Sponsored by the Defense Threat Reduction Agency

Automatic seismogram interpretation and onset estimation are desirable goals facilitating the rapid location and discrimination of seismic events, automation also relieves the human operator from onerous repetitive tasks. At regional distances the onsets of seismic phases are associated with gradual, rather than sudden, changes in the statistical properties of the seismic time series, i.e. changes in autoregressive models and mean square amplitudes. The work presented combines two aspects of the onset time estimation, the detection of the onset as a change in the trend of the cumulative sum (CUSUM) of a suitable test statistic, and a randomized search for its location, simulated annealing (SA). The latter is applied repeatedly, thus, resulting in multiple onset time estimates. These typically form clusters around the arrival times of the known seismic phases, but there are some scattered values as well, that have to be discarded. The uncertainties in the onset times estimated can be quantified in terms of the spread in the values of the best defined clusters. Various standard cluster analysis methods can be applied to define individual clusters. The system is presently developed using MATLAB, utilizing its graphical user interface capabilities. 2.0. OBJECTIVES The objective of this work is to develop automated methods for interpreting regional and teleseismic shor