LithoSI Guide 1

Guide to LithoSI

Introduction

LithoSI is an option within the Geoview program from Hampson-Russell with which you can
use inversion volumes created by pre-stack inversion to predict lithology zones. For LithoSI,
you will need:

• One or more wells, containing lithology logs.

• A set of pre-stack inversion output volumes, either 2D or 3D, which tie the wells.
These have usually been created in a previous application of pre-stack inversion in
Geoview. For details on how to create these volumes, see the document, Guide to Pre-
Stack Inversion.

The output data from LithoSI are a series of volumes containing information about the predicted
classifications, as well as probability volumes associated with each class.

This tutorial takes you through the process of applying LithoSI to a data set from Western
Canada, which exhibits an AVO anomaly. Although this example is a 2D line, the flow that we
will go through is almost identical for 3D data.

The typical workflow consists of the following steps:

(1) Start Geoview.

(2) Load wells into Geoview.
(3) Load seismic data into Geoview.
(4) Load horizons (if available) or pick horizons.
(5) Cross plot selected well log curves and define lithologic zones.
(6) Create lithology logs.
(7) Run LithoSI to determine stochastic parameters.
(8) Apply LithoSI to detect lithologic zones from seismic inversion volumes.

In this tutorial, we will go through each of these steps. Note that during the tutorial, we will not
be discussing the theory behind LithoSI. However, if you are interested in the theory behind the
program, we have included an Appendix that discusses the key concepts.

June 2011
2 LithoSI Guide

Starting Geoview

Geoview is the name of the Hampson-Russell program which contains all available options,
including LithoSI.

To start this tutorial, first start the Geoview program. On a Unix workstation, do this by going to
a command window and typing:

geoview

On a PC, start Geoview by clicking the Start button and selecting the Geoview option on the
Programs > HRS applications menu, or by clicking the Geoview icon on your desktop if one
has been set up.

When you launch Geoview, the first window that you see contains a list of any projects
previously opened in Geoview. For example, the figure below shows a single previous project,
which you can open. Your list will be blank if this is the first time you have ever run Geoview.

June 2011
 LithoSI Guide 3

For this tutorial, we will start a new project. Before doing that, you can set all the data paths to
point to the location where you have stored the tutorial data. To do that, click the Settings and
Path tabs:

Now you can see a series of default locations for the Data Directory, Project Directory, and
Database Directory. We would like to change all of these to point to the directory where the
tutorial data is stored.

To change all of the directories to the same location, click the Set all default directories button
and then click the 3-dot button to the right:

Then, in the File Selection Dialog, select the folder which contains the tutorial data (note that
yours will probably be different from what we show below. It is important to make note of where
your data is loaded on your computer):

After setting all three paths, the Geoview window will now show the selected directories

June 2011
4 LithoSI Guide

When you have finished setting all the paths, click Apply to store these paths:

Now click the Projects tab and choose the option to create a New Project:

A dialog appears, where we set the project name.

June 2011
 LithoSI Guide 5

We will call it Lithosi guide, as shown below. Enter the project name and click OK on that
dialog:

Now a dialog appears, asking you the name of the database to use for this project:

The database is used to store all the wells used in this project. By default, Geoview creates a
new database, with the same name as the project and located in the same directory. For example,
this project is called Lithosi guide.prj, so the default database would be called Lithosi guide.wdb.
That would be desirable if we were starting a new project, intending to read in well logs from
external files.

June 2011
6 LithoSI Guide

To save time in this tutorial, we have already created a database which has the wells already
loaded. To use that database, click Specify database. You will see a pop-up menu appear
below this button. Choose Open:

Then, select the database Lithosi_database, as shown, and click OK:

The previous dialog now shows the selected database and the new project name. Click OK to
accept this:

June 2011
 LithoSI Guide 7

The Geoview Window now looks like this:

Now, we will start looking at the well data.

June 2011
8 LithoSI Guide

Displaying the wells

The left part of the Geoview window (called the Project Manager) shows all the project data so
far. The tabs along the left side select the type of project data. Right now, the Well tab is
selected and we can see the two wells from the external data base. Click the “+” sign near one of
the wells to see a list of curves in that well:

To see more details about the wells, click the Data Explorer tab to the right:

The Geoview window now changes as shown:

June 2011
 LithoSI Guide 9

Click the arrow next to the first well name to get more information about the curves in that well:

Clicking this arrow will bring up a more detailed display, shown below. Note that clicking on the
arrow which points to the left will bring you back to the previous display.

