Using PEST with AnAqSim
You can use the parameter estimation software PEST to optimize and calibrate AnAqSim models. PEST runs multiple AnAqSim models from the command line (read about automated runs in the User Guide) to home in on optimal model input parameters. PEST can simultaneously optimize a variety of AnAqSim input parameters including aquifer properties, recharge rates, well or line element discharges, specified heads, and river conductances. PEST modifiles the text *.anaq input files and reads text outputs from AnAqSim to compare modeled to observed conditions. The AnAqSim outputs that can be matched or optimized include:
- discharges of head-specified wells
- heads at discharge-specified wells
- discharges of head-specified line boundaries
- discharges of river line boundaries
- calibration target heads at observation wells
- calibration target velocity
- vertical leakage rates
- the number of pathlines captured by specified wells or line boundaries
With all these outputs available for optimization, you can make PEST runs for a range of objectives. Here are some common applications:
- calibrate aquifer conductivities, storage parameters and recharge rates to match regional head and discharge patterns
- calibrate aquifer conductivities and storage parameters to observed pumping test records (useful for cases with multi-layers, partial penetration, and irregular boundary conditions)
- Optimize the discharge and placement of wells or drains in a dewatering problem.
- Optimize the number of pathlines captured from source areas and minimize the discharge required of wells in contaminant remediation design.
Example PEST calibration of a model to observed heads.
A well-documented simple application of PEST-AnAqSim modeling has been posted at this link on the FlexAEM blog. FlexAEM is a nice set of tutorials for learning analytic element modeling with AnAqSim, created by McLane Environmental. This is a simple steady, one-domain model with 6 calibration target observed heads. PEST adjusts the discharges of three pumping wells and the horizontal K of the aquifer to match those observed heads. The blog post clearly describes the files that control the PEST simulation and how to set them up. Thank you to Jim Montague of McLane Environmental for supplying this nice example.
Example PEST optimization of a dewatering system
This is more complex example that was created to optimize the discharges of three dewatering wells in a multi-layer model with partially-penetrating dewatering wells, a partially-penetrating barrier (e.g. sheet pile wall), and discharge from the dewatering wells injected back into the aquifer in a shallow trench. The model has two time periods and PEST optimizes the well discharges in each of the two time periods, minimizing discharge while maintaining heads below a threshold value within the barrier system at specified times. This problem was solved by Dave Dahlstrom of Barr Engineering in Minneapolis, as he participated in the fall 2018 AnAqSim in Depth webinar series. This is the session 6 homework problem solved using PEST instead of by trial and error. This example makes use of two PEST utilities. The PAR2PAR utility synchronizes the total well discharge with the injection rate at the recharge trench, and the OBS2OBS utility was used to set the constraint that head be below a specified threshold value. Here is a link to a pdf file that describes the PEST setup and files for this optimization. and here is a link to a zip file that contains all the input and output files for these PEST runs. Thanks to Dave for going the extra mile with this homework problem and for sharing it with us.