SCEC Award Number 10089 View PDF
Proposal Category Collaborative Proposal (Integration and Theory)
Proposal Title Detection of Anomalous Strain Transients Using Principal Component Analysis and Covariance Descriptor Analysis Methods (JPL Task Plan 82-15261)
Investigator(s)
Name Organization
Sharon Kedar California Institute of Technology Danan Dong California Institute of Technology Jay Parker California Institute of Technology Robert Granat California Institute of Technology
Other Participants Chan, Robert (JPL Contract Administrator authorized to submit this Proposal on behalf of Dr. Sharon Kedar)
SCEC Priorities A5, A2, A6 SCEC Groups Geology, Seismology, Geodesy
Report Due Date 02/28/2011 Date Report Submitted N/A
Project Abstract
N/A
Intellectual Merit We have applied and evaluated several anomaly detection methods that were
developed under various NASA projects, to the detection of simulated strain transient cases put
forth before the exercise participants. We have tested algorithms that were based on Principal
Component Analysis (PCA) techniques, as well a Covariance Descriptor Analysis (CDA)
method that originated in image processing. Both classes of techniques have shown promising
results. In phase III od the Community Transient Detection Exercise, we have applied these methods to the new set of cases (A-G), using an
iterative procedure, in which the output of one technique was used to fine-tune the input to the
other. Finally, a forward model based on uniformly slipping rectangular fault patches was
applied.
Broader Impacts Exploring geodetic transient detection methods within a community exercise enables us to share our results and technologies with others, and learn from our colleagues' experience how to best optimize our detection methods to the vast geodetic data set in Southern California, with the ultimate goal of creating alerts based on detection of events that currently go unnoticed.
Exemplary Figure N/A
Linked Publications

Add missing publication or edit citation shown. Enter the SCEC project ID to link publication.