At a few minutes' shy of midnight on August 23, 11km below Rolleston on the Canterbury Plains, seismometers detected a tiny earthquake.
With a magnitude of just 1.98, the tremor was still significant.
It was the 14,600th aftershock since the 7.1 jolt that, five years ago, changed Canterbury forever.
The thousands of tiny tremors of that early morning quake, which tore open the earth near a little rural hub called Darfield in the Selwyn District, serve as a shaky legacy that scientists suspect will keep rolling on.
The misery, death and destruction from the rupture, at 4.35am on a Saturday, can be counted in the 185 people killed in the Christchurch Earthquake five months later, held to be an aftershock despite the unforgettable havoc it wrought, and the tens of billions of dollars in damage.
READ MORE: Canterbury's $40b recovery going 'pretty well' says PM
Without it, Building and Housing Minister Nick Smith might not have this week ordered a billion-dollar strengthening of building and housing; falling parts of unreinforced masonry like parapets and facades that killed 35 people in the February 22 earthquake.
Some of those who happened to be awake when it hit on September 4, 2010, reported a bizarre sight of lightning streaming not from the sky, but into it from the ground.
Amid the 40-second rumble, the ground acceleration created by the thrusting caused stress-induced electrical currents deep within the Earth's crust to rise rapidly and fire from the surface.
GNS Science seismologists were quickly able to identify its cause as strike-slip faulting, where two blocks of the crust violently tear past each other, near the eastern foothills of the Southern Alps, at the western edge of the Canterbury Plains.
They described it an "extremely rare seismic recording near a fault rupture".
People in Christchurch, 40km east, likened it to a train, a hurricane, a battle tank rolling down the street.
More than 6000 people reported it.
At Christchurch Airport, chaos descended as the lights were knocked out; at Kaiapoi motorcamp, caravans sank to their axles as craters opened; shaken revellers stepped out of city bars into blackness and debris.
A Sydenham dairy stayed open selling battery packs for just $2 and giving away milk cartons; a young Christchurch student screamed from beneath her bedcovers, coiled in the fetal position, later telling how "it seemed like it went on forever".
In those 40 seconds, billions of dollars of damage resulted as sewer lines were broken, roads opened up, chimneys collapsed and residents of low-lying communities like Bexley were introduced to the slushy nightmare of soil liquefaction.
By comparison, the level of shaking approximately coincided with the strength of a one-in-500 year quake event that our building code is now tested against, and its force released about 30 times more energy than the July 2013 Cook Strait earthquake.
In recent times, only the Christchurch earthquake, which erupted with the energy of 15,000 tonnes of TNT from a relatively shallow depth, caused a higher peak ground acceleration.
(App users tap here to see graphics)
A quake striking in the heart of Canterbury didn't come completely out of the blue for seismologists, who always regarded the region as a place of "moderate hazard".
But 50 years of no earthquakes had created a cosy perception among many locals that they simply didn't happen in Christchurch.
Top GNS scientist and director of the joint Natural Hazards Research Platform, Dr Kelvin Berryman, told the Herald this week that the location of the epicentre was nonetheless surprising as an active fault hadn't been mapped there before.
A visible rent across the landscape allowed scientists to measure the movement of the longest fault segment, the Greendale Fault.
"From very early in the investigations, it was apparent that rupture of the Greendale fault was very rare because there was no evidence of clear earlier events across river terraces at least 15,000 years old," Dr Berryman said.
His colleagues also had big questions about the complexity of the rupture - as many as eight small faults connected up to rupture in that 15 to 20 seconds represented by the strong shaking.
"So mapping the detail of the surface rupture, and accurately the aftershocks were of prime interest."
In the five months before the Christchurch quake, a large team of scientists tracked aftershock locations with a dense array of freshly installed seismometers and GPS stations.
"The long series of aftershocks have provided new insights into the way stress is transferred in the crust of the earth. We now understand better the 'influence distance' from one earthquake," Dr Berryman said.
"Did the Darfield earthquake trigger the February 22 event? Now we can probably be pretty certain that it did.
"Would the February 22 event have occurred if the Darfield event did not occur? This is a much more difficult question for which there is no clear consensus."
For scientists like Dr Berryman, the key lessons were found in its sheer intricacy, and the detail of data captured by seismographs already set up has made it perhaps the best instrumentally-recorded earthquake event in the world.
