Best-case scenario: 5% survive Milford Sound tsunami

A recent Master’s thesis has found a Milford Sound tsunami – triggered by a landslide – may leave no survivors, with as many as 3500 dying if the wave hits during the peak of the tourist season.

The new modelling, which builds on more than a decade of research, shows the best chance of survival relies on people running for higher ground before the shaking stops. 

The best-case scenario shows 5.2 percent of people would survive the wave, and in this case the tsunami would have to hit at night, during the winter offseason, when only a few hundred people would be in the area.

Though researchers are working to raise awareness and tourism operators in the area know about the potential for disaster, there is a lack of broader public awareness. And there is currently no tsunami warning signage in Milford Sound.

Some 40 percent of the base of Milford Sound is covered in landslide deposits, most of which were caused by large earthquakes. If one were to strike the fiord, as the Alpine Fault is expected to within our lifetimes, there is a 44 percent chance it will trigger a landslide into the fiord and thus a tsunami. If this happens, there is very little anyone on the ground could do to escape, for a number of reasons: the wave would arrive too quickly, safe areas are unmarked or too difficult to access, and public awareness that a tsunami in Milford Sound is even possible is scant.

Across all modelled options in a recent Master’s thesis by University of Canterbury student Livvy Harris, the best-case scenario was a survival rate of 5.2 percent. 

This data has built on years of research into Milford Sound’s hazardscape, and though the general public is largely unaware of the risks, the scientific community is working with local government and businesses to prepare as tourism rates in the area continue to climb.  

At the same time, groups across the country are preparing for Rū Whenua: a simulated Alpine Fault response taking place this Wednesday. It’s a chance to test the nation’s capacity to respond and work out any kinks in the system, but it’s unclear if this scenario will be included in the plan. 

Milford Sound is Aotearoa’s most iconic tourist draw, but in a country dominated by faults and volcanoes, it could also turn out to be the most risky. The fiord’s capacity for tsunami generation has been known for at least a decade; in 2012 and 2014 a pair of PhD theses were completed at the University of Canterbury, both of which addressed this specific hazard. One of these, by Jesse Dykstra, focused entirely on “mass wasting” in the fiord: large slips of land calving from the sheer valley walls. 

Dykstra’s thesis identified more than 30 landslide deposits in the area over the last 18,000 years. 22 of these landed in the fiord, producing waves with an average height of 17 metres – a number still used in simulation modelling today. Dykstra’s work attributed much of the mass wasting to seismicity, primarily because of the influence of the Alpine Fault, which runs for 600km along the Southern Alps before heading offshore at the mouth of Milford Sound. 

Since these initial findings, two things have developed in tandem: research into the fiord’s tsunami potential and development of Milford Sound tourism, focused on the Cleddau Delta. Together, these two factors have raised the overall risk to a level that would be unacceptable in comparable countries such as Norway.

At the time of Dykstra’s thesis, Milford Sound hosted about half a million tourists annually and the scientific consensus was that the Alpine Fault had a 30 percent chance of rupturing in the next 50 years. This led Dykstra to calculate that an individual who spends four hours in the fiord has about a one-in-a-million chance of being there for a tsunami, a number he said was “about the same” as the acceptable level in other countries.

But that 50-year likelihood has more recently been refined to 75 percent, considerably raising the risk of a landslide in the fiord. At the same time, increased investment in Milford Sound tourism saw visitor numbers double by 2019. The pandemic slowed things down, but once borders opened back up tourism operators were eager to get back to business. The individual risk is now significantly higher than it was 10 years ago.

Over the years, scientists have tried to increase public awareness while simultaneously ironing out the data. On the one hand, public awareness demands urgency, on the other, the rigours of academia demand a more glacial pace. The consequences of a false alarm are surely part of the equation. 

So though the known risk grew over the last 10 years, our understanding of what this event might actually look like grew too. Dysktra’s thesis hinged on calculations and landslide deposit mapping, but said little about what could happen on the ground on a given day. A decade later, students at the University of Canterbury are still working on this problem, such as Master’s student Livvy Harris. 

One of her supervisors, Tom Robinson, was responsible for the other Canterbury PhD that helped put Milford’s tsunami risk on the map. Robinson, like so many other scientists included in this work, was deep in preparation for Rū Whenua and was unable to comment in time for this story.

