Changes

===Do not display (Blank) bad portions of the model===

===Do not display (Blank) bad portions of the model===

Most computer modeling algorithms do not extrapolate adequately. Many mapping systems automatically extrapolate to the map edge, and extrapolations are often not needed. Parameters usually exist for constraining extrapolations to near the data during model construction (Figure 1). If these do not exist, then a polygon can be drawn around the data area and the model displayed only inside that polygon. Polygon blanking is also used to blank bad data areas for rush projects when no time exists for making corrections.

[[file:using-and-improving-surface-models-built-by-computer_fig1.png|thumb|left|{{figure number|1}}Contour maps of the same surface data. (a) Unconstrained extrapolation into nondata areas. (b) Contours constrained to areas near data.]]

[[file:using-and-improving-surface-models-built-by-computer_fig1.png|thumb|{{figure number|1}}Contour maps of the same surface data. (a) Unconstrained extrapolation into nondata areas. (b) Contours constrained to areas near data.]]

Most computer modeling algorithms do not extrapolate adequately. Many mapping systems automatically extrapolate to the map edge, and extrapolations are often not needed. Parameters usually exist for constraining extrapolations to near the data during model construction ([[:file:using-and-improving-surface-models-built-by-computer_fig1.png|Figure 1]]). If these do not exist, then a polygon can be drawn around the data area and the model displayed only inside that polygon. Polygon blanking is also used to blank bad data areas for rush projects when no time exists for making corrections.

===Apply filters to the model===

===Apply filters to the model===

A common problem with computer-generated surface models are surface structures that are not supported by data. That is, the structures are by-products of the surface-modeling algorithm. Filters such as least squares, biharmonic, laplacian, and others can be applied to existing surface models to remove these unsupported structures. Data should be honored while filtering so that data values continue to fall on the correct side of contours.

A common problem with computer-generated surface models are surface structures that are not supported by data. That is, the structures are by-products of the surface-modeling algorithm. Filters such as least squares, biharmonic, laplacian, and others can be applied to existing surface models to remove these unsupported structures. Data should be honored while filtering so that data values continue to fall on the correct side of contours.

Sometimes undesired contour wobbles are caused by data. For example, shot point values for a seismic line that parallels strike and whose shot point z values fluctuate about a contour value will cause contours to wobble through the data. This wobble is due to noise in the seismic data and is usually undesirable. Many filter programs allow surface models other than the one being filtered to act as upper and lower constraints within which the resultant surface model must stay. By shifting the original surface model up and down by small amounts (magnitude of the wobbles) and using these new surfaces to constrain filtering, a smoother model that still honors surface form can be created (Figure 2).

Sometimes undesired contour wobbles are caused by data. For example, shot point values for a seismic line that parallels strike and whose shot point z values fluctuate about a contour value will cause contours to wobble through the data. This wobble is due to noise in the seismic data and is usually undesirable. Many filter programs allow surface models other than the one being filtered to act as upper and lower constraints within which the resultant surface model must stay. By shifting the original surface model up and down by small amounts (magnitude of the wobbles) and using these new surfaces to constrain filtering, a smoother model that still honors surface form can be created ([[:file:using-and-improving-surface-models-built-by-computer_fig2.png|Figure 2]]).

 

[[file:using-and-improving-surface-models-built-by-computer_fig2.png|thumb|{{figure number|2}}Cross section showing the output from filtering being constrained between models built by shifting the initial surface model up and down slightly.]]

===Rebuild using different modeling parameters===

===Rebuild using different modeling parameters===

Most mapping systems have many parameter switches to adjust surface modeling algorithms. Only a few are frequently used to improve the model, but the ones used often make drastic improvements. Parameters found to be useful affect items such as amount of smoothing, number of data points required, number of sectors (compass directions from which data are used) that must have data, distance to look for data, type of algorithm used, type of filter used, size of grid increment, and whether a coarse regional model is built and refined to the desired detail.

Most mapping systems have many parameter switches to adjust surface modeling algorithms. Only a few are frequently used to improve the model, but the ones used often make drastic improvements. Parameters found to be useful affect items such as amount of smoothing, number of data points required, number of sectors (compass directions from which data are used) that must have data, distance to look for data, type of algorithm used, type of filter used, size of grid increment, and whether a coarse regional model is built and refined to the desired detail.

===Rebuild using multiple step modeling methods===

===Rebuild using multiple step modeling methods===

====Regional trend assist====

====Regional trend assist====

This technique is sometimes used when the surface is tilted, projection up and down dip is desired, and algorithms that project slope do not produce acceptable results (Figure 3). The general steps are (1) build a first- or second-order trend surface through the data, (2) back interpolate from the trend surface a ''z'' value at each data location, (3) subtract the back interpolated values from the original data creating difference values, (4) build a surface model of the difference, and (5) add the difference surface to the trend surface.

