See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/272488469

Analysis of Linear Magnetic Anomalies

Technical Report · January 2007

DOI: 10.13140/RG.2.2.16763.31522

CITATIONS READS
3 2,091

1 author:

Bruce W. Bevan
Geosight
135 PUBLICATIONS 925 CITATIONS

SEE PROFILE

All content following this page was uploaded by Bruce W. Bevan on 19 February 2015.

The user has requested enhancement of the downloaded file.

 Analysis of Linear Magnetic Anomalies

4 December 2006

by: Bruce W. Bevan

Geosight
356 Waddy Drive
Weems, Virginia 22576-2004
USA

Magnetic anomalies that extend along a straight line may be caused by buried walls or
filled trenches. Techniques that have been developed specifically for the interpretation of
these types of anomalies are described here.
The goal of an interpretation is an estimate of the location of the magnetic material that
is underground. There are three aspects to this location:
Position along a profile (the horizontal line of measurement),
depth underground, and perhaps even
the shape of the magnetic feature.
The interpretation can also estimate the quantity of magnetic material. Depth and quantity
are particularly valuable findings; these will aid one's understanding of underground features.
The techniques that are described here can also be applied to anomalies that are not
long, but instead are small and somewhat circular. While this report is written for
archaeological surveys, much of the information may aid other applications. The examples in
this report illustrate data that may be measured at mid-latitudes (rather than at equatorial or
polar locations). The procedures that are described here are primarily for total-field magnetic
surveys, and not for gradiometers; some publications discuss specific procedures for the
analysis of gradiometer data (Keating and Pilkington 1990; Nelson 1988).
Three computer programs that aid the analysis of linear anomalies are described here;
all of these programs are free. At the end of the report, three case studies illustrate how
these programs can be applied to the analysis of some magnetic measurements.
The electronic version of this report has hyperlinks, primarily to the figures at the end
of the report; these links are marked with blue text. While the figures are colored, most of
their information will be apparent if they are printed with black ink.

Page 1
 Introduction

The sections of this report are outlined below

: Figure
Introduction
The half-width rule, a simple graphical procedure 1
Introduction to the three programs
Mag2dc, manual and automatic prismatic models 4
Pdepth, automatic search for prism edges 7
Euler deconvolution 8
Analytic signal 8
Werner deconvolution 9
Multiple-source Werner deconvolution 9
Prismatic analysis 10
Saki, automatic prismatic models 11
Units for magnetic susceptibility
Examples of analysis of linear anomalies
Example 1, interference between shallow and deep features
The data 13
Magnetic parameters of the Earth
Initial model 15
Analysis with Pdepth 17
Analysis with Saki 21
Example 2, oblique features with remanent magnetization
The data 23
Rotation of the magnetic map 26
Analysis with Mag2dc and Saki 28
Causes of the strong magnetic lows 31
Quantifying the magnetic material 34
Example 3, distinguishing magnetic from nonmagnetic features
The data 35
A simple model 36
Other software
Conclusion
Annotated bibliography

Introduction
Anomalies that are linear are only a small fraction of the anomalies that are detected
by magnetic surveys at archaeological sites; most anomalies are either somewhat circular or
they are amorphous and blob-like. However, where a linear anomaly is found, the readings
on a single line (also called a profile) can be studied; this speeds and simplifies the analysis,
and it also allows the sources (magnetic features) to be plotted clearly on a two-dimensional
cross-section. When it is necessary to analyze detailed magnetic maps, rather than just

Page 2
 Introduction

profiles, the calculations are more complex and the display of three-dimensional sources is
more difficult.
A linear magnetic anomaly is caused by a long magnetic body whose cross-section
remains rather constant for the length of the body; this type of body is called a prism. While
the analysis may assume or require that the body be infinitely long, bodies that are elongated
only slightly, or not at all, will probably still be suitable for this linear-anomaly analysis. Figure
1 shows that the magnetic anomaly of a feature that is 5 m long is almost the same as the
anomaly of a feature that is infinitely long. Compact objects, which may cause rather circular
anomalies, can even be analyzed with these procedures.
If a magnetic prism can be assumed to be infinitely long, a two-dimensional (2-D)
analysis is suitable for it. The programs that are described here can be applied to 2.5-D
analysis (Cady 1980; Campbell 1983). This just means that the prism need not be very long,
and that the data are to be studied with a single line of measurements; this line may cross
over the middle of the prism or it may be closer to one end of the prism.
There are two general procedures that may be applied to a detailed analysis of linear
anomalies. With the first procedure, the edges of magnetic prisms are moved until the
calculated field of the prisms is similar to the measurements; in the second procedure, the
edges of magnetic bodies are sought with a direct analysis of the magnetic measurements.
This first method of analyzing linear anomalies assumes that the magnetic features
can be approximated by a group of prisms; each of these prisms is moderately long and has
a polygonal cross-section. Each of the prisms can also have a different magnetic
susceptibility. These parameters of magnetic shape and susceptibility describe what is called
a magnetic model.
The magnetic susceptibility of the prisms can be changed, along with the locations of
the edges of the magnetic prisms; in the cross-section of each prism, these edges are the
corners of a polygon (in this text, edges of prisms may sometimes called corners). Each time
a change is made to these parameters, the magnetic field of the model can be calculated
instantly and then compared to the measurements. This comparison can suggest how the
susceptibility and prism edges can be changed again so that the calculations match the
measurements more closely. These changes or iterations to the magnetic model can be
selected by the person who is studying the data, or they can be done automatically by the
computer program. The program Mag2dc makes it easy for an interpreter to manually
change magnetic parameters; the program Saki is excellent for automatic iteration of a
magnetic model. Figure 4 shows the computer's display when the Mag2dc program is run.
The second, or edge search, procedure does not create a magnetic model. Instead, in
a typical analysis, the profile of magnetic measurements is analyzed section by section with a
computer program in order to estimate the locations of the corners of magnetic polygons that
could cause the anomalies on the magnetic profile. Only some of the corners of the polygons
will be revealed, and the lines between the corners (the edges of the polygons) will not be
determined. While the shapes of magnetic features cannot be defined clearly with these
methods, they do allow a fast analysis and one that is suitable for the study of large amounts
of data. The main disadvantage of these procedures is that they provide little feedback to the

Page 3
 The half-width rule

interpreter about the reliability of the findings; this feedback is more certain with a prismatic
analysis. The program Pdepth provides four different types of corner searches for studies of
magnetic profiles; Figure 9 illustrates two displays from this program.
Procedures for the analysis of linear magnetic anomalies are described in several
books. The most modern and complete discussion is given by Blakely (1995, p. 238-250,
350-356); this book is technical, but it describes two procedures for the analysis of linear
anomalies that are not discussed in this report. Good summaries of the principles of
magnetic interpretation are found in the books by Parasnis (1997, p. 27-51) and by Dobrin
and Savit (1988, p. 701-749).

The half-width rule

Before turning to the detailed methods of analysis, first consider a simple procedure
for analyzing magnetic anomalies that requires no computer. All that is needed is a plotted
curve that shows the magnetic field as it was measured along a line; this line is perpendicular
to the direction of the linear feature. Figure 1 has examples of calculated profiles; the
average or background value of the Earth's field has been subtracted; therefore, the curves
show the magnetic anomaly rather than the magnetic field. The magnetic susceptibility of the
prism in Figure 1 has a value of one part per thousand (1 ppt); this means a susceptibility of
0.001 in SI units.
A horizontal line can be drawn across the magnetic high at a level that is halfway
between the anomaly's peak and its zero level (average or background). The length of this
line is about the same as the depth to the middle of the underground object, measured from
the elevation of the magnetic sensor. This is called the half-width rule (Breiner 1973 p. 31;
Telford, Geldart, and Sheriff 1990 p. 87). The dashed lines in Figure 1 are each about 1.5 m
long, and this is the depth to the middle of the magnetic square.
Figure 1 shows that this approximation is good even if the feature is not long; a
magnetic cube gives about the same result as a prism that is infinitely long. The curves in
Figure 1 have been drawn for total-field magnetic data; however, the half-width analysis may
also be applied to data from a magnetic gradiometer. As seen in Figure 2, the estimates of
depth from a gradiometer may be somewhat shallow; an increase in the interpreted depth by
20 per cent will give a more accurate estimate if the inclination angle (Ie) of the Earth's
magnetic field is about 68°.
If a linear object does not have a cross-section that is compact (that is, it is not rather
square or circular), but instead it has a width that is much greater than its thickness, the
half-width rule will give an interpreted depth that is too large; see Figure 3. Unfortunately, it
may be difficult to tell from a magnetic profile that the buried feature is broad. The change in
slope along the curve of the magnetic high in Figure 3 gives a clue, but this may not be
apparent in magnetic measurements, which may be noisy and have a wide spacing. Even
though the depth from the half-width rule may be too great, it still gives on a limit on the
depth, and this is valuable to know.
Perhaps you noticed something curious in Figure 1: The infinite-length prism causes a
lower anomaly than the 5-m long box. This appears to be an error; it seems reasonable that

Page 4
 Introductions to the three programs

the anomaly should steadily increase as the length of the box or square cylinder increases. If
an explanation of this is not of interest to you, skip to the next section.
Two general effects may be clear without doing any calculation. For short prisms (they
would then actually be plates or disks), it seems reasonable that the anomaly should increase
linearly with the length of the prism. As the prism gets very long, it is reasonable that the
anomaly should rise to a constant value or limit.
Both of these effects are indeed found, but there is also a third and perhaps
unexpected effect: At a moderate length, the anomaly from the prism rises to a peak value,
but then it drops as the prism becomes even longer. For the prism in Figure 1, the highest
anomaly is found when the box is about 4.15 m long.
This anomaly peak can be understood from the magnetic map of a cube; there is a
broad area with a negative anomaly around the cube. In Figure 1, the anomaly is negative
for north coordinates greater than 0.97 m. When the anomaly of a cube is calculated along
an east-west line, the anomaly remains negative for distances that are greater than 1.88 m
from the middle of the cube. The magnetic anomaly is therefore positive for a somewhat
elliptical area that is near the cube, but negative everywhere else.
The long box or square cylinder can be constructed by placing cubes along a line.
When this line is more than 3.76 m long (2 * 1.88 m), then additional cubes begin to reduce
the magnetic anomaly at the midpoint of the cylinder, and the anomaly keeps dropping as the
cylinder gets longer. The reduction of the anomaly is caused by the fact that anomalies are
negative at a distance from magnetic objects.
If a prism is vertical, the anomaly does not reach a peak and then decrease with
increasing depth extent; it simply approaches a constant anomaly. This is because the
magnetic material is always below the measurement point, and deeper magnetic material
always adds to the magnetic anomaly.

Introduction to the three computer programs

Several free computer programs are available for the analysis of linear magnetic
anomalies; these programs are found on the internet; they may be used without cost and they
may be given to anyone. I appreciate the contribution to the geophysical community that has
been made by the authors of these programs.
If you have this report as a digital copy, these programs should be included in the
same folder; that folder also contains data for a few of the figures that are in this report.
However, there can be advantages to getting a program more directly from its author: It may
be more current, and it may be less likely to contain malware.
Introductions to these programs are included below; a full description is not given since
it is easiest to find out the details of operation of these programs by simply using them and
also by studying the instructions that are provided by the authors of the programs.

