EHB Bulletin

The EHB Procedures

Although useful for seismic hazard assessment, global compilations of earthquake hypocenters and associated phase arrival times and residuals often are too inhomogeneous to be confidently applied, for example, to problems such as Earth structure determination. The main problem is the varying level of mislocation, particularly focal depth, introduced largely by errors in the reference Earth model, unaccounted for effects of lateral heterogeneity, and phase misidentification. The result is loss of structural signal in the residuals. The bias in hypocenter determination can be significantly reduced and at least part of the lost structural signal recovered by:

• Using a proper reference Earth model;
• Improving usage of the data;
• Limiting the events of interest only to those that are well-constrained teleseismically.

The Engdahl et al. (1998; EHB) algorithm has been used to significantly improve routine hypocenter determinations made by the ISS, ISC and PDE. Elements of the algorithm that contribute to these improvements include:

• Use of an improved 1-D global travel time model (ak135);
• Iterative relocation with dynamic phase identification;
• Use of first arriving P, S and PKP phases;
• Use of the teleseismic depth phases pP, pwP and sP (with PDFs and bounce point corrections);
• Ellipticity corrections for the ak135 model;
• Empirical teleseismic station patch corrections (for 5 x 5 degree patches);
• Weighting by phase variance as a function of distance;
• Selection criteria for events having ten or more observations at teleseismic distances (> 28 degrees) and a teleseismic secondary azimuth gap < 180 degrees.

The EHB algorithm does not recalculate magnitudes; the vast majority of the mb and Ms values are taken from the ISC bulletin, and the provenance of Mw values is the global CMT catalogue. The selection of magnitudes for events prior to 1964 is described in Engdahl and Villasenor (2002).

Engdahl et al. (1998) have shown that the model ak135 (Kennett et al., 1995) provides a very good fit to a wide range of seismic phases. The mantle S wave bias of the earlier iasp91 (Kennett and Engdahl, 1991) has been removed. Most core phase times are quite well matched and a baseline problem with Jeffreys-Bullen (JB; Jeffreys and Bullen, 1940) travel times for PKP phases removed. Thus, for global earthquake location there has been convergence on a global, radially symmetric, P- and S-velocity Earth model that provides a good average fit to reported phase arrival times.

A direct method to improve seismic event locations is by better utilization of the data. Until very recently standard teleseismic catalogs (ISC, NEIC) relied almost entirely on first arriving P phases for locating events. Many studies have shown that the inclusion of later arriving phases can provide greater constraints on hypocenter parameters, especially focal depth. For events having a poor azimuthal distribution of P-wave arrivals, epicenter constraints are improved by the inclusion of S- and P-wave core phases at critical azimuths. Additional constraints are often provided by these phases because their travel-time derivatives differ significantly in magnitude from those of direct P. Depth to origin time trade-off is avoided by the inclusion of depth phases (pP, pwP, sP) because their travel time derivatives are opposite in sign to direct P. However, a problem with the use of depth phases is that their correct identification often requires knowledge of the event depth and distance. Hence, depth phase arrivals are re-identified after each iteration in the EHB procedure using a probabilistic association algorithm. Probability density functions (PDF) for depth phases, centered on their theoretical relative travel times for a given hypocenter, are compared to the observed phase arrivals. When PDFs overlap for a particular depth phase, phase identification is assigned in a probabilistic manner based on the relevant PDF values, making sure not to assign the same phase to two different arrivals.

The travel times predicted by recently developed, radially symmetric, Earth models (such as ak135) are extremely valuable for earthquake location and phase identification. Nevertheless, most earthquakes occur in or near subducted lithosphere where aspherical variations in upper mantle seismic wave velocities are large (i.e., on the order of 5 - 10%). Such lateral variations in seismic velocity, the uneven spatial distribution of seismological stations, and the specific choice of seismic data used to determine the earthquake hypocenter can still easily combine to produce bias in earthquake locations of several tens of kilometers. Tests of location bias globally using a new archive of reference event information and the EHB location algorithm (Bondár et al., 2003) show that most explosions and earthquakes are mislocated by less than 20 km if the secondary azimuth gap (the largest azimuth gap filled by a single station) to observing stations at all distances is less than 180 degrees. Hence, at least in the case of events well constrained azimuthally by reporting stations, mislocation errors introduced by lateral heterogeneity can be minimized. For smaller and/or poorly recorded events, however, there is not much hope of significantly reducing the resulting mislocation error until we can somehow better account for aspherical Earth structure in 1-D earthquake location procedures.

Map of Stations used in EHB Bulletin

Magnitude Completeness, EHB and ISC Bulletins