Using the ROC Based Gini Coefficient to Quantify Spatio-Temporal Clustering of Earthquakes

Loading...
Thumbnail Image

Authors

Bladis, Natalie Teva

Issue Date

2023

Type

Thesis

Language

Keywords

Earthquakes , Gini Coefficient , Receiver Operating Characteristic , ROC , Statistical Seismology

Research Projects

Organizational Units

Journal Issue

Alternative Title

Abstract

The purpose of this thesis is to explore a method for measuring the correlation between two measures or random variables using the ROC (Receiver Operating Characteristic) diagram and Gini coefficient. This procedure is then applied to seismic data to produce a metric of earthquake clustering in various regions. Earthquake clustering is a fundamental component of seismicity that reflects various forms of earthquake triggering mechanisms. Zaliapin and Ben-Zion (2021) introduced a simple and robust measure of space-time clustering, using the ROC diagram, that allows disentangling effects related to concentration of events around a heterogeneous regional fault network (marginal space distribution of events) from coupled space-time fluctuations (joint space-time distribution). This work describes the mathematical and statistical foundation of their approach. Specifically, this study:- examines and illustrates seismic clustering in multiple seismically active regions, including the Reno area, - explores several general measures of seismic rate that can account for the number of events, the total area of faultbreaks, seismic moment, and more, - and systematically examines general and coupled space-time clustering of raw and declustered catalogs.Conclusion of this analysis are that the overall observed earthquake clustering is high, for a variety of regional catalogs and global seismicity. At the same time, when the marginal clustering is removed, different catalogs show different degrees of coupled space-time clustering, reflecting a variety of specific triggering conditions and mechanisms.

Description

Citation

Publisher

License

Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 United States

Journal

Volume

Issue

PubMed ID

DOI

ISSN

EISSN