"In the past five years there has been an excellent joining together of science and policy in regard to natural hazard risks in New Zealand," he said.
"This has been one of the most important lessons learned from the Canterbury Earthquake Sequence."
Other lessons included the importance of good land-use planning, the bolstering of legislation and civil preparedness, and the strengthening of communities, where more people now knew their neighbours.
It proved that rare events, even those occurring once in every 20,000 years - still had to happen some time.
This week, when seismographs in the region notched up the 14,614th earthquake over a magnitude of 1.1, earthquake activity in the aftershock zone is still about 10 times higher than before the event.
Dr Berryman expected Canterbury was likely to keep feeling them for many years, and may perhaps be destined to the same fate as the South Island's Buller Region, which has had an elevated level of seismicity since the Murchison earthquake 86 years ago. As recently as the early 1990s, several magnitude 6 earthquakes shook the lower Buller gorge.
This month's earthquake forecast for Canterbury put the probability of quakes measuring between 5 and 5.9, 6 and 6.9, and over 7, at 50 per cent, 6 per cent and less than 1 per cent respectively.
But there was really no telling when their next big one would come.
"The fundamental message to Canterbury residents remains - be prepared."
Two of New Zealand's leading earthquake scientists, Associate Professor John Townend (JT) of Victoria University and Professor Jarg Pettinga (JP) of Canterbury University, discuss the Darfield earthquake and what it taught us.
JT:
The Darfield earthquake and the aftershocks that ensued have underlined the fact that almost all parts of New Zealand are prone to earthquake hazards, unfortunately.
It's unlikely that we will ever be able to predict exactly where earthquakes will strike or how large they will be, and I'm skeptical that we'll ever know the locations of all the faults that pose a hazard. That being the case, we need to focus on continuing to develop resilient infrastructure and social networks on the basis of good science, good engineering, and good policy.
JP: When this earthquake sequence first kicked off with the M7.1 Darfield Earthquake in September 2010 it ruptured a fault line which had not previously been identified and mapped, primarily because it had no visible surface expression.
There also was no associated seismicity - smaller earthquakes along the line of Greendale Fault which subsequently broke through to the ground surface during the M7.1 event.
Seismicity levels beneath the Canterbury Plains were pretty low and insignificant in the decade prior to 2010.
While we did not know about the Greendale Fault we did know that such earthquakes and associated fault ruptures were possible beneath the Canterbury Plains.
In a benchmark study in the late 1990s, two reports were completed for ECan and subsequently formally published in the Bulletin of the Earthquake Engineering Society of NZ.
The first report identified all known active faults in the Canterbury region - more than 100 at that time - and assessed capable of generating moderate to large earthquakes accompanied by ground surface rupture.
The reports also summarised other known information about the faults identified, types of fault movement, timing of last one or more earthquake ruptures, estimated return time between large earthquakes and estimated ground displacement during earthquakes.
As part of this study, and subsequent EQC commissioned research reports by University of Canterbury staff and students, the style of earthquake fault rupture and fault interactions associated with active faults hidden beneath and just starting to break the ground surface of the Canterbury Plains were outlined in and reported on.
The report that perhaps most graphically captured this was on the Ashley Fault Zone immediately northwest of Rangiora.
In the summary we outlined the earthquake scenario most likely when a series of faults rupture interactively - much like what actually happened in 2010-2011.
So, we had the likelihood of large earthquakes covered, we knew that there were active faults in the region, and that many faults while present remained hidden beneath the plains, and we understood the rupture style and fault activity that would likely happen at some time.
Separately, but building on the data already available, a project led by scientists from GNS Science and supported collaboratively by other researchers continued to refine and improve the NZ National Seismic Hazard Model, developed on available regional seismicity data extending back over historic times.
This data includes historic large earthquakes and instrumental records, augmented by the detailed field investigations to document the pre-historic large earthquake ground rupture record.
These data combined are then used in the model to develop a probabilistic analysis of the return times of moderate to large earthquakes.
The net result is that the published 2008 review of the NZ National Seismic Hazard model identified that there was scenario earthquake of up to magnitude 7 possible anywhere under the Canterbury Plains, but that the probability levels were low, with return times of thousands of years between such events.
What such models can not provide is the actual timing of events, so if the last earthquake was many hundreds or even thousands of years ago we are likely to be much closer to the occurrence of the next event.
