FOCMEC Page 1

FOCMEC: FOCal MEChanism Determinations

J. Arthur Snoke
Virginia Tech, Blacksburg, VA, USA
[email protected]

Abstract
This report accompanies a package for determining and displaying double-couple earthquake
focal mechanisms. Input are polarities (P , SV , SH) and/or amplitude ratios (SV /P , SH/P ,
SV /SH). The main program, Focmec (coded in Fortran 77), performs an efficient, system-
atic search of the focal sphere and reports acceptable solutions based on selection criteria
for the number of polarity errors and errors in amplitude ratios. The search of the focal
sphere is uniform in angle, with selectable step size and bounds. The selection criteria for
both polarities and amplitudes allow correction or weightings for near-nodal solutions. Pub-
lished applications include determinations of best-constrained fault-plane solutions for suites
of earthquakes recorded at local to regional distances (e.g., Chapman et al., 1997), analysis
of large earthquakes observed at teleseismic distances (Snoke, 1990), and the use of recorded
polarities and relative amplitudes to produce waveform synthetics (James and Snoke, 1994).
Program Focmec produces two output files: a complete summary of information about
all acceptable solutions, and a summary file that can be used as an input to other programs
for further analysis or display. Another program in the package, Focplt, produces focal-sphere
plots based on the Focmec summary file of input data (polarities and/or ratios) alone or
superimposed on solutions (fault planes, compression and tension axes, SV and SH nodal
surfaces). Other auxiliary programs include one to create input files for program Focmec,
a program that converts from among various ways of presenting a double-couple solution,
a program that calculates the radiation factors for an input mechanism, and programs for
displaying and printing plot files. Instructions and scripts are included for compiling and
running the programs, and there are two data sets with scripts and documentation for running
the programs.
Earlier versions of this software package could be built and run on IBM/CMS, PDP/RSX,
VAX/VMS,and Sunos. Sun Solaris, Windows 98, Linux, and Mac OS X (PPC and i686). The
programs are written in Fortran 77 and compile successfully using gfortran, g77, and Suns f77.
Two programs written in C (gcc) transform graphics plot files from the Sac Graphics Format
(SGF) to postscript (sgftops) and swap the byte order of the binary SGF files (sgfswap).
Unix scripts (csh) are included for displaying plots on the screen and for converting plot files
from postscript to EPS or PDF. The plot-conversion scripts require the external program
ghostscript. If the user has or prefers a different postscript viewer for displaying plots, it is
easy to modify the plot-display script. Except for these plot-conversion scripts, X-Windows
capability is not required by the FOCMEC package.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 2

I. Introduction and Overview

Given excellent station coverage for an earthquake in a region for which the earth (seismic ve-
locity) structure was well determined, and with well recorded, impulsive P -wave first arrivals,
one could get a reliable estimate for the fault-plane solution (focal mechanism) for the event.
More typically, the station density is low, the velocity structure is known only approximately,
and the polarities of the first arrivals are often ambiguous. For imperfect data, one may
need to assume a certain number of errors to find satisfactory solutions, and in many cases
the solutions are not well constrained. The range of possible solutions can often be further
constrained if one can read, to within 10%, the P and S amplitudes on vertical-component
seismograms. If one has data from three-component seismographs, the SH to P amplitude
ratio provides what could be considered almost independent data. For some events, one may
observe clear SH and/or SV polarities at one or more stations, and the SV /SH amplitude
ratio may be better constrained than a ratio including P .
The first version of program Focmec was written in 1984 (Snoke, 1984), and since then
advances in both seismology and computer technology have led to the introduction of routine
determinations of the full moment tensor using digitally recorded waveforms (e.g., Dziewonski,
et al., 1981; and Sipkin, 2001). However, these techniques can generally be applied only to
large earthquakes recorded at teleseismic distances or at regional distances for earthquakes
large enough to have well-recorded surface waves (Randall et al., 1995; Dreger et al., 2000).
Programs such as Focmec are still useful tools for getting double-couple focal mechanisms at
local to regional differences, and, for the Sakhalin Island event discussed below, such programs
can provide useful insights about the focal mechanism for large events using only a small
number of stations.
The source code included with this distribution has been compiled and linked by the
author on several Unix platforms: Sun Solaris, Mac OS X (both PPC and i686), and Linux.
The coding is in Fortran 77, and it has been built and tested using several different compilers:
gfortran (4.1 and 4.3), Sun f77, and GNU g77.
The FOCMEC software package including sources, documentation, and examples is in a
compressed tar file, focmec.tgz. On Unix, one can uncompress and expand this file with a
single command:

tar -xzf focmec.tgz

or, if has only an older version of tar that does not recognize -z, one can use the GNU gtar or
a line as follows:

gzip -d -c focmec.tgz tar -xf -

This operation produces a directory ./focmec. Online documentation for the package can be
accessed by opening the HTML file ./focmec/doc/focmec.html and following the links to other
HTML files in that directory.
To build the libraries, tables, and executables, it should suffice to go to ./src and enter
build package. The default assumes the gfortran Fortran compiler, so if one is using a dif-
ferent compiler, one has to edit the first few lines of build package so that it defines FCMP

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 3

appropriately. If the script runs with no errors, the last line displayed on the screen is a
compile/link line for program vwbeetle (a test program for the graphics package). Executables
are in directory ./bin. To erase all executables and libraries, cd to ./src and enter make clean.
Scripts for testing parts of the package are in ./tests.
The next section in this report has a discussion of the programming essentials and data
preparation. The following section has a description of the programs that make up the package.
The final section includes descriptions, input, and output for sample runs. Appendix A
contains detailed instructions on how to prepare an input file for program Focmec. Appendix
B contains definitions of terms used when discussing focal mechanisms. Appendix C has
cautionary comments for those who want to use SV polarities and amplitudes, and Appendix
C discusses graphics in the context of the FOCMEC package. README files, which contain
additional details regarding running the programs, are included with the distribution in HTML
format. To access the HTML files, cd to ./doc and open focmec.html and follow the links. Also
in ./doc are subdirectories with the output files from scripts in the sample runs subdirectories.
Among the HTML files is one with an update history of the package.