This technique is sometimes used when the surface is tilted, projection up and down dip is desired, and algorithms that project slope do not produce acceptable results ([[:file:using-and-improving-surface-models-built-by-computer_fig3.png|Figure 3]]). The general steps are (1) build a first- or second-order trend surface through the data, (2) back interpolate from the trend surface a ''z'' value at each data location, (3) subtract the back interpolated values from the original data creating difference values, (4) build a surface model of the difference, and (5) add the difference surface to the trend surface.

[[file:using-and-improving-surface-models-built-by-computer_fig3.png|thumb|{{figure number|3}}Cross sections through the same data. (a) Extrapolated values for a weighted average algorithm tend to “come back” to the average of near data values. (b) Acceptable surface extrapolation achieved by creating a first-order trend, modeling residuals between data and trend, and adding the trend and residual surfaces.]]

[[file:using-and-improving-surface-models-built-by-computer_fig4.png|thumb|{{figure number|4}}Cross sections through the same data. (a) The surface model does not honor the data. (b) Surface is shifted to data by modeling the error between the data and the original surface and then adding the original and error models.]]

====Error correction====

====Error correction====

This technique is used to correct a surface model that does not honor the data. It is also used to update a surface model when new data are added and when it is undesirable to rebuild the model, only to adjust it to the new data. The general steps are (1) back interpolate from the original surface model a z value at each data location, (2) determine the error at those locations by subtracting the interpolated value from the data value, (3) model the error, and (4) add the error to the original model (Figure 4). This procedure is commonly used to shift a surface model built from seismic data so that it passes through (honors) well data.

This technique is used to correct a surface model that does not honor the data. It is also used to update a surface model when new data are added and when it is undesirable to rebuild the model, only to adjust it to the new data. The general steps are (1) back interpolate from the original surface model a z value at each data location, (2) determine the error at those locations by subtracting the interpolated value from the data value, (3) model the error, and (4) add the error to the original model ([[:file:using-and-improving-surface-models-built-by-computer_fig4.png|Figure 4]]). This procedure is commonly used to shift a surface model built from seismic data so that it passes through (honors) well data.

[[file:using-and-improving-surface-models-built-by-computer_fig4.png|thumb|{{figure number|4}}Cross sections through the same data. (a) The surface model does not honor the data. (b) Surface is shifted to data by modeling the error between the data and the original surface and then adding the original and error models.]]

[[file:using-and-improving-surface-models-built-by-computer_fig5.png|left|thumb|{{figure number|5}}Contour maps of the same data. (a) Most algorithms weight data isotropically and creat circular surface forms. (b) Single direction bias forces elliptical weighting, allowing surface form to stretch in one direction.]]

====Directional bias====

====Directional bias====

Surface models that stretch along axes of anticlines and synclines are easily produced if directional bias capabilities are available in the surface-modeling algorithm (Figure 5). If these capabilites are not available, this effect (single direction bias) can be built using a multi-step approach<ref name=pt08r11>Jones, T. A., Hamilton, D. E., Johnson, C. R., 1986, Contouring geologic surfaces with the computer: New York, Van Nostrand Reinhold Company, 314 p.</ref>. The general steps are (1) rotate the data so the bias direction is north-south, (2) divide the ''y'' coordinate by the bias magnitude, (3) build a grid, (4) convert the grid to data, (5) multiply the ''y'' coordinate by the bias magnitude, (6) rotate the grid data back to the original coordinate system, (7) merge the original and new data, and (8) build the final grid.

Surface models that stretch along axes of anticlines and synclines are easily produced if directional bias capabilities are available in the surface-modeling algorithm ([[:file:using-and-improving-surface-models-built-by-computer_fig5.png|Figure 5]]). If these capabilites are not available, this effect (single direction bias) can be built using a multi-step approach<ref name=pt08r11>Jones, T. A., Hamilton, D. E., Johnson, C. R., 1986, Contouring geologic surfaces with the computer: New York, Van Nostrand Reinhold Company, 314 p.</ref>. The general steps are (1) rotate the data so the bias direction is north-south, (2) divide the ''y'' coordinate by the bias magnitude, (3) build a grid, (4) convert the grid to data, (5) multiply the ''y'' coordinate by the bias magnitude, (6) rotate the grid data back to the original coordinate system, (7) merge the original and new data, and (8) build the final grid.

 

===Interactively edit the surface model===

===Interactively edit the surface model===