When big earthquakes hit New Zealand, how violent they feel often has less to do with magnitude, and more with geology closer to the surface.
Earlier this year, scientists revealed how readings taken at the North Canterbury town of Waiau during last November's Kaikoura Earthquake proved a new record for vertical ground acceleration, reaching 3g - or 30 times the force an airliner passenger feels at take-off.
One local pub owner described the quake, which shook bricks from homes and buildings and raised a natural, metres-high wall along the nearby Kekerengu Fault, as the most frightening he'd experienced.
Scientists suggested shaking in Waiau may have been worsened by soft soil near the surface interacting with stiffer material beneath.
Such localised factors also made the 2010 and 2011 Canterbury Earthquakes seem a lot worse than their actual magnitudes, and left parts of Christchurch awash with slimy liquefaction.
Ultimately, earthquake scientists failed to accurately predict these effects, partly because the current models were too simplistic for most locations and only looked at one dimension of the seismic waves.
Now, a new study led by the University of Canterbury's Dr Christopher McGann will try to fill in these gaps and give planners a better idea of how soil structures in different places could compound earthquake risks.
"The current ways we use to make site-specific characterisations are fairly simplified - it does the job from a general standpoint, but it's missing a lot of key things," McGann said.
"So our ultimate goal is to create a new way of looking at it that's just as easy to use as the current way, yet allows you to capture something that we can't at the moment."
Interestingly, the answers to the problem won't be found here, but in Japan.
McGann and his team will make use of a wealth of data collected by Japanese instruments that record not just shaking at the ground surface, as our GeoNet seismometers do, but also deeper in the earth, almost at bedrock.
"So if you have those two observations, by comparing them you can tell exactly what did the soil change about the shaking," he said.
"If you know what it was like before it got to the soil, and you know what it was like after it went through the soil, you have a dataset that tells you precisely the level of amplification or de-amplification provided by a particular site."
The team will use information from more than 5000 ground shaking measurements, collected during 100 Japanese earthquakes, to look at local geology and ground motions in three dimensions.
"Using those sites, we can really validate our own modelling framework."
Was there enough in common between Japan and New Zealand's geology to make the science transferable?
"To a degree," McGann said.
"But the real benefit is that we will have a very good understanding of how good or how bad our approach is, and we can quantify it."
More accurate predictions could be used to inform the design and construction of safer urban structures, such as high-rise buildings, bridges, and other infrastructure.
"Essentially, having that quantification allows us to have more confidence that what we predict for a location is going to be in line with what actually happens during the event."
The three-year study is being supported with a $300,000 Marsden Fund grant.