted to occur by the New Zealand PSH model in the same time period. One actual test uses the 169 year historical record of felt intensities (intensity being shaking described in terms of human experience) as the observed record, and the other test uses the measured record of shaking from seismic instruments over the last 20 years or so. The first test is a long, but uncertain record of the strength of shaking, and the other is a short but instrumentally-measured record of shaking. The first test suggests that the New Zealand PSH model does predict the right number of events. The results are different to those of earlier, preliminary studies which suggested the PSH model was predicting too many strong shaking events. In future efforts to formally install these tests within a Centre for formal verification of earthquake hazard models we will evaluate which of our two methods is the most reliable.