Testing and development of earthquake forecasting models

Technical Abstract

The New Zealand Earthquake Forecasting Testing Centre, based at GNS Science, is one of several regional testing centres now being established. The aim of the Centre is to encourage the development of forecasting models for New Zealand earthquakes and to conduct tests of their performance over a period of five or more years. It is intended that the New Zealand testing centre will develop in parallel to other regional testing centres through ongoing international collaboration, with sharing of methods and computer software.

The test region, data-collection region and requirements for testing are described in this report. Statistical tests will measure how closely the expected earthquake occurrence under each model conforms to the actual earthquake catalogue. Short-term forecasting models will be tested over a series of 24-hour time-intervals. Intermediate-term models and long-term models will be tested over a series of three-month and five-year time intervals respectively. Short-term and intermediate-term models may use only the catalogue of past earthquakes to make their forecasts.

Three long-term, two intermediate-term and two short-term forecasting models have so far been installed in the testing centre, with tests commencing at the beginning of 2008. Submission of models is open to researchers worldwide. New models may be submitted at any time.
 

For any model submitted to the testing centre, we can calculate the probability of a given level of ground shaking occurring at any location as a consequence of earthquakes expected under the model. This allows the time-varying hazard indicated by each forecasting model to be compared with existing standard models in which the hazard is constant over time.

This project included several studies which may help us to refine short-term forecasting models. These were concerned with different aspects of earthquake clustering and in particular, foreshock occurrence. A foreshock is an earthquake which is followed soon after by a larger one.
The first study examined the decay of foreshock activity in time and space and the difference in magnitude between foreshocks and main shocks in New Zealand. It identified clear differences in foreshock occurrence between the Taupo Volcanic Zone and the rest of New Zealand. The second study showed that there is little difference in the tendency of foreshocks and main shocks to trigger other earthquakes in both the Californian earthquake catalogue and a world-wide catalogue. The third study examined differences between the catalogue and a world-wide catalogue. The third study examined the differences between the “spontaneous” earthquakes that initiate clusters and the “triggered” earthquakes that follow them. It found that the triggered earthquakes themselves have a greater tendency to trigger other earthquakes, and are more likely to be foreshocks, than the spontaneous earthquakes.