Just off New Zealand's eastern seaside, kilometres from the shores of Gisborne and Hawke's Bay, lies a monster.
A sleeping taniwha that could, at any moment, be awakened by Ruaumoko, the atua or god of earthquakes.
Most always quiet, but ever capable of suddenly pouncing to life and sending a wall of water barrelling into the coast.
This is the major fault that is the Hikurangi Subduction Zone.
The devastation wrought by two other such systems - in Japan in 2011, and Indonesia in 2004 – show why emergency planners take the threat so seriously.
But for scientists, it's something of a puzzle box whose secrets - complex geological processes that play out kilometres below the seabed - aren't unlocked easily.
Its infrequent activity means those seeking to reconstruct past episodes must turn to sediments buried at beaches from Hawke's Bay to Marlborough, or to the invaluable indigenous knowledge system that is Mātauranga Māori.
Much of Dr Laura Wallace's time at GNS Science has been devoted to investigating another mysterious characteristic of the system.
These are "silent" or "slow slip" earthquakes that unfold over days to weeks, and sometimes causing large displacements along the plate boundary fault, without ever being felt at the surface.
Since 2016, these quiet stirrings have been brought into sharper focus by a series of studies and multimillion-dollar scientific expeditions along the margin, and the deployment of sophisticated sensors capable of picking up the tiniest movements.
But sometimes, luck's mattered too.
On one Thursday evening in May, just as Wallace had arrived home, the US-born geodetic scientist casually decided to check the latest data flowing in from GPS stations along the coast.
Some sites near Porangahau, in Central Hawke's Bay, showed there'd been a few millimetres of persistent easterly movement over the past few days, in step with some small, localised earthquakes.
Insignificant as that development may have appeared to the rest of us, for Wallace and her colleagues, it pointed to the opportunity to capture a silent phenomenon that scientists are quickly coming to see as the key to doing a better job forecasting some of the largest earthquakes on the planet.
She recalled thinking "wow, one's just started", before dialling her colleagues to tell them a fresh "slow-slip" earthquake was underway.
Her colleagues Dr Emily Warren-Smith and Dr Neville Palmer immediately cancelled their weekend plans, to scramble more seismometers to the area.
To understand the quake-making potential of the zone - part of the yet larger 3500km Hikurangi-Kermadec-Tonga system that fans up into the Pacific - is to understand the pivotal place it has in Aotearoa's local tectonic environment.
Our country straddles the point where the Australian and Pacific plates meet in an eternal sliding, colliding, geological scrum, that explains why GeoNet sensors register thousands of earthquakes every year – even if the plates themselves moved just tens of millimetres in that time.
We know the on-land segment of this boundary better as the Alpine Fault – demarcated dramatically by the tectonically-uplifted Southern Alps – which has now been predicted to have a 75 per cent probability of unleashing a major quake, likely measuring over 8.0, within the next 50 years.
A rupture of that size could transform the South Island as we know it, while blocking highways in more than 120 places, leaving perhaps 10,000 people cut off, and costing the economy about $10 billion.
Even so, Hikurangi could still claim the title of our largest geological threat, for the sheer impact a combined subduction earthquake and tsunami could wreak on coastal communities from Eastland to Wellington, and beyond.
If we drained the ocean, the subduction zone would appear as a vast mountain range, rising up from the seafloor.
On a map, it appears as a long line curving from waters well north of the East Cape to the northeastern South Island, beneath which the Pacific plate dives – or subducts – beneath the Australian plate.
These zones, a common feature of the Pacific "Ring of Fire", have been responsible for some of the largest cataclysms ever recorded – including the most powerful tremblor in history, the 9.5 Chile earthquake of 1960.
That owed to both the enormous amount of energy that could be pent up by two vast chunks of the planet's crust being constantly mashed together – and the physical scale of subduction zones themselves.
When they unleash these monster quakes- often called a "megathrust" earthquake - the force can abruptly shift the seafloor and ocean above, which in turn creates a tsunami that could measure many metres high.
The tsunami that was triggered by Japan's 9.1 Tohuku earthquake, and killed nearly 16,000 people, reached heights of more than 40m in some places.
On Boxing Day in 2004, a similar-sized megathrust earthquake drove a tsunami measuring 30m high into Indonesia's Aceh Province, where around half of the disaster's 230,000 fatalities were recorded.
Could Hikurangi produce earthquakes and tsunami that large?
Scientists have found local events of that scale – which would likely require the entire margin to rupture at once – are relatively rare events amid others that we have evidence of.
But that wasn't to say any other major subduction zone earthquake didn't pose a major risk.
Headline projections in one EQC-commissioned report estimated worst-case scenario impacts from a one-in-500 year event could include 33,000 fatalities, 27,000 injuries and $45b worth of property loss.
That might be an odd case to consider, given the fact only one person has ever been killed in the 10 tsunamis higher than 5m that New Zealand has experienced since 1840.
But centuries back in our past, sediment core samples have highlighted events that likely would have been highly destructive if they occurred tomorrow.
