← Back to Sessions

Towards a Prospective Test of the Foreshock Traffic Light System

Session: Constructing and Testing Regional and Global Earthquake Forecasts II

Type: Oral

Date: 4/21/2021

Presentation Time: 10:30 AM Pacific

Description: 

Traffic Lights are a tool for quantitatively recognizing risk to initiate risk reduction measures and they have been used to manage risk behaviour in various contexts. Gulia and Wiemer (2019) proposed a traffic light classification, known as Foreshock Traffic Light System (FTLS), based on the b-value space-time variations as a first-order discriminator between normal aftershocks and likely precursory sequences. The model, to be implemented after the occurrence of an M>= 6 to assess if this event was the mainshock, is based on the widely established inverse dependence of the b-value on differential stress. The authors define three levels of alert based on the b-value difference in percentage between the background value and the sequence-specific values.

So far, the FTLS had been tested retrospectively on 58 sequences, reaching an accuracy of 0.95 in a confusion matrix analysis. After the publication of the model, we tested it in a pseudo-prospective application to the 2019 Ridgecrest sequence. Results are fully in line with the FTLS hypothesis: After analyzing carefully the magnitude of completeness of the sequences, we find that in the hours after the first Mw6.4 Ridgecrest event, the b-value drops by 23% on average, when compared to the background value, resulting in a red foreshock traffic light. After the second, M7.1 mainshock, the b-value subsequently increased by 26% over the background value, triggering a green traffic light setting.

We are currently preparing the next step in testing the FTLS model: a fully prospective test embedded in a CSEP style testing framework, as provided by the EU project RISE. Implementing such a fully prospective test is highly challenging since several data quality analysis steps need to be fully automated. For example, the magnitude of completeness varies strongly with time in an aftershock sequence, robust procedures to access Mc(t) are thus needed to assess b-values. The automatic selection of the faults plane is also challenging. We discuss in our contribution the current status and future direction of testing the FTLS model.

Presenting Author: Stefan Wiemer

Student Presenter: No

Authors