Using Cluster Analysis to Optimize Tsunami Evacuation Zones

Using Cluster Analysis to Optimize Tsunami Evacuation Zones William Power, Biljana Luković GNS Science, Lower Hutt, New Zealand

New Zealand tsunami sources Distant/Regional Earthquakes Local Earthquakes Background Figures from: Integrated Tsunami Database for Pacific GNS Science

Source: Mw 9. 1 Southern Peru (1868) Tsunami warnings • Divide the coast into zones • Assign a threat level for each zone, based on maximum predicted water level • Example is based on shipping forecast zones – not optimised for tsunami Tsunami threat levels GNS Science

Source: Mw 9. 1 Southern Peru (1868) Tsunami threat levels GNS Science

GNS Science

Max water level Fault 2 The Basic idea Fault 1 Cluster 2 Cluster 1 Fault 2 Max water level Fault 1 GNS Science

Source locations GNS Science

GNS Science

Colours indicate clusters GNS Science

GNS Science

GNS Science

Problem • The standard algorithms for computing clusters do not require the members of the cluster to be contiguous GNS Science

No relationship between separated clusters of the same colour GNS Science

Conclusions • New Zealand is exposed to tsunami from many directions • Different parts of the coast are more/less susceptible to different source regions • In a warning system based around zones it is beneficial if the coast within each zone has a similar pattern of susceptibility • Cluster Analysis is one route for classifying stretches of coast according to their susceptibility to different sources • A drawback of conventional cluster analysis is that it does not constrain the clusters to be contiguous around the coast • Approaches to adding the contiguity constraint are possible, but more work is required GNS Science

Acknowledgments • NOAA – use of MOST and FACTS • Diana Greenslade (BOM) – discussions about warning zones • David Rhoades (GNS) – discussions about statistical analysis GNS Science