This is an MVS module only.
General Module Function
This works on an external grid, preserves nodal data, and creates cell data. The input cell sets are not preserved, as it will create two cell sets per material, allowing you to remove specific materials with select_cells or Explode_and_Scale. Also, there are some known, albeit rare, stability issues inside the adaptive gridding loops.
adaptive_indicator_krig is an alternative geologic modeling concept that uses the indicator method for assigning each node's material in a grid based on data in a pregeology (.pgf) file. It is an extension of the technology in Indicator_Geology for several reasons:
Material assignments are done on a nodal versus cell basis providing additional inherent resolution
Gridding is handled by outside modules. This allows for assigning material data based on a PGF file after kriging chemistry or other parameter data with Krig_3D.
Though it does not provide material boundaries that are as smooth as Krig_3D_Geology or Spline_Geology, it does provide much smoother interfaces than Indicator_Geology's Lego-like material structures.
There are two fundamental differences between Indicator_Geology and adaptive_indicator_krig
Module Input Ports
adaptive_indicator_krig has three input ports.
The leftmost (blue/black) field port can accept a data field from Krig_3D, 3D_Geology_Map or other modules that have already created a grid containing volumetric cells.
The second (Orange-Blue-Yellow) port receives the pre-geology file name.
The third port (gray-magenta) port receives the refine distance.
Module Output Ports
adaptive_indicator_krig has three output ports.
The first (Orange-Blue-Yellow) port supplies the pre-geology file name to Pre_Geology.
The second (gray-magenta) port outputs the refine distance to Pre_Geology.
The third port (blue-black) is the primary output field containing nodal data and a refined grid representing geologic materials.
Module Status: Interruptible
This module's computational processes can be terminated (interrupted) using the "C Tech" icon in the Windows Notification Area (aka System Tray) in the lower right corner of your desktop. If you hover over the icon, it will tell you the status of the module and expected completion time. Double-Right-Clicking will terminate the process. Note that if you do stop any process, the output of the module is corrupted and any downstream module's results are not usable. You will need to re-run the module.
Module Control Panel
The control panel of adaptive_indicator_krig is shown in the figure above. The Read Data File button opens the File Browser panel shown in the figure below.
The file browser is a standard window browser that allows the user to specify the directory in which the files reside, and the extension used to filter the displayed available file names. For Indicator_Geology, the only file extension is .pgf. The format of .pgf files is described in the pgf_file_format Help topic.
The "Run" toggle controls whether the module will run when applications are loaded or data changes. When this is on, the module runs when applications are loaded or the "Accept" button is pushed. When it is off, the module will not run.
The Quick Method toggle determines the kriging method.
The Quick Method is similar (but improved) to the previous Geologic Indicator Kriging that was in Krig_3D. It assigns the geologic material data based on the nearest geologic material (in anisotropic space) to your PGF borings. This is done on a nodal basis and an enhanced refinement scheme for the PGF borings.
The Kriging Parameters subpanel with Quick Method selected is shown in the figure below.
For the more rigorous probabilistic approach to geologic indicator kriging, the individual material probabilities for each node is computed. The material with the highest probability is assigned to the node. All of the individual material probabilities are provided as additional nodal data components. This will allow you to identify regions where the material assignment is somewhat ambiguous. Needless to say, this approach is much slower (especially with many materials), but often yields superior results and interesting insights.
The Kriging Parameters subpanel with Quick Method deselected is shown in the figure below.
Note that the Quick Method toggle removed Semivariogram Parameters as an option and utilizes a substantially simplified algorithm for indicator kriging.
The Reach input field defines the radial distance (in user units) from any given model node that the kriging module will look for data points to be included in the estimation of the model parameter at that node. The default value of reach is 0, which results in the module calculating a reach value which is approximately two-thirds of the longest distance between any two data points in the data set.
