An intriguing discovery about one of the many faults that ruptured in the Kaikōura earthquake may prompt a rethink about how scientists calculate seismic hazard.
The magnitude 7.8 event, which struck just after midnight on November 14, 2016, was one of the most complex earthquakes observed anywhere in the world.
That was largely due to the sheer number of faults that ruptured in just a single event.
More than 24 faults were activated - 14 of which ruptured violently enough to displace land by more than a metre.
Nowhere was this effect perhaps more dramatic than along Marlborough's Kekerengu Fault, where the land was offset by as much as 12m.
In some places the fault was visible with raised-up folds of earth stretching across the countryside.
But now scientists have learned something new about another one of the faults.
The Papatea Fault, unmapped before the quake and running along a similar path to the lower Clarence River in Marlborough, produced a 19km-long surface rupture and shunted a large area of mountainous country up by 8m in a matter of seconds.
A study published this week in the journal Science Advances indicated that the fault ruptured even though it hadn't accumulated stress normally associated with fault rupturing.
Co-author and earthquake geologist at GNS Science Rob Langridge said it appeared the fault suddenly became squeezed for room by the rupture of neighbouring faults causing it to break in "a very emphatic way".
"The rupture of the Papatea Fault stands out for being one of the most dramatic elements of what was an unusual rupture sequence in the first place," he said.
"It produced the largest vertical movements of all the faults that ruptured during the earthquake and it has puzzled scientists because its rupture could not be fitted to standard models of fault rupture."
However, the new study, led by Anna Diederichs and Ed Nissen of Canada's University of Victoria, used computer analysis of LIDAR (Light Detection and Ranging) images to come up with a solution to its unusual behaviour.
"We discovered a number of unusual characteristics to this fault," Nissen said.
"Most unusually, the standard elastic rebound model of earthquake faulting did not fit the observed ground deformation.
"We've concluded that the Papatea Fault did not release elastically stored tectonic strain as faults normally do during a rupture."
Nissen said the findings indicate that some faults may fall outside typical fault behaviour and conventional modelling may not capture the hazard they pose.
Earthquake forecasting was based on the elastic strain cycle model, where faults gradually accumulated strain until they failed, and then the cycle was repeated.
"However, the Papatea Fault does not seem to follow this model, and such faults may still need to be accounted for in earthquake forecast models."
Going forward, Nissen said this research finding could be considered when assessing the risk from faults that might have a low or unclear strain accumulation signal.
The research was based on computer analysis of LIDAR images of the fault rupture area captured before and after the quake.
Fortuitously, Environment Canterbury collected LIDAR of the Clarence Valley area several years before the Kaikōura earthquake, mainly for flood protection purposes.
These images were compared with LIDAR images collected in the wake of the earthquake in 2016.
The analysis enabled the scientists to derive the 3D geometry of the fault which was a basis for numerical modelling of the rupture.