II. Programming Procedure, and Data Preparation

A. Programming Procedure
The data (station identifiers, azimuths, takeoff angles at the source, polarities, and/or ampli-
tude ratios from among P , SV , and SH arrivals) are read in from a file and stored. Selection
parameters are entered through run-time prompts, or from user-customized script files. These
include the number of allowed polarity errors for a solution, an acceptable range for devia-
tions between the observed and calculated amplitude ratios, the number of ratio errors that
are allowed to be outside that range, the region of the focal sphere to be searched, and the
fineness/coarseness of the search. Using these search criteria, program Focmec systematically
tests all possible focal mechanisms and lists those that fit the selection criteria. It stores these
possible mechanisms in two files: a short file with one line per solution that can later be used
as an input file for a plotting program, such as Focplt, and a more complete listing file that
includes four representations of the solutions (dip, strike, rake for either possible fault plane;
the A and N axes trends and plunges, the P and T trends and plunges; and the moment
tensor), the station identifiers for those with a polarity error, and complete information on the
calculated and theoretical amplitude ratios for each station. The conventions used are as in
Aki and Richards (2002, pp. 101113) and Herrmann (1975), except that our A and N axes
correspond to Herrmanns X and Y axes. (See Snoke (1989) for a review of the terminology
and conventions related to focal mechanism determinations.)
Two sets of input parameters deserve special mention: relative weighting for polarities,
and, for ratios, P or S cutoffs when the numerator or denominator radiation factor is near
a nodal surface.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 4

1. The default polarity weighting is unity, so one gets a 1 for every modeled polarity
that does not match the observed polarity. An option in the program is to use weighted
polarities. Enabling this option means that a mismatch for data near a nodal surface
counts less than a mismatch near the middle of the quadrant. (See Figure B3.) The
weight is the theoretical radiation factor for the trial solution above a chosen threshold.

2. If the numerator or denominator radiation factor is near a nodal surface, the result is a
large amplitude for the calculated log10 (ratio). Further, the velocity structure may not
be so well constrained that polarity errors near a nodal surface should be rejected. If the
factor is less than the chosen cutoff, it gets replaced by the cutoff when calculating the
ratio. Flags in the complete-solution output file indicate if the numerator, denominator,
or both were below the cutoffs and to let the user know that those ratios may have to
be examined more closely. As of the 2009 update of the package, if both the numerator
and denominator are below the cutoffs (N&D), that station is rejected and the number
of allowed solutions is decreased by one. For a trial solution, if the number of allowed
solutions is not greater than the number of ratio errors, the solution is rejected.

B. Data Preparation
The convention for the takeoff angle is that 0 is down. The polarity convention and the nota-
tion for amplitude ratios are included in Appendix A. For phases to be used in an amplitude
ratio, an implicit assumption is that the source process is simple and the travel paths similar
(negligible multipathing). If the frequency content is essentially the same and the seismome-
ters on which they are recorded are matched, one can use the raw traces. Before using SV
data, read Appendix C below. In general, if one has 3-component data, polarities and ratios
involving SH are more likely to be more reliable than those involving SV .
For calculating focal mechanisms from polarities, one needs event-based azimuths and
takeoff angles for each phase from the source. If one uses amplitude ratio data, one needs the
emergence angles for free surface corrections, and, if a correction for attenuation is to be used,
the travel times for the phases used. These data are generally provided along with the event
locations e.g., from an earthquake-location program such as Hypoellipse (Lahr, 1999). If so,
one could write a program (such as program Hypo2foc provided with this package) to take
such data from the earthquake-location output file and put it directly into a Focmec input file.
Programs Ratio prep and Focmec prep, also included with this package and discussed below,
provide an alternative procedure to prepare a Focmec input file that are particularly useful if
one is including amplitude-ratio data. Finally, one could simply enter the data by hand, using
the Focmec input-file format outlined in Appendix A and by looking at the input data files
included with the two sample runs (see Section IV).
For crustal events recorded at local or regional distances, the first arrivals may be the
refracted phases P n and Sn. If the Poissons ratio is constant in the crust and uppermost
mantle, P n and Sn will leave from the same point on the focal sphere and have identical travel
paths.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 5

III. Programs

A.Focmec
Program Focmec calculates all possible focal mechanisms for an input set of polarities and
amplitude ratios subject to user-specified constraints. The program requires an input file that
contains the ratio and polarity data for a single earthquake. Program options can be done
interactively or run from a prepared script. (Using a script is useful if one wants to test the
effects of varying options for a given data set.) See the subdirectories of ./sample runs/ for
examples of both input files (called focmecXXX.inp) and scripts (rfocmecYYY). Output is
two files:

1. a listing file containing detailed information about each acceptable solution (focmec.lst,
but renamed in the script to something else to identify it with that run); and

2. an output file with a header summarizing the data and run parameters followed by a
single line per solution (focmec.out is the default filespec). This second file can be used
as input for the plotting program Focplt that displays data and/or solutions.

B. Other Analysis Programs

Three subsidiary analysis programs are included: DSRetc,, Radiat, and Freesurf. Source (.f)
files are in ./src/. No input files are required.