Mag2dc
One program is called Mag2dc; this Windows program was written by G. R. J. Cooper
(Department of Geosciences, University of Witwatersrand, South Africa). The name of this

Page 5
 Mag2dc

program may be an abbreviation for MAGnetic analysis of 2-Dimensional Cylinders. The

program is available at www.rockware.com; Rockware is a company in the USA that sells
geological and geophysical software. The web site for this company includes a number of
free programs; in addition to Mag2dc.zip (a compressed file), the site has two other programs
by Cooper (GeoModel and SignProc).
Of the files that are included with the software, three are necessary:
Mag2dc.exe is the program itself
Mag2dc.cfg stores the user's general field parameters
MagHelp.hlp has instructions that are displayed by the program
The instructions that have been supplied by Cooper are clear; while these include no mention
of the fundamental algorithms of the program, the instructions are otherwise complete. Since
the internal Help system for this program is so good, almost no additional information is
needed here.
No complex installation is needed in order to run the program; just put the files in any
convenient folder. The version of the program that I have is 2.10, and it is dated 2003; the
program does not add anything to the Windows registry, and it may have been written for
version 3.1 of Windows.
Rather than repeating the program's information about starting a magnetic model, this
introduction will simply skip to a few illustrations of the computer's displays, such as those in
Figure 4. One part of the Mag2dc program is particularly good: It is very easy to move one
edge of a polygonal prism and immediately see the new calculated profile. This is very
educational, for it allows one to understand how the location and depth of a corner affects a
magnetic profile.
In Figure 4, the measurements to be modeled are drawn with a red line. The goal of
the analysis is to change the shape of the prism until the calculated curve (green line) is as
close to the measured curve (red line) as possible; the colors of these curves can be
changed.
With the initial attempt at a magnetic model (in the upper panel of Figure 4), the
measured and calculated curves are quite different. By moving two edges of the prism, a
better match is found; see the lower panel of Figure 4. The amplitudes of the calculated
curve is still too low, so increase the magnetic susceptibility of the prism. This can be done
by double clicking on the prism and changing the value for susceptibility; the top panel of
Figure 5 shows the result. An even better match between the two curves is seen at the
bottom of that figure.
While the correct magnetic model is the rectangle shown in Figure 3, many other
prismatic shapes can also yield calculations that are very similar to the measurements.
Figure 6 has some illustrations. Except for the initial model, the other three models have
calculated curves that are similar to each other and to the measurements. In practice (with
actual measurements), it may not be possible to distinguish among distinctly different models.
This is because of several factors:
1: The measurement spacing may be large
2: The magnetic features may be complex

Page 6
 Pdepth

3: The anomalies of other features may cause interference

4: The magnetic measurements may be affected by temporal noise
5: Ambiguities that are inherent in magnetic modeling
The prisms of all of the magnetic models that are described here have abrupt changes
in magnetic susceptibility across their faces. In practice, many magnetic bodies have graded
interfaces, and the susceptibility will slowly change through a layer. The assumption of
abrupt interfaces is one of many approximations that are made in a magnetic analysis.
The program Mag2dc can speed an analysis by automatically changing the edges of a
magnetic prism and also its susceptibility. This automation is best for changing the depth or
width of prisms, or for changing their susceptibility. While the program does allow for
automatic changes to the locations of a prism's individual edges, this part of the program can
create errors. It is important to save the current best solution before attempting this
automatic iteration of corners; this is because the iteration may move the corners too far, or
to otherwise incorrect locations.
A few general notes about the program: The automatic analysis or iteration that is
done in the Mag2dc program must be applied to a single prism at a time; if a magnetic model
has nearby prisms whose fields affect each other, the analysis is slower if prisms must be
studied one at a time. The program has shortcuts, which allow one to use a keyboard rather
than the mouse to select an operation (by pressing the Control key and another letter); these
keyboard shortcuts speed the operation of the program. The Ctrl-M (move a body corner)
and the Ctrl-I (invert a model) shortcuts do not work correctly; these operations must be
started with a pull-down menu. There is another minor fault in the program: The Field Data
Entry / Remove Average operation fails if the average is less than zero; this operation can be
done correctly with the Edit Model / Modify the Regional Field selection.
Further details on the Mag2dc program: It does allow for the automatic iteration of the
magnitude and direction of remanent magnetization in the model. A total of up to ten
parameters can be iterated at a time. The program may halt if more than ten parameters are
selected for iteration; alternatively, it may then be found that the measurements that are
stored in the program have changed. Magnetic models can be saved in two formats. One
(*.mod) can be read only by the program; the other is a simple text file (which cannot be
re-entered to Mag2dc). The program assumes that the measurements have been made on a
flat and horizontal line. Data entry from a file of measurements is easy; these files are
assumed to have at least one line of text before the start of the numerical data. It is not
necessary to enter magnetic measurements; one can simply calculate the magnetic field of
prisms that are entered. Susceptibility can have either its cgs or SI value. The text that is
displayed in some windows may be truncated, but it will still be readable. A precision of 0.1
for a line coordinate is the smallest that is possible with *.dta files.

Pdepth
The program Pdepth was written by Jeffrey Phillips (U. S. Geological Survey, Denver,
Colorado); this program has many features and some of them are based on programs that
have been written by the colleagues of Phillips at the USGS. It is likely that the name of the

Page 7
 Pdepth

program is an abbreviation for Prism DEPTH. This program is part of a large group of
software that is available through:
http://pubs.usgs.gov/of/1997/ofr-97-0725/pfofr.htm
This web page is Open-File Report 725 for the year 1997, and it is titled "Potential field
geophysical software for the PC, version 2.2". This version 2.2 is the most current; version
2.11 of this group of programs is also on the web site of the USGS, but that version of the
software does not include the Pdepth program. Three files contain the compressed
(archived) programs and text. The file pf22bin.exe is a self-extracting file with all of the
programs and help files. The file pf22src.exe has the source code, which is mostly Fortran; it
is valuable to have this source code, for it can be modified in order to change the programs, if
that is wished. The third file (pf22bndy.exe) is not necessary.
The programs from this group all operate with DOS; if they do not operate correctly by
double-clicking on the executable (*.exe) files in Windows, it may be necessary to start them
from a DOS window or Command prompt (in most versions of Windows), exit to DOS (in
Windows 98 or earlier), or boot to DOS (perhaps with a floppy disk). While the Pdepth
program has an excellent graphics interface, it is still a DOS program. It also requires a
mouse; while the driver software is always available in Windows, it must be added if one
boots to DOS. If the mouse will not operate, it will not be possible to exit the program without
turning off the computer.
The program is the file called Pdepth.exe, which is dated 2 July 1997. The associated
file Pdepth.hlp is an unformatted text file that has instructions on the operation of the
program. Unlike the Mag2dc program, the help file for Pdepth is a separate file; it may be
read with any text editor or word processor.
While the operation of the Pdepth program is more complex than the Mag2dc
program, it is well worth the time that is needed to learn how to operate it; this is because
Pdepth allows many types of analysis that are probably not available in any other single
program.
Figure 7 shows how the program may be started in a DOS window inside the Windows
operating system. The colors that will be seen on your computer screen will be the
complement of the colors here; most importantly, the background here has been changed
from the screen's black to white (this will save ink if you print the figures). As indicated at the
bottom of Figure 7, lines may be drawn in the display by entering a "cross-section file". The
numerical format that is required is described in the help file for the Pdepth program;
however, the format that is needed is rather complicated.
One unusual feature of most of the potential field programs from the USGS is that they
can operate with what is called a command file. This file has a list of many of the numerical
parameters that are to be input to the computer for the analysis of specific magnetic data; an
example is illustrated at the top of Figure 7. These command files are a remanent of the
days before personal computers, when programs were run on large and centralized
mainframe computers. However, the command files also apply the same idea that is in the
huge Windows registry file or in the earlier *.ini files that set up Windows and its programs.
The disadvantage of a command file is that it may be necessary to refresh one's memory

Page 8
 Euler deconvolution

about the values that may be entered; for example, one may forget that an input like
"edec=30" means that the declination of the Earth's magnetic field is assumed to be 30°
clockwise from grid north. There are two advantages to command files: They keep a record
of some of the processing that was done on the data; and they allow easy changing of one or
a few parameters by using a text editor to modify a command file.
After the Pdepth program has started, it is controlled by clicking the left mouse button
when the mouse cursor is on square areas on the right-hand side of the display; if text or
numbers must be entered, that is indicated at the top of the display. After the initial display
shown in Figure 7, one can click on Model, and then Depth (in a menu after that of Figure 7).
Then, one will have a choice of five different types of analysis by corner searching; these are
listed on the right side of Figure 8.
These analysis procedures are very ingenious but their findings are not always clear.
They are excellent for an initial study of a magnetic profile, before starting a prismatic
analysis. They are also suitable for two other cases: When the profiles are very complex or
when time is too short. If your own time is short, you may wish to skip on to the next major
section (Prismatic analysis with the Pdepth program); this is because it is necessary to
describe and understand many details in order to operate most of these procedures; without
this understanding, one may create misleading results. While the analyses are quick, their
understanding may be slow.
These analysis procedures can automatically estimate the locations of some of the
edges of polygonal prisms. As an illustration of the findings of these analyses, calculated
data from the simple prism of Figure 3 can be studied. Figure 8 and Figure 9 show how four
of these procedures suggest the upper edges of the known rectangular prism; these
estimates are indicated by clusters of points in the lower parts of each panel. The fifth
analysis procedure in Pdepth is called Autocorrelation; unfortunately, there is not enough
information about this procedure to be able to operate it with confidence, and so it is not
discussed further here.
The findings of these analyses are plotted on the computer's display, and they are also
written to a normal ASCII text file, called Pdepth.dep. Every time one of these procedures is
used, the results of the analysis are appended to that file; before a new analysis is started,
any old Pdepth.dep file should be deleted or renamed. This cumulative Pdepth.dep file
includes the X coordinates (along traverse positions) and Z coordinates (depths) of the prism
edges that have been found by the analysis. The file also lists the dip angles of the
interfaces that have been interpreted. Finally, the file lists contrasts in susceptibility across
the interfaces; these values for dip and susceptibility are not indicated in the graphic displays
from the program.
Introductions to the four procedures are given next. They are not complete because
the developers of these procedures have written clear descriptions of their basis and their
operation. References to these publications are included.

Euler deconvolution
Several procedures for the analysis of magnetic maps are based on Euler's differential

Page 9
 Euler deconvolution

equation; Leonard Euler was an 18th-century Swiss mathematician. His differential equation
describes how gradients of the magnetic field are related to the total field, and his equation
applies to many magnetic measurements. Deconvolution is a mathematical term; it might be
considered equivalent to the phrase "remove the blurring of features that is inherent in
magnetic measurements in order to reveal the sources more clearly".
An interesting application of Euler's equation was first pointed out by Hood (1965, p.
899). He showed how the depth of a magnetic feature could be calculated by knowing the
peak anomaly of the magnetic field at a location if the vertical gradient of the field at that
same point was also known. The Euler equation is then simply:
Depth (below sensor) = n * anomaly / gradient.
The parameter n is called the structural index; it quantifies how the magnetic field drops with
increasing distance above a magnetic feature. For a magnetic dipole, n = 3, and this is the
exponent in the denominator of the equation that describes the magnetic anomaly above a
small feature:
Anomaly = constant / depth3
For features with other geometries, there are other values for the structural index. These
have been summarized by Reid (2003):
Structural Magnetic feature
index
0 Flat and large interface between magnetic materials
1 Thin magnetic stratum
2 Linear feature with a small diameter (wall or pipe)
3 Compact feature, approximated as a dipole or sphere

All of these features are thin; they are points, lines, or planes. For broad features, Euler's
equation does not apply, for the structural index changes with distance from the feature; while
the Euler method can still be used, the accuracy then will be lower.
In the Euler equation, the anomaly is the total magnetic field after the constant (that is,
background) field of the Earth has been subtracted. The gradient is its true or mathematical
value; this assumes that the two measuring sensors are very close to each other relative to
the distance to the magnetic feature. The actual readings that are measured with a practical
magnetic gradiometer approximate this value.
The only other study besides Hood's where this particular Euler procedure has been
used appears to be a later paper by Slack and others (1967). However, a more general
procedure is widely applied; this is called two-dimensional Euler deconvolution. This method
was developed by Thompson (1982) and a good explanation of the procedure is given in that
paper. The analysis in the Pdepth program is basically the same as the one described by
Thompson.
With this procedure, depth is estimated from several magnetic measurements along a

Page 10
 Analytic signal

line; one needs the total magnetic field and the vertical gradient of that field (as with Hood's
method), and also the longitudinal magnetic gradient. Both of these gradients are calculated
by the Pdepth program from the total field data. Since these depth estimates are made at
many locations along a profile, Euler deconvolution generates a large number of points which
may mark the corners of magnetic features; clusters of points are assumed to reveal the best
solutions.
The upper panel in Figure 8 shows an example of a Euler analysis with the Pdepth
program; for this example, the source geometry is known (the rectangle in the lower half of
that panel). The Euler dimension was set at "n" so that the analysis would be made with the
whole range of structural indexes above. The letters "c" in the plot indicate a contact solution
(structural index about 1); these symbols are found at the upper edges of the rectangular
prism. The letters "p" indicate pipe solutions, and the structural index is about 2 for them;
these symbols are found near the lower edges of the prism. The letter "s" indicates sheet
solutions, with an index of about 1; these symbols are on the left and right sides of the prism.
For the analysis in Figure 8, the window length was set at 10; this parameter
determines the length along the curve over which each analysis will be made; the larger the
number, the greater the length. Since the window length can only be set over a moderately
short range (5 - 20), it is probably easiest just to try a few values. It is likely that the actual
distance along the profile may be equal to the window length parameter times the
measurement spacing times a factor of seven (the likely number of equally-spaced
measurements along each span that are used for each analysis). It is best that this analysis
length approximately cover the entire length of each anomaly that is to be analyzed.
The tolerance parameter was set at 0 so that all solutions would be plotted. The
Pdepth.hlp file explains this parameter further, and the tolerance in the program appears to
be the same as that defined by Thompson (1982).
The original Euler method was later generalized so that it could be applied to magnetic
maps, for a 3-D analysis, rather than simply to profiles (Reid and others 1990). Like other
procedures for magnetic interpretation, a Euler deconvolution gives false results where
anomalies are so close together that their patterns interfere; Hansen and Suciu (2002)
developed a procedure that allows several features to be analyzed at a time, and this can
reduce the effect of overlapping anomalies. The source code of Pdepth suggests that the
programming of a multiple-source Euler procedure was started, but not completed.
Additional information about Euler deconvolution can be found in the papers by Ravat
(1996), Stavrev (1997), Barbosa and others (1999), Mushayandebvu and others (2001), and
Mikhailov and others (2003).

