When the deadly tsunami that caused the Fukushima Nuclear Power Plant meltdown in 2011 was observed spreading across the Pacific Ocean, it travelled about 2% slower than conventional computer simulations predict. This meant it arrived at distant coasts up to 20 minutes later than forecast.

Now researchers have used the NCI supercomputer to find out why. The findings could help us understand how such deadly tsunamis originate.

Dr Sebastien Allgeyer from ANU used Raijin to simulate the giant tsunamis that hit Chile in 2010 and Japan in 2011. When he used the standard model, the tsunamis travelled across the ocean more quickly than in real life.

The key, says Dr Allgeyer, is that the standard model doesn't take into account the immense weight of a tsunami wave, which is enough to 'bend' the ocean floor.

"The old models assume the tsunamis were moving over a rigid ocean floor," explains his colleague Professor Cummins.

"But we know that the ocean floor isn't rigid – it's elastic."

With this in mind, Dr Allgeyer tweaked the model to include the earth-bending effects of the tsunamis on the ocean floor.

"We also included factors like water compressibility, because seawater near the ocean floor is denser than it is at the surface," Professor Cummins says.

The new simulations matched the behaviour of the real tsunamis very closely.

"We've shown that this new model is a much better match to the real-world observations," he says.

"These findings are important for understanding how tsunamis originate. The quality of tsunami data has improved tremendously over the past decade, but our deep-ocean tsunami modelling hasn't.

"In many cases the only data we have to model the tsunami excitation is far-field data, and now we can use that far-field data to better understand the tsunami source."

Read the paper in Geophysical Research Letters.