Dsretc allows one to convert from among the various ways to represent a focal mechanism
dip, strike, rake; plunge/trend for P and T or A and N; and the moment tensor
representation (for a mechanism that has a significant double-couple component). Also
included is a beach-ball printer plot showing the P -wave nodal planes (Sipkin, 1993,
personal communication. Sipkin gives credit to Bob Uhrhammer.)

Radiat produces radiation factors and amplitude ratios for input dip, strike, rake, takeoff
angle, and azimuth.

Freesurf calculates free-surface corrections, which can be useful when preparing ratios
input to Focmec.

C. Focplt
Program Focplt is the plotting program in the FOCMEC package. Input can be from the
Focmec input file (for data) and/or from the Focmec output file (for solutions). The plots are
equal-area Lambert-Schmidt projections (usually lower hemisphere) of the focal sphere. Plots
can include any or all from among the following: data (polarities and/or ratios), fault planes,
P, T, B axes, and SV or SH nodal surfaces for solutions. There are options for adding
time stamps a title, and labels. The program can be run interactively or from a script. As
mentioned above, programs/scripts are included that allow one to display output plot files

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 6

on the screen and to convert SGF files to postscript, EPS, and PDF. Further information on
these programs and scripts is included in Appendix D.
D. Traveltime Tables
For one of the examples discussed below, takeoff angles and traveltimes are calculated using
iaspei-tau traveltime tables. (Traveltimes are needed if one wants to do a Q correction for
amplitudes.) Software for producing such tables is included in this package. As in the earlier
version of this package, the iasp91 velocity model is used (Kennett et al. 1991), but tables
are produced here for the ak135 model as well. The traveltime tables, which are endian and
compiler dependent, are built and placed in directory ./lib/. Files with ASCII versions of
the two models are in directory ./iaspei-tau/build-tables. Program ttimes is a mini-tutorial
on using the traveltime tables, and ./tests/run ttimes is a script that does a run for that
program. For more information about the iaspei-tau package, see Snoke (2009) and/or go to
URL http://www.iris.edu/software/downloads/processing/.

IV. Sample Runs

Directory ./sample runs has two subdirectories that contain real-data applications of the
FOCMEC package. One is for a magnitude 4.2 intraplate earthquake for which P -wave
polarity data were collected from stations at local to regional distances. The second one
is a revisit of an earlier study (Snoke, 1990): focal mechanism are calculated for a M=6.5
deep-focus (611 km) earthquake using two data sets 190 P -wave polarities, and body-wave
polarities as well as amplitude ratios from eight broadband stations. In the next two sections,
I discuss these examples. In addition, there are HTML files for each example that can be
accessed from ./doc/focmec.html. Output listing files and plots from all the runs are included
in subdirectories of directory ./doc/.
IV A. A data set of P-wave polarities
Directory ./sample runs/ETNEvent includes drivers, and input files for an application of the
FOCMEC package to find the focal mechanism for a magnitude 4.2 event using P waves
recorded by 37 stations on short-period, vertical-component seismographs. The eastern Ten-
nessee event, with a focal depth of 19 km, occurred on 27 March 1987 and is number 12 in
Chapman et al. (1997).
The azimuths and takeoff angles for the observed P waves were produced by the earth-
quake location program Hypoellipse (Lahr, 1999), and the Focmec input file was created using
program Hypo2foc that gets input from a Hypoellipse output file. The Focmec run with up to
four unity-weighted errors had 13 solutions with very little spread. Using relative weighting
for the errors produced 27 solutions with slightly more spread for a cut-off of 2.0 errors. Plots
produced by program Focplt for the unity-weighted errors are shown in Figure 1.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 7

Figure 1: Lower-hemisphere projections of the focal sphere from Focplt runs. (left):
Fault planes of all solutions with four or fewer polarity errors; (right): The fault planes
for the acceptable solution used by Chapman et al. (1997) is superimposed on the polarity
data. For polarities, circles represent compressions and triangles represent dilatations.

IV B. The Sakhalin Island Event: both polarities and amplitude ratios

This example includes two data sets: (1) 190 P -wave polarities, and (2) P and S polarities
plus amplitude ratios from eight broadband stations. The event is the 12 May 1990 M=6.5
z=611 km Sakhalin Island earthquake. The first data set came from the NEIC within a week
or two of the event. Seven of the eight broadband stations came from the IRIS DMC through
their recently introduced Gopher quick retrieval system. The eighth broadband station was
a GSETT-2 station that had been installed a few months earlier in Blacksburg (BLA)
at the observatory of my university, Virginia Tech. I used these data sets to teach myself
about the quick retrieval system, the GSETT-2 station BLA, and as a test of my FOCMEC
package. As is often the case, body-wave arrivals from this deep-focus earthquake are very
impulsive even SV arrivals. What is presented here is an update of my presentation at
the Eastern Section of the SSA meeting (Snoke, 1990), and I include a copy of the abstract
of that talk (./doc/snoke-esssa90.pdf). The real purpose here is to provide a worked example
of the package that includes data beyond P -wave polarities, but it is also instructive because
it shows that the smaller data set in this case produces a better solution. (Better because
it differs insignificantly from the CMT solution, which includes full waveforms.) For this
particular station/event geometry, the locations of some of the broadband stations to the SV
nodal surfaces provide a significant constraint on the focal mechanism. (Figure 2).
Included in directory ./sample runs/sakhalin are Focmec runs comparing solutions based
on (1) only the NEIC P -wave polarities, (2) only the broadband data, and (3) the combined
data set. (The P -wave polarity from one of the broadband stations was also in the NEIC data

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 8

set.). The best solutions for (1) and (2) are significantly different, while those for (1) and (3)
and the final best double-couple CMT solution are essentially the same (Figure 2).
The best fit for the NEIC P -polarity data set had 20 errors out of the 190 picks, and
there were four solutions with 21 errors. (The data set includes nine emergent arrivals, but
they were not included in this analysis.) Using the relative-weighting option with a threshold
of 0.1 and 6.5 allowed errors produced 13 solutions that were effectively the same as the
unity-weighted solutions. The unity-weighted solutions are included in the left-hand panel of
Figure 2. A PDF plot from the Focplt run with relative weighting is included in directory
./doc/sakhalin-doc.

