Monday's 7.8 quake was bafflingly complex in some ways, but simple in others. Science reporter Jamie Morton delved into the data to look at what we know so far.
Why did the earthquake start small and get larger?
When the 7.8 mainshock struck at Culverden at 12.02am last Monday, it ripped along a fault segment some 50km long.
From there, the quake surged northward, spreading out over a distance between 150km to 200km, growing more powerful as it jumped from fault to fault.
South to north, at least six faults, among them the Hundalee Fault, the Hope Fault, the Waipapa Fault and the Kekerengu Fault, joined the chain in quick succession.
"When you look at the map of how much of New Zealand has ruptured, it's mind-blowing, really . . . it's just incredible," GNS Science seismologist Dr Caroline Holden told the Herald. The largest amount of energy release, for instance, didn't occur at the epicentre but more than 100km away at Seddon.
"What we are seeing from very simple modelling is just the rupture starting to the south and unzipping all the way to the north - it just kept piling up as it travelled."
It may seem strange that a quake could get bigger as it continued, rather than occurring as a one-off bang firing out progressively smaller shockwaves.
But Holden wasn't at all surprised.
"To move such a big structure, you need that little trickle at the beginning to push it over the edge, so a big one will actually get going."
This wasn't unusual for New Zealand.
The 7.1 quake that hit Darfield in Canterbury in September 2010 began on a small fault - the Charring cross fault - before pushing on to the famous Greendale fault.
"Since we don't know quantitatively how close to failure major faults are, it is not impossible for a small fault to change the surrounding stress enough to trigger that bigger one nearby."
Multiple-fault earthquakes often occur in the upper crust in continental interiors, especially in regions where there were multiple faults bunched close together, as with the Marlborough Fault System.
Scientists were now developing a model for last Monday's sequence, with teams inspecting on-land fault ruptures and feeding other data into the model, which would also help understand fault stress.
Stress didn't always increase stress on the neighbouring faults and could sometimes ease it - an effect that obviously wasn't the case with this quake.
2. Are the different faults part of a single "super fault"?
One interesting question posed by Dr Ake Fagerang, an earthquake scientist at Cardiff University, was whether the various faults were all just part of a much larger, single fault.
Fagerang was intrigued by the variety of fault types - the quake produced both strike-slip and thrust faulting - as well as the number involved.
"A question is whether the faults represent a mixture of strike-slip and thrust faults, or a population of variably oblique-slip faults; another question is whether the fault surface traces are linked to form a large fault at depth," he wrote in earth science magazine EOS.
But Holden said the possibility of a centralised "super-fault" causing the havoc wasn't supported by the evidence gathered so far.
Rather, it was clear that the quake had been caused a combination of individual crustal faults.
"If there was any component deep under the surface - so, where all the faults are branching from - it would be pretty small, because we didn't get the signal from the ground shaking."
Satellite images also appeared to dismiss the theory.
"If there was significant movement in the interface, it would have had a signature on the ground deformation at the surface - and we don't seem to see that, either.
"So, while we can't rule out that nothing happened at the interface, it would have been very small compared to the action going on at the surface."
Despite its complexity, the sequence aligned well with scientists' understanding of tectonics systems in New Zealand, and the area where the quake hit hardest was where much upper crust surface faulting could be expected.
3. Why did the quake last so long?
Fagerang also pointed out the quake's long length - a two minute rumble, with the most severe shaking lasting 50 seconds.
He noted the quake slip lasted longer than the one-minute slip duration typically considered to calculate local magnitudes.
This was intriguing, as such slow propagation was typical of earthquakes originating from a relatively shallow subduction thrust interface, and not faults in a continental crust.
But Holden said she hadn't seen any unusually slow rupture in her modelling, which explained most of the quake's signal.
When one considered the fault was rupturing at a speed of 2km per second, across a length of up to 200km, there was "nothing unusual" in the duration, she said.
"With a multiple crustal source model, supported by geodesy and strong motion, we are able to explain the effects of this earthquake without including any significant subduction component," she said.
"It's worth noting that the proximity in depth and location of these crustal faults to the subduction interface does have an impact on the subduction area nonetheless."
Holden said it would be unusual if the shaking was small but long-lasting, something typical of a low-energy subduction quake.
4. How has New Zealand shifted?
Scientists are still in Marlborough collecting data, but GPS stations and satellites have already revealed dramatic movement - all which happened in seconds.
GeoNet uses specialised satellites which collect radar data used to track these land movements in great detail.
A technique called InSAR (Interferometric Synthetic Aperture Radar) draws on radar satellites orbiting around 700km above Earth to precisely measure the distance between the ground and the satellite.
