Calibration of Seismographic Stations for Improved Earthquake Location in 43 States

During this two-year project, we began by engaging in two somewhat separate research activities

that came together in the third and final stage of the project. First, we developed a 3D model of the

crust and upper mantle beneath North America, for which we could quickly compute travel times

between any two points. Such travel times were regarded in practice as a sum of two terms,

namely: the travel time as calculated for a standard Earth model, plus a correction term, resulting

from the differences between our 3D model and the standard model. Second, we developed a list

of seismic events, called ground truth or GT events, for which we had confidence that the actual

hypocenters were known to within about 5 km. For these so-called GT5 events (or GT10 in some

cases, or GT2…) we acquired travel time data and determined the station set for which we needed

travel time corrections.

The first of these activities was by far the most labor-intensive component of the whole project. As

described further, below, we combined information from extensive surface wave studies and body

wave studies, and in particular we developed a 3D model called NETTLESAK135G that very

significantly improved upon the standard Earth model ak135 of Kennett et al (1995), in its fit to

observed travel times from a series of chemical and nuclear explosions whose signals have been

reported from hundreds of stations across the continental US and parts of Canada. These particular

seismic sources have locations known to within 1 km or better (hence, are called GT1 events, and

in some cases are GT0) and also have accurate origin times, measured or derived from special

information obtained for each source by the organizations that carried out the explosions. The

combined dataset of first-arriving P-wave travel times from these events is an important component

of our research, and is likely to find future uses in the development of better 3D models for North

America.

In the second activity we began with the set of 16 GT5 events listed in our proposal. This was

expanded to a set of 29 events during the project. We gathered the travel time picks for these

events and determined the set of stations for which we needed travel time corrections based on our

3D Earth model NETTLESAK135G. We computed these corrections for a grid of half-degree by

half-degree source locations. At the final stage of our work, when we began relocation of these

events using a variety of travel time models, and compared the resulting locations with the PDE

and GT locations, we made a few changes in our event selection.

In the third and final stage of the project, in order to make our own location estimates we applied

the LOCSAT code developed over a period of several years for operational use in data centers that

analyze seismic signals to monitor for nuclear explosions. Of course most of the events located by

such centers are earthquakes. This code can accept data of three types, namely slowness (S),

azimuth (A), and travel time (T), hence its name. Slowness and azimuth can be measured by

arrays. In this project we used travel time data from numerous single stations, not from any arrays,

hence our input data consisted only of travel times.

In sections below, we describe:

(1) our development of the 3D Earth model NETTLESAK135G and the dataset of travel

times from GT1 and GT0 events, all of them explosions, used to validate it (this section

also describes our approach to computing travel times);

(2) our choice of GT5 events and a summary of the reasons we consider our GT

information to be of high quality; and

(3) the preliminary results of various relocation estimates, and their performance for

particular events.

About 95% of the effort into this project went into the stages (1) and (2) described above.

Although the funding period for this project is now over, and this is our formal final report, we

welcome opportunities to continue to work with NEIC/USGS personnel to augment the work we

have done so far in stage (3). It will be very helpful to conduct joint work, in order for our location

estimates for specific events — based upon our general approach to quantifying the effects of 3D

structure, and upon details of our methods — to be compared to location estimates used in current

operational procedures at NEIC. Some of the assumptions built into operational procedures, and

decisions about what data are selected (or left out), and how the data are weighted, are unknown to

us.

In other sections below we give a non-technical summary, a report on presentations we have made

during this work, and a description of datasets that are available as a result of this project. A final

section comments upon the project as a whole.