Active oral histories of tsunami are part of the Mātauranga - a holistic and complex knowledge system paralleled with cultural values, principles and practices - of many hapū and iwi across coastal Aotearoa.
Some tell of taniwha that suddenly rise from the sea, sweeping away people in their path, bringing sands from across the sea and changing the environment.
They've been passed down to explain events, record loss of life, and to serve as warnings about the nature of particular places – including the stretch between Cape Campbell and the Wairau River.
And it was near Marlborough's Wairau Lagoon that scientists only recently unearthed evidence of two tsunamis: one that occurred between 800 and 900 years ago, and another that struck half a millennium ago.
Using ancient records, researchers have been able to discover more about the frequency of past events – and even give forecasts for the future.
A just-published study, led by Victoria University PhD student Charlotte Pizer, drew on tsunami deposits preserved at Kapara te Hau/Lake Grassmere in Marlborough.
These showed that three out of the last four earthquakes on the subduction zone beneath the Wellington and Wairarapa regions generated large tsunamis, and the recurrence interval – or time between events – had been about 500 years.
From this, the study team was able to calculate a 26 per cent likelihood of a large subduction earthquake occurring within the next 50 years.
"This is the first time we have been able to calculate a well-constrained recurrence interval for large subduction earthquakes at any location along the Hikurangi margin, and the forecasting is time-dependent," study co-author and GNS paleoecologist Kate Clark said.
"So, rather than saying what the likelihood of a large earthquake is at any point in time, we can actually say what the likelihood is from this point in time.
"We have always assumed that subduction earthquakes would produce tsunami, but having the physical evidence of this is really important for increasing the confidence of our tsunami hazard models."
Two years earlier, Clark and colleagues reviewed all evidence for subduction earthquakes along the length of the margin to find that, while they were able to locate records of 10 events in the past 7000 years, recurrence intervals varied by location.
"The latest study is a nice follow-up to that showing that with very detailed, targeted studies at ideal sites, we can get recurrence intervals," Clark said.
"Our current work along these lines is now focused on Hawke's Bay."
As it stood, scientists' knowledge of past earthquakes and tsunamis along the subduction zone offshore the Tairāwhiti region remained poor – although there had been some moderate-sized subduction episodes, such as a pair of magnitude 7.2 events that struck around Gisborne in 1947, which generated large tsunami.
"There is quite a lot to untangle in this area, and there are big implications for whether subduction zone ruptures could extend from the Hikurangi subduction zone up into the Kermadec subduction zone."
Wallace believed that accurately calculating Hikurangi's potential to produce major earthquakes could only come with unravelling the mysteries of slow-slip earthquakes.
"And once we can better understand how slow slip events influence earthquake occurrence, we will eventually be able to use slow slip events to improve earthquake forecasts."
As there's been increasing evidence to suggest such movements can shift stress within the Earth's crust - and in very rare cases trigger large earthquakes - scientists have been watching slow earthquakes around the world all the more closely.
Their interest only grew after these earthquakes were found to precede the Tohuku earthquake, along with the 8.1 Iquique earthquake in Chile, and a 7.2 shake off the coast of Mexico, three years later.
This year, researchers reported how the slowest earthquake ever recorded - lasting 32 years - eventually led to the catastrophic 1861 Sumatra earthquake in Indonesia.
Yet, because of their regular frequency in New Zealand, scientists now know the events to be part of normal behaviour in our subduction zone – and recording one didn't mean a major rupture was on the way.
Here, slow earthquakes tend to occur at shallow depths off Gisborne and Hawke's Bay, and at deeper levels observed off the Manawatū and Kāpiti regions, releasing pent-up energy equivalent to a Magnitude 7.0 earthquake.
More specifically, they played out – sometimes over days, weeks, or even months – in an area where the subduction zone was transitioning from being "stuck" beneath the southern North Island, to an area where the subduction zone was "creeping" further north, around Gisborne and Hawkes Bay.
And because they happened too slowly to be picked up by seismometers – or to be felt by humans – they could only be recorded using special GPS equipment measuring the slow movement of land.
Near Porangahau, the timing of their occurrence – happening every five years – was relatively predictable.
Wallace and her colleagues had some working theories for this apparent cycle.
"Probably part of the answer is that the slow slip area is being loaded steadily by motion between the tectonic plates, which is pretty constant," she said.
"So there may be some kind of threshold - dependent on the strength of the fault - that is reached every five years that causes them to occur so regularly."
Scientists were also exploring ideas that fellow GNS scientist Emily Warren-Smith had developed around the role of build-up of water in the fault zone, which might similarly influence the timing.
In May, the scientists found themselves in a prime position to answer such questions.
Expecting an event to unfold this year, they'd already deployed 26 sensors around the Porangahau area to detect any offshore earthquakes and vertical displacement of the seafloor, along with extra seismometers also being deployed onshore.
Only weeks later, another event began near Gisborne, giving the scientists another trove of data to sift through.
