2-D Signal Detectors For 3-Component Seismic Recordings

Despite much ingenuity in research of optimal signal detector design the simple and universal used STALTA detector has a performance literally second-to-none to other often more advanced detectors. In other words, continued research in this field may be considered a bit futile. Our undertaking here has been motivated by some strange observational phenomenon in analysis of seismic recordings from national Norway network stations namely that apparently strong events were not recorded albeit easily visible in corresponding local records. The other, mentally puzzling attitude among seismologists, is that only 1-D detectors Z-component are used while at local and regional distance ranges horizontal components often exhibit the strongest not necessarily P-waves signal amplitudes. We are investigating both of these signal phenomena with goal of improving signal detection in a CTBT IMS context. The two major reasons for non-detected signals are apparently i a gradual P-amplitude increase typical of many mid-oceanic events and ii spiky records. In the former case, a lag of 5 - 10 sec between the STA and LTA windows respectively often suffice while in the latter case the STALTA detector per se is not robust enough to handle spikes. These may naturally be removed by a simple spike filter preceding in time the detector operation but this is not always done. Anyway, the station TRO Tromsoe, N. Norway has a poor signal detection performance which is obviously related to a spiky operational environment housed in a big building. Introducing a Kolmogorov-Smirnov K-S type detector where the test statistic is tied to the non-parametric K-S-distribution not power as for the STALTA signal detection increased significantly. However, not unexpectedly in a non-spiky environment the latter detector had a superior performance.