Figure 2: Lower-hemisphere projections of the focal sphere from Focplt runs. (left):
Combined data set showing all observed P -wave polarities. The eight broadband stations
have larger symbols. Also shown are the fault planes for the CMT final solution (solid),
composite solutions (long dashes; 27 polarity errors), and the NEIC P -polarity-only so-
lutions (short dashes; 21 allowed errors). (right): Observed SV polarities (arrows away
from center up on the radial component) and the SV nodal surfaces for the CMT final
solution (solid), all acceptable solutions for the eight broadband stations (long dashes;
0.6 SV polarity errors), .and a NEIC P -polarity-only solution (short dashes; 20 polarity
errors).

Two programs were used in preparing the Focmec input data file for the Sakhalin Island
event from the broadband stations. Program Ratio prep, takes as its input a hypocenter plus
a list of stations for which it calculates epicentral distances and azimuths. Using the iasp91
velocity model and traveltime tables (Kennett and Engdahl, 1991), it calculates takeoff angles
and arrival times for both P and S. The output file from Ratio prep, which has one line per
station, is then read by the companion program Focmec prep. Program Focmec prep reads
in an event-specific file with one line containing the data for each polarity and/or ratio. For

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 9

polarities, the program reads in the station name and the polarity key (see Appendix A).
(Emergent arrivals for which there is a polarity, can be used if desired. We chose not to use
any of the nine emergent arrivals.) For a ratio, the line has the station, the ratio key, the key
for the polarity of the numerator (not used, but displayed in the output listings), and either
two or six numbers separated by commas the numerator and denominator amplitudes plus
(if used) attenuation data in the form Qnum , Qden , freqnum , freqden , where the freq terms
are the frequencies at which the amplitudes were determined. (If an amplitude for SV is used,
Focmec prep assumes it is from the radial component. If the SV amplitude was measured
on the vertical, it must be converted to what it would be on the radial component: program
Freesurf can be used to calculate the conversion.) The output of Focmec prep is a file that
can be used as the input file for program Focmec.
The NEIC database and the 1990 analysis of the eight broadband stations used the JB
velocity model and traveltime tables to calculate takeoff angles and, for the amplitude ratios,
the free-surface corrections. For the runs included with this package, the iasp91 velocity
model and traveltime tables were used to allow demonstrations of programs Ratio prep and
Focmec prep. The differences in takeoff angle are never more than a couple of degrees, which
leads to negligible differences in the Focmec solutions. (Not done here is any fine-tuning
of the free-surface corrections for individual stations, but, using the supplementary program
Freesurf, it would be easy to modify the input for program Focmec prep.)
Amplitudes used for amplitude ratios are calculated at the source. At higher frequencies, S
waves lose amplitude relative to P waves because of attenuation, so if the dominant frequency
is high (e.g., 1 Hz), one must correct for the relative attenuation for amplitude ratio analysis.
For these data, amplitudes were measured on matched records for low-pass filtered settings
with corners at 0.1 Hz (Butterworth 4 pole). Comparing the ratios with ones for which the
corner was 0.01 Hz gave the same results, so no correction was used for attenuation beyond
the low-pass filtering. For completeness (and to give an example of a more complicated input
data file), a Focmec prep run is included that has corrections for attenuation in the S/P ratios.
Because one amplitude ratio can be calculated from the other two amplitude ratios, all
three are not independent. I choose the two for which the amplitudes are most easily read.
When SH can be read, SH/P is generally the first choice, followed in general by SV /P .SV
arrivals are picked on the radial component. (See Appendix C for a discussion about working
with SV data.) For stations BLA and HRV, phase SKS arrived less than 20 seconds before
direct S. If low-pass filtering is used on the radial component, SKS interferes with S resulting
in a loss in accuracy for estimating the SV amplitudes. For these two stations, SV /P ratios
were therefore not used. There was no interference on the unfiltered radial component, so
SV /SH was used for these stations. (Q is the same for SV and SH, so one can work with
the unfiltered records.)
For the analysis of the broadband data, there were 24 polarities and sixteen (independent)
amplitude ratios, and in this update of the 1990 study, I corrected for the KIP and HRV
polarity flips. The final Focmec run assumed at most 0.5 (relative weighting) SV polarity
errors, zero P or SH polarity errors, and no ratio errors (with an acceptable error of 0.50
for the log10 ratio). This resulted in 10 acceptable solutions, which are all included in the
right-hand panel of Figure 2. One can see why there might be SV nodal surfaces, as three of
the stations are very near the nodal surfaces. Small changes in the location and/or velocity