Analytic signal
Like Euler convolution, the analytic signal technique is also an automated procedure
that is suitable for the analysis of large quantities of magnetic data. Both the Euler and
analytic signal techniques provide good estimates of the lateral boundaries of magnetic
features, and fair estimates of the depth of those features; neither procedure appears to be
particularly suitable for an estimation of the mass or volume of features. Both techniques

Page 11
 Analytic signal

began as two-dimensional methods for the analysis of magnetic profiles and were later
generalized so that magnetic maps could also be studied.
The analytic signal technique was developed by Nabighian (1972, 1974). While he
modified the original 2-D analysis to 3D (Nabighian 1984), practical procedures for this type
of 3-D analysis of maps were first described by Roest and others (1992). The name "analytic
signal" is a mathematical term that is associated with the Hilbert transform.
The analytic signal method transforms the measurements of a magnetic map or profile
in order to reveal the locations of magnetic boundaries and to suggest the depths of those
boundaries. From the original measurements of the total magnetic field, gradients are
calculated in the horizontal and vertical direction. These two or three gradients are squared
and then added, and the square root of the sum is calculated. The result is the amplitude of
the analytic signal, which can be plotted as a profile or as a map. In the program Pdepth, the
analytic signal of the profile of measurements and its horizontal gradients can be plotted (with
dashed lines at the bottom of Figure 8).
Peaks or ridges on plots of the analytic signal are directly over the edges or corners of
magnetic contrasts that are underground. The analytic signal has a symmetrical or
bell-shaped cross-section at each of these boundaries. The width of each peak or ridge is
proportional to the depth of the corner of the body that is underground. An important
advantage of the analytic signal technique is that this transformation of the original magnetic
map is independent of the direction of the earth's magnetic field or the direction of remanent
magnetization in the bodies.
While several parameters must be selected during the operation of Euler
deconvolution, none are needed for the analytic signal procedure in the program Pdepth. An
illustration of the output of this analysis is shown at the bottom of Figure 8. The letters C and
D locate corners if the feature is assumed to be a Contact (a flat interface between materials
that have a contrast in susceptibility) or a Sheet (a thin layer of magnetic material). While
sheet solutions are to be expected for some mining applications (they could be mineralized
veins), these features are rare in archaeology; except for iron plates, thin archaeological
layers are generally not magnetic enough to be detected. However, common features in
archaeology, such as broad layers (with their breadth greater than their depth) and also
interfaces, may be detected as contact solutions.
The lower part of Figure 8 shows a number of symbols at the right end of the plot.
While I do not know the cause of these, they clearly do not mark interfaces.
In the plots from Pdepth, short line segments give estimates of the range of the dip
angles of the surface of the features. In the selection menu on the right side of the display,
the text is intended to say "ansig", for "analytic signal", but the I may look like an L because of
the font that is used to write the text.
Unlike Euler deconvolution, one does not need to estimate a structural index for the
body that is underground; this is only because the analytic signal technique assumes that
features are elongated. Features that are not elongated will have errors in their interpreted
depths (Zhang 2001). However, a combination of Euler deconvolution and the analytic signal
methods (Salem and Ravat 2003) allows compact bodies to be analyzed; an archaeological

Page 12
 Werner deconvolution

example of this analysis has been published by Abbas and colleagues (2005). If features
extend to a large depth, then the depths determined by the analytic signal technique are to
the top of the features; if the features are thinner, the depths from the interpretation are
somewhere within the thickness of the features (Roest and others 1992, p. 119).
At the current time, probably the greatest amount of geophysical development is being
applied to this analytic signal technique; important new information is discovered each year.

Werner deconvolution
This method for the analysis of linear anomalies assumes that the magnetic features
have cross-sections that are rather polygonal. This procedure attempts to locate the corners
of those polygons, both as distance along a line of profile and as depth underground.
Werner deconvolution is based on the finding that the magnetic anomaly of the edge
of a polygonal prism is determined by only four parameters. In principle, these four
parameters may be calculated from four measurements along the magnetic profile; this is
exactly what is done. The operation is therefore similar to the general procedure called curve
fitting.
Many or most of the calculations will give polygonal edges that are very far from being
correct; however, if many calculations suggest corners in about the same location, one might
assume that these could be correct. The calculations are most likely to yield a correct corner
when the four measurements are evenly spaced across the width of a single anomaly. If the
four measurements are very close together, they will fall on almost a straight line, and this will
clearly cause an error in the calculated corner. If the four measurements are very far apart,
then the four values probably include the effects of more than one anomaly; here again, the
calculated corner must also be in error. These errors will plot as points that are scattered
over a wide area; it is the clustering of solutions that can suggest the most reasonable
locations for the edges of prisms.
The method of Werner deconvolution is valid for vertical gradient data (Hartman and
others 1971). Probably the best introduction to Werner deconvolution has been given by Ku
and Sharp (1983). The name Werner honors the Swedish geophysicist who originated the
idea for this procedure.
The Werner procedure displays solutions for both the input measurements and also
the horizontal gradient of those measurements. This is done because the solutions for the
total magnetic field locate the upper edges of thin magnetic sheets; when the horizontal
gradient is analyzed, the solutions locate the edges of magnetic polygons. Therefore, both
solutions are valuable, but polygonal features are more likely than inclined, sheet-like
features in archaeology.
An analysis with the Werner procedure, like all other types of magnetic analysis and
measurement, can accentuate shallow features. The deeper corners of polygons may be
invisible because the amplitudes of their anomalies can be faint; they can therefore be
obscured by the anomalies of shallower features.
A Werner analysis can provide estimates of the magnetic susceptibility contrasts of
bodies, and also the slope angle of polygonal sides or faces; while the slope angles are

Page 13
 Multiple-source Werner deconvolution

plotted on the Pdepth display, the susceptibilities are only listed in the Pdepth.dep file that is
generated automatically by the program.
An illustrations of a Werner deconvolution is plotted at the top of Figure 9. If there are
broad anomalies that are unimportant (perhaps because of they have a deep and geological
origin), these can be removed from the data with the Interference Polynomial. While a
second degree polynomial was found to be most suitable for the analysis of the calculated
data in Figure 9, I would have thought that no interference polynomial would have been
needed (a degree of 0).
In Figure 9, the asterisks near the upper edges of the known prism are the most
important, for these locate the interface solutions. Sheet solutions are indicated with dashes.
A separate program for Werner analysis is included with the group of potential field
programs from the U. S. Geological Survey; this program is called just Werner. This Werner
program has additional options or parameters that are not available in the Werner section of
the Pdepth program.

Multiple-source Werner deconvolution

This is a modification of the Werner procedure that is described above. If the edges of
prisms are close together, the original Werner analysis will be confused by interfering or
overlapping anomalies. The multiple-source Werner procedure (Hansen and Simmonds
1993) is an improvement that allows complex prisms to be analyzed.
The findings of an analysis with multiple-source Werner deconvolution in the Pdepth
program are shown at the bottom of Figure 9. While the result is much simpler and clearer
than the Werner analysis that was described in the prior section, a greater number of
parameters are needed for this analysis. The Window Length Code is selected similar to the
procedure for the Euler method. Several of the other parameters control how many solutions
are plotted and how nearby solutions are combined into one symbol. The "O" symbols in
Figure 9 indicate how many solutions are combined there; my understanding is that these
"O's" mean that 24 solutions were found near the symbols at the two upper edges of the
prism (these symbols count from 2 - 9, and then continue from A through as high as Z). As
with the other corner searching procedures in Pdepth, contact and sheet solutions can be
selected; the short bars in the figure are plots of the directions of the contact (interface)
solutions.
One input to the program is called "enter min, max number of edges". One or more
spaces must separate the two numbers; if a comma separates the numbers, the program will
restart at the multiple-source Werner section.

Prismatic analysis
This option allows one to manually change the parameters of the model prisms, and
then compare the new calculated profile to the measurements. The general operation is very
similar to the Mag2dc program. Figure 10 illustrates how a edge of a prism can be moved in
order to decrease the error between the measurements and the calculations. The labels on
the right-hand side of each panel list the operations that may be applied. While a single

Page 14
 Saki

prism ("body") cannot be moved as it can in the Mag2dc program, individual corners are
easily shifted to new locations. The magnetic susceptibility of each prism, and also its length,
can be changed with the "prop" (properties) section.
After a moderately good model has been created with the Pdepth program, its
parameters can be saved to a file. This program does not allow automatic iteration of either
the edges or the susceptibilities of prisms. That operation can be done with another USGS
program called Saki, and the file of parameters from the Pdepth program can be entered into
that follow-on Saki program.

Saki
This program does an automatic iteration for the parameters of a group of prisms until
the calculated field closely matches the measurements. This program was written by Michael
Webring (1985) at the U. S. Geological Survey. The name Saki appears to be an
approximate abbreviation for "Semi-Automatic marQuardt Inversion", where a K has been
substituted for the Q. An excellent description of the Marquardt method, which is the basis
for iteration in the Saki program, is found in Johnson (1969).
The program is available with the group of potential field software that was described
in the earlier section on the Pdepth program. The file Saki.exe is the program itself; it is
dated 16 April 1996. A second file, Saki.hlp, is a simple and unformatted text document that
has detailed directions for the operation of the program.
Like Pdepth, this program was also designed to operate with DOS, but it also should
work fine with most or all Windows computers. Unlike Pdepth, the input to the program is
entirely from the keyboard; no mouse is used. However, the Saki program does have a good
graphics display for viewing the measurements, calculations, and magnetic prisms; see the
bottom of Figure 11. The axes on this plot only show the units of kilometers, but any
consistent units may be applied when the program is used. For example, if the sensor
height, measurement spacing, and other values are entered in meters, the calculation will be
correct even if the scales say kilometers (mentally change them to meters). All of these
programs will also work correctly if the unit for all distances is always feet.
Initial estimates for the model prisms can be generated with the Pdepth program,
using the corner searching and prismatic analysis parts of that program. It is best that this
initial model be as good as is practical, for this will speed the analysis with the Saki program
and also make the findings more certain. If an initial model is far from being correct, the Saki
program will probably not be able to determine a reasonable solution. Like the Mag2dc
program, the Saki program never adds or subtracts prisms, or edges from prisms.
While the Saki.hlp file with the program has a complete description of the operation of
the program, some additional notes might help (many of these notes apply to the Pdepth
program also).
The Saki program has three additional files that can be needed for its operation; all of
these are standard ASCII files:
1: The command file lists program settings; these values can be set here so that they
do not need to be entered each time the program is run. The program will also run without a

Page 15
 Saki

command file, and the user can enter the parameters when the program requests them. The
Saki.hlp text file explains the options that can be set in the command file. Note that if a
parameter is not numeric, but is text, that text must have a single quote symbol ' before and
after it; a double quote symbol " will cause the program to stop with an error. An example of
a part of a command file is printed below:
&parms
mfile='\saki.bod'
hfile='\saki.dat'
... middle section is not included here
istatn=1,
labelv=0,
numbod=1, &
The file must start with the text "&params" and end with the text "&". The "azimuth"
parameter is the angle clockwise from true north to the bearing of the line of measurements.
The "edec" parameter is the angle clockwise from true north to the bearing of magnetic north.
If the "plenth" parameter is set to zero, the magnetic prisms are assumed to have a length of
1000 (km or m).
2: The data file is described in the Saki.hlp file. Three numbers are on each line of
the file, in this order:
Line coordinate
Depth of the magnetic sensor at that coordinate
Measurement of magnetic field
Note that depth is used rather than elevation throughout the program. If the measurement
surface is flat, then that surface can be called zero depth. However, if surface topography is
included, it is probably best to set the zero depth surface at a convenient value above the
greatest elevation; this may be called the maximum elevation. Then, depths below this
surface are:
(Maximum elevation) - (Actual elevation)
3: The model file describes the prisms whose calculated field is to approximate the
measurements. The Saki.hlp file describes this file. The numbers in this file (like the other
files) can be separated by commas or spaces. Note that there must be non-alphanumeric
symbols in this file (the examples use a pair of << characters); these symbols mark the ends
of some of the lines of data and there must be a space before these symbols. The
susceptibility is entered in cgs units; the conversion is: k(cgs) = k(SI) / 4B. The declination of
remanent magnetization is positive if clockwise from true north. While remanent
magnetization may be entered into the program, the parameters of this magnetization are not
iterated; instead, the remanent magnetization remains a constant. Since the format of this
model file is somewhat strict, it can be easiest to edit an existing and working file for new
parameters.
After the files have been read by the program, there are many operations that can be
done on the data when the display shows "function:". These operations are listed in the
Saki.hlp file; only the first two letters of these commands need to be typed.

