WinPEST

in Visual MODFLOW

Nonlinear Parameter ESTimation and Predictive Analysis

What Is PEST?

PEST is a model-independent parameter estimator that has become a groundwater industry standard. With its powerful inversion engine, the ability to set bounds on parameters while minimizing the discrepancy between model results and field observations, PEST has achieved outstanding results in the calibration of models of all kinds. Its unparalleled ability to work with large and complex models is further enhanced by the possibilities it offers for user interaction in the calibration process. WinPEST is a fully functional version of the PEST program that has been compiled and optimized to run as a native 32-Bit Windows application and is seamlessly integrated with the Visual MODFLOW modeling environment. Visual MODFLOW's innovative and seamless integration with WinPEST makes this powerful technology accessible to every groundwater modeling professional.

Applications of PEST in Groundwater Modeling

Due to the popularity of programs like PEST, nonlinear parameter estimation technology is quickly becoming the preferred method for assisting in the calibration of groundwater models. Not only does the application of this technology have the potential to reduce calibration time enormously, but it can open up new possibilities for creative, yet scientifically rigorous, model design. By applying parameter estimation methods to the calibration of environmental models of all kinds, the modeler gains valuable insights into the strengths and weakness of the dataset. Armed with this knowledge, future data gathering can be undertaken in an optimal manner. Furthermore, the data analysis made possible by use of nonlinear parameter estimation methodologies allows you to make judgments about the degree of uncertainty associated with model predictions and make decisions regarding appropriate levels of model complexity.

With PEST you can turn any model into a powerful data interpretation package. Model calibration is no longer the time-consuming, frustrating and often fruitless exercise it used to be. With PEST, the user is free to unleash his/her creativity in the calibration and data interpretation process while the program carries out the numerically intensive calculations required to implement his/her ideas. PEST allows a modeler to truly understand the capacity that a dataset possesses for the estimation of parameters governing the workings of a system and how supplementary data are most efficiently gathered in order to increase that capacity.

Here's How it Works

If you have ever tried to use PEST with a different MODFLOW software package, please don't be discouraged by your experiences. Visual MODFLOW is the first MODFLOW processor to make this powerful technology accessible, understandable and easy-to-use for every groundwater modeling professional.

In order to use the WinPEST program, you must first create a Base Case model containing best-guess estimates of the model boundaries, boundary conditions and properties values (just like you would normally do). Then you need to run this model to make sure these values are "reasonable" enough to produce a model that will converge to a "reasonable" solution. Once you have done this, you can then let WinPEST perform the most time-intensive task of calibrating these parameters to obtain a best-fit between the calculated model values and values you observed/measured in the field.

Selecting Model Parameters

This is where you will appreciate the true power of the Visual MODFLOW interface to WinPEST. To set up the WinPEST data file you simply select the model parameters you would like to adjust using the expandable model property tree as shown in the following figure:

Defining PEST Parameters

Once you have selected the model parameters that will be adjusted during the PEST iterations, you must define several PEST parameters associated with each of the selected model parameters as shown in the following figure.

These PEST parameters are used to control the adjustments being made to each of the selected model parameters. The ability to have this much control over the way PEST adjusts the parameters from one iteration to another is what gives PEST a huge advantage over other automated model calibration methods. For example, PEST allows you to set reasonable minimum and maximum values for each of the selected model parameters to avoid unnecessary interations and unreasonable results. In addition, the ability to "tie" one model parameter to another allows you to maintain the ratio of these values from one iteration to the next.

Selecting the Objective Functions

Once the PEST parameters have been set, you need to define the "Objective Function." For most cases when you are calibrating a groundwater model, the Objective Function is a fancy term for "Observed Values." The Objective Function tab lists all of the available observations points that have been defined in the Visual
MODFLOW model (as shown in the following figure).

The Objective Functions can be either observed heads, observed concentrations, or observed flow rates in/out of defined zones of the model (e.g., rivers, lakes, wells).

Setting PEST Controls

The PEST Controls include the Marquardt Lambda function settings, parameter change constraints, the PEST run convergence and termination criteria, and other general settings that control the PEST simulations (see the following figure). The default PEST Control settings provided by Visual MODFLOW are fine for most situations.

Running the PEST Simulation

To run the PEST simulation for automatic model calibration, you must Save & Exit the PEST Control window and return to the Visual MODFLOW - Run Menu.

Select Run from the top menu bar and a pop-up window will appear prompting you to select the models you would like to run (e.g., MODFLOW, MODPATH, MT3D/RT3D and PEST).

Once you have selected the models to run, press the [Translate/Run] button to start the model simulations. If PEST is selected, Visual MODFLOW will automatically create ALL the necessary PEST data files. Load the WinPEST program and start the PEST simulations.

Monitoring the WinPEST Simulation

WinPEST runs the PEST simulations and provides a real-time graphical display of the model parameter values and the various PEST analysis parameters for each PEST iteration. The PEST analysis parameters include:

• Objective Function
• Composite Sensitivity
• Parameters History
• Marquardt Lambda
• Calculated vs. Observed Values
• Jacobian Values
• Correlation Values
• Covariance
• Eigenvectors
• Eigenvalues
• Uncertainties
• Residuals

The following figures show some of the available plots.

Normalized Eigenvectors of Covariance Matrix

Parameter Value vs. Iteration Number

Objective Function vs. Iteration Number

Sensitivity vs. Iteration Number

The PEST simulation will be completed once it has determined the best combination of values for each of the selected model parameters to achieve the best fit between the observed and calculated values. The model parameter values determined by PEST can then be automatically imported into the Visual MODFLOW model for further interpretation and presentation of the modeling results.

Back to Visual MODFLOW Page.

Global Enviro Software   1204 W South Jordan Prkwy Ste B    South Jordan, Utah 84095
Phone (801) 208-3011   Toll Free (U.S.) 1-866-620-9214    Fax (801) 302-1160    E-mail info@scisoftware.com