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 10

models could produce small changes in the takeoff angles. This shows both that SV data has
the potential of producing a powerful constraint, and relative weighting can be useful.
I did Focmec runs with the composite data sets. (As one of the broadband stations was in
the NEIC data set, there was a total of only 197 P -wave polarities.) Assuming a total polarity
error of 27, I got two solutions: one was also one in the acceptable solutions for the run with
the eight broadband stations and the other would have been included in the SV relative error
weighting threshold were 0.6 rather than 0.5. A run with 15.0 relative polarity errors included
those two solutions plus three other similar solutions.
In the abstract for my 1990 presentation, I note that the composite solution was very
similar to the preliminary CMT best-double-couple solution. The final CMT solution can
be found using an event search from http://www.globalcmt.org/CMTsearch.html, and it
is included in both panels in Figure 2. It is even closer to the composite solution than the
preliminary one was in 1990.
Two more tests were made: forcing Focmec to accept (1) the best solution for the NEIC
data set, and (2) for the CMT solution. For (1) there were 24 P polarities, 6 SV errors, and 6
SH errors. For ratios, Focmec excludes a station from the set of ratios to be considered if both
the numerator and denominator are near their nodal surfaces. For this solution, this criterion
dropped four from the 16 ratios. Of the remaining 12, only two were within the allowed
error range. For (2), there were 14.4 P errors (from 29 stations) and 0.4 SV errors (from
three stations). No SH polarity or ratio errors, and none of the P errors were from among
the broadband stations. If the grid search in Focmec had included this solution, it would
have been found to be acceptable in either the broadband analysis or the relative-weighting
composite data analysis. There is a discussion in the HTML file about how to do a Focmec run
to a fixed location on the focal sphere when that solution was not generated from an existing
Focmec run.
The seismograms for KEV (at 54.58 ) show numerous body-wave arrivals in addition to
direct P and S. The iasp91 package gives arrival times and takeoff angles for (almost) all these
phases, so in principle, these polarities could be used as additional input to program Focmec
after allowing for any polarity flips on reflection or conversion. If such arrivals are well
recorded, one could use amplitude ratios from among them (but one would have to include
conversion/reflection coefficients). However, in this particular case, one arrival is probably not
the one predicted by the iaspei-tau package, so care must be taken. A figure showing the KEV
seismograms and a discussion are included in a PDF file ./doc/sakhalin-doc/KEV phases.pdf
refracted from Snoke(2009).

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 11

V. References

Aki, K. and P. G. Richards (2002). Quantitative Seismology: Theory and Methods,

W.H. Freeman, San Francisco, CA, Second edition.

Booth, D. C. and S. Crampin (1985). Shear-wave polarizations on a curved wavefront at an

isotropic free surface, Geophys. J. R. astr. Soc., 83, 3135.

Chapman, M. C., C. A. Powell, G. Vlahovic and M. S. Sibol (1997). A statistical analysis of

earthquake focal mechanisms and epicenter locations in the eastern Tennessee seismic
zone, Bull. Seism. Soc. Am., 87, 15221536.

Dreger, D. S., H. Tkalcic, M. Johnston (2000). Dilational processes accompanying earth-

quakes in the Long Valley Caldera, Science, 288, 122125.

Dziewonski, A. M., T. A. Chou, and J. H. Woodhouse (1981). Determination of earthquake

source parameters from waveform data for studies of regional and global seismicity, J.
Geophys. Res., 86, 28252852.

Goldstein, P., D. Dodge, M. Firpo, and. S. Ruppert (1998), Whats new in SAC2000?
Enhanced Processing and Database Access, Seism. Res. Letters, 69, 202204

Gubbins, D. (1990). Seismology and Plate Tectonics, Cambridge University Press,

Cambridge

Herrmann, R. B. (1975). A students guide for the use of P and S wave data for focal
mechanism determination, Earthquake Notes, 46/4, 2939.

James, D. E. and J. A. Snoke (1994). Structure and tectonics in the region of flat subduction
beneath central Peru. Part I: Crust and uppermost mantle: J. Geophys. Res., 99, 6899
6912.

Jost, M. L. and R. B. Herrmann (1989), A students guide to and review of moment tensors,
Earthquake Notes, 60/2, 3757.

Kennett, B. L. N., and E. R. Engdahl (1991). Traveltimes for global earthquake location and
phase identification, Geophys. J. Int. 122, 429465. Source code for building the tables
for both iasp91 and ak135 can be gotten from URL http://www.iris.edu/pub/programs/iaspei-t

Kisslinger, C. (1980). Applications of S to P amplitude ratios for determining focal mecha-

nisms from regional network observations, Bull. Seism. Soc. Am., 70, 9991014.

Kisslinger, C., J. R. Bowman, and K. Koch (1981). Procedures for computing focal mecha-
nisms from local (SV /P ) ratios, Bull. Seism. Soc. Am., 71, 17181729. Correction for
both this and the preceding paper was published in BSSA volume 72 on page 344.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 12

Kisslinger, C., J. R. Bowman, and K. Koch (1982). Determination of focal mechanisms

from (SV /P ) amplitude ratios at small distances, Phys. Earth and Planet. Inter., 30,
172176.

Lahr, J. C. (1999), HYPOELLIPSE Y2K: A computer program for determining local earth-
quake hypocentral parameters, magnitude, and first-motion pattern, U.S. Geological
Survey Open-file Report 99023, 112p. On-line: http://greenwood.cr.usgs.gov/pub/open-
file-reports/ofr-99-0023. http://geohazards.cr.usgs.gov/iaspei pgms/hypoellipse/ is the
URL for the current version of the package, including documentation and executables
for both PC and Sun/Unix.

Randall, G. R., C. J. Ammon and T. J. Owens (1995). Moment-tensor estimation using

regional seismograms from a Tibetan Plateau portable network deployment, Geophys.
Res. Letters, 22, 16651668.

Sipkin, S. A. (2001). USGS Moment tensor software and catalog, this volume.

Snoke, J. A., J. W. Munsey, A. C. Teague, and G. A. Bollinger (1984), A program for focal
mechanism determination by combined use of polarity and SV -P amplitude ratio data,
Earthquake Notes, 55, #3, 15.

Snoke, J.A. (1989). Earthquake Mechanisms, Encyclopedia of Geophysics (D. E. James,

Ed.), Van Nostrand Reinhold Company, New York, 239245.

Snoke, J.A. (1990), Clyde and the Gopher: a preliminary analysis of the 12 May 1990
Sakhalin Island event, Eastern Section of the SSA meeting (Blacksburg, VA, October),
Seism. Res. Letters, 61, 161.

