There is a bug in the implementation of frozen profiles
(fictitious sites beneath which the model is held fixed at its
starting value). This feature works properly ONLY if the
frozen profile lies WITHIN the sites with actual data.
Thus you cannot, at present, freeze the model outside of
the data. This will be fixed in the next release.
If this is a critical problem for anyone, I will explain
how to fix it yourself.
The paper: Rapid Two-Dimensional Inversion of COPROD2 Data,
Wu, N, Booker, J.R. and Smith, J.T.,
J. Geomag. Geoelec., 45, 1073-1087, 1993.
describes useful strategies for inverting for static shifts
using inv2d. If you simply set all the shifts free, you are
likely to converge to a local minimum of the object function
in which the resistivity data are well-fit, but the phase
data are ignored. The basic idea behind all strategies that
we have found useful is to weight phase data more strongly
until their misfit is small and then bring in the resistivity.
***NOTE*** This problem noted in last section can often be
reduced by damping the inversion for static shifts
(see new release notes below).
***NOTE*** The following comment CONTRADICTS a statement made in
file READ_ME supplied with inv2d.
When trying to fit TM mode data alone, experience has shown that
varying the parameter that controls the depth-dependence of the
norm ("eta") with iteration number can be very useful. For instance
setting eta to 0.0 for the first iteration and letting it
increase to 1.0 or 1.5 has the effect of forcing inv2d to fit
the data with very shallow structure first and incorporating
deeper structure only when shallow structure cannot do the job.
A new release of inv2d is planned for early 1995. It will
have significant advantages over the current version:
a) Discontinuties at which the model smoothness condition
is suspended can be declared anywhere in the model. This is
very useful for incorporating geological information, such
as independently determined depth to conductive or resistive
basement, mapped faults, etc.
b) Individual model nodes can be frozen at their starting value.
This is again useful for incorporating a priori information
or testing hypotheses.
c) Inversion of static shift coefficients can be damped. This is
useful when you want the statics to change only if the data
cannot be fit without changing the statics.
d) Pseudo (or virtual) sites can be declared. These are sites at
which an inversion is performed, but there is no measured
data. They allow model freedom between sites with data. Normally,
the model will simply be smoothed into such a site from adjacent
sites. However, sometimes model changes beneath a pseudo-site
can improve data fit at other sites. With no pseudo-site, the
routine which interpolates structure between data sites would
not allow such structure to exist.
e) Outputs surface magnetic and electric fields as well as MT data.
f) Outputs file containing the fields inside the model. These
can be used to restart the inversion for further iterations.
They will greatly speed up the first iteration of the restart
and in most cases, will permit using 32 bit rather than 64 bit
arithmetic for the inversion. This will halve memory usage and
substantially improve speed on many machines.
g) Versions for 486, Pentium, PowerPC and Mac machines as well
as Unix workstations.
Features that are in testing. Some may make it into the next release.
h) Inversion of vertical to horizontal magnetic field transfer
function data (sometimes called "tipper" by those trying to
confuse people).
i) Surface topography. This will include both the air-earth interface
and the seafloor.
a) Discontinuties at which the model smoothness condition
is suspended can be declared anywhere in the model. This is
very useful for incorporating geological information, such
as independently determined depth to conductive or resistive
basement, mapped faults, etc.
b) Individual model nodes can be frozen at their starting value.
This is again useful for incorporating a priori information
or testing hypotheses.
c) Inversion of static shift coefficients can be damped. This is
useful when you want the statics to change only if the data
cannot be fit without changing the statics.
d) Pseudo (or virtual) sites can be declared. These are sites at
which an inversion is performed, but there is no measured
data. They allow model freedom between sites with data. Normally,
the model will simply be smoothed into such a site from adjacent
sites. However, sometimes model changes beneath a pseudo-site
can improve data fit at other sites. With no pseudo-site, the
routine which interpolates structure between data sites would
not allow such structure to exist.
e) Outputs surface magnetic and electric fields as well as MT data.
f) Outputs file containing the fields inside the model. These
can be used to restart the inversion for further iterations.
They will greatly speed up the first iteration of the restart
and in most cases, will permit using 32 bit rather than 64 bit
arithmetic for the inversion. This will halve memory usage and
substantially improve speed on many machines.
g) Versions for 486, Pentium, PowerPC and Mac machines as well
as Unix workstations.
Features that are in testing. Some may make it into the next release.
h) Inversion of vertical to horizontal magnetic field transfer
function data (sometimes called "tipper" by those trying to
confuse people).
i) Surface topography. This will include both the air-earth interface
and the seafloor.
j) Semi-automatic mesh control. This will warn user that the mesh
may be inaccurate and suggest appropriate modifications.
k) Interface for GEOTOOLS-MT.
How to get inv2d, the rapid 2D MT inversion program (commonly called RRI). use command: ftp www.geophys.washington.edu -or- ftp 128.95.16.50 at login prompt type: anonymous at password prompt type: your last name or userid then type commands: cd pub/out/RRI get README.FIRST mget Inv2d.news* bin <-- VERY IMPORTANT get newrri.tar If you want the imagetool executable (see README.FIRST) get imagetool.Z finally, logout using bye You now need to unpack the Unix tar file inv2d.tar using the command tar -xf inv2d.tar This will set up directory structure. There are several readme files and a runme file that will get you started. Good Luck! John Booker
You can also download the source directly using the following ftp link:-