Nankai-Tokai subduction hazard for catastrophe risk modeling

The historical record of Nankai subduction zone earthquakes includes nine event sequences over the last 1300 years. Typical characteristic behaviour is evident, with segments rupturing either co-seismically or as two large earthquakes less than 3 yrs apart (active phase), followed by periods of low seismicity lasting 90 - 150 yrs or more. Despite the long historical record, the recurrence behaviour and consequent seismic hazard remain uncertain and controversial. In 2005 the Headquarters for Earthquake Research Promotion (HERP) published models for hundreds of faults as part of an official Japanese seismic hazard map. The HERP models have been widely adopted in part or full both within Japan and by the main international catastrophe risk model companies. The time-dependent recurrence modelling we adopt for the Nankai faults departs considerably from HERP in three main areas: ■ A “Linked System” (LS) source model is used to simulate the strong correlation between segment ruptures evident in the historical record, whereas the HERP recurrence estimates assume the Nankai, Tonankai and Tokai segments rupture independently. The LS component models all historical events with a common rupture recurrence cycle for the three segments. System rupture probabilities are calculated assuming BPT behaviour and parameter uncertainties assessed from the full 1300 yr historical record. ■ An independent, “Tokai Only” (TO) rupture source is used specifically to model potential “Tokai only” earthquakes. There are widely diverging views on the possibility of this segment rupturing independently. Although all historical Tokai ruptures appear to have been composite Tonankai -Tokai earthquakes, the available data do not preclude the possibility of future “Tokai only” events. The HERP model also includes “Tokai only” earthquakes but the recurrence parameters are based on historical composite Tonankai -Tokai ruptures and do not appear to recognise the complex tectonic environment in the Tokai area. ■ For the Nankai and Tonankai segments only, HERP assumed Time-Predictable (TP) recurrence behaviour. The resulting calculated 30 and 50 year rupture probabilities are considerably higher than standard renewal model estimates as used in the adopted model. While perhaps more contentious, the weight of evidence available does not appear to be consistent with TP behaviour. For the adopted modelling the estimated probabilities of no Nankai segment rupture within the next 30 & 50 years are 56% & 27% respectively. The disparity between the models is highlighted by the much lower estimates obtained by HERP (2.5% & 0.039% respectively as at 2006). Even for just the Nankai and Tonankai segments (ie. ignoring Tokai), HERP estimated only 1.7% probability of no rupture in 50yrs. These estimates can be contrasted with the fact that in 2056 (50 yrs from 2006), the elapsed time since the start of the last rupture cycle (112yrs) will still be 5 yrs short of the historical mean recurrence interval since 1360. Net effects on nation-wide catastrophe risk estimates for all earthquake sources depend on modelled exposure distributions but can be as much as a factor of two. The differences are important as they impact on multi-billion dollar international risk transfer programs.