Snoke, J.A. (2009), Traveltime Tables for iasp91 and ak135, Seism. Res. Letters, 80(2),
260262.

Stein, S. and E. Klosko (2002), Earthquake Mechanisms and Plate Tectonics, Interna-
tional Handbook of Earthquake and Engineering Seismology (W. H. K. Lee,
H. Kanamori, P. C. Jennings, and C. Kisslinger, Eds.), Academic Press, San Diego,
Chapter 7.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 13

Appendix A: Input conventions and format for Focmec and Focplt

This is a text file describing the input format for Focmec (done in subroutine focinp.f)
and which is used in program Focplt (through subroutine prplot.f) for displaying first motions
and/or ratios.

READ(1,(A)) COMMNT ! First line input file

200 READ(1,5,END=300,ERR=500) STA,AZIN,TOANG,SENSE,
. RATLOG,SVSH2,STOANG,INFO
5 FORMAT(A4,2F8.2,A1,F8.4,1X,A1,1X,F6.2,1X,A)

STA is a name for the station. Up to four characters. Not used

AZIN is the azimuth (degrees) of the station measured from the epicenter

TOANG is the takeoff angle (degrees) of the S if S polarity or a SV /P , SH/P , or

SV /SH ratio. It is a P takeoff angle for a P polarity

SENSE is a one-character key for a polarity or a ratio

Virginia Tech Symbol conventions:

C = compression at station (plotted as a hexagon symbol number NS = 1)
U = same as C (up on vertical)
D = dilatation --- down on vertical (plotted as a triangle NS=2)
+ = emergent compression (NS=3) (optionally included in Focmec)
- = emergent dilatation (optionally included in Focmec
e = emergent P arrival
> = SH first motion to right (back to event, facing station) impulsive
< = SH first motion to left (back to event, facing station) impulsive
L = same as <
R = same as > (Note: earlier versions used R for SV/P ratio)
F = SV first motion away from event impulsive
B = SV first motion towards event impulsive
The SH and SV first motions are plotted as oriented arrows (NS=6)
l = SH first motion to left (back to event) emergent
r = SH first motion to right (back to event) emergent
u = emergent SH arrival
V = Log10(SV/P) (NS=4) (plotted as an X)
H = Log10(SH/P) (NS=4) (plotted as an X)
S = Log10(SV/SH) (NS=4) (plotted as an X)

If one wants to have errors marked, follow the polarity or ratio with a duplicate line with an
E in the symbol position. The error will then be flagged by a square superimposed on the
polarity or ratio. If SENSE is R, V, or H:

RATLOG is the log10 (S/P )

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 14

SVSH2 is the polarity of numerator S. It is not used, but is shown in the listing file
along with the calculated values. Convention used for the calculated ones are L and R
for SH, B and F for SV

TOANG1 is the S takeoff angle

INFO is a field of up to 40 characters. If the input file was prepared by Focmec prep, it
summarizes information used in the amplitude ratio. (See ./sample runs/sakhalin for
an example.)

No terminator line is needed. MAX in FOCMEC.INC is the maximum number of allowed

ratios plus polarities.

Appendix B: Glossary for Focal Mechanisms

Seismic Moment Tensor: The ground displacement from an earthquake can be written
as the convolution of a moment tensor and the gradient of a Greens function. The Greens
function includes propagation effects, while the moment tensor contains information about the
source properties. The moment tensor can be interpreted as the volume integral of the stress
release associated with the earthquake. The most general moment tensor has six independent
components, but the focal mechanisms for the vast majority of earthquakes are well described
by a double couple, for which the moment tensor has only three independent components. See
Jost & Herrmann (1989) for a comprehensive review of moment tensors.

Figure B1: Schematic of the foot-wall side of a fault showing the conventions for
the fault parameters: strike (), dip (), rake (), and slip direction (~a). The normal
to the fault is ~n. (After Snoke, 1989, Figure 2, which was in turn adapted from Aki
and Richards, 2002, Fig. 4.13.)

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 15

Strike, Dip and Rake: One way to parameterize a double-couple solution is to specify,
with three angles, the direction of a vector along the direction of slip of the earthquake along
a fault surface. The strike () is the azimuth of the fault, with the convention that if one
faces down-dip, the strike direction is to the left. The dip () is measured down from the
horizontal and is bounded by 0 and 90 . The rake () is along the direction of slip on the
fault surface and is bounded by 180 and +180 . The convention for rake angle : Reverse
fault if 0 < < 180 , normal fault if 180 < < 0 , right-lateral strike slip if = 180 ,
and left-lateral strike slip if = 0 . Note that with z down (Figure B1), both (x, y, z) and
(North, East, z) are right-handed coordinate frames.

Figure B2: Two representations of the double-couple point-source model for strike-
slip motion on a vertical fault. (a) moment tensor is evaluated in a coordinate frame
such that x1 lies along ~a. the slip direction, and x2 along ~n, the normal to the fault
plane. (Direction x3 is into the page.) (b) normal forces corresponding to the same
moment tensor in a coordinate frame rotated clockwise from the one in (a) By 45
about x3 . T is the tension axis, and P the compression axis. (After Snoke, 1989,
Figure 3.)