Harris’s MSc thesis used agent-based modelling to simulate how people might move around the scene in the event of a landslide-driven tsunami. And though any scientist would be the first to tell you that no model is perfect, the results from her simulations left little room for debate: “If the wave arrival time is within 172 seconds of the onset of shaking, there are unlikely to be any survivors.”

172 seconds is less than the expected shaking time for an Alpine Fault event, meaning that evacuees would have to start moving before the earthquake ends. But even if they are able to stand and run, they may not have time.

Dykstra’s work identified several sheer faces in the fiord with evidence of geomorphic weakening. These included Rover Peak, The Lion, northeast-facing parts of Mitre Peak, and Sheerdown Peak. Even a small landslide from Sheerdown Peak into village-adjacent Deepwater Basin would leave no time for consideration, while further down the fiord, “Waves caused by landslides from either Mitre Peak or The Lion will reach the Cleddau Delta in ~2 minutes.”

Assuming evacuees had both the time and ability to seek shelter, Harris’s model identified two potential safe havens: one about a kilometre up the highway, and another up an unmarked path to a viewing platform measuring just nine square metres. Even including the surrounding bush, the platform is too small to accommodate more than 158 people, and the highway is too far away to offer sanctuary to your average visitor. Both of these areas are above the 17-metre modelled inundation height, but evidence from Dysktra’s thesis suggests a given wave could be well above or well below this height.

Because of the limited routes and size of the safe zones, Harris’s research said “the maximum possible evacuation rates for most scenarios is <10 percent of the exposed population”, even if the modelled wave arrival times were “too conservative”.

Even with low odds, when it came to survival, timing was everything. The model considered everything from time of day and time of year to how quickly people began to move towards safety, and found that the best time for a tsunami to hit would be at night during the winter off-season, when only a few hundred staff would be on site. This gave the highest rate of survival for any scenario: 5.2 percent.

For every metre an individual was able to climb, their risk of fatality decreased by 11 percent. For some, though, their starting location precluded escape: “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.” All of these scenarios were “optimistic”, according to the thesis. Nobody was able to evacuate in the shortest wave arrival times.

In a worst-case scenario, this tsunami could claim more than 3500 lives, making it Aotearoa’s worst disaster by an order of magnitude. And though this number is dwarfed by the estimated 20,000 casualties attributed to a tsunami generated by the Hikurangi Subduction Zone, it is somewhat unique in the sense that the majority of the victims may not be New Zealanders.

Depending on the time of year, the victims of such an event would likely be from overseas, with none of the cultural awareness that your average Kiwi may have of earthquake hazard. “Long and strong, get gone” is not a universal catchphrase, but knowing what to do in an earthquake might actually hamper self-evacuation efforts rather than encourage them. Typically, people expect to “drop, cover, hold” in an earthquake, unless they’re aware that a tsunami is possible in the area. As Harris noted, there is currently no tsunami warning signage in Milford Sound.

The apparent safety of New Zealand makes it an attractive option for tourists who may otherwise be risk-averse. Active geology breeds spectacular vistas, but inherently carries the risk of resuming activity while people are on-site. And unlike the prospect of heli-skiing Tititea, your average non-geologist visitor may not assess a visit to a well-patronised fiord as particularly risky.

On Wednesday, agencies from across New Zealand will confer for Exercise Rū Whenua: a national simulation of an Alpine Fault event. Organised by National Civil Defence Emergency Management, the three-day exercise will span six weeks and involve local response groups from around the country, the same players who are expected to respond to the real deal. 

The event focuses on the Alpine Fault, but the CDEM noted that the chance to conduct a nationwide test of emergency response procedures also offered “an opportunity to test or practise some of the lessons identified following the response to the January and February 2023 North Island severe weather events”.

The simulation has been developed by Project AF8 (Alpine Fault 8), a group at the forefront of community resilience and planning for this exact scenario. This team has conducted national roadshows for the last several years in an effort to bolster community preparedness for an Alpine Fault earthquake. And stakeholders in the Milford Sound tourism sector are aware of the hazard, and are working with research groups to address the problem.

What is not clear is if Wednesday’s planned scenario includes a Milford Sound tsunami. Research into the hazard has pushed the associated risks higher, but it is still a relatively new concern. However, given that the potential impacts of the tsunami represent an order of magnitude increase in Alpine Fault fatalities, the decision to include it in the simulation is the biggest “what if” on the table. 

As Robinson noted in the opening of his PhD, “The scale of a disaster is a direct function of the pre-event actions and decisions taken by society.”