Testing the effect of deficient real-time earthquake catalogs on non-Poissonian earthquake likelihood models: Examples from the Canterbury earthquake sequence

Annemarie Christophersen, David A. Rhoades, David S. Harte, & Matthew C. Gerstenberger

Published August 14, 2016, SCEC Contribution #6789, 2016 SCEC Annual Meeting Poster #308

The still on-going Canterbury earthquake sequence has provided us with a wealth of seismological data. Among them are a series of near-real time earthquake catalogs that show the deficiency of real-time data. For example, while the final GeoNet catalog lists 20 M≥5.0 earthquake within the first 24 hours of the M7.1 Darfield earthquake, the initial earthquake catalog had only 6 M≥5.0 in the same time period. It took more than 18 months for the data to be finalised. Other changes in the final catalog are relocation of hypocenter and changes in earthquake magnitudes. We are interested in the effect of the deficient real-time data on earthquake forecasts. We calculated forecasts of durations varying from 1 – 365 days from different implementations of the Epidemic-Type-Aftershock-Sequence (ETAS) model and the Short-Term-Earthquake-Probability (STEP) model. Some models produce CSEP-type grid forecasts, others simulated catalogs, and some both of these forms of forecasts. We employ different testing methods to evaluate the effect of the changing input data on the model forecasts and performance while avoiding the Poisson assumption of the standard CSEP consistency tests.

Key Words
statistical seismology, earthquake forecasting, earthquake likelihood models

Citation
Christophersen, A., Rhoades, D. A., Harte, D. S., & Gerstenberger, M. C. (2016, 08). Testing the effect of deficient real-time earthquake catalogs on non-Poissonian earthquake likelihood models: Examples from the Canterbury earthquake sequence. Poster Presentation at 2016 SCEC Annual Meeting.

Related Projects & Working Groups
Earthquake Forecasting and Predictability (EFP)