Simulation by Indication
This window presents the
Simulation by Indication which aims at obtaining a regular grid of values
acquired from point sample data. This grid is generated from a set of
simulated grids following the procedure of sequential simulation described by
Deutsch e Journel, 1992.
This module was
implemented in SPRING based on the indicator kriging using the "sisim"
code from GSLIB - Geostatistical Software Library (Deutsch and Journel, 1992). This module relies on measures of spatial variability according to
regular grids, of spatial attributes, continuous or categorical. Along with
an attributes map, an uncertainty map is generated with an estimation of a
regular grid representation.
Executing Simulation by Indication :
-
Select on the "Control Panel" a numeric IL that contains point samples;
-
On the SPRING main menu choose
Analysis - Geostatistics - Simulation by Indication...; Two windows will be presented:
Model Parameters and Kriging by Indication. The window Model Parameters
can also be activated by pressing the button Models/Probability...
inside the window Kriging by Indication. Follow the
instructions described at
Models/Probability...to fill out the fields;
-
On top of the window
"Simulation by Indication"
you will see the variable correspondent to the active IL;
-
the structural parameters can be changed at any time by clicking
the button Models/Probability...;
-
define the attribute nature by choosing
Continuous or Categorical on the field Variable;
-
choose between
Ordinary
or Simple at the field Kriging Type;
-
Define an option for Kriging by Indication:
Full IK or Median.
If the Option isMedian, type the median value at the
field Threshold;
-
Press the button
Prior Prob. ... to define a set of prior
probabilities valid for the points of the grid to be created. This option is not
implemented yet;
-
the button
Indirect Data... must be activated in case you
want to choose a file with auxiliary data of attributes that are
correlated with the attribute being studied. These indirect data must be
in a file with Geo-EAS ( Deutsch e Journel, 1992 ) format, X and Y coordinates on the columns 1 and 2
respectively, and indirect attribute values already transformed by
indication, placed on the column 3.
-
Press the
Bounding Box... button in case you want to define a new area for the grid to be
generated. The
default value is the active Information Layer;
-
The fields X Res. and
Y Res. contain predefined values for the active IL. Change these
values in case you want to use different resolution. Notice that the
values defined for the bounding box and resolution determine the size of
the grid and can not be greater than 1000 rows or 1000 columns;
-
The parameters of Maximum
Samples in the Searching Area must be defined by filling out the fields:
Original which
determines the maximum number of original samples of the attribute, Prev. Nodes which defines
the maximum number of previously simulated nodes of the grid and Soft Nodes which
determines
the maximum number of indirect nodes inside the searching area. By
default, these fields have predefined values of 8, 4 and 4, but the user
can change them if he wishes.
-
Next, define the radius and
orientation for the Searching Ellipsoid. The fields R.min,
R.max and Angle have predefined values for the Isotropic case:
R.min
and R.max (in meters) are equivalent to the reach of the
isotropic variogram and Angle is equal to zero. If there is anisotropy,
the parameters must be adjusted accordingly (Deutsch e Journel, 1992);
-
The Simulation Parameters
must be defined in a sequence. Fill out the field
N. Realiz. which corresponds
to the number of times the simulation was executed. Define the minimum (Min. Value) and maximum (Max. Value) values for the
attribute, to be used by the probability distribution function. Provide
an initial value for the simulation at the Seed: field. The type of
search (simple, multiple 2, multiple 4, multiple 6, multiple 8 or
multiple 10) determines the number of previously simulated grids (1, 2,
4, 6, 8 or 10 respectively). Finally you should decide if you want or
not to make a previous association of the sample values to the nodes, by
marking or not the box of Data
in Nodes. This option helps to optimize the original
data search decreasing the simulation total time.
-
As
outputs of Kriging by
Indication we have two Information Layers with regular grid
representations, one with attribute
values and another with estimation
uncertainties. To define these ILs it is necessary to choose
a Category,
type in a name for the IL, IL values (different from the active IL),
choose the type of value
and uncertainty.
For continuous attributes you should choose for Value mean or median and for
Uncertainty
confidence intervals based on standard deviations or quartiles. For
categorical attributes, the distribution mode is chosen for Value and the maximum
probability or Shannon entropy for Uncertainty . The uncertainty Information Layer
has a _Inc suffix added to the IL values field.
[
Procedures] [
Concepts] [
Exploratory Analysis ] [ Semivariogram
Analysis ]
[
Semivariogram Modeling] [
Model Validation ] [
Kriging]
See also:
SPRING - Spatial Analysis
Theoretical Models for adjusting experimental variogram
|