Nearshore tsunami variability from stochastic earthquake models. Tests at Australian tide gauges 1960-2021.

Tsunami hazard assessments from earthquake sources often rely on simulating numerous stochastic earthquakes on source zones of interest. However, current approaches lack standardization, and different modeling choices can significantly affect the simulated tsunamis. Poor modeling choices may introduce bias into hazard assessments, making it crucial to test models against a wide range of historical tsunami data. This study evaluates three stochastic earthquake-tsunami models by comparing them with 65 tide gauge records for 12 tsunamis observed in Australia since 1960. The models, included in the 2018 Australian Probabilistic Tsunami Hazard Assessment, range from a simple fixed-area uniform-slip model (FAUS) to a variable-area uniform-slip model (VAUS) and a heterogeneous slip model (HS). For each historical tsunami, and each of the three models, 60 random tsunamis with similar earthquake location and magnitude as the observed event are simulated from source to nearshore. These random tsunamis characterize the modeled tsunami distribution for each event and provide a basis for testing the models against data. Firstly, we assess whether the random scenarios include synthetic tsunamis similar to the observations. While all models perform well for many historical events, the FAUS model performs poorly for events such as Chile 1960, Sumatra 2004, and Solomon 2007. Secondly, we evaluate biases in the stochastic tsunami wave size. All models perform adequately for small tsunamis, but a clear downward bias emerges for the FAUS model in larger observed tsunamis. These results highlight the importance of rupture area variability for realistic tsunami simulation in the far-field and provide further evidence that the VAUS and HS models produce reasonable random tsunamis. Presented at the American Geophysical Union (AGU24) Meeting