The Points parameter defines the maximum number of data points (within the specified reach) that will be considered for the parameter estimation at a model node. The default value for points is 20, which generally provides reasonably smooth modeled parameter distributions. The effects of decreasing and increasing the values for reach and points on the model output are somewhat similar, but for different reasons. If the data have a fairly even spatial distribution throughout the domain, then increasing these values will generally include more of the input data points that will be used to krige the value for a given model node, and thus will result in smoother modeled data distributions. Decreasing the values of reach and points (in an evenly distributed data set) results in fewer input data points being used to calculate the parameter estimates at a given model node, and result in modeled distributions with greater variations across smaller areas.
The user should consider both the spatial distribution and the range of values in the input data set when deciding upon values for the reach and points parameters. If the specified reach is too small to allow the kriging module to locate at least one point within the search area, then no kriging will be completed at that model node, the nodal value will be set to 0, and the confidence level will be set to <0.1%. Note that this nodal value is generally inappropriate, and the regions of the model receiving the 0 values should be subsetted out of the domain by using an plume_volume module with a confidence isocomponent of 1% or more. If the user specifies a large number of points (that are within the specified reach), then the output will be smoother, but the execution time for the kriging can increase significantly. By posting the input data using the post_samples module, and looking at the characteristics of the resulting kriged data using the Statistics module, the user can quickly analyze the characteristics and distribution of the kriging output for a given set of parameters, and test the effects of changing the kriging parameter values.
The Octant Search toggle changes the method by which data sample points are selected for inclusion in the kriging matrix. If this is on, the "Points" parameter switches to "Max Points in Octant". Searching is performed for each of the eight Octants surrounding the point to be kriged. Within each octant a maximum number of points (up to one-fourth of the total points) are selected. Then, points are taken sequentially from each octant up to the maximum number of total points or until all octant's points have been used. The panel display changes when this option is selected as shown above. Octant Search is applicable for the Statistics and Min-Max Plume kriging modes.
Understanding the Refine Dist(ance) parameter requires an understanding of the basics of indicator kriging with a .PGF file as input. Since a PGF file has only a single point defining the bottom of intervals of a specific material, it is necessary to create an intermediate dataset that is used for the actual kriging. Since each interval represents a line segment with constant material characteristics, we approximate it by refining the interval into a set of points. The spacing between these points affects accuracy. Each segment will receive at least two points near the end points of the interval, but longer segments will broken up into segments no larger than the refine distance. Setting this value too small will result in much greater run times, as well as requiring higher values for the points parameter. Setting this value too large can create artificial inaccuracies in the prediction because the center of the intervals can have a lack of sample points, causing points in other borings to appear closer. If you leave the default setting of 0.00 for this parameter the expert system determines the spacing for you by searching the original PGF file for the shortest segment in any boring. The calculated refine distance will be twice the shortest segment length or one half the height of the first cell in the grid, whichever is larger. This tends to give reasonable results with most datasets. You can also use the Pre_Geology module to visualize the refined points directly by setting its refine distance parameter to be the same as the computed one in Indicator_Geology.
Since a PGF file has only a single point defining the bottom of intervals of a specific material, it is necessary to create an intermediate dataset that is used for the actual kriging. Since each interval represents a line segment with constant material characteristics, we approximate it by refining the interval into a set of points. The spacing between these points affects accuracy. If you leave the default setting of 0.00 for this parameter the expert system determines the spacing for you by searching the original PGF file for the shortest segment. This minimum distance will receive two points surrounding the center of the interval and longer segments will receive many more. Setting this value too small will result in much greater run times.
The Horiz./Vert. Anisotropy Ratio parameter allows the user to consider the effects of anisotropy in the conductivity of soil matrices to fluid flow. In most cases, geologic materials are deposited with platy clay minerals oriented horizontally, and thus flow of water in both the saturated and unsaturated zone can be slower in the vertical direction than in the horizontal direction. Also, ore deposition can occur along horizontal or vertical fault or fracture systems. Chemical constituents being transported with flowing fluids may therefore show a larger degree of spreading in one or the other direction. The Horiz./Vert. Anisotropy Ratio basically tells the kriging algorithm what multiplication factor should be used to apply biased weighting on data points in horizontal and vertical directions away from a given model node. The default value is 10, which allows data points in a horizontal direction away from a model node to influence the kriged value at that node 10 times more than data points an equal distance away in a vertical direction. However, the user can specify any positive number with a magnitude up to 100,000. When the property being modeled is not related to fluid flow or other processes that might be affected by matrix anisotropy, then this parameter should be set to 1.