Page 16
 Units for magnetic susceptibility

The most common operation will be "plot", which creates a drawing that shows the
prismatic bodies (green lines), and a graph of the measurements (magnetic values are
plotted with a triangle) and the calculations (blue line). Hit the Enter key to clear the graphic
plot and return to the text of the program.
The first operations will usually be "calculate" or "synthetic"; these will allow one to do
forward calculations and these may help one to decide on an initial model.
The main operation of the program will be "invert", and the parameters of the prisms
are then iterated so that the calculated field approaches the measurements. Note that no
new edges are added to the prisms; however, an edge may be temporarily removed if the
iteration straightens the angle between two adjacent sides of a prism. The programs can
change each of the edges of each prism and also the magnetic susceptibility of each prism.
A maximum of 20 parameters can change on each iteration.
There appears to be a minor bug in the Saki program: Part of the calculated field will
plot as zero; these zero values are distant from the span where the model parameters are
being changed. This fault can be corrected by running "calculate" at the function prompt
before plotting the data or saving a file of the calculated values with the "output" command.
The program prompts one when two prisms intersect during the "invert" operation. However,
it does not appear to correct the prism edges so that prisms do not overlap.
The main differences between the Saki program as compared to the Mag2dc program
are as follows: The Saki program can iterate as many as 20 parameters at a time, and these
parameters can be selected from more than one prism; the Mag2dc program can iterate at
most 10 parameters, and these must all be from one prism. My experience is that the Saki
program accurately iterates to a good model; the Mag2dc program may sometimes be
unstable in its iteration and this can yield an impossible or unsuitable model. The magnetic
measurements entered into the Saki program can be on an irregular surface; the line of
readings for the Mag2dc program must be straight and horizontal. The Mag2dc program is
easier to operate than the Saki program. Remanent magnetization can be iterated in the
Mag2dc program, but not in Saki. The measurement spacing with the Mag2dc program can
be irregular, unlike the Saki program. Both programs are excellent.

Units for magnetic susceptibility

All of the computer programs that are described here aid in the estimation of the
magnetic susceptibility of underground features; susceptibility can be denoted with the letter k
or a Greek letter kappa. This quantity can also be called volume magnetic susceptibility,
although the word "volume" is to be assumed if it is not stated. Mass susceptibility is a
different quantity; it is volume susceptibility divided by the density (or sometimes by the
mass) of the material; it is designated with a Greek letter chi (i).
Magnetic susceptibility is a pure number, without a unit name. In about 1980, the
definition of susceptibility changed. The earlier unit can be called its cgs value
(centimeter-gram-second), while the quantity after 1980 can be called an SI value (an
abbreviation for the French words for International System). While current publications will
almost entirely employ SI units for susceptibility, earlier publications will probably have only

Page 17
 Example 1

cgs units. The conversion from cgs to SI units is easy:

k (SI) = 4Bk (cgs)
While the Mag2dc program allows susceptibility values to be entered and displayed as either
SI or cgs, the Pdepth and Saki programs work only with cgs units.
For a period after the change to SI units, there was confusion about the conversion of
cgs units to SI units. I have found this conversion stated in 11 different publications;
however, four of them have the four-pi factor on the wrong side of the equation above. Those
four publications are in error; the most recent publication that I have seen with this error is
dated 1995. The best explanation for the conversion between the units is that of Telford,
Geldart, and Sheriff (1990, p. 64).
It is important to understand that there may be an error in the conversion of some
published susceptibility values since about 1980. The error may result in susceptibility values
in apparent SI units being too small by a factor of about 158; this error is large enough that
one may be able to see if a value is unreasonable. The program Mag2dc uses the correct
conversion between cgs and SI units.
Values of susceptibility may be listed in several different ways. Typical values for
susceptibility will range between about 0.0001 and 0.01 in SI units; written this way, it is
necessary to count zeros in order to get an idea of the quantity. Some workers will write
values with a varying exponent, such as 4.3 (10-3) or 2.7x(10-4), but then one must read both
the multiplier and also the exponent to understand the value; it is therefore difficult to
compare quantities. Many workers use a standard exponent (typically -5), so that a value
may be 835 (10-5); this works well, but the exponent of -5 usually gives large multipliers, the
values before the exponent. Another option is one of borrowing terminology from chemistry,
parts per thousand (an exponent of -3), which can be abbreviated as ppt; this option has
been applied in this report.
It is important to remember that the susceptibilities that are applied to this type of
magnetic analysis indicate the difference between the susceptibility inside the feature as
compared to that from the surrounding soil. That is, if the model susceptibility of a feature is
2 ppt, and the susceptibility of the surrounding soil is 3 ppt, the actual susceptibility of the
feature is 5 ppt.

Examples of analysis of linear magnetic anomalies

The following three examples illustrate different applications of the analysis of linear
anomalies.

Example 1
The total field magnetic measurements for the first example are plotted in Figure 13.
This survey was done across the shallow ridge of a large prehistoric earthwork at the
Hopewell Cultural National Historical Park in southern Ohio (central USA). The earthen ridge
of this earthwork outlines a large square and a circle, each over 200 m in diameter. The
geophysical data in Figure 13 were measured as part of a project that is directed by Mark

Page 18
 Magnetic parameters of the Earth

Lynott (Midwest Archeological Center, National Park Service).

John Weymouth (University of Nebraska, Lincoln) directed a magnetic survey that
explored the entire area (32 hectares) of this earthwork (Weymouth 2004). His magnetic
map reveals that there are a pair of magnetic bands that extend along the ridge of the linear
earthwork; these bands are marked with X's in Figure 13. Figure 14 shows the orientation of
these magnetic bands relative to the line of measurement in Figure 13. While the line of
measurement is not perpendicular to the line of the magnetic bands, the difference is only
18°, and so the error in the analysis that is caused by ignoring this angle will be small. John
Weymouth's magnetic map also reveals many other parallel lines throughout this farmed
field; these magnetic lines appear to be caused by plowing. In Figure 13, the measurements
with a sensor height of 0.3 m probably reveal these plowing marks by the many small
undulations on the curve.
A detailed topographic map of these earthworks has been prepared by Jarrod Burks
and Karen Royce (Hopewell Culture National Park, Chillicothe, Ohio). When this is combined
with the magnetic map from John Weymouth, it is seen that the two magnetic bands are
rather symmetrical on the two sides of the shallow ridge that remains from the original high
earthwork. That original ridge has been smoothed and spread out by erosion and by farming.
Its height now averages about 0.6 m and its width is about 50 m.
While there have been no archaeological excavations along the line of the
measurements in Figure 13, several trenches were excavated at other places across the
earthen ridge by Mark Lynott. The general stratigraphy of the soil has been revealed by his
excavation drawings; the magnetic susceptibility of the soil along the faces of some trenches
has been mapped by Rinita Dalan (Minnesota State University, Moorhead). The findings
from these colleagues allow a general approximation of the magnetic cross-section of the
earthwork; this is plotted at the bottom of Figure 15. The archaeological strata dip away from
the central ridge of the earthwork, which is midway between the two magnetic bands. The
modern agricultural features (plow marks) are shallow and they are also less magnetic than
the deeper archaeological features.
The calculated magnetic field of this magnetic model is plotted in Figure 15 for a
sensor height of 0.3 m; the shallow agricultural patterns seriously distort the anomalies from
the important archaeological features. At a sensor height of 1.3 m, this distortion is minor,
although all anomalies are weaker at this height. Even though the spatial resolution at a
height of 1.3 m is lower, an analysis of the data at this height will be much easier than a study
of the readings at a height of 0.3 m. Only the data from a height of 1.3 m will be analyzed
further here.

Magnetic parameters of the Earth

The calculation of the magnetic model needs the magnetic parameters of the Earth at
the location and the year of the survey. There are three parameters that are required: The
magnitude of the flux density of the Earth's field, in nT; the inclination of the field below the
horizontal, in degrees; and the declination of the Earth's field, east of grid north, in degrees.
The magnitude of the Earth's field (B) is simply the average or typical magnetic field in

Page 19
 Initial model

the area of survey; this can be determined from a map of total magnetic field in the area. An
error of even more than 100 nT will probably cause no difficulty. The declination (D) of
magnetic north relative to grid north is easily measured with a normal magnetic compass; an
accuracy of a few degrees is acceptable. The inclination (I) or dip angle of the Earth's field is
not usually measured in the field; instead, it can be determined on the internet, with a
computer program, or from published maps.
On the internet, the parameters of the Earth's field are available at a web site of the
National Geophysical Data Center of the National Oceanic and Atmospheric Administration:
http://www.ngdc.noaa.gov
From this main page, one can follow the link to Geomagnetic Data and Models, and then to
the link to Field Models. One can also go directly to this page at:
http://www.ngdc.noaa.gov/seg/geomag/models.shtml
At this page, you may run the on-line calculator, which will list the magnetic parameters of the
Earth at a given location, time, and elevation.
It will probably be easier to download a computer program from this web page so that
an internet connection is not needed for future determinations. This program is Geomag
version 6.0, and it is a small DOS program that will run fine in Windows. The file
geomag_v60.zip can be uncompressed, and the result will then include two files that will
allow one to estimate the Earth's magnetic parameters until 2010 (new parameters will be
available before then).
When the program Geomag60.exe is run, one can select either of the two coefficient
files (IGRF10.cof or WMM2005.cof). Select Geodetic coordinates also. After the input is
completed, the program will list the parameters: D = declination (relative to true north); I =
inclination; and T = total magnetic field. Unless the coordinates for the magnetic survey are
aligned with true or geographic north, the declination value will not be suitable for the
magnetic analysis.
If total field measurements are not available for the site, perhaps because a
gradiometer was used, then a calculated value for T can substitute. These geomagnetic
parameters have also been published in maps, for example by Garland (1971 p. 222); while
these published parameters may be less accurate because they are older, the dip angle does
not change very much over the decades.

Initial model
The next step in the magnetic analysis will be an exploration of the calculated curve of
the initial magnetic model with the four different procedures for corner searching. One can
get greater confidence in these procedures if they are first tested with the calculations of an
anticipated or approximated magnetic model. This will allow the best parameters for the
searches to be selected. The calculated data from Figure 16 for the combined effect of
archaeology and agriculture have been examined with the four procedures for corner
searching that are in the Pdepth program; see Figure 17 and Figure 18, which list the
parameters for each procedure. There appears to be moderately good detection of the
corners of the archaeological polygons, and minor detection of the shallow agricultural

Page 20
 Analysis with Pdepth

polygons (because of the sensor height). The analytic signal and multiple-source Werner
methods appear to have detected the desired corners better than the other two methods.

Analysis with Pdepth

These four techniques for corner searching were next applied to the actual
measurements along the line. Figure 19 and Figure 20 show the findings; it was necessary
to change some of the parameters of the procedures from their values in Figure 17 and
Figure 18. The depth section at the bottom of each of the four panels has a curve that shows
surface topography. The multiple-source Werner analysis (Figure 20, lower panel) appears
to have been the most informative, and the corner locations and dip angles at 31 and 46 m
along the traverse are reasonable.

Analysis with Saki

The line of measurements was also analyzed with the Saki program. An initial guess
at the locations of four rectangular prisms is indicated in Figure 21; the calculated magnetic
field of these prisms is very different from the measurements, although the calculated peaks
are close to the locations of the peaks on the measurements.
Starting with this initial guess, a much better magnetic model was determined with the
Saki program. Figure 22 shows a result that is adequately good for the four main anomalies
that are seen in the measurements. The two most magnetic prisms, near n35 and n45,
suggest the character of the magnetic material in the two magnetic bands of archaeological
importance. There is no known archaeological significance to the two prisms near n15 on the
profile.
For the analysis with Saki, the topography of the surface was entered into the
program; this surface is plotted at the bottom of Figure 22. Since the slope is rather small,
the analysis would still have been very good without this refinement.
The analysis was done with four prisms, and each prism had four edges. This means
that each prism is defined by nine parameters (X and Z at each of four edges, and one value
for susceptibility), so the entire model has 36 parameters, which is more than can be iterated
at one time. In order to limit the number of changing parameters to 20 or less, the
approximate sequence of iteration that I tried was as follows:
1: Iterate 4 susceptibilities (k) alone.
2: Iterate X, Z, and k of left two prisms.
3: Iterate X, Z, and k of right two prisms.
4: Iterate Z and k of all prisms.
5: Iterate X and k of all prisms.
6: Iterate X, Z, and k of two middle prisms.
It is valuable to "plot" the calculations after every two iterations or so to see how the
calculations are approaching the measurements.

Example 2
Figure 23 shows a magnetic map with several linear features. This map was

Page 21
 Rotation of the magnetic map

measured at the site of the Bullock House at the Fredericksburg and Spotsylvania National
Military Park in Virginia; this was the location of a battle in the US Civil War during 1863. The
linear pattern of the anomalies suggests that they might be caused by refilled trenches from
the battle, and trenches are known to have been in this vicinity.
It would be easy to extract a line of measurements from the data of the map, as long
as this line was parallel one of the two sides of the map. However, the analysis procedures
assume that the line of data that is studied is perpendicular to the line of the anomalies.
Figure 24 shows the error that could result.
Fortunately, some mapping programs (such as Surfer from Golden Software) allow
one to extract lines of readings that are angled in a grid or matrix of measurements. The
green line in Figure 23 crosses a linear anomaly along a perpendicular path; this line has
been extracted and the curve is plotted in Figure 25. A half-width analysis of this anomaly
indicates a depth that would be reasonable for an archaeological feature.
Since the linear anomalies change along their length, the readings along a single line
can be somewhat irregular and perhaps not typical. The unwanted variations can be reduced
by averaging the readings along a band of parallel lines that cross the anomaly. This
averaging is easiest to do if the magnetic lines are parallel to one of the two sides of the
rectangular map. Figure 26 illustrates how part of the map in Figure 23 has been rotated.