Effectively, the NZ National Seismic Hazard Model provides a platform for planning and engineering codes and design.
The 2010-2012 Canterbury Earthquake Sequence has provided new information on the presence of more active faults which prior remained hidden beneath the Canterbury Plains.
However, such faults were absolutely expected to exist, and in that sense to we geologists studying the earthquake activity and active earth deformation in the region this was not a surprise or unexpected.
Their likely existence was also recognised in the regional earthquake hazard models.
JT:
What was important from the outset was to understand how the sequence of earthquakes might develop.
This is an enormous challenge, based on what geophysicists know about how earthquakes interact.
Working out how one earthquake affects the next requires input from many different disciplines: seismology, geology, hydrogeology, geomorphology, and geodesy.
The most useful lessons to be learned from the Canterbury earthquake sequence will be those that tell us about earthquake processes generally, and not just about what happened in this case.
JT: In the case of Canterbury, the most damaging earthquake was not the first one, or the biggest.
In fact, the largest earthquake in our part of the world in recent years was in July 2009, in Fiordland.
It was magnitude 7.8 or so, and the second largest earthquake in the world that year, but due to its location and the way the fault ruptured, that earthquake caused very little damage and is now largely forgotten.
What's important, I think, is to understand whether or how the Fiordland earthquake related to what happened in Darfield 14 months later.
It's extremely important to bear in mind that what we have seen in Canterbury may be totally different from what occurs when, say, the Alpine Fault or the Wellington Fault, or the subduction zone beneath the eastern North Island ruptures, in terms of the duration, amplitude, and effects of shaking.
JP: Yes, the world community has learnt a great deal from the Canterbury Earthquake Sequence.
Firstly, it is probably one of or perhaps the best instrumentally recorded earthquake event globally.
This is because a network of recorders was already deployed across the Canterbury Plains by Geonet.
This deployment was actually done for a different purpose and conceived by a former University of Canterbury Engineering Professor John Berrill.
He was acutely aware of the high probability that the Alpine Fault would rupture during a large earthquake and the instruments were to "capture" that event and provide data for analysis of the way the seismic waves would transmit through the crustal rocks underlying the southern Alps and Canterbury Plains, and in turn affect Christchurch city.
In the event, the network received a "bulls-eye" hit from the Darfield Earthquake, and provided the most extraordinary detailed record for seismologists and engineers to analyse.
Results are still being used to provide new insights and models which will profoundly impact future analysis and modelling of large earthquake events globally.
JT:
Big fleas have little fleas upon their backs to bite them - earthquakes are much the same in that, generally speaking, you get lots of small earthquakes every time you have a big one.
But what controls the timing of the aftershocks that follow a large earthquake like the Darfield one is still not clear, and much of the research that's being done is focused on this question.
A real challenge in understanding earthquakes is reconciling long-term estimates of seismic hazard with much shorter-term and rapidly changing hazards posed by aftershocks.
At the heart of this challenge is improving our understanding of earthquake physics - determining what processes control the stresses acting on faults, and how those stresses change in space and time.
JT: What makes earthquakes so vicious is that it's not only the big ones that are damaging.
Geophysicists often compare earthquake sequences to piles of sand.
With a sand pile, it's easier to describe what will happen to the sand pile as a whole if you jiggle it a bit than it is to account for the behaviour of individual grains.
Similarly, it's easier to account for patterns of seismicity in an aggregate sense than it is to predict individual earthquakes, and that's why statistical analysis remains at the heart of most seismic hazard calculations.
When it comes down to it, though, it's the relationships between individual earthquakes that matter: to really understand how sequences of earthquakes are triggered and how they develop, I think it's crucial that we supplement earthquake statistics with earthquake physics.
JP: For communities, businesses and decision-makers and leaders the February 22 earthquake was a disaster with far-reaching consequences.
In that sense, the fact that it was an aftershock is probably less significant, but for scientists it forms part of an important complete sequence of events, and so the relationship between the Darfield M7.1 and subsequent aftershock sequence provides important insights into the earthquake process and how faults can interact and rupture sequentially.
Having said that, it really is important to think of this as an earthquake sequence, one where one fault rupture has in turn triggered further fault ruptures through subsurface crustal stress changes in the region which in turn has triggered other adjacent faults to rupture.
The board removed Nigel Simpson as Hawke's Bay chair just one month into the role.