Testing An Earthquake Early Warning Algorithm For New Zealand

Earthquake early warning systems (EEW) are valuable tools for minimising the impact of seismic events on communities and infrastructure. While some countries, like Japan, Taiwan, South Korea, Mexico, and the United States, have successfully implemented EEW systems, New Zealand has yet to establish a national EEW system. This thesis explores and evaluates a potential EEW algorithm for a national EEW system for New Zealand.

PLUM (Propagation of Local Undamped Motion) is a ground motion based EEW algorithm that offers several advantages over point source and finite fault EEW algorithms due to its robustness, speed, and simplicity. It does not rely on an earthquake source model, making it less computationally expensive. The PLUM algorithm has been successfully implemented in Japan and tested using United States West Coast data.

To evaluate the PLUM algorithm's performance and usability in the New Zealand context, it is necessary to test it using a range of earthquakes that could happen over the next century. Because there is insufficient recorded data from past damaging earthquakes, we used a 220000-year-long physics-based synthetic earthquake catalogue of New Zealand and generated synthetic seismograms using a ground motion simulator. We also analysed the latency of the current New Zealand GeoNet sensor network since the latency reduces the warning time of alerts.

We focused our evaluation on the Canterbury and Wellington regions. The Canterbury region has experienced several damaging earthquakes in the past 15 years, and a dense network of seismic instruments monitors part of it. The Wellington region is affected by a very complex set of faults, is considered very earthquake-prone and has a population centre adjacent to Cook Strait. The algorithm was tested for the Canterbury and Wellington regions in terms of alert correctness and timeliness, algorithm configuration settings, and the impact of some sensor failures.

The results showed that the PLUM algorithm works well in highly populated areas of Canterbury, but not as effectively in some southern parts of Wellington, due to the lack of stations in Cook Strait to provide warnings for earthquakes centred south of Wellington. However, it does not perform well in rural areas of either region because the sensor density is too low, even with an increased warning radius.

For the highly populated areas of Canterbury, PLUM could provide Correct Timely Alerts (warning time > 0 sec) for more than 90% of the expected shakings in a 100-year interval. Sparsely populated regions with a low density of sensors received fewer Correct Timely Alerts. The most populated areas of Wellington could also receive more than 70% Correct Timely Alerts. However, PLUM also generated a large number of precautionary alerts (an alert is received, but only weak shaking is felt), which would need to be addressed. The results suggested an appropriate choice of PLUM warning radius in the New Zealand context would be more than the original 30 km radius designed for the Japan context. The analysis also demonstrates the robustness of the algorithm under simulated sensor or communication breakdowns, particularly in regions with high sensor density.

The thesis shows that PLUM could be a valuable component of a New Zealand EEW system, and the results will be beneficial for designing future earthquake early warning systems for New Zealand. The testing framework in the thesis would also be useful for testing the suitability of earthquake early warning algorithms using synthetic earthquake catalogues.