Rotation of the magnetic map

The rotation of the measurements is not done with the matrix or grid of data; instead it
is done on a data file that contains the coordinates (X and Y) along with each measurement:
X, Y, B. If the measurements are to be rotated by a clockwise angle of 2, the new
coordinates (Xr and Yr) are:
Xr = (X - Xp) cos 2 + (Y - Yp) sin 2 + Xp
Yr = (Y - Yp) cos 2 - (X - Xp) sin 2 + Yp
where Xp and Yp indicate the coordinate of the pivot or rotation point; if this point is placed
near the middle of the important area, then the coordinates of the rotated map will change
only a little. This rotation may be done with a simple programming language (such as BASIC)
or with a spreadsheet program. After the coordinates have been rotated, it will be necessary
to grid the data with their new locations. This gridding must be done with a procedure that
does not smooth the measurements; a short search radius in the gridding procedure will
allow this.
A span can be selected in the rotated map where the anomaly does not change very
much with distance along the length of the anomaly; one selection is indicated in Figure 26.
The average anomaly across this band is plotted in Figure 27.

Analysis with Mag2dc and Saki

This line of readings was studied with the Mag2dc and Saki programs, and Figure 28
shows similar magnetic models from both.

Causes of the strong magnetic lows

Page 22
 Causes of the strong magnetic lows

While the calculations are similar to the measurements with both models, the
calculated magnetic lows are fainter than the measurements in both studies. There are four
possible causes for this: 1, Remanent magnetization; 2, Non-magnetic features; 3, Wide
features; and 4, Demagnetization.
1: Remanent magnetization can affect the ratio of magnetic highs to lows if the
direction of the remanence differs from that of the Earth's present field. The actual inclination
of the field at the location of the survey was about 67°; if the inclination is instead assumed to
be only 37°, then a magnetic model can be found that provides a much better match between
the measurements and calculations. This result is plotted with a dashed line in Figure 27.
The inclination of 37° suggests the direction of total magnetization within the prisms; this is
not the direction of remanent magnetization since there is an unknown amount of induced
magnetization. The remanent magnetization could have an angle that is even shallower than
37°. Since the inclination of the Earth's field at this site has never been as shallow as even
37° in historical times, perhaps these linear features were formed at an earlier geological
period. The susceptibility (about 45 ppt) in the model of Figure 27 is an effective
susceptibility, and not the actual susceptibility; this effective susceptibility includes the effect
of remanent magnetization.
2: Non-magnetic features usually have anomalies whose magnetic lows predominate
over their highs. These types of feature are less magnetic than the surrounding soil. One
example could be a cavity filled with air or water; another could be a limestone wall in soil that
was formed by the decomposition of limestone bedrock. Figure 29 illustrates how a magnetic
feature in the northern hemisphere can cause the same anomaly as a non-magnetic feature
in the southern hemisphere; the angles of inclination differ by 90°. If remanent magnetization
contributes to induced magnetization, a non-magnetic feature in the northern hemisphere can
also be revealed as a magnetic high; a lava tube could be an example (this is an air-filled
tunnel in magnetic basalt). Because of the shallow depth of the features that are modeled in
Figure 27, cavities are unlikely; limestone is not found in the vicinity, and so this type of
feature is also unlikely.
3: Wide features also have strong magnetic lows. Figure 30 and Figure 31 show how
magnetic lows become stronger for wider prisms. However, the analysis of the anomalies in
Figure 28 did not suggest that the sources were wide compared to their depth, and the shape
of the anomalies does not indicate that the features are wide. A magnetic map that was
measured over a wide magnetic feature is illustrated in Figure 32; the magnetic low is
stronger than the high. This rectangular feature is composed of a number of perforated steel
plates; these types of plates were used in World War 2 for the construction of temporary
aircraft runways.
4: Demagnetization can also lower the apparent inclination angle of magnetization and
therefore amplify a magnetic low. Figure 33 summarizes the origin of demagnetization, and
this figure shows how the direction of magnetization will tend to be aligned with the length of
magnetic features. This effect is significant only for features that are quite magnetic. While a
pair of horizontal steel plates might cause the anomaly of Figure 27, this type of feature
would be unlikely for the historical period of interest (19th century).

Page 23
 Quantifying the magnetic material

The name "demagnetization" comes from another effect of strongly-magnetic features:

The magnetic anomalies that they create are somewhat less than expected. A detailed
explanation of demagnetization has been given by Dunlop and Ozdemir (1997, p.83); the
practical effects of demagnetization have been summarized by Guo and others (2001). It is
difficult to calculate the effect of demagnetization, and this calculation is not often necessary;
therefore, it is rarely done.

Quantifying the magnetic material

The depths that have been interpreted by this magnetic analysis are reasonable for
archaeological features; are the magnetic properties reasonable also? The model in Figure
27 indicates that the prisms have a susceptibility that is about 45 ppt greater than that of the
surrounding soil.
In the USA, a typical topsoil will have a magnetic susceptibility of 0.2 ppt (Cook and
Carts 1962); this is 200 times less magnetic than the susceptibilities of the magnetic model in
Figure 27. While the soil in the area may be more magnetic than the average, it appears
unlikely that the magnetic features could be caused by soil contrasts. However, if there was
a good fraction of magnetite in the soil, the susceptibility could easily match the values of
about 45 ppt of the model.
My magnetic measurements on brick have found susceptibility values ranging from 1
to 20 ppt. Since the remanent magnetization of brick can be about ten times its induced
magnetization, brick could readily cause the magnetic anomalies that were measured.
However, it is unlikely for brick (fired in place or not) to have such a shallow inclination of total
magnetization. In principle, bricks could have been fired while resting on their medium-sized
faces (or edges), and then set down on their large faces (rotating each the same way); the
large remanent magnetization of this mass of brick would then be oriented to an unusually
shallow angle.
Errors in an analysis that increase the cross-sectional area of a prism will often
correlate with a decrease in the magnetic susceptibility of that prism. It may be beneficial to
calculate the magnetic moment per unit of prism length, for this has the product of these two
parameters as follows:
Ml = AkBe/(400B) where
Ml = magnetic moment per unit of prism length (= 0.47 Am)
A = cross-sectional area of prism (= 0.28 m2)
k = magnetic susceptibility contrast, SI units (= 0.04)
Be = flux density of the Earth (= 53,000 nT).
The values in parentheses at the ends of the lines are for the prism on the right side of Figure
27. The cross-sectional area of an irregular prism can be calculated with the procedure
shown in Figure 34.
The magnetic moment of steel is roughly 0.3 Am2/kg; this means that the mass of steel
per unit length for this prism could be
0.47 Am / 0.3 Am2/kg = 1.6 kg/m
Therefore, the interpretation of a sheet of steel does not appear unreasonable.

Page 24
 Example 3

While the magnetic model indicates that the features are very shallow, one must still
consider the possibility that they have a geological origin, perhaps outcroppings of a
hydrothermal mineral deposit. Other evidence suggests that this is the most likely
explanation for these linear anomalies. A century ago, mines were active in the region; a
nearby town is called Goldvein. Also, magnetic maps that have been measured at other
locations in the vicinity show many distinctive linear patterns, too many to be caused by
humans. Finally, nothing unusual was detected by two conductivity instruments from
Geonics (EM31 and EM18) and also by a ground-penetrating radar when these features were
profiled.
This example illustrates that one can interpret a magnetic anomaly in great detail, and
still not be certain of an identification of the cause of that anomaly. However, since the
evidence suggested a natural origin for these features, they have not been tested by
archaeological excavation.

Example 3
A geophysical survey was done at the West House on the Richmond National
Battlefield at Malvern Hill; this was the location in Virginia of a US Civil War battle in 1862. A
ground-penetrating radar readily traced six parallel trenches of a drainfield; these modern
features allow household sewage to seep into the soil over a wide area. A magnetic map
also included the drainfield; see Figure 35. While the trenches tend to be marked by
magnetic lows, the patterns are complex and the correlation is unclear. However, the pattern
is much clearer if a band of measurements is averaged; this has been done in Figure 36.

A simple model
A suitable magnetic model is also included in Figure 36. Prisms were placed where
the radar survey found the trenches, and the prisms were each given the same negative
susceptibility. Even with this very simple model, the calculated anomalies are very similar to
the measurements.
The negative susceptibility of the model just means that the soil within the trenches is
less magnetic that the natural soil. Perhaps part of the natural and somewhat magnetic soil
was replaced by non-magnetic gravel in the bottoms of the trenches. Even without the
addition of non-magnetic material, simply refilling a trench with the soil that was removed can
cause the soil to become less magnetic. Topsoil is usually more magnetic than subsoil and
this topsoil may have been replaced at a greater depth underground; also, any remanent
magnetization of the soil would be almost entirely lost after the trench was refilled.
Without the evidence of the radar, one would probably have guessed that the prisms
had a positive susceptibility, such as in the calculation of Figure 37 (if the susceptibilities were
negative there, the anomaly curve would just be a mirror image about the horizontal line).
This error in interpretation would place the trenches in the wrong locations, and excavations
might fail to find the trenches. Perhaps the two polarities of susceptibility in the magnetic
model could be distinguished by the measurements that were made more distant from the
features; the broad low at the right side of Figure 37 might help. In environmental surveys,

Page 25
 Other software

parallel trenches are often found where debris has been disposed in the soil; since these
trenches likely contain iron in containers, these are detected as magnetic highs.
The trenches that are studied with the data of Figure 35 must not contain iron pipes,
for iron would cause much stronger anomalies than those of the measurements. Ceramic
pipes are possible, but plastic pipes are most likely for the drains. Also, there cannot be
aluminum pipes in the trenches, for no metal was revealed with a conductivity survey (using a
Geonics EM38 and EM31).

Other software
Several companies probably supply computer software that is suitable for the analysis
of linear anomalies. One of these companies is Interpex (www.interpex.com). While the
company sells a Windows program for this magnetic analysis, they have a demonstration
version of an earlier (DOS) program that is free. This is available as file MagixSL.zip on the
company's web site.
Other commercial developers of geophysical software, such as Geosoft and PetRos
EiKon may also sell programs for the analysis of linear magnetic anomalies.
The central part of a computer program that calculates the anomaly of a cluster of
magnetic prisms has been published by Won and Bevis (1987). This subroutine can be the
basis of a larger program for the analysis of linear anomalies.
The group of software from the U. S. Geological Survey includes additional programs
that are valuable for the analysis of magnetic data; some of these include: FFTfil, Magpoly,
and Chess. The potential field division of the US Geological Survey was closed in about
1995 (in order to reduce the cost of government) and most of the individuals there have
moved to other jobs. The USGS no longer develops any significant amount of geophysical
software.
An improvement of the Euler deconvolution has been proposed by Cooper (2002), and
he has written a computer program that applies those principles (listed below as Euler).
Several computer programs are included within the folder MagLin that contains this
document; these programs include:
Mag2dc and Euler, from G. R. J. Cooper
Pdepth and Saki, from the U. S. Geological Survey
Geomag60 from the U. S. National Geophysical Data Center
In addition, this folder contains data for Figure 4, Figure 7, Figure 11, and Figure 17; the
auxiliary files that are needed for an analysis of the data with these programs are also
included.

Conclusion
A prismatic analysis can give one the greatest confidence in an interpretation. This is
because the calculated field can be compared to the measurements. While many different
magnetic models may yield about the same calculated curve, these differences in the models
give one a good understanding of how much of the detail in the actual underground feature
might reliably be determined by the measurements.

Page 26
 Annotated bibliography

There are a good variety of corner searching procedures (Euler deconvolution, analytic
signal, Werner deconvolution). These procedures are best applied if the data are so complex
that it is not practical to try a prismatic analysis. The ideas behind these procedures are very
clever. For the examples here, the multiple-source Werner method appears to have given
results that were clearer and more distinct than the other methods.
Because of natural irregularities along the length of linear anomalies, it is valuable to
average the anomaly across a band of data, rather than extracting a single line of
measurements. Using this average, the analysis will be more certain and more
representative of the actual feature.
It is important to take the time to understand the operation and limitations of all of
these methods. One cannot just push a few keys on a computer without thinking, and then
expect to get correct or valuable findings; it is necessary to understand how these programs
work. If the process was simple, it would not be very interesting. As it is, the analysis of
linear magnetic anomalies is a fine intellectual challenge.

You are welcome to copy this document and you may give it to anyone else. Should
you distribute any part of this widely (such as on the internet), please notify me.

Annotated bibliography
Not all of these publications are referenced in the text above.

