Global Ground Truth Data Set with Waveform and Improved Arrival Data (original) (raw)

The main objective of the three-year research project is to produce a quality controlled global GT0-5 event set, accompanied with waveform and groomed arrival time data sets. Our efforts are directed toward developing and refining methodologies for generating new GT events through multiple event location analysis. Multiple event location techniques, such as Hypocentroidal Decomposition (HDC) (Jordan and Sverdup, 1981; Engdahl et al., 2004), provide precise relative locations within an event cluster. However, the absolute locations could still be biased. In order to get accurate absolute locations, independent GT information is needed. We have developed a novel multiple-event location technique, Reciprocal Cluster Analysis (RCA), which combines local data with regional/teleseismic HDC results and uses local stations as GT0 constraints to obtain accurate absolute locations. We have validated the HDC-RCA methodology using an event cluster of GT0 nuclear explosions and GT5 earthquakes which occurred within the Nevada Test Site (NTS). We demonstrated that the HDC-RCA method requires neither dense local networks, nor prior GT information. It relies on a few local stations, provided that the station centroid is inside the event cluster. We showed that absolute locations obtained from the HDC-RCA analysis are consistent with the true GT locations as RCA reduces the regional/teleseismic bias to less than 5 km. Monte Carlo simulations demonstrated that the RCA error ellipses are conservative estimates of the absolute location uncertainties. This allows us to identify GT5 events based on the semi-major axis of their error ellipses scaled to the 95% confidence level. Using this criterion, we identified 21 out of 24 GT events in the NTS cluster. The size of the 95% confidence error ellipses is mainly driven by the reading errors. We utilize waveform crosscorrelation to reduce reading errors, and possibly identify phases not reported in bulletins. Waveform correlation also offers a way to flag and correct phase identification errors. We follow a rigorous statistical approach by using the significance of the cross-correlation to assess the similarity of waveforms. Arrival times are automatically adjusted according to the optimal alignment derived from the waveform cross-correlation, thus resulting in accurate phase picks with reduced measurement errors. We further demonstrate the potential of the HDC-RCA approach on selected event clusters (Chi-Chi, Taiwan; Afar triangle, Africa) and we are prepared to process candidate event clusters selected from an updated EHB (Engdahl et al., 1998) bulletin. 27th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies Extreme Value PDF Extreme Value Cumulative Distribution Z-transformed max(correlation) Maximum correlation Normal Distribution Measured Correlations Z-transformed max(correlation) Significance=0.995