"Slow-slip has probably been the biggest thing to hit the field of seismology in the last 20 years," Wallace said.
"And it only seems to be increasing in importance as a research topic, as scientists are recognising the insights they can give us into earthquake occurrence, and the potential for using them to improve earthquake forecasts."
A major goal of scientists has been to develop a locally based capability to investigate offshore geohazards through seafloor sensor deployments that can record vertical and horizontal movement of the seafloor, and offshore earthquakes.
In 2014, scientists made their first step by planting ocean-bottom sensors and seismometers near Gisborne, which revealed thousands more earthquakes were happening on the offshore subduction zone than can be detected by the land-based GeoNet network.
The 7.8 Kaikōura Earthquake that struck two years later also led to some fascinating new insights.
"One of the new insights that Kaikōura gave us is that a large, regional earthquake can actually trigger widespread slow slip over large regions of the Hikurangi subduction zone," Wallace said.
"The very widespread triggering of slow slip that we observed had never been seen at other subduction zones before, and really opened our eyes to how earthquakes hundreds of kilometres away can influence activity on subduction zones."
In 2017 and 2018, Wallace was part of an international team that drilled boreholes 500m beneath the East Coast seabed to install what she called sub-seafloor observatories.
Since then, these two listening posts, containing high-tech measuring and monitoring equipment inside steel casing, have been constantly gathering data on how rocks are strained during slow-slip events, as well as changes in temperature and flow of water through the fault zones.
The scientists were also able to retrieve sediment cores to give them a first-time look at the region's geological record and the types of rocks that hosted slow slip events.
"The data we recovered a few months ago suggests that some of the big slow slip events start deeper – and closer to the coast - and then propagate offshore towards the trench, where the Hikurangi plate boundary emerges at to the seafloor."
A mission on Niwa's Research Vessel Tangaroa in March this year involved sending down a remotely operated vehicle to the seafloor observatory to download the data that was recorded during a slow-slip quake in 2019.
This yielded yet more evidence to suggest these events were more frequent on New Zealand's offshore faults than scientists first thought, with many going undetected by our on-shore seismic network.
"We will continue to visit the observatories every few years, and we are working with our US collaborators to get a mission underway two years from now to recover the next data two years from now, and to replace two of the instrument strings there," Wallace said.
"An ultimate dream would be to have a cable linking the observatories to the land, so we can get the data in real-time, but that may be some years away."
Wallace couldn't over-state the impact that global partnerships and investment – totalling more than $70m to date – had made.
"It's been pretty cool to see the international science community take so much interest in the Hikurangi subduction zone, and to see how they have been able to bring some world-class data acquisition capability and cutting-edge techniques to bear on this problem," she said.
"Many scientists have realised that New Zealand's Hikurangi margin has a number of characteristics that make it a very important natural laboratory to understand earthquakes, tsunamis and slow-slip event occurrence on other subduction zones globally."
While New Zealand has a warning system for tsunamis caused by distant earthquakes, it doesn't have one for those caused by local events, such as Japan's billion-dollar, state-of-the-art network.
That was simply because tsunami generated by local earthquakes could potentially arrive at the nearest coast before scientists could calculate the location of the earthquake and issue a warning.
A 2013 GNS Science report used a scenario similar to the March 1947 tsunami earthquake off the coast north of Gisborne to assess GeoNet's detection capabilities and potential required updates to the network.
After testing a range of detection and classification algorithms with the simulated data, the report authors concluded such an event could be detectable by the network in real time.
However, it found a large portion of the geodetic sensor network would need to be upgraded to stream the data and provide accurate information.
More recently, a Google trial which recently woke many people along the East Coast with a phone alert that incorrectly reported the location and magnitude of an earthquake highlighted how far local, smaller-scale technology has to come.
For the time being, emergency planners are hammering home a single message - "long, strong, get gone" - to encourage people in seaside communities to immediately move to higher ground in the event of a major earthquake.
Civil Defence's latest survey suggests that 85 per cent of people would know to do this, despite just two per cent of coastal residents taking part in 2019's "Shake Out" tsunami drill.
Work focused on the subduction zone also led to the formation of East Coast Lab (Life at the Boundary) - a collaboration between scientists, emergency managers and other experts.
Along with public talks, school field trips and roadshows, the group last year helped develop a planning toolbox to better prepare residents.
Another recent collaboration between Ngāti Kahungunu, Wairoa Taiwhenua Incorporated and GNS Science brought together hapū from 37 marae and involved sharing mātauranga such as earthquake and tsunami memories.
"The six wananga attended by our marae highlighted local knowledge about Ruaumoko from ancient to recent," Taiwhenua chairman Nigel How said.
"These conversations generated recorded knowledge and emergency preparedness resources for our marae and wider community."
East Coast Lab chairman Lisa Pearse said all of the planning had been "very beneficial", as understanding the risk enabled communities to identify potential consequences ahead of time.
"We hope ultimately, it will save lives."
How a resilient Kiwi village hopes to bounce back from a big drop in visitors.