Abbas, Abbas Mohamed, Tareq Fahmy Abdallatif, Fathy A. Shaaban, Ahmed Salem, and
Mancheol Sul, 2005. Archaeological investigation of the eastern extensions of the
Karnak temple using ground-penetrating radar and magnetic tools. Geoarchaeology
20(5):537-554. Applies ratios of derivatives of the analytic signal for estimates of
depths.
Atchuta Rao, D., H. V. Ram Babu, and P. V. Sanker Narayan, 1980, Relationship of magnetic
anomalies due to subsurface features and the interpretation of sloping contacts.
Geophysics 45(1):32-36. Shows that the magnetic anomaly of a horizontal cylinder is
the same as the first horizontal derivative of the anomaly of a thin sheet, which is also
the same as the second horizontal derivative of the anomaly of a flat boundary.
Atchuta Rao, D., H.V. Ram Babu, and P.V. Sanker Narayan, 1981. Interpretation of
magnetic anomalies due to dikes: The complex gradient method. Geophysics
46(11):1572-8. In addition to the amplitude of the analytic signal, this method also
analyzes the phase.
Barbosa, Valeria C.F., Joao B.C. Silva, and Walter E. Medeiros, 1999. Stability analysis and
improvement of structural index estimation in Euler deconvolution. Geophysics
64(1):48-60.
Bastani, Mehrdad, and Laust B. Pedersen, 2001. Automatic interpretation of dike parameters
using the analytical signal technique. Geophysics 66(2):551-61.
Blakely, Richard J., 1995, Potential theory in gravity and magnetic applications. Cambridge
University Press (Cambridge). Werner deconvolution is described on p.245-250, and

Page 27
 Annotated bibliography

is based on the article by Ku and Sharp. This is the most complete description of the
process that is found in a book.
Breiner, S., 1973. Applications Manual for Portable Magnetometers. Geometrics
(Sunnyvale, California). Available on the web site: www.geometrics.com. A general
introduction to the principles and procedures for magnetic surveys.
Brinker, Russell C., 1969. Elementary Surveying, 5th edition. International Textbook Co.,
New York.
Cady, John W., 1980. Calculation of gravity and magnetic anomalies of finite-length right
polygonal prisms. Geophysics 45(10):1507-12.
Campbell, David L., 1983. BASIC programs to calculate gravity and magnetic anomalies for
2.5-dimensional prismatic bodies. U.S. Geological Survey Open-File Report 83-154,
Denver, Colorado.
Cook, John C., and Stanley L. Carts, Jr., 1962. Magnetic effects and properties of typical
topsoils. Journal of Geophysical Research 67(2):815-828. Soils at 250 sites across
the USA were tested; little correlation was found between color and susceptibility.
Cooper, G. R. J., 2002. An improved algorithm for the Euler deconvolution of potential field
data. The Leading Edge 21(12):1197-1198.
Davis, Raymond E., Francis S. Foote, James M. Anderson, and Edward M. Mikhail, 1981.
Surveying Theory and Practice, sixth edition. McGraw-Hill, New York.
Debeglia, Nicole, and Jacques Corpel, 1997. Automatic 3-D interpretation of potential field
data using analytic signal derivatives. Geophysics 62(1):87-96. Allows a better
distinction between the sources of anomalies that are near each other.
Dobrin, Milton B., and Carl H. Savit, 1988. Introduction to Geophysical Prospecting, 4th
edition. McGraw-Hill (New York).
Dunlop, David J., and Ozden Ozdemir, 1997. Rock Magnetism: Fundamental and Frontiers.
Cambridge University Press (Cambridge).
Garland, George D., 1971. Introduction to Geophysics: Mantle, Core, and Crust. W. B.
Saunders (Philadelphia).
Godson, Richard H., and Margaret R. Mall, 1989. Potential-field programs for IBM
compatible microcomputers, version 1.0. Open-File Report 89-197, U. S. Geological
Survey (Denver, Colorado). This was the original release of the potential-field
software. An updated version was released as OFR 92-18. This publication has a
detailed description of the format of the files that are read by the USGS programs.
Green, R., 1980, Field processing of magnetic data. Geophysical Prospecting 28(3):384-391.
Shows that a programmable calculator may be used for the analysis of Werner
deconvolution.
Grauch, V. J. C., and others, 1993. Materials provided at the workshop "Geophysical map
interpretation on the PC", convened April 21-22, 1993. Open-File Report 93-560, U. S.
Geological Survey. Part of this report includes the updated potential field programs
from the USGS.
Guo, Wanwu, Michael C. Dentith, Robert T. Bird, and David A. Clark, 2001. Systematic error
analysis of demagnetization and implications for magnetic interpretation. Geophysics

Page 28
 Annotated bibliography

66(2):562-570.
Hansen, Richard O., and Marc Simmonds, 1993, Multiple-source Werner deconvolution.
Geophysics 58(12):1792-1800. Generalizes the procedure for the analysis of adjacent
bodies. Points out that the process is not actually deconvolution. Includes a
comparison with CompuDepth. This is the most recent publication on this method.
Hansen, R. O., 2002. Multiple-source Euler deconvolution. Geophysics 67(2):525-35. For
the analysis of complex magnetic maps, but includes a description of 2D analysis.
Hartman, Ronald R., Dennis J. Teskey, and Jeffrey L. Friedberg, 1971, A system for rapid
digital aeromagnetic interpretation. Geophysics 36(5):891-918. The first publication of
the method of Werner deconvolution in the US. Has a description of how clustering of
calculated corners can imply a magnetic boundary. States that the vertical gradient
can be used for this process (p.909).
Hood, Peter, 1965. Gradient measurements in aeromagnetic surveying. Geophysics
30(5):891-902. Describes Euler analysis at anomaly peaks.
Hsu, Shu-Kun, Jean-Claude Sibuet, and Chuen-Tien Shyu, 1996. High-resolution detection
of geological boundaries from potential-field anomalies: An enhanced analytic signal
technique. Geophysics 61(2):373-86. Rather than just gradients, higher derivatives of
the magnetic field are calculated; this allows closely-spaced anomalies to be analyzed
with less interference and error.
Jain, Sudhir, 1976, An automated method of direct interpretation of magnetic profiles.
Geophysics 41(3):531-541. Werner deconvolution. Analysis of calculated data shows
the effect of nearby bodies. Discusses the difference between a magnetic sheet and a
magnetic interface.
Johnson, William W., 1969. A least-squares method of interpreting magnetic anomalies
caused by two-dimensional structures. Geophysics 34(1):65-74. This has a complete
description of how the magnetic anomalies of 2D prisms are calculated and how the
Marquardt method can be used to change the locations of prism edges for an
inversion of a measured profile.
Keating, Pierre B., and Mark Pilkington, 1990. An automated method for the interpretation of
magnetic vertical-gradient anomalies. Geophysics 55(3):336-43.
Kilty, Kevin T., 1983, Werner deconvolution of profile potential field data. Geophysics
48(2):234-237. Describes the subtraction of a regional anomaly, and also data noise
and the correlation between parameters.
Ku, Chao C., and John A. Sharp, 1983, Werner deconvolution for automated magnetic
interpretation and its refinement using Marquardt's inverse modeling. Geophysics
48(6):754-774. See also the discussion of the article: Geophysics 49(7):1119. This
article has the most complete description of the process. Operator width is discussed
on p. 756 and 765.
Ku, Chao C., and John A. Sharp, 1983. Werner deconvolution for automated magnetic
interpretation and its refinement using Marquardt's inverse modeling. Geophysics
48(6):754-774. The last part of this article has a description of Marquardt's method.

Page 29
 Annotated bibliography

Mikhailov, Valentine, Armand Galdeano, Michel Diament, Alexei Gvishiani, Sergei Agayan,
Shamil Bogoutdinov, Elena Graeva, and Pascal Sailhac, 2003. Application of artificial
intelligence for Euler solutions clustering. Geophysics 68(1):168-80. Eliminate some
improper solutions.
Mushayandebvu, Martin F., P. van Driel, Alan B. Reid, and James Derek Fairhead, 2001.
Magnetic source parameters of two-dimensional structures using extended Euler
deconvolution. Geophysics 66(3):814-23. Estimate susceptibility contrast and
interface slope also.
Nabighian, Misac N., 1972. The analytic signal of two-dimensional magnetic bodies with
polygonal cross-section: Its properties and use for automated anomaly interpretation.
Geophysics 37(2):507-517. The first description of the procedure, and it is a clear
description of the 2D analysis.
Nabighian, Misac N., 1974. Additional comments on the analytic signal of two-dimensional
magnetic bodies with polygonal cross-section. Geophysics 39(1):85-92. A
restatement and extension of the principles and procedures with clarification of the
original paper.
Nabighian, Misac N., 1984. Toward a three-dimensional automatic interpretation of potential
field data via generalized Hilbert transforms: Fundamental relations. Geophysics
49(6):780-6. First description of the study of magnetic maps; complex.
Nelson, J. Bradley, 1988. Comparison of gradient analysis techniques for linear
two-dimensional magnetic sources. Geophysics 53(8):1088-95.
Parasnis, D.S., 1997. Principles of Applied Geophysics, 5th edition. Chapman and Hall
(London).
Press, William H., Brian P. Flannery, Saul A. Teukolsky, and William T. Vetterling, 1986.
Numerical recipes. Cambridge University Press (New York). Pages 523 - 528
describe the Levenberg-Marquardt method and include source code.
Radhakrishna Murthy, I. V., 1998, Gravity and magnetic interpretation in exploration
geophysics. Geological Society of India (Bangalore), Memoir 40. A Fortran program
for Werner deconvolution is described on p. 328-332; a listing and an executable
program are included on a floppy disk.
Ravat, D., 1996. Analysis of the Euler method and its applicability in environmental magnetic
investigations. Journal of Environmental and Engineering Geophysics 1(3):229-238.
Reid, Alan B., 2003. Euler magnetic structural index of a thin-bed fault. Geophysics
68(4):1255-6. Lists other indices also.
Reid, A.B., J.M. Allsop, H. Granser, A.J.Millett, and I.W. Somerton, 1990. Magnetic
interpretation in three dimensions using Euler deconvolution. Geophysics 55(1):80-91.
Includes an illustration of excellent data and analysis.
Robinson, Edwin S., and Cahit Coruh, 1988, Basic exploration geophysics. John Wiley (New
York). Includes a summary of the paper by Hartman, Teskey, and Friedberg on
Werner deconvolution.
Roest, Walter R., Jacob Verhoef, and Mark Pilkington, 1992. Magnetic interpretation using
the 3-D analytic signal. Geophysics 57(1):116-25. Provides practical procedures for

Page 30
 Annotated bibliography

Nabighian's method and adds new details about the analysis; plots depths with circles
whose diameters increase with interpreted depths. See also the discussion of this
paper in (1993) Geophysics 58(8):1214-6 and in (1998) Geophysics 63(2):667-70.
Salem, Ahmed, and Dhananjay Ravat, 2003. A combined analytic signal and Euler method
(AN-EUL) for automatic interpretation of magnetic data. Geophysics 68(6):1952-61.
The combination of the two techniques allows for the analysis of compact bodies,
where the analytic signal method otherwise fails.
Slack, Howard A., Vance M. Lynch, and Lee Langan, 1967. The geomagnetic gradiometer.
Geophysics 32(5):877-92. A restatement of Hood's (1965) finding is on p. 833-5. See
also the discussion by Nelson C. Steenland (1968) in Geophysics 33(4):680-3.
Stavrev, Petar Y., 1997. Euler deconvolution using differential similarity transformations of
gravity and magnetic anomalies. Geophysical Prospecting 45(2):207-246. A detailed
description of the procedure.
Telford, W. M., L. P. Geldart, and R. E. Sheriff, 1990. Applied Geophysics, second edition.
Cambridge University Press, Cambridge.
Thompson, D.T., 1982. EUDELPH: A new technique for making computer-assisted depth
estimates from magnetic data. Geophysics 47(1):31-37. A clear description of 2D
Euler deconvolution.
Thurston, Jeffrey B., and Richard S. Smith, 1997. Automatic conversion of magnetic data to
depth, dip, and susceptibility contrast using the SPI method. Geophysics
62(3):807-13. An elaboration of the analytic signal method that allows susceptibility
contrast to be estimated.
Wagini, Alexander, 1985. An automatic program for the interpretation of 2D gravity and
magnetic anomalies. Open-File Report 85-377, U. S. Geological Survey (Denver,
Colorado). This appears to be a similar program to SAKI, although this program does
not appear to have been used further at the USGS. This program was derived from an
earlier program called Hypermag (USGS OFR 83-241).
Webring, Michael, 1985. SAKI: A Fortran program for generalized linear inversion of gravity
and magnetic profiles. Open-File Report 85-122, U. S. Geological Survey (Denver,
Colorado). The operation of the original version of the program is described here;
changes in later versions are described in the SAKI.hlp file. The source code of the
original program is listed with this report.
Weymouth, John W., 2004. Hopeton geophysical survey: The 2004 season. Report
prepared for the Midwest Archeological Center, National Park Service, Lincoln,
Nebraska, for Purchase Order P6115040044.
Won, I.J., and Michael Bevis, 1987. Computing the gravitational and magnetic anomalies
due to a polygon: Algorithms and Fortran subroutines. Geophysics 52(2):232-8.
There are several special or difficult points in a calculation, and the procedure here
gives correct calculations at those points.
Zhang, Jisheng, 2001. An analysis of the accuracy of magnetic source-body geometry
determined from the 3-D analytic signal. Geophysics 66(2):579-81. Shows how the
error in the estimate of depth increases as features are less elongated.

