[S10-P] Poster

[S10-P-02] Application of earthquake forecasting models in central New Zealand following the November 2016 Kaikoura earthquake

^○David Rhoades, Annemarie Christophersen, Matthew Gerstenberger, David Harte (GNS Science, Lower Hutt, New Zealand)

Following the M7.8 Kaikoura earthquake of 14 November 2016, GNS Science, through the GeoNet website, provided forecasts of future earthquake occurrence, similar to the practice that has evolved during major earthquake sequences in New Zealand since 2011. The forecasts were expressed in three forms: (1) Tables of the number of earthquakes expected during future time periods in different magnitude ranges within a defined aftershock region, with associated uncertainties; (2) Maps of the probability of earthquake shaking exceeding certain Modified Mercalli intensities; and (3) Future earthquake scenarios with associated probabilities. Initially the short-term earthquake probabilities (STEP) model was used to estimate the expected number of earthquakes of magnitude 5.0 and above for time periods of 1, 7 and 30 days. A week after the mainshock, the 1-day forecasts were discontinued and yearly forecasts were added. After five weeks, only monthly and yearly forecasts were produced. The latter were based on a hybrid model, with long-term, medium-term and short-term components derived from well-established models. The long-term component was provided by a smoothed seismicity model, the medium-term component by two versions of the Every Earthquake a Precursor According to Scale (EEPAS) model, and the short-term component by the STEP and Epidemic-Type Aftershock (ETAS) models. The Kaikoura earthquake presented some particular challenges: (1) Careful communication of changes to the estimated mainshock magnitude, the contributing models and the method for estimating uncertainties in the expected number of earthquakes (a switch from the Poisson to the negative binomial distribution); and (2) Amendment of forecasts to allow for increased slow slip on the North Island subduction interface after the Kaikoura earthquake. In the absence of any established models to incorporate slow slip into the forecasts, the probability of a triggered major event was assessed by expert elicitation.