Please enable JavaScript to view this site.

ModelMuse Help

The Pilot Points tab is on the PEST Properties dialog box.

This pane is used to define the locations of pilot points and to specify whether or not pilot point locations will be displayed.

For an in-depth discussion of pilot points, see https://gmdsi.org/blog/pilot-points/.

A PEST parameter can be associated with pilot points (Doherty, 2003; Doherty and others, 2010). Pilot point approaches used in ModelMuse are explained in the manuals for PLPROC and the PEST Groundwater Utilities available from the PEST home page (https://pesthomepage.org/). ModelMuse uses PLPROC to implement pilot points. Section 4 of the documentation for the PEST Groundwater Utilities provides a conceptual description of pilot points and how they are used. In brief, pilot points are a parameterization device that facilitates increased parameter flexibility by estimating properties at user-specified locations in the grid; the remainder of the grid is then filled by kriging.  Pilot points can be grouped with zonation, but such zonation is not required. For instance, suppose the hydraulic conductivity of an aquifer is being estimated. From aquifer tests, it may be known that the hydraulic conductivity varies from place to place but the spatial variation might not be well defined. One way to approach this would be to specify zones within the aquifer and to have PEST estimate separate uniform parameter values within each zone. A potential drawback of this approach is that the chosen zonation might not be optimal. Pilot points provide a way to get around this problem. The modeler designates a number of points and assigns a value to each point. Between points, values are assigned to each cell via kriging. Within PEST, the value assigned to each pilot point is a parameter that can be adjusted by PEST to improve the fit between the observed and simulated values. If all pilot points are tied in the PEST control file, they will act like a zone. Likewise, a zone with zero or on associated pilot points will also produce zone-like results. Such aspects can be useful in stepwise modeling, where model complexity is added in sequential steps in response to model performance.

Each pilot point value is treated as a separate parameter in PEST. PEST will typically run the model at least once for each parameter during each parameter estimation. Thus, the number of pilot points can have a large effect on the time required for parameter estimation. You can estimate the time required for one parameter estimation by multiplying the number of parameters by the time required to run the model one time. You can find the time required to run the model by determining the time difference between when the last model input file was created by ModelMuse and the time the last output file was created by the model. The last input file for the model created by ModelMuse is typically "RunModel.bat". The number of parameters is the variable NPAR in the PEST control file. NPAR is the first number outside of a comment in the control data section of the PEST control file.

Pilot points are typically used with regularization where constraints are applied to the pilot point values so that pilot points that are close to one another will be assigned similar values.

If desired, pilot points can be combined with zones. In such cases, a set of pilot points will be selected from all the pilot points that have been defined. A pilot point will be included in the set used for a particular zone if it falls within the zone or if it is no farther away then the Pilot point buffer from a cell center (MODFLOW) or node location or element center (SUTRA) that is part of the zone. If you wish to use pilot points, the Pilot point buffer must be greater than zero.

The locations of pilot points can be defined in several ways. If desired, the various methods can be combined.

A common method is to have a series of Regularly spaced pilot points in which the pilot points form either squares or equilateral triangles. For this method, the user selects the Pattern and Pilot point spacing.

The user can also specify the locations of individual pilot points. This is often used to supplement the pilot points specified in a regular pattern when there is reason to  believe that more detail is desirable in a specific region. The table within the Individually specified pilot points panel displays the location of these points. The data can be typed into the table or imported from text files. With text files, each line must contain the X and Y coordinates of a point and the coordinates must be separated by a tab character. If the data are in a spreadsheet, the data can be copied from the spreadsheet to the clipboard and then pasted into the table by selecting a cell of the table and pressing the keyboard combination Ctrl-V. Finally, if Show candidate pilot points is checked, then when the user can closes the PEST Properties dialog box, the user can select Edit|Add Pilot Point. Then, when the user clicks on the top view of the model with the mouse, a new pilot point will be created at the location of the mouse click. This command can also be accessed by clicking the Add pilot point Screen capture of the Add pilot point button. button.

Yet another way to define pilot points is to have ModelMuse create them Between point observations. If you have defined two or more calibration observations with point objects, you can click the Generate pilot points between point observations button and ModelMuse will create pilot points between the observations. If three or more point calibration observations are defined, ModelMuse will create a triangulation of the point observations and put new pilot point on the midpoints of the triangulation lines. If there are numerous closely spaced point observations, this process may result in too great a concentration of pilot points in a small area. To lessen this problem, the user can specify a Minimum separation. If Minimum separation is set to a value greater than zero, some of the newly created pilot points will be removed to ensure that no two of them are closer together than the Minimum separation.