Page 31
Publication history:
4 May 2007: Added appendix illustrating the program Pdyke and corrected four
typographical errors.
4 December 2006: Original report
 4
5-m long box
Calculated magnetic anomalies
prism, infinite length 1-m square cross-section, k = 1 ppt
3 Be = 53,500 nT; Ie = 68o; De = 0
Sensor height = 1 m above top

cube
2 1.64 m
1.56 m Half-widths
Depth = 1.50 m
Magnetic anomaly, nT

1.56 m
1

0 0
sensor line

Depth below sensor, m

-1 1

magnetic
square

-2 2

-3 3
-3 -2 -1 0 1 2 3
North coordinate, m

Figure 1: The half-width rule. The width of a magnetic anomaly at half of its peak
amplitude is approximately the depth of a compact object below the level of the
magnetic sensor. This approximation is good for a cube, a prism, and an elongated
box-like feature. These calculations plot the total-field magnetic anomaly of a feature
with a square cross-section, shown in the bottom half of the figure.
 4
Calculated magnetic anomalies
1-m square prism, k = 1 ppt
Be = 53,500 nT; Ie = 68o; De = 0
Sensor height = 1 m above top
3
Total field
Sensor spacing on Vertical gradient of total field
gradiometer = 1 m Vertical gradient of vertical component

2
1.56 m
Magnetic anomaly, nT and nT/m

Half-widths
1.24 m Depth = 1.50 m
1 1.26 m

0 0
sensor line

Depth below sensor, m

-1 1

magnetic
prism

-2 2

-3 3
-3 -2 -1 0 1 2 3
North coordinate, m

Figure 2: The half-width rule and gradiometer measurements. The estimates of depth
will be somewhat shallow but may still be helpful. The vertical gradient of the vertical
component of the Earth's magnetic field is typically measured with a fluxgate
magnetometer. The narrower widths of the anomalies from gradiometers indicate the
greater spatial resolution of these instruments.
 12
Calculated magnetic anomaly
Rectangular prism 1 x 4 m, k = 2 ppt
Be = 53,500 nT; Ie = 68o; De = 0
Sensor height = 1 m above top

6 Half-width = 3.1 m
Depth = 1.5 m
Magnetic anomaly, nT

sensor line
0

Depth below sensor, m

-6
1

2
magnetic rectangle
3

-12 4
0 5 10 15
North coordinate, m

Figure 3: Failure of the half-width rule. If the cross-section of a feature is not compact,
for example it is a rectangle rather than a square, then the half-width rule gives an
estimate that is too deep. While this magnetic anomaly is not symmetrical about the
peak, it may not always be easy to see from an anomaly that the source object is wide
and not compact. The curve here shows a total-field calculation.
 Calculation of initial model

Red curve =
measurements

Green curve =
calculations

Initial model
Susceptibility = 1 ppt

Recalculation after moving two right-hand corners to right

Amplitudes of the
curves differ

Broadened
model

Figure 4: The application of the computer program called Mag2dc. In the top panel,
calculations from a simple magnetic model can be compared to the measurements. In
the lower panel, the model has been widened, and now the locations of both the
magnetic high and the low agree between the measurements and calculations. Note:
The "measurements" here are calculations from the simple model in Figure 3.
 Recalculation after doubling susceptibility

Calculated and
measured curves
are very similar

Susceptibility = 2 ppt

A later model and its calculation

The two curves

appear to be
identical

A good model

Figure 5: A refinement of the magnetic model. In the upper panel, the susceptibility of
the polygon has been doubled, and now the measurements are very similar to the
calculations. The model error is listed as a misfit at the bottom of each panel. In the
lower panel, a different model gives an even lower error. These displays illustrate how
manual (non-automatic) changes may be made to the parameters of the magnetic
model.
 Models of a magnetic prism
Initial, k = 1 ppt, error = 40.8
12 Poor, k = 5.6 ppt, error = 10.66
Fair, k = 1.2 ppt, error = 3.42
Good, k = 1.9 ppt, error = 0.32

Prismatic models
Magnetic anomaly, nT

Be = 53,500 nT; Ie = 68o; De = 0

sensor line
0

Depth below sensor, m

-6
1

2
cross sections
3

-12 4
0 5 10 15
North coordinate, m

Figure 6: Reducing the error of a model. The final (Good) model, with a minimum error,
has a calculated curve that is essentially identical to the measurements. However, the
models that are called Fair and Poor have calculations that do not differ very much from
those of this final model, although the model shapes and depth are quite different. That
is, rather different models can appear about equally good.
 Command file: Band.cmd
&parms Necessary at start
hfile='band.dat' File with magnetic data
compon='t' Total field, not component
efield=53500 Earth’s field = 53,500 nT
einc=68.0 Inclination = 68 degrees
edec=0 Declination = 0
jave=0 Do not remove average
azmuth=0 Prism is perpendicular to line
iplotr=10 Use VGA resolution
& Needed to mark end of data

Pdepth Start program

enter command filename :band.cmd File contains parameters above
enter gravity profile file name (or return): No gravity data; hit Enter key
Plot like that below is seen
Click on “geology” square
enter name of geology file: Text at top of plot; hit Enter key
enter name of cross section file:band.cor See body rectangle below:

Plot of measurements

Rectangular body
added with
geology option

Figure 7: Starting another analysis program, called Pdepth. General parameters for the
program can be entered from a command file; an example is listed at the top. However,
these parameters may also be entered with the keyboard when the program requests
individual values. If a file of measurements (band.dat) is read, these values will be
plotted, as seen at the bottom. The rectangular body is optional.
Black text is displayed by the program, red text is entered by the operator of the
program, and the green text has explanations.
 Euler: Dimension = n; Window = 15; Tolerance = 0

Euler

Analytic Signal: No parameters to set, but plot curves

Analytic Signal

Figure 8: Two analysis procedures in the program Pdepth. Note how these
procedures may sometimes locate the corners of magnetic bodies from an analysis of
the measurements: Symbols cluster at the two upper corners of the rectangles in the
lower half of each panel. The parameters that were applied to the two procedures are
listed in green at the top of the panels.
 Werner: Degree = 2; Cluster = 1; Individual = y

Werner

Multiple-Source Werner: Window = 0; Cluster = 2; Minimum = 5; Edges = 2 2

Multiple-Source Werner

Figure 9: Two different types of Werner analysis with the Pdepth program. The earlier
and simpler analysis is shown in the top panel; as with Figure 8, estimates of the
corners of the magnetic body cluster near the upper corners of the rectangle that
locates the body. The Multiple Source Werner (MSW) analysis in the lower panel is
more complex, but may give more accurate results.
 Initial calculation; error = 2.13

No symbols =
measurements

Symbols =
calculations

Four-sided
polygon

Recalculation after moving right-hand corner; error = 1.00

Green polygon =
model

Black rectangle =
actual source

Figure 10: A second part of the program, Pdepth. Like Mag2dc, this program can also
allow one to move a corner of a polygonal prism, and quickly see how the calculations
change. These two panels illustrate how model error (listed at the top) can be reduced
by moving the corner of the polygon; this move is started by placing the mouse cursor in
the "move cor" (move corner) square on the right, and clicking there.
enter command filename : Hit Enter: No control file
model filename: bandpo.bod Initial magnetic prism
gravity data filename (optional): Hit Enter: No gravity data
magnetic data filename (optional): band.dat Magnetic data
number of vertices = 4 Corners
number of bodies = 1 Prisms
present earth`s field vector Desired values below:
intensity= 0.000000E+00 Magnitude = 53,500 nT
inclination= 0.000000E+00 Dip angle = 68 degrees
declination= 0.000000E+00 Line goes magnetic north
enter 3 new values Enter changes next
53500,68,0 Three values entered
change the profile azimuth ? n Do not change direction
do you want total field or a component :t Total magnetic field
function:setup Change a program parameter
setup parameters :datum Background field
set floating magnetic datum ? n Set constant at zero
setup parameters :return Finished with Setup
function:calc Calculate the magnetic field
magnetic rms error 1.01 Error: Measured - Calculated
function:plot Plot the data
Hit Enter key to see plot below

Magnetic
anomaly

Measured
= Red

Calculated
= Black

Surface

Magnetic
prism

Hit Enter key to quit plot

Function:exit Quit program

Figure 11: Starting the program called Saki. This is a companion to the Pdepth
program, and it allows automatic iteration (inversion) of the parameters of a polygonal
prism. As before, black text is displayed by the program, while red text is entered by the
user; additional explanations are noted with green text.
function:invert Automatically change parameters
enter number of free x vertices:4 Move all corners in traverse direction
enter indices : 1 2 3 4 List of corner numbers
enter number of free z vertices:4 Move all corners in depth direction
enter indices : 1 2 3 4 List of corner numbers
enter number of free susceptibilites :1 Change susceptibility of all prisms
enter indices : 1 Number of the single prism
active magnetic range (km) 5.000000E-02 15.000000
starting rms error 1.0093 Initial root mean square error
print parameter correlations ? n Not needed
Several lines of print are not shown
parm change new value List of new parameters
x 1 -.1803 3.757 Along-traverse coordinates
x 2 .1799 8.275 of the prism
x 3 .1550 7.263
x 4 -.1202 4.655
z 1 1.1624E-02 .7333 Depth coordinates
z 2 -4.7295E-02 1.262 of the prism
z 3 .1489 2.635
z 4 5.8967E-02 1.636
s 1 3.8138E-06 1.5381E-04 Susceptibility, in cgs units
iteration 1, rms error is .4422 Error: Measured - Calculated
percent improvement is 56.18 Compared to initial error above
save new parameter values ?y Keep them, for they are better
another iteration ?y Continue to improve the model
After 5 - 15 iterations:
another iteration ? n When the error is small
function:plot Look at the result, below:

Measured
and
Calculated
curves
are the same

Model prism
is exact

Figure 12: Concluding the operation of the Saki program. This shows how the corners
of the prism (X and Z vertices) and its magnetic susceptibility can be changed
automatically by the program until the model calculations are very close to the
measurements. For this simple illustration, the program found essentially an exact
solution for the shape and susceptibility of the prism.
 Magnetic fields (-53,500nT)
40 Measurement interval = 0.2 s and 0.02 m
Survey 27 June 2004
Sensor height = 0.3 m

30
steel
rebar
Sensor height = 1.3 m
at n75
20
Magnetic anomaly, nT

10

0
magnetic bands

-10

-20
0 25 50 75
north coordinate, m

Figure 13: Magnetic measurements along a line. The two major magnetic anomalies
near the middle of this profile are the most important. As is reasonable, the anomalies
at the lower sensor height show the greatest variability. The anomalies at the upper
sensor height are smoother, and some of the features that cause these anomalies may
be deeper underground.
Figure 14: Magnetic features along the line. These two types of features may be the
sources of the patterns in Figure 13. The two magnetic bands are important to
archaeology. The many small and linear features are caused by plowing in this farmed
field. These shallow features are modern and they interfere with the sought-after
patterns. The line of magnetic measurements in oblique to these features.
 25
Effect of shallow agriculture
Archaeology alone
20 Archaeology and agriculture

15
Sensor height = 0.3 m
10

5
Magnetic anomaly, nT

-5

-10

-15 Calculated magnetic profiles

Total magnetic field
-20 Be = 53,500 nT; Ie = 68o; De = 0o

Agriculture, k = 2 ppt
-25 surface

Depth, m
0
Archaeology, k = 5 ppt 1
-30 2
0 5 10 15 20 25
North coordinate, m

Figure 15: The effect of the magnetic features when the sensor is at a low height. The
bottom section shows a simple model or cross-section of the deep archaeological and
the shallow agricultural features that are believed to be along the line of survey. The
agricultural features have a large effect on the readings. Only the middle 25-m span of
the line in Figure 13 is plotted here.
 12

Effect of shallow agriculture

Archaeology alone
Archaeology and agriculture
8

Sensor height = 1.3 m

4
Magnetic anomaly, nT

-4

Calculated magnetic profiles

Total magnetic field
-8 Be = 53,500 nT; Ie = 68o; De = 0o

Agriculture, k = 2 ppt
surface

Depth, m
0
Archaeology, k = 5 ppt 1
-12 2
0 5 10 15 20 25
North coordinate, m

Figure 16: The effect of magnetic features when the sensor is higher. With the sensor
at a height of 1.3 m, the shallow agricultural "noise" is reduced relative to the anomalies
from deeper archaeological features. While the spatial resolution is lower at this greater
height, the simpler pattern of the anomalies aids the analysis of larger features.
 Euler: Dimension = 0.3; Window = 5: Tolerance = 50

Euler

Analytic Signal: No parameters to set, but plot curves

Analytic Signal

Figure 17: Analysis of the simple model with the Euler and analytic signal methods.
The blue symbols locate the estimates (from Pdepth) of the corners of magnetic
features. While some of these match the corners of the magnetic models, some
corners have not been detected. The analytic signal of the data and its horizontal
derivative are plotted in the upper part of the lower panel with dashed lines.
 Werner: Degree = 0; Cluster = 20; Individual = y

Werner

Multiple-Source Werner: Window = 3; Cluster = 5; Minimum = 5; Edges = 3 3

Multiple-Source Werner

Figure 18: Analysis of the simple model using the two Werner procedures in the Pdepth
program. Again, some of the known corners have been approximated with these
methods. Since the data from a height of 1.3 m are being studied, there is little
detection of the shallow agricultural features.
 Euler: Dimension = 0.3; Length = 20; Tolerance = 15