If the ground moves between two subsequent satellite passes, due to an earthquake or volcanic eruption, then the distance between the ground and the satellite changes.
Observing these changes in the position of the land with InSAR enabled scientists to generate detailed maps of ground movement, often with centimetre-level accuracy.
So far, satellite images had revealed huge changes in land movement across the Hope and Kekerengu faults, as well as several other faults in the region.
To the east of these faults, the land went mostly southwest, while land to the west of these faults the land moved mostly northeast.
The satellite images clearly showed the earthquake was one of the most complex earthquakes ever observed - and backed scientists' conclusions that the earthquake ruptured at least four different faults, and probably more.
The biggest displacements were seen on the Hope, Hundalee, Waipapa Bay and Kekerengu faults - the latter accommodating up to 10m of slip.
GPS stations on the ground similarly showed astonishing shifts.
Cape Campbell, on the northeast tip of the South Island, was shunted north-northeast by more than 2m- bringing it that much closer to the North Island - and was also pushed up vertically by almost 1m.
Similarly, Kaikoura had moved to the northeast by nearly 1m, and had been lifted upwards by 70cm.
Hanmer Springs, the closest GPS site to the quake epicentre, jumped eastward by approximately 50cm.
A key observation was that, although the earthquake fault rupture began near Culverden, by far the largest motions of GPS sites occurred at Cape Campbell, supporting the idea that the quake ruptured north over a very long distance from where it started.
And not only did the quake shift landmasses in the northern South Island, but it also caused movements across most of the country.
In the lower North Island, the east coast had shifted west by up to 5cm, while the Wellington and Kapiti regions were shunted 2 to 6cm to the north.
Christchurch and Banks Peninsula were now approximately 2cm further south than they were the day before the quake.
Some parts of the west coast of the South Island have been shifted eastward by as much as 1cm, although northern North Island and southern South Island only moved a few millimetres.
Previous surveys had shown the country was normally moving by 4cm a year but, rather than shifting as a whole, tectonic forces were deforming the land surface - stretching, slimming and sliding it southward.
5. For how long can we expect aftershocks to continue?
More than 2000 aftershocks have followed the quake - 20 of them over 5.0 - and we can expect more of the same for some time.
Holden pointed out the 5.9 Valentine's Day jolt that shook Christchurch this year - another legacy of the sequence activated by the 2010 Darfield quake.
"They just go on for years," she said.
"Aftershocks are not fun, but they're now going to become part of our daily life here."
GeoNet calculates its aftershock probabilities based on data from previous earthquake sequences that have occurred around the world.
Most earthquake sequences fade away over time, with spikes of activity and occasional larger earthquakes.
In the short-term, the odds of another quake over 7.0 are becoming increasingly smaller, while aftershocks between 5.0 and 5.9 are a virtual certainty.
The latest forecast, issued by GeoNet yesterday, put the probability of another quake over 7.0 occurring within the next 30 days at 21 per cent, compared with chances of 10 per cent this week and two per cent within the next 24 hours.
The probability of one or more quakes between 6.0 and 6.9 hitting had also decreased, but was still relatively high, at 90 per cent this month, 63 per cent this week and 16 per cent in the next day.
That of "felt" aftershocks between 5.0 and 5.9 was more than 99 per cent this week (an average 10 such-sized events) and this month (an average 24).
The probability of one or more in the next 24 hours was 85 per cent, with an average of 1.9 quakes.
GeoNet spokesperson Caroline Little said aftershocks were occurring throughout a broad area that surrounded the faults that ruptured in Monday's 7.8 earthquake.
"Most of these aftershocks are occurring near these faults, but a small number of aftershocks have occurred as far away as the lower North Island," she said.
"Our current forecasts indicate that it is likely for that aftershocks near these faults will continue, but for the frequency of aftershocks to decrease with time."
The area nearest the faults had a probability of 80 per cent or more for damaging shaking in the next 30 days.
In comparison, the probability of damaging shaking in the Wellington area was less than 10 per cent in that period.
"While this probability is considerably lower than in other regions, it is possible for shaking similar to what occurred on Monday to happen again in Wellington."
The capital was already a high seismic risk area and Monday's earthquake had increased this risk, Little said.
Christchurch's aftershock probabilities were not greatly affected by the 7.8 earthquake.
The most likely scenario was that aftershocks would continue to decrease in frequency over the next 30 days.
Felt aftershocks would occur from the Culverden epicentre right up along the Kaikoura coastline to the Cape Palliser and Wellington area.
Much less likely was the probability of a quake smaller than Monday's mainshock and between 7.0 to 7.8 - a 20 per cent chance over the next 30 days - and even more unlikely, with a probability of less than one per cent, was the possibility of one larger than the mainshock.