Poster #088, Earthquake Forecasting and Predictability (EFP)

Measuring robust b-values from magnitude differences

Nicholas J. van der Elst
Poster Image: 

Poster Presentation

2020 SCEC Annual Meeting, Poster #088, SCEC Contribution #10404 VIEW PDF
The earthquake magnitude-frequency distribution is characterized by the b-value, which describes the relative frequency of large vs. small earthquakes. There is evidence to suggest that the b-value for aftershocks is larger than for background events, and that exceptions to this rule tend to occur exclusively in foreshock sequences, with low b-values indicative of larger earthquakes to come [Gulia and Wiemer, 2019]. However, the measurement of b-value during an active aftershock sequence is strongly biased by short-term incompleteness of the earthquake catalog. Since smaller earthquakes are obscured by larger ones, the b-value in an active aftershock sequence tends to be biased low by incomp...lete detection across the entire range of magnitudes. Here I develop a new estimator of the b-value that is relatively insensitive to transient changes in catalog completeness and allows for the real-time, unbiased measurement of b-value during an ongoing sequence. The estimator is based on the differences in magnitude between successive earthquakes, which are described by a double-exponential distribution with the same b-value as the exponential magnitude distribution itself. The new estimator broadly confirms the findings of Gulia and Wiemer [2019], showing a decrease in b-value for events between the foreshock and mainshock for several prominent foreshock sequences over the past decade – with some caveats. The new estimator greatly improves the robustness of b-value measurement during active earthquake sequences, as well as in historical catalogs with variable completeness.