• Antelope Release 5.5 Linux 2.6.32-220.el6.x86_64 2015-04-21

 

NAME

pmelavg - average grid-based solutions produced by pmel

SYNOPSIS

pmelavg db [-pf pffile]

SUPPORT


Contributed code: NO BRTT support.
THIS PIECE OF SOFTWARE WAS CONTRIBUTED BY THE ANTELOPE USER COMMUNITY. BRTT DISCLAIMS ALL OWNERSHIP, LIABILITY, AND SUPPORT FOR THIS PIECE OF SOFTWARE.

FOR HELP WITH THIS PIECE OF SOFTWARE, PLEASE CONTACT THE CONTRIBUTING AUTHOR.

DESCRIPTION

The implementations of pmel (dbpmel(1) and pmelgrid(1)) focus on geographically defined gridding described in cluster(1). That is, they aim to satisfy the approximations of multiple event location methodologies, of which pmel is an example, by only utilizing groups of events located close together in space. In this process, however, it is inevitable that many if not most events in a catalog get associated with more than one target point in space. As a result, the output of pmel often has multiple origin estimates for the same event from overlapping constellations of events associated with different grid points. pmelavg averages these multiple estimates in a fairly simple way and defines its estimate as the prefor for that evid.

The averaging algorithm is the local OR of two fairly simple algorithms. (i.e. you can use one or the other or both cascaded together) The first approach weights each location estimate by the reciprocal of the rms residual for that event (sdobs found in the origerr of CSS3.0). The second uses a distance dependent weight. Each event is compared to it's distance from the hypocentroid of the group with which it is associated. Below a cutoff distance set in the parameter file (see below) an event is given full weight. If the distance of an event from the hypocentroid of the group is larger than the cutoff a 1/r weight is applied (correctly normalized to have a unit weight at the cutoff distance.). To make the algorithm a simple mean simply make the cutoff distance a large number. To make the average approximate a nearest neighbor estimate, make r smaller than the nominal grid spacing.

If both averaging methods are selected, the result is a weighted sum formed by multiplying the weights of the two terms together. If only one method is selected, the weights are as described above.

OPTIONS

Use the -pf option to specify an alternative pf to the stock one defined by the program name (i.e. pmelavg.pf).

PARAMETER FILE

pmelavg has one major parameter that controls it's behaviour. The real parameter full_weight_distance sets the cutoff parameter described above. Units are assumed to be kilometers.

The two weighting methods are controlled by two boolean parameters called use_inv_r_weighting and use_ssr_weighting.

This program also uses dbprocess to form a working view. The sequence passed to dbprocess is defined by the input Tbl pmelavg_dbview. This should not normally be altered. It was made a parameter instead of being hard wired into the program to make it less susceptible to schema changes and changes in attribute naming conventions. One exception is that the following line:

    dbsubset algorithm=~/pmelgrid/
is variable depending on the version of pmel used. The above form is correct for pmelgrid. If you used dbpmel, change pmelgrid to dbpmel.

ATTRIBUTES

NOT MT-Safe

SEE ALSO

dbpmel(1), cluster(1), makegclgrid(1)

BUGS AND CAVEATS

Making the cutoff very small could cause problems in an obvious way with a divide by 1/r weighting scheme. Anything rational for this parameter (e.g. larger than 0.1) should not cause a problem.

AUTHOR

Gary L. Pavlis
Indiana University
pavlis@indiana.edu

Antelope User Group Contributed Software
Printer icon