Predicting earthquakes is generally seen as not possible - but in the 1970s some eventual flops sparked a brief stir in the scientific community.
The most famous failure may have been what is known as the Parkfield case, when US scientists used a prediction of a magnitude-six earthquake to occur within a few years on the San Andreas fault near Parkfield, California.
GNS principal scientist David Rhoades said this forecast was based on the idea of repeated occurrence of characteristic earthquakes on the same fault segment at regular intervals.
"The previous record of events at Parkfield was used in support of this idea, but in retrospect it is easy to see that there was a lot of wishful thinking involved in the interpretation of the past data to support this idea," he said.
No earthquake occurred within the time-window stated, but much later, in 2004, a magnitude-six earthquake was finally recorded near Parkfield.
This had since steered many US researchers away from trying to forecast quakes.
In another case in the 1970s, US physicist Dr Brian Brady made a precise public forecast of a large earthquake to occur near Lima, Peru, in 1981.
"This caused a lot of consternation for quite a few months, until a few days before the day of the predicted event the US Earthquake Prediction Evaluation Council announced that in their view it had no merit," Dr Rhoades said.
Of course no earthquake came - but there had been some flukes.
In 1975, Chinese authorities successfully evacuated the city of Haicheng before a 7.3-magnitude quake struck that could have otherwise killed tens of thousands of people.
Dr Rhoades said this led to a widespread expectation that it could be repeated elsewhere.
"Now we realise that was an exceptional circumstance, aided by a very pronounced sequence of foreshocks in the days before the main shock," he said.
"It was a well-founded hunch that paid off, rather than a scientific prediction."