The Semivariogram Parameters subpanel is shown in the figure above. The pair search range specifies the radial distance from any input data point that will be searched to assemble the data pairs that are used in the variance analysis. Clicking in the window and using standard windows style editing procedures changes the values in the data windows. The default value for the pair search range is set to 0, which if left alone, results in the value being set to approximately 2/3 of the largest distance between data points in the data set. The user must consider the spatial characteristics of the data set when setting or revising the default calculated Pair Search Range. If large areas exist in the data domain that do not have data points within them, the user must set the Pair Search Range to a value that will allow a pair of data points to be identified, if these outlying data are going to affect the characteristics of the semivariogram. Data sets with large variations over short distances can be modeled most accurately using smaller pair search ranges, as this effectively limits the distance over which the semivariogram will search for and include data points. If a large number of data points exists in the data set which are in close proximity to each other, the user should set the pair search range to the shortest distance that will allow trends in the data to be included in the semivariogram production.
The semivariogram Pair Search Range is an important factor in the execution time required for calculating the semivariogram, as the number of data pairs to be considered is on the order of n squared, for n data samples. However, including a greater number of data pairs in the semivariogram analysis will generally produce kriged distributions that more accurately represent trends and other larger scale characteristics of the data set. If a large data set is being kriged using a large Pair Search Range, the number of pairs the best-fit variogramming procedure must consider also gets very large. EVS implements a deterministic random pair selection algorithm to limit the total number of pairs that are considered in the semivariogram production when the number of potential pairs exceeds 50,000. This algorithm speeds execution and allows the user to input very large data sets. The flexibility of EVS's modular design allows the user to experiment with different search ranges (preferably starting small and getting larger), to obtain the desired results with reasonable execution times.
The Semivariogram Symmetry parameter describes the degree to which EVS's expert system is allowed to distort the geometry of the semivariogram in calculating the best fit to the data. The valid range for this parameter is from 0 to 1. The default value is 1, which forces the semivariogram to be symmetrical in all axes of the data set. Symmetrical variograms run the fastest in EVS, and give reasonable results for many data sets. Unless the data being kriged shows a very high degree of asymmetry, good results are generally obtained by setting this parameter to a value between 0.5 and 1. When utilizing symmetry values less than 0.5, the user should post the original data set and examine areas of the resulting model in which only sparse data are available to fully understand the effects of the asymmetric model semivariogram.
The XY Minimum Range parameter defines the smallest distance in the XY plane at which the semivariogram procedure can set the sill of the semivariogram. In essence, this parameter constrains the minimum distance between data points beyond which EVS's best fit algorithms will consider all points to have an equal and minimum influence on the kriged model node value. The default value for the Minimum Range is 0, which allows the best fit procedure to calculate the range that produces the best fit to the data. The valid range for this parameter can be any number up to the largest distance between points in the data set. However it is generally not meaningful to set the minimum range to values less than approximately five times the shortest distance between data points. When this value is changed from 0, the user should check the calculated range to see if the specified value constrained the semivariogram production, which is indicated when the calculated range is set to the specified value (the output of the semivariogram and kriging procedures is displayed in the console window). If so, the user may want to experiment with different range values, or allow the default value to be used, and compare the kriged results.
The Z Minimum Range parameter defines the smallest distance in the Z axis of the data at which the semivariogram procedure can set the sill of the semivariogram. This parameter is similar to the XY Minimum Range in function, and the same considerations apply.
The user can now set the semivariogram by checking the Set Semivariogram toggle This feature is useful for incorporating a semivariogram that was not calculated with the expert system. Type in boxes are supplied for Major Axis Rotation, Major Range, Sill, and Minor Range. Rotation angle requires an angle measured in degrees from East (equals 0) in a counterclockwise direction. The Major Range refers to the long axis of anisotropy (if any) and the Minor Range refers to the perpendicular of that axis. Note that both ranges refer strictly to the horizontal direction.