Recent Experiences in Aftershock Hazard Modelling in New Zealand

Abstract:

The occurrence of several sequences of earthquakes in New Zealand in the last few years has meant that GNS Science has gained significant recent experience in aftershock hazard and forecasting. First was the Canterbury sequence of events which began in 2010 and included the destructive Christchurch earthquake of February, 2011. This sequence is occurring in what was a moderate-to-low hazard region of the National Seismic Hazard Model (NSHM): the model on which the building design standards are based. With the expectation that the sequence would produce a 50-year hazard estimate in exceedance of the existing building standard, we developed a time-dependent model that combined short-term (STEP & ETAS) and longer-term (EEPAS) clustering with time-independent models. This forecast was combined with the NSHM to produce a forecast of the hazard for the next 50 years. This has been used to revise building design standards for the region and has contributed to planning of the rebuilding of Christchurch in multiple aspects.

An important contribution to this model comes from the inclusion of EEPAS, which allows for clustering on the scale of decades. EEPAS is based on three empirical regressions that relate the magnitudes, times of occurrence, and locations of major earthquakes to regional precursory scale increases in the magnitude and rate of occurrence of minor earthquakes. A second important contribution comes from the long-term rate to which seismicity is expected to return in 50-years. With little seismicity in the region in historical times, a controlling factor in the rate is whether-or-not it is based on a declustered catalog. This epistemic uncertainty in the model was allowed for by using forecasts from both declustered and non-declustered catalogs.

With two additional moderate sequences in the capital region of New Zealand in the last year, we have continued to refine our forecasting techniques, including the use of potential scenarios based on the aftershock forecasts. These scenarios provide example of how the sequence might eventuate, including the understanding of nearby faults and the Hikurangi megathrust. They have been developed with input from social scientists and have been provided to the public and government officials, and have proven useful in aiding the interpretation of the aftershock probabilities.