Abstract
We compare the on-fault seismicity and probabilistic seismic hazard in Aotearoa New Zealand produced by a synthetic earthquake catalog generated in the physics-based Rate and State Earthquake Simulator (RSQSim) and the inversion fault model (IFM) from the 2022 New Zealand National Seismic Hazard Model (NZ NSHM 2022). Because both models implement the New Zealand Community Fault Model (NZ CFM), the similarities and differences between their magnitude–frequency distributions (MFDs) and seismic hazard estimates can provide insights into how they simulate on-fault seismicity. We find that the NZ NSHM 2022 IFM indicates that faults participate in larger magnitude ruptures than the NZ CFM-RSQSim catalog, although the latter’s MFD is still within the range of the NZ NSHM 2022 IFM’s 5th and 95th percentile uncertainty. For a 10% probability of exceedance in a 50 yr hazard level, the normalized discrepancies in peak ground acceleration between the two models are <50%. However, at 0.5% probability of exceedance in 50 yr, larger differences (>250%) emerge for sites around low-slip-rate faults or for sites along the eastern North Island that lie above the Hikurangi–Kermadec subduction zone interface. Because both models are incomplete representations of seismicity and seismic hazards in New Zealand, we also incorporated the NZ NSHM 2022 distributed seismicity model (DSM) into these comparisons. In low-seismicity regions, the normalized hazard discrepancies are reduced once the DSM is included, but they persist in sites overlying the Hikurangi subduction zone. Cumulatively, our study highlights that although there is an obvious scientific appeal to adding “physics” into earthquake rupture forecasts, there are challenges and opportunities with incorporating uncertainty into earthquake simulators, and developing meaningful tests for them in regions with short historical earthquake catalogs.