MTINV

1-D interpretation of magnetotelluric EM soundings

Version 1.3 (c) 2004, Markku Pirttijrvi

Introduction:
The MTINV program can be used to model and to interpret geophysical electromagnetic (EM)
magnetotelluric (MT) measurements using a horizontally layered earth model. The MTINV
program uses either the apparent resistivity and phase components of the MT data together or the
apparent resistivity data alone. Parameter optimization, which is based on linearized inversion
method, can be utilized in 1-D interpretations.
The MTINV program is a 32-bit application that can be run on Windows 9x/NT4/2000/XP with a
graphics display of at least 1024800 resolution. Memory requirements and processor speed and
are not critical, since large data arrays are not used and the EM solution is fast to compute even on
older computers. The MTINV program has a simple graphical user interface (GUI) that can be used
to change the parameter values, to handle file input and output, and to visualize the EM response
and the model. The user interface and the data visualization are based on the DISLIN graphics
library (http://www.dislin.de).

Installing the program:

The program requires the following two files:
MTINV.EXE
DISLIN.DLL

the executable file

dynamic link library for the DISLIN graphics

The distribution file (MTINV.ZIP) contains also a short description file (_README.TXT), this
user's manual (MTINV_MANU.PDF) and an example data file (EXAMPLE.DAT). To be able to
read this user manual you should have Adobe's free Acrobat Reader program. To install the
program just unzip (Pkzip/Winzip) the distribution files into a new folder. If you wish, create a
program shortcut on the desktop. To be able to start the program from a shortcut that locates in a
different directory move or copy the DISLIN.DLL file into the WINDOWS\SYSTEM folder or
somewhere where a path always exists.

Starting the program:

On startup the program reads the model parameters from the MTINV.INP file and the DISLIN
graphics parameters from the MTINV.DIS file. If these files cannot be found, new ones with
default parameter values are automatically created. The program then computes the MT response of

the initial model and builds up the user interface shown in the Appendix. The MT response is
plotted in the graph area along with a model view, i.e, a resistivity-depth curve of the model and a
text description of the model parameters. The parameters of the layered earth model and the
inversion method can be altered using the controls on the left side of the MTINV program window.
As shown in Appendix the main window of the MTINV application contains two menus. The File
menu contains the following nine options:
Open Model
Save Model
Read Data
Save Results
Read disp.
Save Graph as PS
Save Graph as EPS
Save Graph as PDF
Save Graph as WMF

open an existing model file (*.INP).

save the model into a file (*.INP).
read in measured data for interpretation (*.DAT).
save the results into a file (*.OUT).
read in new graph parameters from a file (*.DIS)
save the graph in Adobe's Postscript format (*.PS).
save the graph in Adobe's Encapsulated Postscript format (*.EPS).
save the graph in Adobe's Acrobat PDF format (*.PDF).
save the graph in Windows metafile format (*.WMF).

Selecting these menu options brings up the Windows file selection dialog that can be used to
provide the file name for the open/save operation. Model file, results file, and the graph parameter
file are stored in text format. The graphs are saved as they appear on the screen in landscape A4
size.
The Edit menu contains following items:
Comp.->Meas.
Remove Meas.
Phase in/out
Weights in/out
Time periods
Axis min/max
Error bars

put and use the computed response as measured data.

remove all information about measured data.
select and to deselect the phase component.
include or exclude the data weights in the inversion.
redefine the time periods in forward modeling.
redefine the min & max values of x axis and y axes.
include/exclude the error bars from model view.

The Comp.->Meas. item can be used to test the program and the inversion method with synthetic
field data. The Remove Meas. item is used to remove any information about measured data
(synthetic or not). The Phase in/out item provides a possibility to include/exclude the phase
component in/from the computation and visualization. If the phase component is omitted the
inversion will use only the apparent resistivity data. The Weights in/out option is mainly used to
test how the weights affect the inversion.
After applying the Time periods menu item, the program will ask the user (on the console window)
if the periods should be computed automatically or if they are to be provided manually. After this
the program will ask the number of periods. Giving a value less than two will ignore the task and
the program will use the old values. If the periods are to be computed automatically, the program
will ask the minimum and maximum values. The periods are then computed so that they are evenly
distributed on logarithmic scale. If the periods are to be given manually, the program will ask the
value of each period separately. The periods must be given in an ascending order.