Euler

Analytic Signal: No parameters to set, but plot curves

Analytic Signal

Figure 19: Analysis of the actual measurements with the Euler and analytic signal
methods. The full 75-m span of data is studied here and the surface topography is
plotted in each of the two depth sections. Some of the symbols that are at a depth of
0.5 - 1.5 m in the two sections are reasonably caused by archaeological features.
 Werner: Degree = 0; Cluster = 10; Individual = y

Werner

Multiple-Source Werner: Window = 3; Cluster = 5; Minimum = 5; Edges = 4 4

Multiple-Source Werner

Figure 20: Werner analysis of the measurements. Because of the complexity of the
archaeological features, the multiple-source Werner method is more suitable for
separating and distinguishing them. The symbols to the right of the 50-m point are
probably less significant than those on the left side.
 Magnetic anomaly: Calculation from initial model
Sensor height = 1.3 m
30 Be = 53,500 nT; Ie = 68o; De = -6.4o 116

20 measurements
112
Calculated anomaly, nT

10

108

Elevation, m
0
Measured anomaly, nT

calculations
104

-10
vertical
surface exaggeration = 9
C
1.0 D 100
B 1.0
-20 1.0

A
1.0 Initial magnetic prisms
Susceptibility in ppt

-30 96
0 25 50 75
north coordinate, m

Figure 21: The initial model that was entered into the Saki program. The RMS
(root-mean-square) error of this model is 8.69. First estimates of the four prisms are
shown in the cross-section at the bottom. While the locations of the calculated
anomalies are nearly correct, the amplitudes are much lower than the measurements.
The height of the magnetic sensor is 1.3 m above the ground surface.
 Magnetic anomaly: Calculation from final model
Sensor height = 1.3 m
30 Be = 53,500 nT; Ie = 68o; De = -6.4o 116

20
112
Calculated anomaly, nT

measurements
10

108

Elevation, m
0
Measured anomaly, nT

104
calculations
-10

vertical
surface
exaggeration = 9

A C 100
6.2 D
-20 2.1 5.0
B
2.0

Final magnetic prisms

Susceptibility in ppt
-30 96
0 25 50 75
north coordinate, m

Figure 22: The final magnetic model that was determined from the Saki program. The
RMS error has been reduced to 3.36. Except at the right side, the calculated curve is
quite similar to the measurements. While the features extend to a greater depth than is
reasonable for archaeological constructions, the bottoms of magnetic features are
always more difficult to determine than their upper surfaces.
 Magnetic map, measurement interval = 2.5 ft, sensor height = 2.5 ft (0.76 m), contour interval = 10 nT
80

N (mag)

60
North coordinate, ft

-472 nT
40

971 nT

20

-20
50 75 100 125 150 175 200
East coordinate, ft

Figure 23: Magnetic measurements of an area that is 46 m wide. Linear features that are about 12 m long are visible on
the western side of this map. An important historic cellar was revealed near E165 N35, where the contour lines are not
completed. Lines of traverse went east-west during the survey and the line spacing was 2.5 ft (0.76 m). This survey was
done on 3 May 2000.
 Calculated magnetic anomaly
Rectangular prism 1 x 4 m, k = 2 ppt
Be = 53,500 nT; Ie = 68o; De = 0
12 correct: error:
perpendicular oblique

prism
6
Magnetic anomaly, nT

line of measurement

sensor line
vertical
-6 exaggeration = 2

Depth below sensor, m

0

2 ppt correct model

2
k = 3.9 ppt model has error
-12 4
-20 -15 -10 -5 0 5 10 15 20
North coordinate, m

Figure 24: Errors caused by profiles that are not perpendicular to magnetic features. If
the line of analysis is oblique to the line of the feature, then the anomaly is broadened.
Analysis of this broadened anomaly can result in a magnetic model that is too deep, too
wide, and too magnetic. However, small angles from the perpendicular (such as in
Figure 14) cause only small errors in an analysis.
 20
Anomaly perpendicular to a linear feature
Data: E75 N30 to E100 N45
Sensor height = 2.5 ft (0.76 m)

Half-width = 3.8 ft (1.1 m)

0
Magnetic anomaly, nT

-20

-40
0 10 20 30
Distance northeasterly from E75 N30, ft

Figure 25: The extraction of a profile that is perpendicular to a linear feature. The
measurements that were made near the green line in Figure 23 (near E85 N30) were
extracted from the grid of data with a mapping program. The half-width of the magnetic
high suggests that the feature could be about 1.3 ft (0.4 m) underground.
 Rotated by 36 degrees clockwise about E100 N20

Rotated magnetic map, contour interval = 10 nT

60
N (mag)

50

40
northerly coordinate, ft

30 averaging
span

20

10

-10

-20
80 90 100 110 120 130
easterly coordinate, ft

Figure 26: A rotation of part of the magnetic map in Figure 23. This aligns the
anomalies with the sides of the new map; it then allows an average magnetic anomaly
to be calculated in a span across the readings. This averaging can reduce the effect of
changes in the anomaly along the length of the feature; therefore, the resulting analysis
may be more accurate.
 40

Average magnetic anomaly across linear features

Rotated magnetic map (36o clockwise)
Average in span n20 to n35
Sensor height = 2.5 ft (0.76 m)
Earth: 53,000 nT; Ie= 67.1o; De= 44.3o
20

Half-width
= 5.2 ft
Magnetic anomaly, nT

Magnetic profiles
(1.58 m)
measured
calculated
0
measured
Average (background) = -5.7 nT

calculated
-20

surface
0

Depth, ft
prisms: Assumed 2
k = 50 ppt k = 40 ppt inclination = 37o 4
-40 6
80 90 100 110 120 130
easterly coordinate, ft

Figure 27: An averaged magnetic anomaly across the two linear features. The
measurements are drawn with a solid red line. The dashed black line of the calculations
is a good match to the measurements if the Earth's field is assumed to dip at an angle
of 37°, rather than its actual dip of 67°.
 Calculation with Mag2dc program

Solid black curve =

calculations

Dashed green curve =

measurements

Magnetic prisms

Calculation with Saki program

Red curve =
calculations

Black symbols =
measurements

Magnetic prisms

Figure 28: Analysis of the linear features with the Mag2dc and Saki programs. While
there is moderate agreement between the measurements and calculations, the
calculated magnetic lows are fainter than the measured lows. The models in the two
programs are somewhat different, but both are quite magnetic. The susceptibilities of
the two prisms in the Saki program are 26 and 28 ppt (SI units).
 Calculated magnetic anomalies
Rectangular prisms 1 x 4 m
12 Be = 53,500 nT; De = 0
Sensor height = 1 m above top

6
Magnetic anomaly, nT

sensor line
0

Depth below sensor, m

-6
1
k = 2 ppt magnetic prisms k = -2 ppt
2
vertical
3
exaggeration = 2
-12 4
0 5 10 15 20 25 30
North coordinate, m

Figure 29: Identical anomalies from very different features. A magnetic feature in the
northern hemisphere can have the same anomaly as a non-magnetic feature in the
southern hemisphere. The directions of assumed magnetization here are exactly 90°
apart. The feature on the right side could also be a lava tube (an air cavity in volcanic
basalt) in the northern hemisphere.
 15
Prism goes east-west
Calculated magnetic anomalies
Rectangular prism, k = 2 ppt
Be = 53,500 nT; Ie = 68o; De = 0
Sensor height = 1 m above top
10

5
Magnetic anomaly, nT

-5
sensor line
0

Depth below sensor, m

1
-10
2
Increasing width of prism vertical
exaggeration = 2 3

-15 4
-5 0 5 10 15 20 25
North coordinate, m

Figure 30: The effect of the width of a prism that extends east-west. The amplitude of
the magnetic low increases with the width of the prism. Note also that the major
anomalies are found at the edges of the prism, and not at its middle. The amplitudes of
the magnetic highs change with width in a manner that is similar to Figure 1.
 15
Prism goes north-south
Calculated magnetic anomalies
Rectangular prism, k = 2 ppt
Be = 53,500 nT; Ie = 68o; De = 90o
Sensor height = 1 m above top
10

5
Magnetic anomaly, nT

-5
sensor line
0

Depth below sensor, m

1
-10
2
Increasing width of prism vertical
exaggeration = 2 3

-15 4
-5 0 5 10 15 20 25
East coordinate, m

Figure 31: The effect of the width of a prism that extends north-south. For this
orientation also, the amplitudes of the magnetic lows increase with the width of the
prisms. Unlike the anomalies in Figure 30, these anomalies are symmetrical about the
middles of the features. The amplitudes of these anomalies are somewhat lower than
those in Figure 30, although the parameters are otherwise the same.
 Magnetic map of a steel plate, contour interval = 500 nT

-450 N (magnetic)

-500
North coordinate, ft

-550

plate
-600

-650

-700
2500 2550 2600 2650 2700 2750
East coordinate, ft
Total magnetic field, sensor height = 8 ft (2.4 m)
Traverse N - S, measurement and line spacing = 20 ft (6.1 m)
Earth's field = 51,400 nT; inclination = 63 degrees

Figure 32: The magnetic anomaly of a large iron plate. It is 49 m wide and it has strong
anomalies (-5100 nT and +4400 nT) at its northern and southern sides. While the
beaded effect on the magnetic high must be caused by the bidirectional traverses,
locational error is not apparent in the map. This survey was done on 17 June 1989.
Figure 33: The effect of demagnetization. The lines of flux around a very magnetic
feature are drawn in red. If a similar feature, also magnetized by induction, is to be
placed at location B, the local direction of the magnetic field at B is no longer that of the
Earth, but is pointed somewhat toward feature A. That is, the field of a strongly
magnetic material tends to be aligned with the length of that material. When feature B
is added, the magnetic field from B will rotate the local magnetic field at A, which will
further affect the field at B.
Figure 34: Calculation of the area of a polygon. The cross-sectional area of a magnetic
prism, along with its magnetic susceptibility, quantify the amount of magnetic material
per unit of prism length. The area can be calculated with an equation that is often
applied by land surveyors (Davis and others 1981 p. 344; Brinker 1969 p. 248). This
equation can be used for polygons with other than four sides by noting that the
X-coordinate of a corner is multiplied by the difference in the Z-coordinates of the
corners that are clockwise and counterclockwise from the corner with the X-coordinate.
 Original map has been rotated 4.5 degree clockwise about E1075 N1030
Difference in total magnetic field: Lower sensor - upper sensor
Lower sensor height = 1.5 ft (0.46 m); upper sensor height = 6.7 ft (2.04 m)
Traverse E-W, measurement spacing = 0.5 ft (0.15 m)
Line spacing = 2 ft (0.61 m) N (mag)

Magnetic map, rotated drainfield, contour interval = 1 nT

1050

1040
Northerly coordinate, ft

1030

1020

1010

1000
1050 1060 1070 1080 1090 1100
Easterly coordinate, ft

Figure 35: The magnetic map of parallel trenches in a drainfield. The paths of the
trenches were easily mapped with a ground-penetrating radar; they are indicated with
dashed green lines. While the anomalies are irregular, it appears that the trenches
were typically located at magnetic lows. The original magnetic map has been rotated
slightly here. This survey was done on 14 December 2001.
 Magnetic anomaly perpendicular to drainfield
Sensor height = 1.5 - 6.7 ft (0.46 - 2.04 m)
Average in span Nly 1010 - 1045
4 Be = 52,300 nT; Ie = 66.1o; De = -6.5o Magnetic profiles
measurements
calculations + 2.08 nT

2
Magnetic anomaly, nT

0
surface
0

Elevation ft
-1 -2

Magnetic prisms, k = -0.7 ppt vertical -4

Depth = 0.5 - 2.5 ft (0.15 - 0.76 m) exaggeration = 2
-2 -6
1050 1060 1070 1080 1090 1100
Easterly coordinate, ft

Figure 36: The averaged magnetic anomaly of a drainfield. The measurements (red,
solid line) reveal a simple, oscillating pattern. With a very simple model (green below),
the calculated anomaly (dashed black) is very similar to the measurements. The filled
trenches of the drainfield are clearly less magnetic than the surrounding soil.
 15

Calculated magnetic anomaly

Square prisms 1 x 1 m, k = 20 ppt
Be = 53,500 nT; Ie = 68o; De = 0
10
Magnetic anomaly, nT

Depth below sensor, m

sensor line
0 0
2
4
4 prisms 6
-5

-10
-20 -15 -10 -5 0 5 10 15 20
North coordinate, m

Figure 37: The magnetic anomaly of parallel features. In this case, the features are
more magnetic than the surrounding soil, so magnetic highs are near each feature. The
general droop of the anomaly toward the north has the same origin as the droop in
Figure 30. Without understanding this effect, one may suspect that the northern prisms
are less magnetic.
 Figure A1: The program Pdyke. The program calculates the magnetic anomaly of a
prism that has the cross-section of a parallelogram (shown in red above). This program
is very simple to operate; the parameters that are listed in the panel on the right side
can be adjusted (with the arrow buttons, or by text entry) and the new anomaly is
immediately calculated and displayed on the left, with a red-line curve. This free
program is available at the web site <www.geoss.com.au> and it is also included in the
folder here called Programs.

View publication stats