Parametric and Nonparametric Discriminants for Regional Earthquakes and Explosions

Abstract

Conventional methods for discriminating between earthquakes and explosions at regional distances have concentrated on extracting specific features such as amplitude and spectral ratios from the waveforms of the P and S phases. We consider here an optimum nonparametric classification procedure derived from the classical approach to discriminating between two Gaussian processes with unequal spectra. Two robust variations based on the minimum discrimination information statistic and Renyi's entropy are also considered. We compare the optimum classification procedure with various amplitude and spectral ratio discriminants and show that its performance is superior when applied to a small population of 8 land-based earthquakes and 8 mining explosions recorded in Scandinavia. Several parametric characterizations of the notion of complexity based on modeling earthquakes and explosions as autoregressive or modulated autoregressive processes are also proposed and their performance compared with the nonparametric and feature extraction approaches.

Open PDF

Document Details

Document Type
Technical Report
Publication Date
Jul 31, 1993
Accession Number
ADA273807

Entities

People

  • Allan D. Mcquarrie
  • Joseph E. Cavanaugh
  • Robert H. Shumway

Organizations

  • University of California, Davis

Tags

Communities of Interest

  • Materials and Manufacturing Processes

DTIC Thesaurus Topics

  • Data Science
  • Databases
  • Detectors
  • Discriminant Analysis
  • Earth Sciences
  • Earthquakes
  • Explosions
  • Feature Extraction
  • Frequency
  • Gaussian Processes
  • Geography
  • Geology
  • Geophysics
  • Information Science
  • Planetary Sciences
  • Statistics
  • Surveys

Readers

  • Seismology
  • Statistical inference.

Technology Areas

  • AI & ML
  • AI & ML - Bayesian Inference