← Back to Sessions

Is Accounting for Spatial Variation of b-Values Useful for Earthquake Forecasting?

Session: Constructing and Testing Regional and Global Earthquake Forecasts II

Type: Oral

Date: 4/21/2021

Presentation Time: 10:15 AM Pacific

Description: 

The b-value is the most studied parameter in the field of statistical seismology. It has been widely used as the proxy of the current state of stress in the earth's crust. Numerous phenomenological studies have been conducted to establish its credibility as the parameter that holds the key to improving our understanding of seismogenesis. Given the importance of this parameter, we want to assess its usefulness for earthquake forecasting. Specifically, we want to understand if accounting for the spatial variation of b-values can significantly improve the Epidemic-Type-Aftershock-Sequence (ETAS) models' performance. In this study, we first consider several methods widely used to infer the underlying spatial variation of the b-value and an additional method that we develop in this study. We then apply these methods to synthetic catalogs, generated using a known underlying spatial variation of b-values, and assess their ability to infer this albeit known but hidden pattern. We then apply the same methods to the Californian earthquake catalog. The usefulness of the inferred spatial variation of b-values by the different methods for the real catalog is then evaluated using numerous pseudo-prospective forecasting experiments. The spatial variation of b-values inferred from the training periods is coupled with the ETAS model to get forecasts for the testing periods in these experiments. The forecasts resulting from the different methods are validated against the earthquakes in the testing periods. The resulting scores for each method are mutually compared, and to the score of a null model, which does not consider the spatial variation of b-values. This comparison allows us to establish if considering spatial variation of b-values is useful for earthquake forecasting and which of the widely used methods leads to the most useful forecast.

Presenting Author: Shyam Nandan

Student Presenter: No

Authors