Tsunamis are a wave or series of waves typically triggered by large earthquakes, landslides or volcanic eruptions that are able to travel significant distances inland and cause extensive damage and loss of life. Landslide-generated tsunamis can be characterised by particularly short wave arrival times, often seconds to minutes, and some of the largest recorded wave heights. Given their rapid and sudden onset, the most effective method for reducing loss of life is evacuation of the exposed population. In the case of landslide-generated tsunamis, self-evacuation - evacuation without an official warning - is essential to increase survival rates.

In order to improve the outcomes for evacuation events, pre-event evacuation planning is used. This is particularly necessary in the case of landslide-generated tsunamis as the short arrival times mean there is no time for planning during an event and thus populations need to be equipped with the knowledge for self-evacuation. Evacuation modelling is a particularly useful tool for the planning process as it can highlight potential issues that may arise during an evacuation and identify factors that influence the efficiency of an evacuation event. Agent-based modelling is particularly useful as it can increase understanding of potential evacuation outcomes as well as the different factors that affect these outcomes.

In New Zealand, Milford Sound has previously been shown to have a high risk from landslide- generated tsunami. There is an estimated 44% chance of a landslide-generated tsunami occurring in Milford Sound during the next Alpine Fault rupture, with the potential for a 17 m wave arriving on shore within 2-7 minutes. Given Milford Sound is one of New Zealand’s most visited tourist attractions with an average of >1500 people visiting a day, there is potential for widespread loss of life. Despite this, there has been limited formal evacuation planning undertaken. This research developed an agent-based model to understand how many people can be safely evacuated in Milford Sound for a local source tsunami and explored the factors that may affect these outcomes.

The results suggest that currently no one can be safely evacuated before the shortest wave arrival time. For longer wave arrival times, the number of people safely evacuated varies considerably across a range of different scenarios. At the longest wave arrival time the percentage of people safely evacuated ranges from 0.1% to 5.2%, with the highest percentage occurring in the scenario with the lowest population exposure during a winter night-time scenario.

The primary factors influencing the number of people safely evacuating were the evacuees’ speed of travel and origin locations, and high population exposure resulting in congestion. Increasing the speed of evacuees increases the number of people safely evacuated. The origin locations of evacuees were a major influence, as those evacuating from the Ferry Terminal and the staff accommodation area are located too far away from evacuation points to be able to reach safety in time, even for the slowest waves. Higher levels of population exposure increased the levels of congestion on key evacuation routes, which increased the time taken for evacuees to reach safety. Increasing the capacity of the evacuation locations made little difference to the number of people safely evacuated as in many of the modelled scenarios the capacity was not reached before even the longest wave arrival times.

To improve the evacuation outcomes in Milford Sound, it is recommended to increase the signage, limit the population exposure, and relocate the staff accommodation. Installing signage to inform visitors to Milford Sound of the tsunami risk and evacuation routes would have the potential to influence evacuation numbers by increasing the number of people that would be aware of the necessity to quickly evacuate. Limiting the population exposure can decrease the total time to safety by reducing the amount of congestion that is present on key evacuation routes. The relocation of staff accommodation is in line with previous recommendations that have been made for Milford Sound, as the current location restricts the ability for staff to reach safety.