The Axis min/max menu item allows changing the minimum and maximum values of the x axis
(period) and the y axes (apparent resistivity and phase). The program will ask the new values on the
console window. Giving a zero value will keep the old axis minimum or maximum value.
The Error bars item can be used to change the appearance of the error bars of the resistivity and
thickness of the layers in the model view. The error bars represent the 95% confidence limits
(minimum and maximum errors) obtained from the singular value decomposition used in the
linearized inversion method.
The Exit menu has only one item, which is used to confirm the exit operation. On exit the latest
model is saved in the MTINV.INP file replacing an existing one, and the results are saved in the
MTINV.OUT file. Before exiting, one can use the Save results item to provide a user defined name
and location for the output files. Errors that are encountered before the GUI starts up are reported in
the MTINV.ERR file. Inside the GUI mode run-time errors are displayed on the screen.
After exiting the program one should take a look at the *.OUT file that contains the results. If data
has been used in the interpretation, the file contains the model parameters, the RMS error, the mean
of the damping factors, the (damped) 95% confidence limits of the parameters, the singular values,
the damping factors, and the parameter eigenvectors. The result file also contains the measured and
computed apparent resistivities (and optionally the phases) and the differences. In addition the file
contains the data values required to create the model curve and the error bars using a third-party
plotting program. If data were not read in the output file would just contain the model parameters
and the computed MT response.
The MTINV program is not an idiot-proof interpretation program. It requires an initial model, the
parameters of which are optimized during an iterative process. The inversion method can easily get
stuck to a local minimum, and therefore, the user should pay attention to the validity of the
resulting model. Special care must be used when interpreting conductive layers, since due to the
equivalence the optimization tends to yield solutions where the layer conductance (conductivitythickness product) remains the same but the individual values may vary. The RMS error and the
mean of the damping factors can be used to assess the validity of the inversion results. Optimally
the RMS error should be zero and all damping factors should be equal to one. Note also that after
optimization the resistivity-depth section shows the minimum and maximum parameter ranges
using dotted lines. These auxiliary curves are derived from the 95% confidence limits.

Program controls:
The text controls in the leftmost control panel define the number of layers and the resistivity and
thickness values of the layers. After editing the parameter values one has the apply for the changes
using the Update parameters push button on the top left corner of the MTINV application window.
When increasing or decreasing the number of layers the unnecessary controls get hidden. The
maximum number of layers is six. The Default push button resets all model parameters to the
default (100 m and 10 km).
The text controls in the right control panel are related to the inversion. Some of these controls, as
well as the Optimize button become active only after data is read in using the Read Data menuitem.
The Iters. control defines the number of successive iteration made after the Optimize button is

pressed. The Thres. control defines the minimum singular value threshold used in the optimization.
This parameter (actually multiplied by 1000) controls the strength of the damping. Decreasing its
value loosens the damping and makes the inversion method work like a steepest descent algorithm.
Decreasing its value can be used to reduce the damping. Increasing its value might be advantageous
if the inversion is not stable. The default values of iterations and threshold are 10 and 0.01 (1.e-5),
respectively. The S.Shift control can be used to manually set a correction factor for the static shift
effect for the measured apparent resistivity curves. A value of 10.0 shifts the apparent resisitivity
curve one decade higher (the default value is 1.0).
The text fields (Free) on the right side of each resistivity and thickness value are used to fix (0) and
free (1) parameters in the inversion. By default all parameters are free and used in the inversion. If
you want to fix some parameter during inversion, set the corresponding parameter from to zero.
The Sel. all push button can be used to fix or free all parameters in the inversion.
Because the inversion can easily get stuck into a local minimum, the inversion should not be
restarted using the previous model as a starting point. Note also that new layers are added to the
bottom and given default parameter values (100 m and 10 m). A single layer can be removed if it
is given a zero thickness and the Update button is pressed.

File formats:
When using the program for interpretation purposes make sure that your input data files (*.DAT)
are formatted properly before running the program. Note also that there is no need to edit model
files manually when interpreting field data.
The following example illustrates the format of the input data file.
Synthetic data
12 2 3 4
4.00
8.00
16.00
32.00
64.00
128.00
256.00
512.00
1024.00
2048.00
4096.00
8192.00

0.12355E+03
0.20821E+03
0.34956E+03
0.54016E+03
0.67331E+03
0.65905E+03
0.65739E+03
0.82007E+03
0.10629E+04
0.11222E+04
0.10261E+04
0.99405E+03

0.74851E+02
0.74662E+02
0.71626E+02
0.64862E+02
0.55826E+02
0.50679E+02
0.53018E+02
0.55906E+02
0.52590E+02
0.46432E+02
0.44160E+02
0.44886E+02

1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.
1.

