Seismogram Picking Error from Analyst Review (SPEAR): Single-Analyst and Institution Analysis

Abstract

Listen

Translate article

Abstract We perform two experiments to define and isolate errors associated with picking seismic phases (the process of naming and measuring the arrivals of seismic phases). The first experiment establishes a method and model to describe picking errors for a single analyst measuring arrivals recorded on short-period instruments. The analyst’s picks are constrained to earthquakes that had similar epicenters with stations at local to near-regional distances. We find that the main source of error for an individual analyst is the signal-to-noise ratio (SNR) measured using a wideband spectral ratio (WSR). The pick error is 0.1 sec for measurements made with a WSR greater than 10. In some cases, errors that occur with high WSR (>10) appear to result from the variability of the seismic signals, caused through magnitude or scaling differences and source specific differences. The second experiment confirms our hypothesis that institutions have approaches in measuring seismic arrivals that inherently cause the pick placement to differ from institute to institute. For the institute analysis, we use data collected from the International Seismological Center (ISC). The SNR is not reported in the ISC catalogs; however, we show that the distance from station to source is a factor that can define the error between picks made at different institutions. The error between picks increases with distance until the transition from regional phases to teleseismic (∼20–30°), where the error is decreased by over 50%. The error then begins to increase after 40°. Another factor of error may be that institutions name seismic phases differently. We also perform a one-way multivariate analysis of variance on the institutions and the pick information to show the institutions pick differently. Measurement differences between institutions are greater than measurement differences of a single analyst. The single-analyst measurement error can be defined as 0.1 sec for an experienced analyst picking local to regional phases on short-period instruments that have a WSR greater than 10. However, this precision level of measurement error is not valid when combining picks from multiple institutions. We need to define a new measurement error model for combining picks to establish velocity and earth models when merging data from multiple institutions; a seismic phase picking error model cannot be established from the current earthquake catalogs.