June 2011
10 LithoSI Guide

Finally, to see a graphical view of the log curves within a well, find the icon for that well within
the Project Data window and double-click it:

This creates a new tab within the Geoview window, called the Wells tab, which displays the
selected well curves:

We can see that this well contains five log curves which are used in LithoSI: the P-wave velocity
log, the density log, the S-wave velocity log, as well as two log curves calculated from the first
three. These are the Vp/Vs ratio and the calculated P-Impedance. In addition, the depth-time
curve for this well has been modified using the process of log correlation. For details on those
processes, including loading logs into the Geoview database, see the tutorial AVO Modeling
Guide.

June 2011
 LithoSI Guide 11

Now double-click the second well icon, called Wet well:

The second well is very similar to the first well, except in the zone of interest, between the two
formation tops:

In fact, this well, the Wet Well, was actually calculated from the Gas Well by using Biot-
Gassmann substitution to replace the gas with brine within the reservoir zone.

June 2011
12 LithoSI Guide

Loading the Seismic Data

We have now loaded the wells which will be used in the LithoSI process. The next step is to
load the seismic volumes.

On the far left side of the Geoview window, click the Seismic tab:

The window to the right of this tab shows all seismic data loaded so far. This is empty. Go to
the bottom of the window (on the left of the workspace) and click the Import Seismic button:

On the pull-down menu, select From SEG-Y File:

June 2011
 LithoSI Guide 13

On the dialog that appears, select the two files vpvs.sgy and Zp.sgy. This can be done either by
selecting one file at a time and clicking Select both times, or simply by clicking Select All as
shown here:

Click Next at the base of the dialog:

On the next page, we are specifying that the volumes are 3D and they are to be treated as
Separate volumes.

June 2011
14 LithoSI Guide

Click Next to accept the defaults on this page:

On the third page, we tell the program several things about the volumes we are reading in. First,
we tell the program the type of data contained in each volume.

Change the Amplitude Type, as shown:

Also on the third page, we are telling the program what information it can use from the trace
headers.

June 2011
 LithoSI Guide 15

In fact, in this data set, there are no Inline and Xline numbers, or X and Y coordinates. That is
why we answer No to both questions:

After modifying these items, as shown, click Next to see the SEG-Y Format page:

By default, this page assumes that the seismic data is a SEG-Y file with all header values filled
in as per the standard SEG-Y convention. For example, it expects to find the Inline and Xline
numbers at the byte locations shown above. (In this case, we are happy with the byte locations.
If you are not sure that is true, you can click Header Editor to see what is in the trace headers.)

June 2011
16 LithoSI Guide

Since we believe that the format information is correct and is the same for both input files, click
the box Apply Format to all files first:

Then click Next to move to the final page.

Now the following warning message appears because the program is about to scan the entire
SEG-Y file:

Click Yes to begin the scanning process.

June 2011
 LithoSI Guide 17

When the scanning has finished, the Geometry Grid page appears:

Because we have not read X-Y information from the headers, the program assumes this is a
single straight line, which is correct. Click OK. After building the geometry files, a new
window appears, showing how the well is mapped into this seismic volume:

In this case, the mapping is correct, so click OK to accept the locations shown on this window.

June 2011
18 LithoSI Guide

Now the seismic data appears within the Geoview window:

Modifying the Seismic Display

The workspace currently shows the impedance volume, Zp. We would like to see the other
volume, VpVs, as well. First, open a second viewing window by clicking this icon on the lower
right:

June 2011
 LithoSI Guide 19

The Geoview window now looks like this:

Now click once to highlight the volume name, VpVs, on the Project Data list:

Holding the left mouse button down, drag the name over to the empty window on the right and
release the mouse button. The Geoview window now looks like this:

June 2011
20 LithoSI Guide

To see the display positioned at the wet well location, go to the Well icon and click the down
arrow as show below:

The drop-down menu shows the two wells in the project. Select the Wet Well and the Geoview
window shows the seismic data in the vicinity of that well location:

June 2011
 LithoSI Guide 21

Importing Horizons

The last data component required is a set of horizon picks. These are not actually used by the
LithoSI process, but they can enhance the display. You can use Geoview to pick the data
directly. Alternatively, you can import horizons which have been previously picked.

To start the import process, select Horizon > Import Horizons > From File:

June 2011
22 LithoSI Guide

From the File Selection Window, highlight the file called Lithosi_horizons.txt and click Select:

Note that, at the lower left corner of the dialog, we are specifying this to be a Free Format file.
Click Next:

The next page of the dialog allows you to specify how the file is organized. In general, this
dialog will not be filled in correctly, as each file will be different. Click the View Files button to
see the ASCII file:

June 2011
 LithoSI Guide 23

The file display shows that there are 3 horizons in the file, and that we need to skip the first
information line.

Fill in the format dialog as shown below:

When you have modified the dialog as shown above, click OK and the imported horizons will be
displayed on the seismic windows:

June 2011
24 LithoSI Guide

Creating the Lithology Logs

Now that all the data has been loaded, we are ready to start the LithoSI analysis. One of the
critical requirements for the LithoSI process is a lithology log within each of the analysis wells.
A lithology log is a special log curve, which has number-coded regions corresponding to each of
the lithology classes within the well. For example, if we wished to analyze three lithologies,
corresponding to Sand, Shale and Carbonate, the lithology log might contain the numbers 1, 2,
and 3 corresponding to the locations of each of the lithologies. Often this lithology log has been
created by a petrophysicist before starting the LithoSI project.

For the purposes of this guide, we will create the lithology logs within Geoview. To do that, we
will cross plot the two inversion attributes, using the well log data. Click on the Processes tab
to see all the processes available in Geoview:

June 2011
 LithoSI Guide 25

To create the cross plot, click the “+” sign next to the Cross Plotting item and double click
Cross plot logs:

The dialog which appears on the right contains all the parameters for creating the cross plot.
Note that this window will disappear if you click anywhere else on the workspace. If this
happens, simply click on the following Parameters button on the far right of the main window:

On the parameters dialog, we will use Multiple Wells. Then we Select All the available wells,
and set the Cross Plot Type to Vp/Vs vs P-Impedance:

June 2011
26 LithoSI Guide

Looking further down the dialog (you may need to use the scroll bar at the right of the dialog,
depending on your screen resolution), set the Window Option to use All data between targets.
Then set the Target Events to be 600 to 700 meters:

Finally, click OK to create the cross plot:

June 2011
 LithoSI Guide 27

The cross plot shows three separate lithology clusters. The one in the lower left, corresponding
to low values of Vp/Vs ratio and P-Impedance, probably corresponds to the gas sand anomaly.
We will pick that zone first.

To draw a zone, first click once on the Polygon icon:

Then move to the cross plot and draw the shape roughly as shown below, using a series of left-
mouse clicks at each of the corners of the polygon. Double-click on the last corner to finish the
polygon. When you are done, the screen should look similar to this:

Note that the polygon can be modified by grabbing the “handles” and dragging them. We can
also see the highlighted zone projected back on the original log curves, by clicking Options >
Cross Section:

This causes a new window to appear, showing the relevant log curves from both wells.

June 2011
28 LithoSI Guide

The new window looks like this:

Note that the red zone appears precisely within the target sand layer for the Gas Well only.

Back on the cross plot, name this zone Gas Sand, as shown:

Now repeat this process for the other two clusters.

June 2011
 LithoSI Guide 29

For each of the other two clusters, click once on the polygon icon. Then draw a polygon around
the cluster of points as shown below:

Note that in these cases, we are not sure about the precise lithologies, so we leave the default
names, Zone 2 and Zone 3.

June 2011
30 LithoSI Guide

Note also that the highlighted zones appear on both well logs:

Now that we have created the cross plot zones, we are ready to transform this information into
Lithology logs.

June 2011
 LithoSI Guide 31

Before doing that, send the floating cross plot window to the Cross Plots tab by clicking the
“airplane” icon on the lower right of the cross plot window:

Finally, to create the lithology logs, go back to the Processes tab and double click Create
curve of Cross Plot Zones, as shown below. Note that this option is under Log Processing >
Log Transform > Lithology, so you have to click three + signs:

June 2011
32 LithoSI Guide

On the dialog which appears, the only parameters to change are to select Multiple Wells, and
then Select All wells:

This process creates new lithology curves in each of the wells called Lithology_trans. Looking
at the Wells tab, we see that the new curve has been added to the display:

June 2011
 LithoSI Guide 33

Starting LithoSI

We are now ready to start the LithoSI analysis. To do that, go back to the Processes tab and
double-click Facies Classification (LithoSI) > New Session:

The dialog which appears is a multi-tab dialog, which will help you set up the LithoSI process.
The first page of the dialog defines which seismic volumes will be used in the process. These
seismic volumes are collected in a “Super Volume”. In this example we select both the Zp and
VpVs volumes as shown below (either click once on each volume and click Select or double-
click each volume):

June 2011
34 LithoSI Guide

When you have changed the first page as shown above, click the Wells tab to see the next page:

This page specifies which wells to use in the analysis. We want to use both wells, so click
Select All:

Then, click the Facies Classification tab:

The Facies Classification page is used to define the set of facies classes in our training data.
First we select the lithology log, which was created in each of the wells in the previous section:

This was called “lithology_trans”, so that is selected.

In the rest of the page, we define the classes, telling the program how many classes and what
their names are. We could do that explicitly, but there is a very convenient connection to the
cross plot zones, which have already been defined.

June 2011
 LithoSI Guide 35

Select the option to use an existing Cross-plot Zone Set:

The cross plot zone set is called “crossplot”, so select that on the pull down menu. Notice that
the Class elements are automatically filled in, corresponding to the cross plot zones we
previously defined.

Now click the last tab, Seismic Attributes-Well Log Mapping:

On this page, we specify how many seismic attributes are used, and how the seismic volumes are
related to the well log curves.

The default number of attributes is 2, and we will accept that:

The mapping of volumes looks like this:

June 2011
36 LithoSI Guide

We associate the seismic inversion volume, Zp, with the well log curve P-Impedance_trans.
This well log curve was created from the sonic and density logs in each well. Similarly, we
associate the seismic inversion volume VpVs with the well log curve VpVs_Ratio_trans.

When you have filled in the dialog as shown above, click OK to create the LithoSI window:

The LithoSI window contains a series of options for creating the probability distributions
associated with each class. The first thing we can do is create more space on the computer
screen by temporarily removing the Project Manager. Click the “x” as shown:

June 2011
 LithoSI Guide 37

Note that we can restore the Project Manager any time later by clicking the rotated name Project
Manager on the left side:

Notice that the LithoSI window is divided into three parts. On the left are the Common
parameters. The top part of this section is just a summary of the data to use:

The Number of attributes to use, Classification attributes and the wells and classes have
already been defined, but you can change them here if you want to. However, we will leave
these options alone. The Number of traces around wells is set to 0, which means only extract
the closest trace. If you set it to a larger number, a composite trace will be built. Again, this
option will not be adjusted in this guide. Use training-set from defaults to Well Logs, but can

June 2011
38 LithoSI Guide

be changed to either Upscaled Well Logs or Composite Traces. This will also be left alone in
this guide.
If the Export to File option is chosen, the program will create an ASCII file consisting of the
well log digits. The beginning of the file looks like this, where -99999 is a null value:

The bottom part of the left window defines how the distributions are computed and smoothed,
and looks like this:

The last option that can be entered on this part of the Common Parameters section is Cutoff on
distributions, which is defaulted to zero. We will use this default. Finally, the last part of this
section tells us where the distributions are being computed, and the default smoothing value.

June 2011
 LithoSI Guide 39

The middle section of the dialog is where most of the work will be done and consists of six tabs,
which are:

Currently the Data Selection tab is chosen. You will see that the top of the dialog looks like
this:

This part of the dialog confirms the data zone over which the calculation is to be made and
defaults to the full volume. An interval defined by horizons can also be used. Also, an external
training-set can be loaded if we decide not to use the litho-logs.

In this guide, we will just use the defaults as shown. The bottom part of the Data Selection
dialog looks like this:

June 2011
40 LithoSI Guide

This part of the dialog allows you to change the litho-log selection and the crossplot to use, and
also displays the proportions of points in each class (to be used for the prior). The proportions
can be recomputed from a subset of the wells if desired.

Under the Data Selection tab, the map on the right looks like this:

Click the Kernel Analysis tab within the LithoSI window:

June 2011
 LithoSI Guide 41

This displays the cross plot once again, color coded according to the lithology classes:

To compute the probability distributions, click the Compute button on the left (on the Common
Parameters section):

June 2011
42 LithoSI Guide

The cross plot is redrawn to show probability distribution associated with each class:

Notice that with the default smoothing parameters, there is a certain amount of separation
between the classes. Also, note that after clicking the Compute button, this button is “greyed
out”. If you want to make the button live again, and to change the smoothing applied to the
distributions, move the Smoothing slider. You may then click Compute again:

June 2011
 LithoSI Guide 43

This changes the distributions to something like the following window, depending on how far
you slid the Smoothing slider.

In this case, since we increased the smoothing parameter, there is less separation between the
classes than before. There is no “right” or “wrong” smoothing value, and each dataset will
require experimentation to find out what gives the optimum result. It is important, however, to
avoid the two extremes. Too little smoothing will result in the classes being too isolated from
each other. Too much smoothing will “smear” them together.

Experiment a bit more with the smoothing option. Once you are happy with the result, click the
QC at Wells tab (note that you can always return to the Kernel Analysis tab and change the
smoother length if you are not pleased with the QC results):

June 2011
44 LithoSI Guide

The display now shows, for each of the two wells, a comparison between the actual lithology log
at the well and the lithology log predicted by applying the statistical distributions to the inversion
volumes:

If the process had worked perfectly, the two color displays for each well would be identical.
Any color mismatch is an error or “confusion” between the classes. The mismatch for the whole
process is summarized in the “Confusion Matrix”, shown in the middle panel:

June 2011
 LithoSI Guide 45

The confusion matrix shows that, for this simple problem, the classes are very well resolved. For
example, when the actual lithology log is a Gas Sand, the classified log is also a Gas Sand 100%
of the time. One the other hand, when the actual lithology log contains a Zone 2 class, the
classified log is correct 84.44% of the time and “confused” with Zone 3 15.56% of the time. In
summary, the confusion matrix is ideal if the diagonal elements are large and the off-diagonal
elements are small.

To get a quick look at the application of the LithoSI process to the seismic volumes, click the QC
on Sections tab:

This is currently showing the Most probable facies calculation, as predicted by LithoSI. We
can also see the probability distribution associated with the Gas Sand by selecting this from the
Section menu:

These quick displays will be enhanced in the next section, when we apply the calculation to the
entire inversion volumes.

Notice that there is also a tab called QC on Maps, which you can certainly click on and look at.
However, this option is really only useful for 3D datasets. Since this is a 2D line, this option is
not very diagnostic to us here. We will next go to the final step, which is to apply LithoSI to the
whole dataset.

June 2011
46 LithoSI Guide

Applying LithoSI

Now that we have determined the probability distributions associated with each class, we would
like to classify the entire volume. To do that, click the tab Run on 3D Volume:

The page which appears allows you to specify the output volume name and the range of data to
apply the process to:

We will accept the default, which is to create volumes whose names are prefixed by the name
“lithosi”. Click Run to run the process. When the process has finished, the information dialog at
the bottom of the page lists the run-time statistics:

June 2011
 LithoSI Guide 47

It also displays two of the resulting lithology outputs on a split-screen mode:

The left panel shows the Most Probable Facies, color-coded according to the classes we defined.
As expected, the Gas Sand (red) zone appears at the Gas Well location and not at the Wet Well
location.

On the right, we see the probability associated with one of the classes – in this case, Zone 3.
This window actually contains the probability volumes for all three classes. To see the
probability corresponding to the Gas Sand, hold the mouse pointer over the probability display
and right click. From the pop-up menu, select Color Data Volume > lithosiGas_Sand_Proba,
as shown below:

June 2011
48 LithoSI Guide

The display now looks like this:

We see there is a very high probability (> 95%) of gas within the layer near the Gas Well and
very low probabilities elsewhere.

Finally, let’s modify the display to include the input VpVs volume. To do that, open a third
display window by clicking the button shown here:

The Geoview window now looks like this, with an empty third view:

June 2011
 LithoSI Guide 49

To display the VpVs volume, we first need to re-display the Project Manager, which was hidden
previously. Click the rotated Project Manager name on the left:

When the Project Manager appears, click the Project Data and Seismic tabs:

We can see the four volumes created by the LithoSI process:

June 2011
50 LithoSI Guide

We can also see original input volumes. To display the VpVs volume, click it once to select it:

Then, holding the left mouse button down, drag the name over to the empty third view and
release. The Geoview window now looks like this:

We can see that the predicted Gas Sand region corresponds to the low value of Vp/Vs Ratio, as
expected.

June 2011
 LithoSI Guide 51

Exiting the project

We have now finished the LithoSI project. Close down the entire Geoview program by clicking
File > Exit on the Geoview window:

June 2011
52 LithoSI Guide

Appendix – LithoSI Theory

A full explanation of the theory behind LithoSI is beyond the scope of this guide, and for a more
complete discussion the interested reader should refer to Chapter 4 of the book: “Seismic
Reservoir Characterization: An Earth Modelling Perspective”, by P.M. Doyen, published by the
EAGE. However, the following discussion will summarize the key concepts in LithSI.

LithoSI uses a Bayesian classification scheme in which we compute both the conditional
probability and the a priori probability for each class using the well logs. For K classes, the
Bayes rule for each class is written

K
p ( X | c i ) p (c i )
p (c i | X ) = , where p ( X ) = ∑ p ( X | ci ) p (ci ).
p( X ) i =1

and where p(c i |X) is the conditional probability that we are in class c i given a value of X . In our
example we have three classes (Gas Sand, Zone2, and Zone3) and X is the two-valued vector
consisting of measurements of Vp/Vs and Zp at a particular integrated time on a well log.

Note that p(c i |X) is what we are trying to determine, but the computations we will use to
compute this value are on the right side of the equation. That is, we take the product of p(X|c i ),
the conditional probability of X given c i , and p(c i ), the a priori probability of class c i , and divide
this by p(X), the probability of X. The reason that we can determine p(X|c i ) is that we have
defined each class earlier. Thus, the conditional probability p(X|c i ) can be found using a non-
parametric kernel density estimate of the points in each class. Although the most common kernel
density estimate is the multivariate Gaussian distribution, LithoSI uses the Epanechnikov kernel
(Doyen, Fig 4.2.9, p 112) because it decays exactly to zero, unlike the Gaussian kernel. The key
parameter in the kernel estimation is the length of the smoother, which is controlled by the
sliding bar in the Common Parameters part of the main LithoSI window.

The a priori probability (or prior) is computed for each class by dividing the number of points in
that class by the total number of points in all the classes.

The probability p(X) can be determined by summing the products of the conditional probabilities
and a priori probabilities for all classes, as shown in above equation. However, as we will see
next, this probability is a constant in each term that we will use to assign X to a class, and
therefore can be dropped.

Finally, we assign the new value X to either the class i or class j, based on the following decision
rules where, as just mentioned, the term p(X) can be factored out:

1. if p( X | ci ) p(ci ) > p ( X | c j ) p (c j ), then X is in class ci , or

2. if p ( X | c j ) p (c j ) > p ( X | ci ) p (ci ), then X is in class c j .

In fact, the above rule is used for the default option, A posteriori proportions. However,
sometimes you may feel that there are not enough data points to warrant using the prior. In this

June 2011
 LithoSI Guide 53

case, the prior is not used. This is called the Maximum Likelihood approach and is set using the
option: A posteriori likelihood. The decision rule is now:

1. if p ( X | ci ) > p ( X | c j ), then X is in class ci , or

2. if p ( X | c j ) > p ( X | ci ), then X is in class c j .

Let us finish with a brief discussion of the confusion matrix, which was used in the option: QC
at Wells. In this discussion, we will use a simple example that will serve to illustrate the concept
behind the confusion matrix.

A confusion matrix is a visualization tool used in supervised learning. The instances in a

predicted class are shown in the columns and the instances in an actual class are shown in the
rows. A confusion matrix can easily show if the system is confusing two classes and commonly
mislabeling one class as the other.

When the different classes have greatly different number of samples, the data set is unbalanced.
Then the error rate of a classifier does not represent the true performance of the classifier. For
example: If there are for example 990 samples from class A and only 10 samples from class B,
the classifier can easily be biased towards class A. If the classifier classifies all the samples as
class A, the accuracy will be 99%. This looks good but it is not a good indication of the
classifier's true performance. The classifier has a 100% recognition rate for class A but a 0%
recognition rate for class B. If you are distinguishing between disarmed bombs (A) and live
bombs (B), this could be serious.

In the example confusion matrix below we are comparing a potential stratigraphic zone in 21
wells. Of the 8 actual tight zones (non-porous sands), the system predicted that one was wet and
3 were oil-bearing, and of the six wet zones, it predicted that one was tight but none were oil-
bearing. It also mislabeled 3 oil zones as tight. We can see from the matrix that this system has
trouble distinguishing between tight zones and oil-bearing zones, but can distinguish between
wet zones and the other zones.

Example of a confusion matrix

Tight Wet Oil-bearing
Tight 4 1 3
Wet 1 5 0
Oil-bearing 3 0 4

June 2011