Statistical classification methods applied to seismic discrimination

Technical Report · Tue Jun 11 04:00:00 EDT 1996

DOI:https://doi.org/10.2172/257361· OSTI ID:257361

Ryan, F M ^[1]; Anderson, D N; Anderson, K K; Hagedorn, D N; Higbee, K T; Miller, N E; Redgate, T; Rohay, A C

---

1. ed.

To verify compliance with a Comprehensive Test Ban Treaty (CTBT), low energy seismic activity must be detected and discriminated. Monitoring small-scale activity will require regional (within {approx}2000 km) monitoring capabilities. This report provides background information on various statistical classification methods and discusses the relevance of each method in the CTBT seismic discrimination setting. Criteria for classification method selection are explained and examples are given to illustrate several key issues. This report describes in more detail the issues and analyses that were initially outlined in a poster presentation at a recent American Geophysical Union (AGU) meeting. Section 2 of this report describes both the CTBT seismic discrimination setting and the general statistical classification approach to this setting. Seismic data examples illustrate the importance of synergistically using multivariate data as well as the difficulties due to missing observations. Classification method selection criteria are presented and discussed in Section 3. These criteria are grouped into the broad classes of simplicity, robustness, applicability, and performance. Section 4 follows with a description of several statistical classification methods: linear discriminant analysis, quadratic discriminant analysis, variably regularized discriminant analysis, flexible discriminant analysis, logistic discriminant analysis, K-th Nearest Neighbor discrimination, kernel discrimination, and classification and regression tree discrimination. The advantages and disadvantages of these methods are summarized in Section 5.

• Research Organization: Pacific Northwest National Lab., Richland, WA (United States)
• Sponsoring Organization: USDOE, Washington, DC (United States)
• DOE Contract Number: AC06-76RL01830
• OSTI ID: 257361
• Report Number(s): PNNL--11192; ON: DE96012316
• Country of Publication: United States
• Language: English

Related Subjects

35 ARMS CONTROL
45 MILITARY TECHNOLOGY, WEAPONRY, AND NATIONAL DEFENSE
CALCULATION METHODS
CHEMICAL EXPLOSIONS
COMPLIANCE
CONTINENTAL CRUST
DATA ANALYSIS
EARTHQUAKES
EXPLOSIVE FRACTURING
MULTIVARIATE ANALYSIS
NON-PROLIFERATION TREATY
NUCLEAR EXPLOSIONS
PATTERN RECOGNITION
PROBABILITY
REGIONAL ANALYSIS
SEISMIC ARRAYS
SEISMIC EVENTS
SITE CHARACTERIZATION
VERIFICATION