GMS

Tips for Working with PEST

By aquaveo on April 17, 2019

Parameter estimation using PEST can assist in data interpretation, model calibration, and predictive analysis. PEST can do a lot for your project.

Because PEST offers so much, it can be a little overwhelming to use at first. We provide tutorials and articles on our wiki to help guide you in using PEST with GMS.

In addition to the resources available in the tutorials and wiki, this post covers some useful tips that have helped some of our users.

General PEST Tips

Here are some general tips for troubleshooting PEST in GMS:

Verify that the parameter key values have been successfully initialized under MODFLOW | Parameters. The key values should match the parameters names. For example, if the key value for hydraulic conductivity is “-30”, the parameter name should be “HK_30”. No key value should be used twice, even across different packages. To continue the example, if you have an “HK_30”, there should not be a “RCH_30”.
Key values are needed for both zones and pilot points.
Key values are expected to be negative integers. Non-integers will not be automatically recognized as key values.
If the key value is not a negative integer, or is a parameter from the WEL package, they will have to be manually added. If you’re missing a value, check that it initialized in the first place.
Parameters included in the run will be written to the *.param file. Once PEST runs, the optimal values are written to the *.par file.
If the iterations are giving identical values for the same parameter, check if the value is either the minimum or the maximum of the range assigned to the parameter, then evaluate if that range needs adjusting.
Adjusting the starting values or the PEST options may help.

Pilot Point Tips

When using PEST and pilot points, consider the following:

Follow the recommended guidelines found here.
You can get a quick 2D Scatter set to use for pilot points with the MODFLOW Layers → 2D Scatter Points tool. This is particularly useful for areal parameters, such as HK or RCH.
If you’re running Null Space Monte Carlo, you must use pilot points and your interpolation method must be set to kriging.
If running with SVD-Assist, you can point to the Jacobian file so it doesn’t re-solve for every iteration.
Depending on the parameter, pilot points often should not start with zero values. This especially applies to log transforms.

We hope these tips help and that your project are improved by using PEST with GMS.

Blog tags:

GMS

Troubleshooting MODFLOW, Part 2

By aquaveo on March 27, 2019

A while back on this blog, we discussed troubleshooting MODFLOW errors. That blog post specifically discussed making use of the Model Checker, the MODFLOW command line output, and the output file. It also gave a few tips on how to fix your model when an error is encountered.

We wanted to expand on this, and specifically discuss what to do when you model doesn’t converge. When the model does not converge, an error message should appear in the MODFLOW command line output.

Essentially, when a model doesn’t converge a component of the model has not been setup correctly. This inaccurate component may only cause the model to not converge when certain conditions have been met, but otherwise the model will converge when those conditions are not present.

As for why your model converges sometimes and not others, there are a wide range of possible causes for instability. Here are a few general suggestions for helping MODFLOW converge:

Check model inputs for reasonableness.
Try running the model with different solvers. There are several solvers to choose from, and each of them have their own strengths and weaknesses. To switch solvers, select MODFLOW | Global Options | Packages and then select a different solver in the lower left area of the MODFLOW Packages dialog.
Try changing the solver parameters.
Check the troubleshooting items for a model that is not converging can be found under item K of the Frequently Asked Questions section of the Online Guide to MODFLOW.
Deselecting the "Enable saving of computed flows for all source/sink objects" option in the MODFLOW Output Control dialog.
Reduce the time step length for the model run.

It can take some time to review the model to discover why it is not converging, but the effort it worth it for an accurate result. For more in-depth assistance with model troubleshooting, please consider reaching out to Aquaveo’s consulting team.

Blog tags:

GMS

Exporting and Importing to MODFLOW 6

By aquaveo on March 6, 2019

In 2018, USGS released MODFLOW 6. This version of MODFLOW uses object-oriented programming to provide support for multiple models within the same simulation. Like many of you, we at Aquaveo were excited to see this new development and started on working on ways for GMS to interface with this new version of MODFLOW.

Did you know that with GMS 10.4 you can export your MODFLOW project for use with MODFLOW 6? This allows you to convert your older or current GMS projects for use with MODFLOW 6. This is just one of the new features in GMS 10.4.

Support has also been added to run MODFLOW 6 from GMS and read the head and flow outputs which may be contoured.

The general workflow process for saving, running, and importing the MODFLOW 6 files is as follows:

After building your MODFLOW model, open the MODFLOW Global/Basic Package dialog.

In the dialog, turn on the Save MODFLOW 6 copy option under the MODFLOW Version section.
Save your project.
Open MODFLOW | Advanced | Run MODFLOW Dialog... to run the MODFLOW 6 files.
Use the Custom MODFLOW option to point to the mf6.exe executable in program files (e.g., C:\Program Files\GMS 10.4 64-bit\models\mf6\mf6.exe).
Browse to and select the NAM file out of the *_MODFLOW_mf6 folder that will have saved in the same directory as your GMS project. This is not the default, so you will need to browse for it at least the first time.
Run MODFLOW.
Use MODFLOW | Read Solution to select the MFN file out of the *_MODFLOW_mf6 folder.

The exported files can also be used directly in MODFLOW 6.

In the future, we plan to add more MODFLOW 6 functionality to GMS including a full interface. For now, get ready by converting your projects to MODFLOW 6 using GMS 10.4.

Blog tags:

GMS

Creating a Confined Aquifer

By aquaveo on February 6, 2019

Does your MODFLOW model contain a confined aquifer? A confined aquifer layer is defined as "an aquifer below the land surface that is saturated with water. Layers of impermeable material are both above and below the aquifer, causing it to be under pressure so that when the aquifer is penetrated by a well, the water will rise above the top of the aquifer."

In MODFLOW, a layer is considered confined when the head in the cell is above the top of the cell. Additionally, any cell located above the water table will be unconfined because the head in those cells will be below the top of the cell.

The layers in your GMS MODFLOW model can be assigned as confined or convertible in any of the flow packages, such as in the LPF package. Other flow packages can be used as well, including the BCF, HUF, and UPW packages. When setting up the MODFLOW model, select the desired flow package in the MODFLOW Packages dialog. The selected flow package must be compatible with the specified MODFLOW version.

In any of the flow package dialogs, under the Layer Type subheading, a layer can be defined as "Confined" or "Convertible". "Convertible" means GMS will automatically assign the layer as confined or unconfined depending on the elevation of the water table in the simulation. Only one layer type can be assigned to each layer. By default, all layers are set to convertible unless specified otherwise.

When a layer is explicitly set to be confined, MODFLOW will use the thickness of the cell, rather than the saturated thickness, to compute a transmissivity value. It will not check for the unconfined condition in the layer.

After defining the layers as confined or convertible, you will have a confined aquifer for your simulation. Try using confined aquifer layers in your GMS models today!

Blog tags:

GMS