GSLIB Help Page: SISIM
The sisim program is for the simulation of either integer-coded
categorical variables or continous variables with indicator data defined
from a cdf.
The following set of parameters are required for each of the
vartype: the variable type (1=continuous, 0=categorical)
ncat: the number of thresholds or categories
cat: the threshold values or category codes (there should be
ncat values on this line of input)
pdf: the global cdf or pdf values (there should be ncat
values on this line of input)
datafl: the input data in a simplified Geo-EAS file.
icolx, icoly, icolz, and icolvr: the column numbers for
the x,y, and z coordinates and the variable to be
simulated. One or two of the coordinate column numbers can be set to
zero which indicates that the simulation is 2-D or 1-D.
directik: already transformed indicator values are read from
this file. Missing values are identified as less than tmin
which would correspond to a constraint interval. Otherwise, the cdf
data should steadily increase from 0 to 1 and soft categorical
probabilities must be between 0 to 1 and sum to 1.0.
icolx, icoly, icolz, and icoli: the columns for the
x, y, and z coordinates, and the indicator variables.
imbsim: set to 1 if considering Markov-Bayes option for cokriging
with soft indicator data, otherwise, set to 0.
b(z): if imbsim is set to 1, then the B(z)
calibration values are needed.
tmin and tmax: all values strictly less than tmin
and strictly greater than tmax are ignored.
zmin and zmax: minimum and maximum attribute values when
considering a continuous variable
ltail and ltpar specify the extrapolation in the lower
tail: ltail=1 implements linear interpolation to the lower limit
z_min ltail=2 power model interpolation, with
w=ltpar to the lower limit zmin and
implements linear interpolation between tabulated quantiles
(only for continuous variables).
middle and midpar specify the interpolation within the
middle of the distribution: middle=1 implements linear
interpolation; middle=2 implements power model
interpolation, with w=midpar
and middle=3 allows for linear interpolation between tabulated
quantile values (only for continuous variables).
utail and utpar specify the extrapolation in the upper
tail of the distribution: utail=1 implements linear interpolation
to the upper limit zmax, utail=2 implements power model
interpolation, with w=utpar, to the upper limit
zmax utail=3 implements linear interpolation between
tabulated quantiles, and utail=4 implements hyperbolic model
extrapolation with w=utpar The hyperbolic tail
extrapolation is limited by zmax (only for continuous variables).
tabfl: If linear interpolation between tabulated values is the
option selected for any of the three regions then this simplified
Geo-EAS format file is opened to read in the values.
One legitimate choice is exactly the same file as the
conditioning data, i.e., datafl Note that tabfl
specifies the tabulated values for all classes.
icolvrt and icolwtt: the column numbers for the values
and declustering weights in tabfl Note that declustering
weights can be used but are not required - just set the column number
less than or equal to zero.
If declustering weights are not used, then the class probability
is split equally between the sub-classes defined by the tabulated
idbg: an integer debugging level between 0 and 3. The larger the
debugging level the more information written out.
dbgfl: the file for the debugging output.
outfl: the output grid is written to this file. The output
file will contain the results, cycling fastest on x then y
z then simulation by simulation.
nsim: the number of simulations to generate.
nx, xmn, xsiz: definition of the grid system (x axis).
ny, ymn, ysiz: definition of the grid system (y axis).
nz, zmn, zsiz: definition of the grid system (z axis).
seed: random number seed (a large odd integer).
ndmax: the maximum number of original data that will be used to
simulate a grid node.
ncnode: the maximum number of previously simulated nodes to use
for the simulation of another node.
maxsec: the maximum number of soft data (at node locations) that
will be used for the simulation of a node. This is particularly useful
to restrict the number of soft data when an exhaustive secondary
variable informs all grid nodes.
sstrat: if set to 0, the data and previously simulated grid nodes
are searched separately: the data are searched with a super block
search and the previously simulated nodes are searched with a spiral
search. If set to 1, the data are relocated to grid
nodes and a spiral search is used; the parameters ndmin and
ndmax are not considered.
multgrid: a multiple grid simulation will be performed if this is
set to 1 (otherwise a standard spiral search will be considered).
nmult: the target number of multiple grid refinements to consider
(used only if multgrid is set to 1).
noct: the number of original data to use per octant. If this
parameter is set less than or equal to 0, then it is not used;
otherwise, the closest noct data in each octant are retained
for the simulation of a grid node.
radius_hmax, radius_hmin and radius_vert:
the search radii in the maximum horizontal direction, minimum
horizontal direction, and vertical direction (see angles below).
sang1, sang2 and sang3: the angle parameters that
describe the orientation of the search ellipsoid. See the discussion
mik and mikcat: if mik is set to 0, then a full
indicator kriging is performed at each grid node location to
establish the conditional distribution. If mik is set to 1,
then the median approximation is used, i.e., a single
variogram is used for all categories; therefore, only one kriging system
needs to be solved and the computer time is significantly reduced.
The variogram corresponding to category mikcat will be used.
ktype: the kriging type (0 = simple kriging, 1 = ordinary
kriging) used throughout the loop over all nodes.
SK is required by theory, only in cases where
the number of original data found in the neighborhood is large enough
can OK be used without the risk of spreading data values beyond their
range of influence. The global pdf values (specified
with each category) are used for simple kriging.
nst, and c0: the number of semivariogram structures and
the isotropic nugget constant.
For each of the nst nested structures one must define it
the type of structure; cc the c parameter;
ang1,ang2,ang3 the angles defining the geometric anisotropy;
aa_hmax, the maximum horizontal range; aa_hmin,
the minimum horizontal range; and aa_vert, the
vertical range. Each semivariogram model refers to the corresponding
indicator transform. A Gaussian variogram with a small
nugget constant is not a legitimate variogram model for a discontinuous
indicator function. There is no need to standardize the parameters
to a sill of one since only the relative shape affects the kriging