Lines 2 and 4 are used for comments and can be left empty. The first line defines a header text
(max. 40 characters). The 3.rd line defines the number of time periods (NOT) and the column
indices (ICO1, ICO2, and ICO3) for the apparent resistivity, phase, and weights. Phase data is not
read at all if ICO2= 0. The next NOT lines define the time periods (TIME, s), and the measured
apparent resistivity (RHOM, Ohmm), and optionally the phase (PHAM, deg.) data. The last column
contains the weights of each time period. Weights are not read if ICO3= 0. The weight values,

which must be positive, are relative: large weight gives that particular period more importance than
others. Zero weight means that the particular measurement is not used at all. As such the weights
can represent, for example, the inverse of measured or estimated data errors or variances. Negative
value of ICO3 is means that the weights are inverted so data errors can directly be used as weights.
Note that if the number of time periods is less than zero (NOT<0), they are interpreted as
frequencies and converted to periods accordingly. If the first column index is less than zero
(ICO1<0), the periods are interpreted as the square roots of the period and converted accordingly.
Finally, if the second column index is less than zero (ICO2<0) the sign of the phase data is
reversed.
Note that the maximum amount of time periods is 60 and the header text is used as a second line in
the response graph title. If the header text line is empty the default title in the MTINV.DIS file is
used instead. The time periods must be either in an ascending or in descending order. Note also that
the data file can contain several data columns, of which one (or two if phase is used) is read for the
interpretation. This means that the same data file can contain, for example, measurements in two
orthogonal directions. Manual editing of the column indices is (currently) required to choose the
correct data column for the MTINV program. Resistivity and phase values equal to zero are omitted
and, thus, used for missing data values.
Model files must (currently) be edited manually to provide the desired frequencies in forward
modeling. The following example illustrates the format of the model files:
MTINV model file:
3 1
0.10345E+04
0.11521E+02
0.99930E+04
16
0.1000E+01
0.6398E+02
0.4097E+04

0.34792E+02
0.10000E+01

0.2000E+01
0.4000E+01
0.8000E+01
0.1280E+03
0.2560E+03
0.5120E+03
0.8190E+04 0.1638E+05 0.3277E+05

0.1600E+02
0.1024E+04

0.3200E+02
0.2048E+04

Lines 1, 2 and 7 (in this case) are used for comments and can be left empty. The 3.rd line defines
the number of layers (NOL), the maximum amount of which is 6, and a parameter that is used to
include (1) or exclude (0) the computation of the phase response component. Lines 4 and 5 define
the resistivity (RH, Ohmm) and the thickness (TH, m) of the two topmost layers and the 6.th line
defines the resistivity of the bottom layer. The 1.st line of the 2.nd paragraph defines the number of
time periods (NOT). The periods are then provided at the end of the file in either in an ascending or
in a descending order.
The results file contains the model parameters, i.e., layer resistivities and thicknesses, as well as
layer depths. If inversion was made before exiting the file contains also the RMS error, the mean of
the damping factors, the (damped) 95% confidence limits of the parameters, the original (W) and
normalized (S) singular values, the damping factors (T) and the parameter eigenvectors (V-matrix).
The result file also contains the measured and computed apparent resistivities (and optionally the
phases) and the difference between the measured and the computed data. Note that for historical
reasons the #-character is used to comment out lines for the Gnuplot plotting program.

Graph options:
Several graph parameters (see Appendix) can be changed by editing the MTINV.DIS file. Note that
the format of the MTINV.DIS file must be preserved. If the format of the file should become
invalid, one should delete the file and a new one with default parameter values will be generated
automatically the next time the program is started. The file format and default parameter values are
shown below.
40
1
370
1.

32
32
26
26
1
7
300 0.60 0.87
10000. 1. 10000.

10.

90.

20.