Pressure and Tension Axes: As shown in Figure B2, an alternative parameterization for a
focal mechanism is to specify the trend and plunge of the P axis (direction of maximum source
compression that is the same as the direction of maximum dilatation at the recording site)
and T axis (direction of maximum source dilatation or recording-site compression). Because
the P and T axes are orthogonal, there are only three independent parameters from among
the two sets of trends and plunges, so there are still only three independent parameters in this
representation. This representation for a focal mechanism is particularly useful for relating
earthquake mechanisms to directions of compression or tension of the regional stress.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 16

Focal Sphere: Modeling the earthquake as a point source, the focal mechanism can be
uniquely specified by specifying the directions of P and T with respect to a sphere of vanish-
ingly small radius surround the focus (Figure B2b). Observations of polarities and amplitudes
used to constrain the solution must be corrected to points on the focal sphere designated by
the takeoff angle and azimuth for the raypath leading to that station.

Figure B3: Radiation patterns for (a) P waves and (b) SH waves for a vertical
strike-slip fault oriented as in the previous figure. In (a) the + and indicate the
direction of first motion of the P with respect to the source, and the arrows in (b)
indicate the first-motion directions for SH. (After Snoke, 1989, Figure 1 except
the direction of the SH polarity arrows have been reversed from that figure)

Radiation Pattern: A double-couple solution produces a radiation pattern for P waves with
four quadrants that are alternatively compressive and dilatational. Figure B3 shows the four
quadrants for both the P and SH radiation patterns.

Fault Plane and Auxiliary Fault Plane: The radiation patterns for SV and SH have
similar four-fold symmetry to that for P , and the result is that there are two equally possible
fault planes consistent with all polarity and ratio data for a single event. Hence for each
solution, two possible sets of dip, strike and rake are presented, and each fault-plane plot
includes two planes. Solutions produced by program Focmec include the trend and plunge of
vectors A and N, which are along the slip directions in the two possible fault planes (~a and
~n in the preceding two figures in this section). These vectors can be gotten from P and T by
a rotation of 45 around the vector B, which is along the normal to the plane formed by P
and T.

Sign Conventions for Components of Ground Motion: Polarities in the FOCMEC

package are designated by directions with respect to an observer facing the station with
his/her back to the epicenter. Accordingly, P arrival first motions are up or down, SH left or
right, and SV forward or backward.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 17

Horizontal components for a three-component seismograph must be rotated from North-

South and East-West into Radial and Transverse to separate SV from SH. The convention
in common usage e.g., in SAC has positive Radial as forward and positive Transverse
as to the right with respect to an observer facing the station. Note that Vertical, Radial, and
Transverse form a left-handed coordinate frame.

North-South

Vertical

East-West

Radial
Radial

Transverse

Vertical Transverse (x 0.3)

Figure B4: Components for P (left) and S (right) arrivals at station COR for the
Sakhalin Island event. The backazimuth is 311 , so the direct arrivals come from
the northwest. Positive for the Radial component is hence towards the southeast,
and positive on the Transverse is towards the southwest. For display purposes, the
P arrivals are lowpass filtered and S on the transverse component is rescaled.

From Figure B4, one can see the basis for the polarity choices for station COR in the focmec
input files in subdirectory ./sample runs/sakhalin: D (down on Vertical) for P , B (down on
Radial) for SV , and R (up on Transverse) for SH.

Upper- and Lower-Hemisphere Projections: The projections of the focal spheres shown
above in Figures 1 and 2 are lower-hemisphere projections. This is the more popular projection,
presumably because most data are teleseismically recorded. Some times an upper-hemisphere
projection is used, so it is instructive to review briefly how these projections are related and
how one constructs them.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 18

The particle motion is identical for antipodal points on the focal sphere, so a projection
(from above) of station positions and focal mechanism onto the upper hemisphere can be
mapped onto the lower hemisphere by simply adding 180 to all station azimuths and trends
for the P and T axes.
To get such plots, one projects the position on the focal sphere of the P and T axes
(trends and plunges) and station polarities or amplitude ratios (takeoff angles and azimuths)
onto a circle. For a station position, if it is an upper-hemisphere projection and the takeoff
angle is up, the mapping is straightforward. If for that projection the takeoff angle is down,
one maps the antipodal position onto the circle the supplement of the takeoff angle and
azimuth plus 180 . For a lower-hemisphere projection, the mapping is the opposite. It is
easiest to derive the nodal surfaces from the directions of the P and T axes: As these axes
project in antipodal directions from the focus, to map a lower-hemisphere vector into its
upper-hemisphere counterpart, one simply adds 180 to the trend.

Beachball Representations of Focal Mechanisms

Figure B5: Shown are beachball representations of the P -wave radiation pattern for
the three fundamental double-couple fault types: (a): pure strike-slip faulting, (b):
pure dip-slip reverse (thrust) faulting, and (c): pure dip-slip normal faulting. Above
each is a sketch of the fault geometry for that mode of faulting. (The auxiliary plane
is shown as a dashed line.) The compressional quadrants are shaded, and the sign
of the direction of vertical-component first motion is included for each quadrant.
(Usually these signs are not included.) Such plots are an efficient method to show
trends in focal mechanisms for groups of earthquakes. See, for example, figures
in the paper by Stein & Klosko, 2002. Such plots are also used to represent non
double-couple mechanisms with nodal-plane projections replaced by nodal-surface
projections. (After Gubbins, 1990, Figure 6.7.)

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 19

Appendix C: SV Polarities and Amplitudes Should be Used with Care

Kisslinger and coauthors (1980, 1981, 1982) describe procedures for using SV /P vertical-
component amplitude ratios recorded by local and regional networks to constrain focal mech-
anisms. The emphasis was on the vertical components because at that time most networks
used single-component seismographs and analog recording.
With modern instrumentation, one can often find well recorded SV arrivals although
generally not as well recorded as for the deep-focus Sakhalin Island event discussed above. We
show below why SV should be used only with great care.

