Signal Quality and the Reliability of Seismic Observations

The ability to detect, time and measure seismic phases depends on the location, size, and quality of the recorded signals. Additional constraints are an analyst’s familiarity with a seismogenic zone and with the seismic stations that record the energy. Quantification and qualification of an analyst’s ability to detect, time and measure seismic signals has not been calculated or fully assessed. The fundamental measurement for computing the accuracy of a seismic measurement is the signal quality. Several methods have been proposed to measure signal quality; however, the signal-to-noise ratio (SNR) has been adopted as a short-term average over the long-term average. While the standard SNR is an easy and computationally inexpensive term, the overall statistical significance has not been computed for seismic measurement analysis. The prospect of canonizing the process of cataloging seismic arrivals hinges on the ability to repeat measurements made by different methods and analysts. The first step in canonizing phase measurements has been done by the IASPEI, which established a reference for accepted practices in naming seismic phases. The New Manual for Seismological Observatory Practices (NMSOP, 2002) outlines key observations for seismic phases recorded at different distances and proposes to quantify timing uncertainty with a user-specified windowing technique. However, this added measurement would not completely remove bias introduced by different techniques used by analysts to time seismic arrivals. The general guideline to time a seismic arrival is to record the time where a noted change in frequency and/or amplitude begins. This is generally achieved by enhancing the arrivals through filtering or beam forming. However, these enhancements can alter the characteristics of the arrival and how the arrival will be measured. Furthermore, each enhancement has user-specified parameters that can vary between analysts and this results in reduced ability to repeat measurements between analysts. The SPEAR project (Zeiler and Velasco, 2009) has started to explore the effects of comparing measurements from the same seismograms. Initial results showed that experience and the signal quality are the leading contributors to pick differences. However, the traditional SNR method of measuring signal quality was replaced by a Wide-band Spectral Ratio (WSR) due to a decrease in scatter. This observation brings up an important question of what is the best way to measure signal quality. We compare various methods (traditional SNR, WSR, power spectral density plots, Allan Variance) that have been proposed to measure signal quality and discuss which method provides the best tool to compare arrival time uncertainty.