for a seismic network
The development of the Experimental Design software is
supported by:
Istituto
Nazionale di Oceanografia e di Geofisica Sperimentale - OGS, Udine,
Italy
The Experimental Design software is developed by:
Anthony
Lomax
in collaboration with:
Alberto
Michelini, Andrew
Curtis, and David
Leslie,
This Experimental Design Applet is an aid for
deciding which of a set
of potential new receivers are likely to most improve
the capability of
a new or existing, local or regional seismic network to monitor given
seismic source
regions.
If you use this algorithm or
software for a
publication or on the WWW, we would most
gratefully receive a reference! Please reference this web page
(http://www.alomax.net/projects/expdesign) and the article:
Curtis, A., A. Michelini , D. Leslie and A.
Lomax, (2004), A
deterministic algorithm for experimental design applied
to tomographic and microseismic monitoring surveys, Geophys. J. Int., 157, 595–606. (view PDF *notice)
If you prefer a C version of the algorithm
as presented in the article, contact Andrew Curtis at
Edinburgh University, UK:
www.geos.ed.ac.uk/homes/acurtis
Illustration:
The image on the left shows the existing Friuli,
Italy network (blue
triangles), target source regions (green circles), and potential new
recievers (grid of grey triangles). The image on the right shows 15
optimal new recievers (red triangles) selected using the default values
in the Friuli Network example: grid w/ all existing stations
link below.
The importance of each receiver is given by a relative quality factor (varying from 0 for poor quality to 1 for high quality).
- The quality factor for a receiver will be higher if it provides information and constraint on event locations that is not provided by other existing or potential receivers. This constraint is determined from the vectors of partial derivatives of travel-time at the receiver with respect to perturbations in the positions and origin times of events in the target source regions. These are the partial derivatives for the event location problem; they depend on the velocity model and the distance and azimuth of the receiver from the sources.
- The quality factor for a receiver will be lower if it provides information on the event locations that duplicates that of other receivers, or if it is far from the source regions (depending on the attenuation relation).
The Experimental Design Applet is based on the algorithms
introduced in:
Curtis, A., A. Michelini , D. Leslie and A. Lomax, (2004), A deterministic algorithm for experimental design applied to tomographic and microseismic monitoring surveys, Geophys. J. Int., 157, 595–606. (view PDF *notice)
The implementation of the design equations in the applet is
presented here:
Experimental Design Applet
- Equations
The examples have preset lists of sources, receivers and all other parameters. To run an example, you can go directly to tabbed pane [5. Run the design] and press the [ Run design > ] button.
You can also edit any of the source or receiver lists or other parameters.
Each tabbed pane has a [HELP] button which gives further information.
- Experimental
Design Applet - Seismic Network example
- Experimental
Design Applet - Big Seismic Network example
- Experimental Design
Applet - Friuli Network example
- Experimental
Design Applet - Friuli Network example: grid w/ all existing stations
- Experimental
Design Applet - Friuli Network example: grid w/o existing stations
- Experimental
Design Applet - NE Italy example: grid w/ all existing stations
(NOTE: Attenuation->Joyner-Boore->Cutoff set to 150km) - Experimental
Design Applet - NE Italy example: grid w/o existing stations
(NOTE: Attenuation->Joyner-Boore->Cutoff set to 150km) - Experimental Design Applet - CMLA paper, microsiesmic, constant velocity (rescaled, no attenuation, P only)
To run your own experimental design, edit the parameters in each tabbed pane, and paste your source, existing receiver, and potential receiver lists into the appropriate text boxes. Then go to tabbed pane [5. Run the design], edit the parameters, and press the [ Run design > ] button.
You can also add a set of map lines in tabbed pane [6. Extras].
Each tabbed pane has a [HELP] button which gives further information.
- Suggestion: If the optimisation runs slowly, try unchecking the [Show map] option in the [5. Run the design] pane.
- For the homogeneous half-space model, the velocity is irrelevant to the results.
- There are two optimisation algorithms:
- Summation over qualities - The algorithm
as in the paper (quality = a summation over each other receiver of [the
sum over each source of the quality of the target receiver with respect
to the other receivers]).
This algorithm favors the selection of receivers that constrain the most isolated sources, but does not penalize strongly closely positioned receivers (thus it favors data redundancy). - >Product of summation over source qualities
- A modified algorithm (quality = a product over each other receiver of
[the sum over each source of the quality of the target receiver with
respect to the other receivers])
This algorithm favors the selection of receivers that constrain many events, and it penalizes strongly closely positioned receivers (thus it avoids data redundancy).
- Summation over qualities - The algorithm
as in the paper (quality = a summation over each other receiver of [the
sum over each source of the quality of the target receiver with respect
to the other receivers]).