Predicting earthquakes in Japan is a numbers game

TOKYO -- Anyone who has experienced a major earthquake, such as the Kobe Earthquake that struck western Japan in 1995, or the March 2011 earthquake and tsunami that devastated northeastern Japan, is aware of their frightening power. What is much harder to know is when the next big quake will occur near you.
In seismically active Japan, the government calculates and publishes the odds of a major temblor occurring in any one place using various methods, but the predictions are not always consistent -- with good reason.
Margin of error
In the early hours of April 16, a magnitude-7.3 earthquake hit the southern Japanese prefecture of Kumamoto. The temblor occurred along the Futagawa fault. Active faults -- cracks in the earth's crust -- are constantly being pushed and pulled by the surrounding tectonic plates. When the stress builds beyond the breaking point, the plates slip suddenly, causing an earthquake.
The government's Headquarters for Earthquake Research Promotion put the likelihood of a magnitude-7 or -8 earthquake occurring in the next 30 years in the area where Kumamoto earthquake took place at between "nearly zero and 0.9%." The research group publishes long-term predictions for big earthquakes by studying 97 major active faults across the country. Their predictions are based on the frequency of earthquakes at a given point, and when the most recent quake has occurred.

An active fault causes a major earthquake once every 1,000 to tens of thousands of years. Thus, a probability of a big temblor shaking any one spot over a 30-year time span is quite low. Although this is a convenient, human-scale time span, it can give people a false sense of security. When the Kumamoto quake struck, the group's predictions for the area were criticized for not properly communicating the risk to residents.
In response to that criticism, the team in August decided to use a probability ranking to indicate the level of risk along active faults. It labels faults with a 3% or higher chance of having a big earthquake within 30 years as "S" (highly probable), those with a probability between 0.1% and 3% as "A" (somewhat probable), and those with a less-than 0.1% chance as "Z" (other). Those that the probability cannot be or has not been determined are classified as "X."
But there are other active faults near the designated faults and movement along any of them can trigger a temblor. So researchers began publishing the probability of fault quakes by region in 2013.  To calculate the likelihood of a magnitude-6.8 or stronger quake taking place somewhere in a given region within 30 years, the probability of an occurrence is calculated for different active faults within a region.
In determining the overall probability, the team also takes into consideration earthquakes that may occur along smaller active faults, since these caused significant damage in earthquakes that hit central Japan's Niigata Prefecture in 2004 and the Noto Peninsula in Ishikawa Prefecture in 2007.
This method led researchers to conclude, before the Kumamoto earthquake, that there was an 18-27% chance of a large quake in central Kyushu, the island in the southern Japan where the Futagawa fault runs.


http://asia.nikkei.com/Tech-Science/Science/Predicting-earthquakes-in-Japan-is-a-numbers-game