MT interpretation
Test measurement
Period (s)
Apparent resistivity (Ohmm)
Phase (deg.)
Rhoa comp.
Phase comp.
Rhoa meas.
Phase meas.
Model description
Resist. (Ohmm)
Thickness (km)
RMS-error:
Damp-mean:
Depth (km)

The 1.st line defines five character heights. The first value is used for the main title and the
graph axis titles, the second is used for the axis labels, the third is used for the plot legend text,
the fourth is used for the model description text, and the last value is used for the axis labels in
the model view.
The 2.nd line defines parameters that modify the graph appearance. The first one can be used to
include (1) or exclude (0) the model information text to/from the top-right corner of the page.
The second one can be used to include (1) or to exclude (0) the model view to/from the bottomright corner of the page. The third parameter is used to define the corner where the legend text is
positioned. Values 1-4 put the legend in SW, SE, NE or NW corner of the page (outside the
graph). Values 5-8 put the legend in the SW, SE, NE, or NW corner inside the graph. The
default values are 1, 1, 7.
The 3.rd line defines first the x- (horizontal) and y- (vertical) distance of the origin of the main
graph (in pixels) from the bottom-left corner of the page. The third and the fourth parameter
define the length of the x and y axes relative to the size of the total width and height of the plot
area (eg. 0.5= 50 % of the width or height), which is equal to 29702100 pixels (landscape A4).
The 4.th line defines axis ranges. These are the minimum and maximum values of the
logarithmic x axis (period) and logarithmic apparent resistivity and linear phase axes, and the
tick-mark step of the phase axis (see Appendix).
The fifth line should be left empty.
The following lines define various text items of the graph (max. 40 characters). These are:
6

Two main titles of the response graph

Three axis names of the graph
Four legend texts (30 chars.)
The title of the model description text
Two column headers of the model parameters (20 chars.)
Texts used for the RMS-error and the mean of damping factors (12 chars.)
One additional axis title for the vertical axis of the model view

Note that the DISLIN graphics uses ISO Latin-1 character codes. See the DISLIN documentation
for further information.

Additional information
The MTINV program was originally made at the University of Oulu in December 2001, when I
worked as a researcher funded by a grant from Outokumpu Foundation addressed to Prof. SvenErik Hjelt. Further modifications were made in September 2002 when I had received a grant from
Tnning Foundation. The forward computation is based on an old AMT interpretation program and
the MT response is computed using the algorithm described by Kunetz (Geophysics, 1972). The
inversion method, which is based on the singular value decomposition and adaptive damping
method, is described in my PhD thesis (M. Pirttijrvi, 2003. Acta Univ. Oul. A 403).
The MTINV program is written in Fortran-90 style using Compaq Visual Fortran 6.6. The
graphical user interface is based on the DISLIN graphics library (version 8.2) by Helmut Michels.
The program distribution includes the DISLIN.DLL that is required for the GUI. The official
WWW homepage of DISLIN is at http://www.dislin.de. Since the DISLIN graphics library is
independent form the operating system the MTINV program could be compiled on other operating
systems (Solaris, Linux) without any modifications. At the moment, however, the source code is
not made available and I do not intend to provide any support for the program. If you find the
computed results erroneous or if you have suggestions for improvements, please, inform me.
Note that the data weighting does not allow separate weights for the apparent resistivity and phase.
It can be used only to emphasize (or deminish) the effect of measurements made at certain period.
In addition, the inversion works with the 10-base logarithm of the data and the parameters. Phase
data is scaled relative to the mean of (logarithmic) apparent resistivity data. The special phase scale
parameter can be used to increase or to decrease the importance of the phase data as a whole.

Terms of use and disclaimer:

You can use the MTINV program free of charge. If you find the program useful, please, send me a
postcard.
The program is provided as is. The author (MP) and the University of Oulu disclaim all warranties,
expressed or implied, with regard to this software. In no event shall the author or the University of
Oulu be liable for any indirect or consequential damages or any damages whatsoever resulting from
loss of use, data or profits, arising out of or in connection with the use or performance of this
software.
7

Contact information:
Dr Markku Pirttijrvi
Division of Geophysics
Department of Geosciences
PO Box 3000
FIN-90014 University of Oulu
Finland

GSM: +358-40-5548959
Tel: +358-8-5531409
Fax: +358-8-5531484
URL: http://www.gf.oulu.fi/~mpi
E-mail: markku.pirttijarvi@oulu.fi

Appendix: