MapPredict Guide

Contents
Introduction to MapPredict .................................................................................................................. 1
Part 1: Depth Structure Example ........................................................................................................ 2
Reading the Seismic Data into MapPredict .................................................................................. 13
Cross Plots....................................................................................................................................... 23
Variogram Modeling ...................................................................................................................... 36
Kriging the Well Log Data............................................................................................................. 46
Cokriging the Well Log and Seismic Data ................................................................................... 53
Analyzing Output Maps ................................................................................................................. 65
Saving the Project and Exiting the Program ................................................................................. 70
Part 2: Porosity Example using Logs and Seismic .......................................................................... 74
Displaying the wells ....................................................................................................................... 75
Loading the Seismic Data .............................................................................................................. 80
Importing Horizons ........................................................................................................................ 87
Reading Well Log Data into MapPredict:..................................................................................... 91
Variogram Modeling ...................................................................................................................... 99
Kriging the Well Log Data........................................................................................................... 100
Reading a Data Slice into MapPredict ........................................................................................ 106
Cokriging with the Secondary Data ............................................................................................ 109
Reading more Data Slices into MapPredict ................................................................................ 116
EMERGE Analysis ....................................................................................................................... 122
Summary of Data Flow .................................................................................................................... 144
 1

GUIDE TO MapPredict

Introduction to MapPredict

MapPredict is an interactive program designed to help the geoscientist integrate multiple sets of
geological or geophysical measurements, using the techniques of geostatistics. An example of
this would be well log and seismic information about a parameter such as event structure or
porosity. To this end, MapPredict contains a range of geostatistical display and analysis tools,
such as cross plotting, histograms, variogram analysis, kriging, cokriging, and conditional
simulation. In this guide section, we will apply these techniques to several real data sets that
have been supplied.

This tutorial is divided into two parts and each one can be run independently. These parts are:

Part 1:
This demo applies the MapPredict program to the problem of depth conversion. Two
data sets are read in from ASCII files. The first ASCII file contains a set of measured
depths from well logs. The second ASCII file contains horizon times picked from
seismic data. The objective is to create a depth map using seismic times. Basic
geostatistical techniques including Variogram Analysis, Kriging and Cokriging will be
covered.

Part 2:
This demo creates a porosity map. In this case, the well log porosities are extracted
directly from a Geoview database. Instead of mapping the sparse well data alone,
multiple attribute slices are extracted directly from seismic volumes. Next, the Multiple
Attribute Transformation, an EMERGE process, is used to optimally combine these
seismic attributes into a single one, and then use this final one as the secondary data to
create a porosity map using Cokriging.

MapPredict October 2018

 2

Part 1: Depth Structure Example

To begin this tutorial, first start the Geoview program. If you are using a Unix-based operating
system, go to a command window and type in the command:

Geoview <RETURN>

On a PC, Geoview is initialized by clicking the Start button and selecting the Geoview option
on the list All Programs / HampsonRussell 10.x applications.

When you launch Geoview, the first window that you see contains a list of projects previously
opened in Geoview. For example, the figure below shows two previous projects. These could
be opened now if you wished. Your list will be blank if this is the first time you are running
Geoview.

For this tutorial, we will start a new project. Before doing that, it will be helpful to set all the
data paths to point to the location where you have stored the tutorial data.
Note that the directories shown may not exactly match the directories your setup will have.

To set the data paths, click the Settings tab at the top of the window:

MapPredict October 2018

 3

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 on the option Set all default
directories. Then click the button to the right:

Then, in the Set Default Directories Dialog, select the folder which contains the tutorial data
and click OK:

After setting all three paths, the Geoview window will now show the selected directories (note
that yours may be different):

MapPredict October 2018

 4

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

MapPredict October 2018

 5

Now select the Projects tab and click the create New Project button:

A dialog appears where we set the project name. Enter the project name Guide and click OK on
that dialog:

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

MapPredict October 2018

 6

The database stores 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 Guide.prj, so the default database will be called Guide.wdb. For this first part of
the tutorial, we will not be using external well logs, so we will allow the project to create the
default database. Click OK to accept the default database. The Geoview / Start window now
looks like this when this is your first project.

MapPredict October 2018

 7

The main features of the Geoview window will be described later in the second part of this
tutorial, where we use the well logs and seismic data. For this first part, we will use only ASCII
files as input to the MapPredict module. To start MapPredict, click the Processes tab and then
double-click the Map Predict option below the line “Map Predict (ISMap Geostatistical
Analysis)”:

Before the MapPredict main window appears, well data is needed to be loaded first.

In this project, we will be creating a depth structure map using both picked well depths and
seismic structure time values. We will start with well log data. Notice that you can read the well
data either from a file or from a well database. In most cases, your choice will be determined by
what data you have available. In this demo, we will use the ASCII Text File option:

MapPredict October 2018

 8

Choose the ASCII Text File option and click OK. You will now see the Read Well Data from
ASCII Text File window. Go to the data directory and select the file named
well_depth_structure.txt, as shown:

Then click Next>. You can then tell the program which columns contain the X and Y coordinates
and the data values.

MapPredict October 2018

 9

To help you to fill in the window, click the View File… button to display the contents of the file:

Notice that the data consists of X and Y locations in columns 1 and 2, and a depth structure value
in column 3. Also, there are 2 lines at the top of the file that do not contain well values. If the
well names were included as one of the columns of the file, they could also be read in. However,
this is not the case for this file. Fill in the window as shown below:

Session name = Reef

Amplitude Type = Depth
Amplitude Unit =m
Number of rows to skip =2
Well name = Number sequentially

When you have completed the window as shown, click OK. The well data will be plotted and
annotated on the screen:

MapPredict October 2018

 10

The MapPredict main window starts with a map viewing area and a table of the input data on the
right. Let’s first look at the map display area where the wells are. To allow more screen space for
the map area, we can close the Project Manager on the left. Besides, we can also temporarily
hide the Input Table and analysis widgets on the right.

The wells have been posted at their correct locations. Their depth values have been color coded
with the color key legend shown on the left of the plot. Notice that the wells are concentrated in
the center of the survey area.

The most commonly used option of the map display is to fit all the data into the available
drawing area on the computer screen. To do that, press the following “extent” icon button on the
map toolbar:

MapPredict October 2018

 11

Note that the wells have been named in the order in which they were read in. This is because
there were no well names in the ASCII file. If the names had been read, they would be displayed
as given in the ASCII file. The colour scale is telling us the values at the wells. Sometimes the
well name and value annotations are blocking the well colors. Click on the Map Display button
in the toolbar on the left of the window to turn off the well names and values:

The options under the Map Display button apply to all map windows within one Map Predict
Session. All the future new maps will have no well annotation. You will see the following
display.

MapPredict October 2018

 12

The basic statistical analysis must include the use of a histogram. In MapPredict, standard
analysis tools are made as “dock-widgets”, i.e. always docking on the right-hand-side of the
main window. The Home and the Open/Close buttons on the very right of the main window
provide easy controls for organizing, hiding or showing all these widgets together.

MapPredict October 2018

 13

In this Input window, we see three tab widgets: Input Table, Input Cross Plot and Histogram.
Next, click on the tab Histogram, and then we see the following histogram plot of the well data
values:

Because the well data is so sparse, the histogram is really not that informative. However, note
that the mean and standard deviation of these values have been computed and shown. If the
number of data points increases, the histogram can be useful. Next, we will read in seismic data.

Reading the Seismic Data into MapPredict

The next step is to read in structure times picked from a seismic survey. To do this, at the
control toolbar on the left click Add/Delete Attributes….

MapPredict October 2018

 14

This brings up the following window:

You will find that there are four ways to bring in a seismic dataset. The first one is using the
ASCII File. The other three are from existing data in the current project. The ASCII file option
will be discussed first. Please select to Add from ASCII and then click OK to get the loader:

MapPredict October 2018

 15

Select the file called seismic_time_structure.txt, as shown above, and click Next>. In the next
page where the file format will be specified, click the View file Contents… button to see the
contents of this file:

From the file listing, we can see that the Crossline (X) coordinates are in column 1, the Inline (Y)
coordinates are in column 2, and the time structure values are in column 3. The first part of the
file contains null values that are written as the number -1.00000E+21, but if you scroll down you
will see live values. Also, the header consists of 2 lines that give valuable information to us but
need to be skipped by the program.

Fill in the seismic name (Time Structure), its amplitude type (Time) and unit (ms):

Skip 2 header lines, and check the data columns XY and values:

MapPredict October 2018

 16

Next, press the OK button. While the software is reading the file, it will report the progress as
shown in the following dialog. If format specifications are all wrong, the reported values can tell.
You have an option to pause, or to stop it anytime and read the file again.

MapPredict October 2018

 17

After the file is read successfully, the final statistics will be shown below. Press the OK button if
the reading is as expected.

A processing grid is needed for proper geostatistical analysis. Since we are reading the seismic
data from an ASCII text file, there is no grid definition and hence a message dialog will appear.

MapPredict October 2018

 18

After pressing OK in the above dialog, a new grid will be estimated from the text file.

The above dialog shows the estimated grid. You have the option to change it or to copy a
different grid from a seismic volume. After OK accepting the grid dimension, an information
window will show the final loading results.

MapPredict October 2018

 19

Since a grid is used in loading the data, multiple points may fall into the same grid cell. Please
pay attention to the number of points repeated in a cell. We only take the last point going so all
the previous ones will be discarded. Close the information window.

The following display of the seismic structure map appears:

There are a number of options for modifying the plot parameters of this window. There are a
few commonly used options available on the top toolbar. The first thing we want to do is change
the map from unscrolled to scrolled mode with a fixed map scale. Fill in the window as shown
below using a map scale of 1:50000 and then press the <Enter> key.

MapPredict October 2018

 20

Now even when the window border is being re-sized, the map scale will still be fixed at 1:50000.
Please note that the map scale will be altered if you intentionally zoom or unzoom the map. If the
desired map scale is lost, any previously entered map scale can be recalled very easily from the
dropdown list:

Since a map scale of 1:50000 works pretty well for this dataset, we set it as the default scale. To
do that, please select the Tie/Untie Scale in the list of Map Display. All display options here are
session-wise options shared by all map windows.

Now we have the input seismic data. One important step is to look at a histogram of the input
data. To do this, go back to the Histogram analysis tab on the right (the dock widget area).

MapPredict October 2018

 21

Please note that the Histogram has now switched from Well Data to Seismic Data. The histogram
by default automatically follows what you are looking at on the map. It will change with your
map selection.

You may open or close the parameter and control area by press the Up-Arrow button below the
plot. The button is highlighted in the following screen capture:

Once the control area is opened, one option in the Data Selection tab is to explicitly select the
data to use for making the histogram. By default, the selection is to follow the map selection.
You can choose to select any other seismic attributes or the well data directly.

MapPredict October 2018

 22

There are more options in this control area. Here is how we add the Normal Distribution curve
in the Distribution tab.

MapPredict October 2018

 23

Cross Plots

The first thing that we always want to do when comparing two datasets is to crossplot them
against each other. To do this, click the Input Cross Plot tab on the right (in the dock-widget
area). This will bring up a display of the cross plotted values as shown below:

This display contains one point for each location on the grid that has both a well data value and a
seismic data value. Of the 25 wells, only 23 are displayed. This is because we have the default
searching distance for the crossplot as 0 units. I.e. we will only cross plot locations where both
the well and seismic exist within one seismic bin. Well 20 and Well 21 are outside the seismic
grid. The searching distance can be changed in the Cross Plot Parameters… in the Input
Table tab.

MapPredict October 2018

 24

To include the two outlying wells, fill in a longer range 500 m maximum searching radius as
shown below:

Click OK. A capturing circle with a radius of 500m will be drawn to the map. The Well 20 and
21 are now included.

MapPredict October 2018

 25

A new Cross Plot appears with all 25 well points. If a plot does not update, we can always press
the Redraw button at the lower right corner.

Below is the new cross plot for all 25 wells.

Well names can be added to the cross plot by using the right-mouse click menu. The well that is
furthest from the seismic coverage at the top right NE corner of the map is the Well 21.

MapPredict October 2018

 26

Selection of wells can also be done on the cross plot by clicking-and-draging the mouse cursor
pointer. The selected wells will be highlighted and zoomed-in at the center of on the map view.

Now go back to the Input Table page. From the Cross Plot Parameters dialog, change the
search radius back to zero.

MapPredict October 2018

 27

The two wells that are too far from the seismic survey will be turned off again automatically.
These two outliner points will be plotted in gray to indicate their current status of not being used
in the analysis. A new correlation coefficient will be re-calculated. In this case, without the two
improves the correlation from 0.89 to 0.95. The correlation coefficient value can be seen on the
Cross Plot.

MapPredict October 2018

 28

Permanently deleting a well can be done in the Options / Input Data / Edit Well Data… dialog.

In the Input Well Data Editor, we can also edit the XY locations, amplitude values, unit and
type.

Modifications in this dialog are not revertible. We don’t need to do it in this demo. Please
Cancel this editor.

The scanning of the input seismic data after changing the Cross Plot Parameters will
temporarily hide the wells which are too far away from any live seismic data. Explicit user
control on hiding a well point from the analysis can also be done manually in the Input Table
using the Hide Wells from Calculation… option as in the following screen capture:

MapPredict October 2018

 29

We do not need to temporary turn off any more wells in this exercise. Please Cancel the dialog.
We will continue to use only 23 wells. The well #20 and #21 are not checked.

MapPredict October 2018

 30

The simplest way to create a map using the seismic data is to apply a regression fit from the
crossplot with the seismic data. Click the Single Attribute tab button at the top.

After seeing the Single Attribute tab, you will notice that the map scale is tied to the previously
used one in the Input tab. The session-wise Tie Scale option is still checked on:

This map display has only the input well data. To generate a linear regression map, let’s press the
Create Single Attribute List from the toolbar menu on the left.

MapPredict October 2018

 31

This will calculate the linear regression line from the well and seismic data pairs. And then apply
the linear transformation to the input seismic data to produce the output. Since the linear
regression is to transform a time horizon to a depth horizon, we can share the same color key of
the input well depths which has been used previously.

Sharing of Color Key: Color and Amplitude Range Selections

To setup a shared color key using the input well depths, we right-click on the Color Key display
area on the left and select Modify Range…

In the dialog, change from Individual to Shared by maps and then press OK:

MapPredict October 2018

 32

To use the previously calculated color key for the output maps, right-click on the Color Key
display area on the right and select Modify Range…

In the popup dialog, change the output map color key from Individual to Shared by maps:

We will choose to use the keep the Color Key previously set by the Well Depths:

Press OK to share the same color key. The following display will be shown with two color keys
exactly the same.

MapPredict October 2018

 33

Notice that both the wells and seismic data are now colored with the same depth scale. Press the
Cross Plot tab to look at the regression line application at the well locations. We call this the
Application Cross Plot.

MapPredict October 2018

 34

The result is as expected from a linear regression method. The predicted values from seismic do
not match perfectly with the target wells. The points were scattered in the cross plot.

The linear transformation or the regression equation can be found in the History text of the
created color slice. Right-click on the map area:

MapPredict October 2018

 35

Here is the informatinon about the equation applied:

As seen, the transformation equation is using about the same one as reported by the Input Cross
Plot in the Input tab. Note that an amplitude normalization or standardization step was added in
the Single Attribute Transform to stablize the process. ( That is the reason why the two slope
numbers are slightly different: 1.565 versus 1.567. )

MapPredict October 2018

 36

Variogram Modeling

The variogram is the mathematical function that measures the spatial continuity in our two data
sets. It contains information about how well correlated the data points are and whether there are
directional continuity patterns. It is implemented in MapPredict under the Kriging/Cokriging
page. Click this Kriging/Cokriging tab at the top.

At the first time this tab is selected, a new Kriging process with the well data only will be started
automatically. The well data is known as the primary data or the hard data for Kriging. We can
create a map from the well data alone. In this example, we do plan to incorporate the seismic
data. The seismic data is known as the secondary data or the soft data for Co-Kriging. We can
go ahead and include the seismic data by pressing the Select Seismic Attribute button. We
currently have only one available seismic data called “Time_Structure”. Please select it.

The secondary data and the well data will be shown together. Please note that the secondary data
in each Co-Kriging run will be annotated as “seismic” regardless of its originally assigned input
name. Its original name, which is Time_Structure in this case, is printed at the tab label as
shown in the screen capture below.

MapPredict October 2018

 37

First we will calculate the variogram for the well data.

Click the Well-to-Well variogram tab button at the lower bottom on the right (the dock widgets
on the right). This will bring up a Variogram Analysis window. The empty Variogram plot
with only the green background indicates that there is no variogram yet. Fill in the analysis
parameters as shown below. We are calculating an isotropic variogram with 10 offsets ranging
from 0 to 1500 meters.

MapPredict October 2018

 38

Click Apply and the following variogram will appear:

On this display, the black points represent the calculated values and the red line shows the
modeled variogram. The histogram is the number of values used to calculate each black dot –
the lag. To look at the model parameters and modify them, click the Set Model Parameters
tab button.

MapPredict October 2018

 39

The default variogram model is a Spherical model with a single structure: with the three key
parameters of nugget, range and sill. The parameters have been computed automatically by the
program in such a way as to fit the measured points, and are annotated below on the figure.

The Parameters window allows you to modify these parameters and see the effect. In this case,
we will not modify anything since the variogram looks good.

MapPredict October 2018

 40

Another way of looking at the same information is to compute the covariance function. To
display this function, select View>Show Covariance. The resulting plot looks like the one on
the right:

Unlike the variogram, which is the sum of the squared differences of the values at different
offsets, the covariance is the sum of the products of the values. Thus, the covariance is large
where the variogram is small, and vice versa. After looking at the covariance, click View and set
the display back to Show Variogram.

Next let us look at the seismic variogram. To do this, select Seismic-to-Seismic. On the
Analyze Data tab, change the Maximum offset to 3500 and the Number of offsets to 20, as
shown below.

MapPredict October 2018

 41

Next, click Apply to get the following display:

Notice the excellent fit of the modeled variogram to the calculated points. Notice also that the
range is much longer than the Well-to-Well variogram.

To look at the model parameters and modify them, click the Set Model Parameters tab button,
as you did for the well variogram. The parameters are as below:

With such a large number of data points, we should test for possible anisotropy, that is, a change
in the variogram with direction. To do this, select the Analyze Data tab. Change the Number
of Directions from One to Two, make sure the maximum offset is still set at 3500, and click
Apply to recalculate a new variogram.

MapPredict October 2018

 42

The result will look like this:

MapPredict October 2018

 43

Notice that two variograms have been computed. The lower points represent the 0 degree (north-
south) direction, and the upper points represent the 90-degree (east-west) direction. The curve
represents the average modeled variogram. The interpretation of these results is that the range
and the sill being different indicate anisotropy. Another way to investigate anisotropy is to look
at the Covariance Map.

Covariance is calculated whenever you use more than one direction in the variogram analysis.
To see this plot, select View>Show Covariance Map on the Variogram Plot:

The covariance map above shows the calculated variances as a function of direction as well as
distance. Notice the bias in the northwest-southeast direction. The black circle shows the
(isotropic) modeled range. The model estimated by the software after analyzing the real data is
only isotropic. In other words, the estimated model is isotropic until you manually change it.

MapPredict October 2018

 44

The map shows an anisotropy with a principal angle of about 45 degrees clockwise from Map
North. To modify the model to reflect the anisotropic properties of this data set, we would first
rebuild the two histograms at a different principal angle i.e. the 45:

Then, after the estimation, we would select the Parameters page and change the Anisotropy
Factor and Principal Direction of the final model manually:

For now, we will leave these parameters alone and continue this guide example using the
isotropic variogram model.

We will compute one more variogram before finishing this section, the Well-to-Seismic
variogram. From the main window, select Well-to-Seismic. Fill in the Analyze Data window
as shown below using a maximum offset of 3500 m:

MapPredict October 2018

 45

Click Apply. The following variogram will then appear:

The Well-to-Seismic Variogram contains the implicit relationship between the sparse (well) data
and the dense (seismic) data. This relationship is also contained within the linear regression fit
on the Input Cross Plot display. As we will see in the following sections, either of these
measures may be used during the kriging of the well log data.

MapPredict October 2018

 46

Kriging the Well Log Data

We now want to make our first geostatistical map, a kriged display in which we optimally
contour the well log data. To do this, click the Kriging tab button at the bottom right to bring up
the following dialog page:

For this map, simply use the defaults. That is, we will do ordinary kriging with a maximum
number of samples that includes up to 20 neighboring wells and a maximum radius over the
complete map surface. Click Apply to compute the kriged map, which should look like:

MapPredict October 2018

 47

The resulting map has colors pretty much the same as those of the wells. The matching at the
wells is much better than that from a linear regression application, which is what we expect.

Besides just looking at the colors on a map, the results at the well locations can be cross plotted
with the target well values using the Cross Plot window:

MapPredict October 2018

 48

The result honors the well value. There is a small error because a Variogram of non-zero Nugget
is used. Detailed information about the point statistics can be found as in the way as described in
the following screen capture. Well 20 and 21 which are outside the seismic grid are not used in
the calculation. The number of wells in the calculation is only 23, not 25.

Another diagnostic map that is useful when analyzing geostatistical results is the Kriging Error
map. To make this plot, click the pulldown list from the bottom of the map window and select
Kriging Error.

This will produce the map shown below:

MapPredict October 2018

 49

Notice that, as expected, the error is smallest around the wells and gets larger towards the edges
of the map.

A different color key can be used for highlighting the error map. Please change to use the color
scheme “Green to Brown”.

MapPredict October 2018

 50

Finally, one more diagnostic plot can be produced, called the Cross Validation display. Cross
Validation is the process of deleting one well at a time from the kriging calculation, predicting its
value from the other wells, and displaying the error associated with that prediction. To produce
this display, in the Kriging page check ON the “Cross Validation only”, as shown below. Click
Apply to calculate. The validation process will create a cross plot showing the cross validation
result.

Again, to find out the name of the well that has highest cross validation error, right-click and
choose to Show well names:

MapPredict October 2018

 51

The validation error will also be shown on the map. When it is the first time to run cross
validation, the Map will show the Validation Error automatically:

MapPredict October 2018

 52

The choice of colors can make the error presentation more effective. Right-click on the “Depth
Error” color bar and select the Blue to Red color scheme.

And then use the Modify Range… dialog to set the ranges from -5 to 5 m, as in the following:

The map shows the relative error with both positive and negative values.

MapPredict October 2018

 53

In the map view, we can spot for any systematic spatial problem. The red and the blue dots
should be evenly distributed spatial on the map. While in the cross plot view, the points should
be nicely distributed around the perfect line in red (the dotted line with a slope of one as shown
in about three pages above.)

In this section, we have produced a series of maps using the well log data alone. In the next
section, the seismic data will be used to enhance the interpolation process.

Cokriging the Well Log and Seismic Data

Now that we have seen the results of kriging the well log data, we will move to the next stage in
our geostatistics analysis, that of cokriging the well log data using the seismic data as a
secondary variable. This will produce a map that honors the well data, but uses both the seismic
data and the variograms to interpolate between the wells.

To perform cokriging, click the Kriging tab and select Conventional Cokriging. This is the
traditional approach to cokriging and will use all three of the variograms that we have calculated.
Also, a range of seismic data around each grid point is used in the calculation (this is different
from Collocated Cokriging, which we will discuss next).

Fill in the window as shown below. Note that we do not select Cross Validation, please uncheck
it. In this test, we are using the One Constraint method.

MapPredict October 2018

 54

Press Apply.

In some other data conditions, you may see the following message, warning you that the three
variograms are inconsistent:

MapPredict October 2018

 55

This can easily happen. Therefore, we often recommend using the Markov-Bayes linear
assumption to avoid this problem. More explanations will be given later. Press Yes if you do see
this warning.

The following map will be produced. This map looks very similar to the kriged result except that
details from the seismic are more evident. Note that this is a result with three independent
variograms.

An alternative to Conventional Cokriging is to use the technique of Collocated Cokriging.

This approach differs from Conventional Cokriging in that as each grid point is calculated, and
only the single seismic data value closest to that point is used, so the speed of the process will be
much faster than for conventional cokriging. Also, we may use the linear relationship between
the well and seismic data to simplify our analysis by requiring only one variogram.

To start the process, in the same Kriging tab choose Collocated Cokriging for the Type of
Kriging. The completed window should look like the following:

MapPredict October 2018

 56

In the Assume Linear Relationship option, we will use the default Markov-Bayes method and
calculate the a and b coefficients from the pairs of nearest well and seismic. Also, we will use
the seismic-to-seismic variogram, since this is the most reliable.

Then, click Apply to produce the following map:

MapPredict October 2018

 57

Notice that this map does a better job of honoring the seismic data than the previous cokriged
result. In fact, the fit is excellent. To look at the cokriging error, click the map data pulldown
list and choose Cokriging Error from the menu.

The error range of this Cokriging is only about 4 to 6. Error from Collocated Cokriging is
expected to be smaller than that from Kriging. Change it by right-clicking on the color key
container on the right side, select Modify Range… / Edit Color Key Range dialog.

MapPredict October 2018

 58

Again, the error should always be small around the well locations. We can turn off the well
symbols to see the background color map better at the well locations:

The error looks like the following.

MapPredict October 2018

 59

As a final example of cokriging, we will explicitly incorporate the trend from the seismic data.
On the left-hand side vertical toolbar of the current Kriging/Cokriging main window, select
Trend >Remove.

Fill in the window as shown here:

As shown above, we are using a square smoother with dimensions of 1000 m by 1000 m. Click
OK to see a new window showing the input seismic residual after subtracting the trend from the
same seismic. Below is the Trend Removed seismic.

MapPredict October 2018

 60

The trend derived from the original seismic can be shown from the pulldown list at the bottom
of the map view. Please select Seismic Trend from the pulldown menu below the color slice.

MapPredict October 2018

 61

There is another option to show the trend transformed which will be applied to the well data.

MapPredict October 2018

 62

Note that the linear transformation equation is the same one we used in the Kriging/Cokriging
process.

We will now redo the variogram calculation by selecting Seismic-to-Seismic to bring up the
existing Variogram. Turn OFF the Covariance Map if it is still ON previously. Select the
Analyze Data tab to bring up the parameter window. Fill in the window as shown below here.
We will analyze up to 3500m with 20 offset bins. The number of directions is now reset to One.

Then, click Apply to get the following display:

MapPredict October 2018

 63

Notice that the variogram has a smaller range than before because we are using the trend
removed residual data.

Now redo the cokriging with the same parameters as before (Collocated cokriging with the
Markov-Bayes assumption). Make sure we continue to use the linear assumption but with the
shorter range variogram from the trend removed seismic.

MapPredict October 2018

 64

The following result will be shown after pressing Apply. In the final result before plotting, the
trend has been added back to the result of Cokriging with trend removed. It should use and share
the same color key range as before.

The previous run without the trend removal applied can be found in the list. It should be named
as “Cokriging_2” (depending on how many Apply you have pressed). Select it to compare. You
can put the mouse pointer on top of the pulldown list and use the middle wheel button to scroll
back and forth the two selections. Below was the map we previously obtained without trend
removal.

MapPredict October 2018

 65

The history of the slice will have a record of the details including the trend removal parameters
used:

Analyzing Output Maps

We have created several maps using kriging and cokriging. At the bottom a map window, there
is a list representing all the results done in this Kriging/Cokriging job.

We can switch to view different results easily. If we also want to compare results
mathematically, we can do it in the Results tab.

In the Results tab, we can view and compare all different results including those from different
operations like Multi-Attribute Transform.

To compare the differences on Cokriging due to Trend Removal, go to the Results tab:

MapPredict October 2018

 66

The Results Table on the right by default will list all the active “Live” data (in the runtime
memory) that you have previously created and displayed.

MapPredict October 2018

 67

The current results in Kriging/Cokriging are just temporary. The software stores only 5 copies
of each type. If you continue to run the operations many times, the older ones will be removed to
save disk space. The results are named sequentially, as in this example, “CoKriging_1”,
“CoKriging_2” and so on while the current one is just “CoKriging” with no suffix.

To compare the last two maps, select the CoKriging and CoKriging_2 as in the following
screen capture and then press Show Selected:

The Results tab will then allow different mathematical operations between maps including the
most commonly used difference plot (A – B).

MapPredict October 2018

 68

The above plot is the difference between Cokriging with and without seismic trend removed.

To save a Live result permanently with your own name, we can do a Save Result from the
control panel on the left in the Kriging/Cokriging tab:

MapPredict October 2018

 69

As an example, select the current CoKriging output which has the trend removed, and enter the
name to save: “CoKriging with Trend Removed”:

This data slice will then be saved permanently in the Results tab with the given name. The saved
slice can then be selected for display or for being used in other map related purposes.

MapPredict October 2018

 70

Saving the Project and Exiting the Program

This completes our look at the MapPredict program using a structure map from the wells and an
isochron map from the seismic data. To exit from MapPredict, select File>Exit on the session
main window. A session will be saved automatically at exit.

Before closing down, MapPredict will prompt you for a confirmation.

The software will save the current session automatically. Multiple sessions can be saved in one
project. A session is a set of all the current parameters and some results. Some other results, e.g.
the linear transformations, can be recalculated very quickly. In that case, we just store the
transformation equations in the project.

A saved session is typically accessible from the project Scenes tab. Click on the check box to
re-open one.

MapPredict October 2018

 71

A session is also listed in the dialog from double-clicking the Map Predict button in the
Processes tab of the Project Manager.

MapPredict October 2018

 72

Double-click on the button will start the following Select Map Predict Session dialog as long
as there is one or more existing sessions.

You have the option to open a previous session using this dialog instead of using the Scenes tab.
To start a second session, use the New… button. A project can have many MapPredict
sessions.

In this Guide, we are going to exit from this project and start a new project with a different well
database. We have prepared that well database ahead of time. Please Close this dialog and then
Exit from Geoview.

MapPredict October 2018

 73

In summary, we can easily generate the following maps of this reef example. The first three plots
honor the well values exactly. The first is an Ordinary Kriging. The middle two compares the
effect of adjusting the normalized correlation parameter in Collocated Cokriging. We promote
the Collocated Cokriging method. The last one is a straight linear regression transformation from
the seismic data for comparison.

Kriging on well data (w/ seismic variogram)

Collocated Cokriging with cc=0.75

Collocated Cokriging with cc=0.95 (default)

Linear Regression using seismic

MapPredict October 2018

 74

Part 2: Porosity Example using Logs and Seismic

This example creates a porosity map. In this case, the well log porosities are extracted directly
from a Geoview well log database. Multiple seismic attributes are directly extracted from two
seismic volumes.

Our first step is to start a new project for this second example. From the Start main tab, click on
New Project… :

Set the new project name as Porosity Guide and click OK, as shown below:

MapPredict October 2018

 75

Now a dialog appears, asking you to specify the name of the Well Log Database to be used with
this project:

For this project, we will be using an existing database, which has already been loaded with logs.
From the list of available databases, please select well_database.wdb. Then click OK.

Displaying the wells

One part of the Geoview window (called the Project Manager) shows all the project data so far.

Click the Project Data tab if it is not shown:

The tabs along the left side select the type of project data. Right now, the Well tab is selected
and we can see the 12 wells from the external data base. Click the triangular sign in front of a
well (01-17 is shown as an example), to see a list of curves in that well:

MapPredict October 2018

 76

To see more details about the wells, click the Data Explorer tab to the top-right. The Geoview
window now changes as shown:

MapPredict October 2018

 77

Click the arrow next to any of the wells (for example, well 01-17) to get more information about
the curves in that well:

The following list of log curves then appears:

To the right of the workspace, we can see a base map, showing the location of the wells by
clicking on the Map icon as circled in the following screen capture:

MapPredict October 2018

 78

To see the curves for the selected well (in our case: well 01-17), we can click on the Single Well
Display icon:

MapPredict October 2018

 79

Finally, to see the most complete view of the log curves within a well, go to the icon for that well
within the Project Data window and double-click. In this case, we will choose well 08-08:

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

MapPredict October 2018

 80

Loading the Seismic Data

We have now loaded the wells which will be used in the MapPredict 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 and click the Import Seismic button. On the pull-down menu, select
From SEG-Y File:

MapPredict October 2018

 81

On the dialog that appears, we see two seismic files in the Emerge data directory. We will load
them both. Click the Select All button and the Next at the base of the dialog.

On the next page, we are specifying two things. First the files are 3D geometry. Secondly, these
are two separate files, which happen to have the same geometry. Click Next to accept these
defaults :

On the third page, you specify more information about the data. We know that the inversion.sgy
file is a post-stack inverted Impedance data with amplitude unit in (m/s)*(g/cc). Please change
its setting as shown in the following screen capture. Click Next after.

MapPredict October 2018

 82

The inversion volume is of impedance type. Should you have forgotten to specify the amplitude
type, the software will make a detection by looking at the amplitude ranges of some traces and
prompt you for a confirmation:

On the Next, you will now see the SEG-Y Format page:

MapPredict October 2018

 83

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. If you are not sure that is true, click Header Editor to
see what is in the trace headers.

In this case, we believe the format information is correct for both files we are reading in. To
confirm that click Apply Format to all files.

Now click Next to move to the next page.

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

Click Yes to begin the scanning process. When the scanning has finished, the Geometry Grid
page appears:

MapPredict October 2018

 84

On this window, we specify the geometry of the 3D volume. Because we have read the
inline/xline numbers from the trace headers, the geometry is correct. Click OK.
After building the geometry files, a new window appears, showing how each of the wells is
mapped into this seismic volume:

MapPredict October 2018

 85

In this case, all the wells are mapped to the correct Inline / Xline locations because the X and Y
locations have been properly set within the Geoview database. If this had not been done
previously, you would type in correct values for the Inline and Xline numbers.
Click OK to accept the locations shown on this window. Now the seismic data appears within the
Geoview window:

We have read in two seismic volumes, but, by default, only one is visible. We will show the
second volume in a split-screen mode. First, turn on the second display panel by clicking this
button on the lower right:

The Geoview window should look like this:

MapPredict October 2018

 86

We would like to plot the inversion volume in the second, empty, window. To do that, select the
inversion volume from the Project Data list:

Then, holding the left mouse button down, drag and drop this name into the empty window on
the right. The Geoview window now looks like this:

MapPredict October 2018

 87

Importing Horizons

The last data component required is a set of horizon picks. You can use Geoview to pick the
data directly. Alternatively, you can import horizons which have been previously picked in other
software. Geoview can read in picks in Landmark, GeoQuest, or General ASCII format.

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

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

MapPredict October 2018

 88

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. Click the View Files
button to see the ASCII file:

The file display shows that there is a single horizon in the file, and that we need to skip the first
information line. Fill in the format dialog as shown below:

MapPredict October 2018

 89

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

MapPredict October 2018

 90

Jump to the line location at well 09-08.

MapPredict October 2018

 91

Reading Well Log Data into MapPredict:

To start MapPredict, click the Processes tab and double-click the Map Predict option as
shown:

The MapPredict main window starts off with loading well log data. Notice that you can read the
data either from a file or from the database. In this case, we will read data from the well log
database. Select the Project Well Database item as shown and click OK.

This brings up the following window:

MapPredict October 2018

 92

We start by entering a session name Porosity and then select log curve type Porosity too.

MapPredict October 2018

 93

The Select Wells area allows you to select all the wells in the database or any subset of them. In
this case, select them all by clicking the Select all >> button.

Now, click Next to bring up the Curves and Data page. We leave it as using Active logs.

Click Next will bring up the Window Selections page:

MapPredict October 2018

 94

MapPredict October 2018

 95

This window allows you to choose from any of the curves or tops that have been included in the
database. You can average these curves over zones around the tops or between tops. In this
case, let us select porosity by averaging from the top TOP to the top BASE. Fill in the window
as shown above, and click OK.

The next page that appears is for editing amplitude name and unit. We keep them as Porosity and
measured in percentage.

Click OK. The MapPredict main window will now show the porosity in a map view, as shown
here:

MapPredict October 2018

 96

We can close the Project Manager area on the left to allow the MapPredict main window to have
more screen space to display.

The dock-widget area on the right of the map view within the MapPredict main window is a
parameter dialogue and analysis QC plotting area. Currently, it is showing the porosity values at
the wells. This is called the Input Table.

MapPredict October 2018

 97

Before doing any geostatistical analysis, we need to define a processing grid. Press on the
Define Grid… button on the left-hand-side toolbar of the current Input tab.

We know that we will use the inversion and seismic attributes later. It is safe to define our
processing grid as that of the seismic. Select the option to set the grid using the seismic, as
below:

Press OK when we confirm the Grid as followed:

MapPredict October 2018

 98

Now, we wish to create a kriged map with the well data and this grid. This involves first
computing a variogram model, and then kriging the data. All these geostatistical analyses are
performed in the Kriging/Cokriging tab. Click on the tab header as shown here:

We are now starting a Kriging or Cokriging job with well data only. When the MapPredict
session has only the well data like in this case, the Kriging/Cokriging run will be automatically
named as Wells Only until you add a seismic attribute:

As you can see, the variogram area is empty and is waiting to be created.

MapPredict October 2018

 99

Variogram Modeling

A variogram is a mathematical function that measures the spatial continuity in the dataset. It
contains information about how well correlated the data points are and whether there are
directional continuity patterns.

We will calculate the variogram for the well data. In the Well-to-Well window, click on the
Analyze Data dialog. Fill in the dialog as shown below (42 units of offset in 6 divisions)

By using the parameters shown above, we are choosing to calculate an isotropic variogram with
6 offsets ranging from 0 to 42 meters. Click Apply and the variogram will appear. The final
display will look like this.

MapPredict October 2018

 100

On this display, the black point represents the average value of all the well pairs within the offset
division which is also called a lag. The vertical histogram bar shows the number of pairs in a
division. The red line shows the modeled Variogram fitting all the points.

To look at the model parameters and modify them, click the Parameters tab. This will display
the following dialog:

The default variogram model is a Spherical model with a single structure. The parameters have
been set automatically by the program in such a way as to fit the measured points. This window
allows you to modify these parameters and see the effect. In this case, we will not modify
anything since the variogram looks good.

Next, we will produce a kriged map of the porosities using this variogram model.

Kriging the Well Log Data

We now want to make our first geostatistical map, a kriged display in which we optimally
contour the well log data. To do this, click the Kriging button at the lower right of the
Kriging/Cokriging tab to bring up the following parameter window:

MapPredict October 2018

 101

For creating our first map, simply use the defaults. That is, we will do Ordinary Kriging with a
maximum number of samples that includes up to 12 neighboring wells and a maximum radius of
142m. The default maximum radius is the diagonal distance of the grid and will cover the
complete map surface.

Thus, click OK to compute the kriged map. This will produce the following display:

MapPredict October 2018

 102

There are several interesting diagnostics that can be displayed to see how well the kriging has
worked. From the first pulldown list of map data at the bottom of the color map viewing area,
select Kriging Error.

This will produce the following plot of the error in the kriged map:

As expected, the absolute error is smallest around the wells. Let’s use the Green-to-Brown
color scheme for the absolute error:

MapPredict October 2018

 103

Finally, click the Kriging tab again, and check the Cross Validation button at the bottom of the
Kriging parameters.

MapPredict October 2018

 104

Click Apply to produce the cross-validation plot. The cross-validation plot shows the effect of
removing the wells one by one in the kriging process, and then comparing the known value of
the removed well to the new kriged map. This plot shows the true or target well value versus the
estimated or predicted value for each of the wells.

To show the validation results on a map format, do the following. First turn off the color slice
map and turn on the well data point to Validation Error. If it is the first time you run Validation,
the plot will be setup automatically for you:

MapPredict October 2018

 105

Modify the color scheme for “Porosity Error” using the right-click button on the key area on
the left. A color scheme of Blue to Red can show error distribution better.

And then press the Default Scan for a better amplitude range.

MapPredict October 2018

 106

Some users prefer to center the amplitude range such that zero error will be shown as white. We
will see the following validation error in a map next to the cross plot:

Reading a Data Slice into MapPredict

It is known that impedance and porosity are related. The next step is to bring in this inversion
result into our MapPredict session and use it as a secondary variable for the porosity estimation.

Go the Input main tab. Go to Add Attributes… :

MapPredict October 2018

 107

We will first bring in the inverted impedance slice. Select Add by slicing data volume in
project and click OK.

First select the inverted impedance volume inversion.

Then we are going to slice the average impedance amplitudes from a 10 ms window below the
channel top picks. It should capture the target porous sand layer from seismic inversion.

MapPredict October 2018

 108

Leave the name the same as the volume and accept the full trace range. Click OK at the end.

MapPredict October 2018

 109

This is what the input slice looks like. The Input Table on the right now lists the inversion
impedances extracted from the slice at the well locations.

Cokriging with the Secondary Data

Let us now perform a conventional geostatistical analysis. Recall that we had already computed
the well-to-well variogram and the kriged map. We now also have loaded the inverted
impedance as the secondary data.

Click on the Kriging/Cokriging tab where we had the Kriged map with Wells Only.

We can add a secondary attribute using the Select Seismic Attribute button on the left. See
below screen capture:

MapPredict October 2018

 110

After that, the name of the tab will be updated to reflect your selection:

The “inversion” impedance is now added for analysis. Previously, your map display could be
showing the Porosity Validation Error from Kriging. Please now set up the map display as
followed showing the well Porosity from the well and the Impedance from Seismic.

MapPredict October 2018

 111

Note that the secondary data used in Kriging/Cokriging is also known and labeled as the
“Seismic”. The “Seismic” in this case is the input named “Inversion”. Now, we need the
seismic-to-seismic variogram. Click Seismic-to-Seismic, enter 80 meters as the max offset,
and change to analyze 25 offset bins, as in the following way:

Click Apply and the following variogram appears:

It is a good habit to always check the variogram model parameters before using it. Turn to the
Parameters page:

MapPredict October 2018

 112

This is a seismic-to-seismic variogram model. We will not need a well-to-seismic variogram,

since we will be using the Markov-Bayes approach, which uses a scaled version of the seismic-
to-seismic variogram for the cokriged and KED (Kriging with External Drift) maps. To produce
the cokriged map, click the Kriging tab button and fill out the window as shown below:

MapPredict October 2018

 113

Make sure the Cross Validation only option is turned off. Click Apply to get the cokriged
result, shown next.

Notice that the higher porosity channel is defined by this map. To see the error associated with
this map, select Cokriging Error. The following map will appear:

MapPredict October 2018

 114

Now, click the Kriging button again and fill out the window as shown below to produce the
KED result:

Click Apply to produce the next map shown.

MapPredict October 2018

 115

To see the error associated with this map, select External Drift KED Error. The following map
will appear:

MapPredict October 2018

 116

Reading more Data Slices into MapPredict

We have 2 SEG-Y volumes. One is an inverted impedance volume that we loaded previously.
The other is a raw seismic stack which we can also make use of. One way is to generate multiple
seismic attributes and then bring them all into our MapPredict session. Use the Emerge
technologies within MapPredict to create different estimations of porosity from all these seismic
attributes together with the seismic inversion. The best seismically estimated porosity map can be
used for the co-kriging process.

We will bring in some seismic attributes. We don’t really know which attribute is better;
therefore, we bring in a few possible ones to select.

Again, all data loading are to be done in the Input tab. Press Add/Delete Attributes to add.

In this example, we will generate some commonly used and basic transformations directly from
the input seismic volume without clustering up the project and taking up extra disk spaces. These
seismic transformations are also known as the seismic internal attributes in our Emerge
technique.

Often, we will map some other advanced seismic attributes, like, the Energy Ratio Attribute. If
you are interested in all different volume attributes that HampsonRussell has, please see our
Processes / Volume Attributes for details. In that case, you will create those advanced attribute
data volumes before running MapPredict.

Regardless how you obtain the attributes, they can be prepared in a project as data slices ahead of
time. Then, we will simply use the “Add from existing Data Slice in project”.

MapPredict October 2018

 117

In this demo, press OK to Add by slicing data volume directly.

Select the input “seismic” SEG-Y volume in the following dialog:

We are going to generate some seismic attributes and create data slices for 10ms below the target
channel_top_horizon:

MapPredict October 2018

 118

One of the options is to apply a series of transforms to the seismic data before extracting the data
slice. Select the Transformation options shown below. Then set the Averaging Option to RMS
Average. This means that we have to explicitly remove Arithmetic Mean from the list. (This
also means that the averaging done for the inversion slice was defaulted to arithmetic mean.)

MapPredict October 2018

 119

Click OK.

The following maps have all been included in the session:

• Raw amplitude: An RMS average over a 10 ms window below the ch_top event of the
seismic amplitudes from the seismic volume, covering the entire survey.
• Amplitude envelope: An RMS average over a 10 ms window below the ch_top event of
the amplitude envelope of the seismic amplitudes from the seismic volume, covering the
entire survey.
• Instantaneous phase: An RMS average over a 10 ms window below the ch_top event of
the instantaneous phase of the seismic amplitudes from the seismic volume, covering the
entire survey.

MapPredict October 2018

 120

• Instantaneous frequency: An RMS average over a 10 ms window below the ch_top

event of the instantaneous frequency of the seismic amplitudes from the seismic volume,
covering the entire survey.
• Integrated: An RMS average over a 10 ms window below the ch_top event of the
integrated seismic amplitudes from the seismic volume, covering the entire survey.
• Inversion: An Arithmetic Mean average over a 10 ms window below the ch_top event of
the inversion volume, covering the entire survey.

The slices will be generated and then added to the MapPredict session. They will be listed in the
map view of the Input window. In the map view, they are overlaid on top of each other. The
pulldown menu allows you to switch the selected one on top:

They are also listed in the Input Table window on the right. If it is a wide table, drag the window
out of the docking area as a floating one to see all the columns.

MapPredict October 2018

 121

The Regression Summary from the Input Table gives a list of their correlations with the well
data.

Click on the Regression Summary button will give you the following window.

The Regression Summary window tells you that the averaged porosity values from the wells
correlate with the inversion data slice with a correlation coefficient of 0.68, which is very good.
Possible good ones are the amplitude envelope at 0.45 and the raw amplitude at 0.42. The other
correlations are not as good. The table can be sorted at increasing or decreasing order by
clicking at the column header. Close this window.

Next, we will use the EMERGE option to pick the right attributes for us and to produce a better
starting map for geostatistical analysis.

MapPredict October 2018

 122

EMERGE Analysis

So far, we have been using a single secondary data set to enhance the mapping of the primary
data set. We have used the Inversion data slice to aid in the prediction of porosity. We now
have extracted quite a number of other seismic attributes. Theoretically, we could use each of
those separately to make independent predictions of porosity. However, we can expect that
using multiple attributes simultaneously should be better than using any of them individually.
This is the process known as EMERGE.

Within MapPredict, this EMERGE technology is separated into three major tabs: Single
Attribute, Multiple Attribute and Neural Network so that their analyses can be done
independently and organized in a logical flow by the tabs from left to right.

Let’s go back and start from the Input tab first.

We have looked at the Regression Summary for all attributes. The values being used in
calculating the correlations with well data depends on the Cross Plot Parameters.

MapPredict October 2018

 123

Click on this Cross Plot Parameters… button to bring up the following window:

This window determines how many samples around each well location will be used in the
analysis. Currently there are two capturing options:

Same areal size for all wells:

The default assumes that all well data points will use the same searching method and same areal
size for calculating the average values from the corresponding seismic. This is well independent
or, in other words, all wells will be handled the same way.

Different size according to the input specification at each well:

This is the option to allow different areal size in capturing the seismic data. This option is mainly
for handling horizontal wells and for production analysis. Each well location can have a different
capturing length along the well bore and different depth into the pay zone. This option will not
be discussed here.

The default Max Search Radius is 0, meaning that the nearest data slice sample is used with
each well sample for the analysis. If you entered a different one, say a 5 m radius over a circle,
then all data slice samples within 5 meters would be averaged to provide the data slice sample.

MapPredict October 2018

 124

This is a way of enhancing the signal to noise ratio, and avoiding spurious noisy samples just at
the well location. For this analysis, we will accept the default to the Nearest Bin and no
searching distance 0, i.e. the only sample within a bin only. Click Cancel.

Again, the Input Table shows each of the well locations, the corresponding Porosity value from
the well, and the corresponding seismic value from each of the attributes. If you scroll the
horizontal slider, you can see all the attributes in the analysis.

MapPredict October 2018

 125

The first step in the analysis is to create the Single Attribute List. First, we switch to the tab
Single Attribute. The correlation table is empty until we start to run the process.

Then press the Create Single Attribute List button as shown here:

We will analysis all the attributes in the project. When the calculation has completed, the result
window looks like this:

MapPredict October 2018

 126

This is a plot after applying a linear transformation. The output amplitude type is Porosity as
seen on the right hand side color key.

The Correlation Table shows each of the attributes correlated with the target, Porosity, ordered
in decreasing correlation. This display shows that “inversion” correlates best with a correlation
coefficient of (-0.68). We can also see a cross plot for this attribute as well as the predicted map
which would come from using the regression line to predict Porosity.

MapPredict October 2018

 127

The layout of the Cross Plot, Histogram and Correlation Table can be rearranged or flown-out
as wish. If you want the Cross Plot window to be shown above the Correlation Table, drag it out
by grabbing the top of the window border, and then drop it to the top docking area. The
droppable area will be highlighted, currently in light blue, but its color may be different for
different user.

The following is a typical layout showing table and cross plot together with the map.

By clicking another row on the attribute list (in the Single Attribute Correlation Table), we can
see both the cross plot and the prediction map change. For example, the above is the cross plot
and predicted map when selecting the row of seismic envelope attribute.

MapPredict October 2018

 128

However, this still does not show us how to use the attribute simultaneously to improve the
prediction. The Multiple Attribute option will do the job. To do that, go to Multiple Attribute
tab and click on the Create Multi-Attribute List button.

The default analysis parameters, as shown below, were used. They are listed in the lower part of
the Multi-Attribute Correlation Table window:

This is telling the program to analyze all the attributes. In this process it looks for combination
of 2 attributes, 3 attributes, 4 attributes and then 5 attributes. In each case it finds the best
combination by Exhaustive Search. That means it examines all possible combinations of each
type and stores the one with the lowest prediction error.

MapPredict October 2018

 129

The layout for this Correlation Table for the Multi-Attribute analysis looks similar to the layout
of the previous Correlation Table for the Single Attribute analysis. The top panel shows a list of
attribute names:

Different from that in the Single Attribute analysis, each line on this list is however a
combination of attributes. The number of attributes in each combination is the number on the
left. To see exactly, which attributes correspond to each line, click on a row (the second row in
the following screen capture) and then select Show transformation equation:

This causes an equation to appear, showing the exact combination for each line. This means that
the best possible combination of 2 attributes are the ones shown on the list. If you use those
exact weights as multipliers for the attributes, and add the constant shown, you will get the best
possible prediction of porosity. The Training Error, printed on the first list is the RMS average
error between the actual Porosity and that predicted by this combination. Note that the result is
~2.06 in the target porosity measurement unit Percentage.

The cross plot between the actual porosity and that predicted by this combination can be seen at
the Cross Plot tab. For example again, if the second row is selected, the cross plot looks like
this:

MapPredict October 2018

 130

Note that some cross plot display details can be viewed or modified by pressing the wide up-
arrow button at the bottom of the plot. As an example, we can show the statistics of those cross
plot points.

MapPredict October 2018

 131

There is an option to show a summary plot of the prediction errors for each attribute
combination:

From the Plot Error versus Attribute Number button, we get the following:

The black curve shows the prediction error when all the wells are used in the analysis. The red
curve shows the Validation Error. This is the average error when each well is left out
systematically from the analysis and predicted using only the other wells. For most cases, this
red curve goes through a minimum and then rises up, as shown above.

We would conclude from this analysis that precisely two attributes are optimum for predicting
Porosity. This is because the red curve goes through its minimum at two attributes. If we use

MapPredict October 2018

 132

three attributes, the average prediction error on a hidden well is actually larger than using two
attributes. It indicates that using three attributes are not helpful.

In summary, we use this plot to determine exactly which combination of attributes is ideal. We
do this by observing at which number of attributes the red validation curve goes through a
minimum. Two attributes have the minimum error.

Except the first row, each row is a combination of multiple attributes. The map shows the result
of applying any particular derived attribute combination. If you select the second row, the plot
looks like this:

The above is the Prediction Result from two attributes. A second useful plot is the Prediction
Interval plot. To see that plot, click on Prediction Interval checkbox, and then press the Apply
button:

MapPredict October 2018

 133

This shows the range of porosity values for which we are 95% confident. The output plot looks
like this:

Most of this plot is light blue and green, indicating a Prediction Interval of 12% porosity. This
means that in these regions, the porosity estimate is:

(Prediction Result) +/- 6% Porosity

A lower number on the Prediction Interval plot means a higher confidence in the result and vice
versa.

When the multiple attribute list is recalculated with different parameters, all the corresponding
Prediction Interval maps will be recreated automatically.

Note that if we wished to see the Prediction Interval corresponding to a different confidence
level, we would run Prediction Interval with a different parameter. If we test a new 80%, we
will have the following available maps after another Apply:

MapPredict October 2018

 134

When we are asking for a lower confidence level, the prediction interval can certainly go lower.
Now the color map is around the orange which is 8%, the porosity estimate is then:

(Prediction Result) +/- 4% Porosity

There is a button to clear them all at once:

CoKriging with an EMERGE Result

The EMERGE Multi-Attribute analysis has told us that using two attribute slices simultaneously
is the best predictor of the Porosity values.

The final step in this process is to pass back the predicted Porosity map from EMERGE and use
it as a secondary data set for cokriging in MapPredict.

To do this, select Save Result:

MapPredict October 2018

 135

Select our best predicted map “MA2_seismic_freq_rms_average” that we want to save and enter
its name “Porosity using 2 attributes”:

Click OK on this window. Then this slice will be saved permanently in the Results tab. The
new item will be listed under the “ResultWindow” section as shown:

MapPredict October 2018

 136

In order to use this newly predicted map for co-Kriging, go back to the Kriging/Cokriging tab,
from the blue [+] icon button at the top-left corner select to create a new run, as shown below:

A new sub-tab is now created within the Kriging/Cokriging main tab. The new sub-tab will have
the data seismic name used.

MapPredict October 2018

 137

The map window will look like this with the Emerged result as “seismic” - the secondary data.

MapPredict October 2018

 138

Our final step will be to redo the geostatistical options with the new secondary dataset. First,
select Seismic-to-Seismic Variogram and fill out the window as follows: 80 meters of
Maximum Offset and 25 Number of Offsets:

Click Apply to get the following variogram:

Note that the nugget is quite large on this variogram, so let’s change from a spherical to an
exponential fit.

MapPredict October 2018

 139

To do this, click on Parameters and change the Type to Exponential, as shown below:

Click on Ok and the new variogram looks like this:

This variogram does not fit the points well because it has not been optimized. Click on Analyze
Data to get the following window:

MapPredict October 2018

 140

Click on Apply to get the final optimized variogram:

We decide to honor the well locations. In the Variogram Parameters page, please enter a Nugget
value 0 and press Apply.

MapPredict October 2018

 141

Don’t forget the Apply button.

Next, click the Kriging button and fill out the window as follows:

MapPredict October 2018

 142

Note that we will perform collocated cokriging with the Markov-Bayes assumption (when the
Assume Linear Relationship button is selected). Click Apply to get the final cokriged result:

This represents our best Porosity prediction, using the well data as well as two seismic attributes.

Since we have chosen to honor the well locations, the application error at any a well is zero. The
application Cross Plot can show this fact.

MapPredict October 2018

 143

The MapPredict session including multiple Kriging/Cokriging jobs will be saved automatically
at exit. Close the Geoview program by selecting File>Exit.

After the session is re-opened or the project is reloaded the next time, only one of multiple
Kriging/Cokriging runs will be re-opened automatically. If you have more than one
Kriging/Cokriging run, all the others can be accessed from the blue corner icon button [+] the
“Open an existing run” option as shown below:

This completes the part 2 of this Guide.

MapPredict October 2018

 144

Summary of Data Flow

As a review, the data flow of making a map combining the latest EMERGE and the classical
geostatistical techniques can be depicted in the following diagram:

In Part 1 of this Guide we have gone through various Geostatistical techniques. In Part 2, the
EMERGE Multilinear Regression techniques were explained. Different data loading methods
were also shown so that you are ready to start your own projects anytime.

This completes our MapPredict guide. Thank you.

MapPredict October 2018