Figure C1: Ray geometries showing directions of positive first-motions for free-
surface incidence of P (left) and SV (right). This conventions differs from that used
by Aki & Richards (2002, Figure 5.5) only for the direction of incident SV . I use
this convention because for both P incidence and SV incidence, there is the same
relationship between propagation directions and particle motion for incoming and
reflected rays.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 20

Figure C2: Free-surface amplitudes (left) and phase shift (right) for emergent P and
SV at the free surface for a vS /vP ratio of 0.6. The critical angle is sin1 (vS /vP ) =
36.9 .

20 Degrees 20 Degrees

40 40

60 60

80 80

Figure C3: Waveforms for the vertical ground motion (left) and the radial ground
motion (right) for plane wave SV incidence angles of 20, 40, 60, and 80 degrees in a
uniform halfspace with a vS /vP ratio of 0.6. Polarities follow the convention defined
in Figure C1.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 21

The input waveform for the ground motion shown in Figure C3 is normalized so that the
area under the curve is the same as the amplitude for that component shown in Figure C2
when there is no phase shift. For P incidence, there is no critical angle and the polarities are
all positive using the convention defined in Figure C1, so there is zero phase shift for all angles
of incidence. The rest of this discussion is for SV incidence.
As discussed by Aki & Richards (2002, page 151), the sign of the phase shift depends
on the Fourier sign convention and on the sign of the frequency. The phase shifts in Figure
C2 use their conventions: positive frequencies and +it for the forward Fourier transform.
Because of the different polarity for SV incidence between them and Figure C1, phase shifts
presented here will differ by 180 from theirs.
Below critical angle incidence, the phase shift is zero for the vertical ground motion and
180 for the radial motion (see waveforms for 20 incidence in Figure C3). At 45 incidence,
the phase shift for the radial component jumps by 180 , which can be seen in the polarity flip
between the waveforms on the radial component for 40 and 60 in Figure C3.
The rapid changes in amplitude within a few degrees of the critical angle mean that neither
the vertical nor radial component amplitudes should be used for arrivals at those angles.
One sees from Figure B4 above that for small emergence angles (22 at COR for S for this
event), SV is much better recorded on the radial than the vertical. In general, below about
30 SV on the radial component is the more reliable. For angles above the critical angle, the
phase shift results in a drastic change in waveform for the radial component, so SV on the
radial should definitely not be used in ratios and used only with great care for polarity.
For the vertical component with SV incidence, the polarity is stable up to about 80 ,
so those polarities can be used if the arrivals are well recorded. Up to about 50 , the small
phase shift results in a negligible change in the amplitude, so the vertical component could
be used in an amplitude ratio. At 80 , the waveform shown in Figure C3 is visibly changed
and the amplitude is reduced to 0.27/0.52 = 0.52 the amplitude predicted by the free-surface
correction plotted in Figure C2.
Kisslinger et al. (1961) write that the free-surface correction for the SV /P ratio on the
vertical, the free-surface correction is close to unity for incidence angles from a few degrees
above the critical angle to about 80 . While that is true based on the amplitudes shown
in Figure C2 (see also their Figure 1), the effect of the phase shifts on the waveforms and
amplitudes leads to a much larger variation for higher angles.
The above discussion is based on plane wave incidence appropriate for teleseismically
recorded events. In their paper on shear-wave polarization, Booth & Crampin (1985) show
that for curved wavefronts associated with local to near-regional events, an SP arrival that
comes in just before direct S affects the observed waveforms. In addition, a low-velocity
zone at the surface can complicate the waveforms particularly for SV incidence. Such
considerations must also be considered when choosing polarities and amplitudes for focal
mechanism determinations.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009

FOCMEC Page 22

Appendix D: Graphics in the FOCMEC package

As described above, program Focplt produces plots of focal-sphere projections of data and/or
fault-plane solutions. Earlier versions of the FOCMEC package used the SAC (Goldstein, et al.
1998) Fortran library that was distributed with pre-1994 versions of SAC. This library allowed
the user to create both plot files and interactive on-screen plots. The screen-plot option has
been dropped, because the 15-year old libraries were obsolete. Program Focplt continues to
use the SAC Graphics Format (SGF) to produce disk files. Program sgftops converts .sgf files
to postscript files. The FOCMEC package can be compiled and run on big-endian systems
(Sun Solaris, Mac OSX PPC) or little-endian systems (Linux, Mac OSX i686/intel) that
have the opposite byte order for binary data. Because a .sgf file is a binary file, program
sgftops will only work on a .sgf file that has one byte order. If there is a byte-order mismatch,
sgftops will exit with the message that one should run program sgfswap on the file. Program
sgfswap is provided with the package.
Also included are four (Unix csh) scripts:

1. plsgf: produces a print of a .sgf file.

2. sgftox: Displays a .sgf plot on the screen. As written, the script uses program gs
(ghostscript), but one can substitute a different postscript-file display program.

3. sgftoeps: produces a EPS file (Encapsulated Postscript file) from a .sgf file. Program gs
is required.

4. sgftopdf: produces a PDF file from a .sgf file. Program gs and the Perl script epstopdf
are required.

To find out more the programs or scripts, enter the program or script name with no arguments.
The file ./doc/graphics.html has more information about how to test the graphics package
and how one can use the symbol routine to label plots.
Several scripts in Section IV run program Focplt to produce SGF files. Note that those
for the Sakhalin Island event superimpose sets of solutions, so studying them may be helpful
regarding how one answers the prompts for adding additional solutions. The screen output
for these scripts is ported into a file named a.junk in the same directory, so if the script exits
with an error message, comparing the contents of that file with the calling script should help
the user find the error.
The scripts have calls to sgftox and it sgftopdf, and the PDF plots are included in the
./doc subdirectories for each application. The figures in this document were produced from
SGF plots using sgftoeps.

Package URL: http://www.iris.edu/software/downloads/processing/ June 24, 2009