PEST++ — The Complete Guide: Free Model Calibration and Uncertainty Analysis
PEST++ is a USGS toolset for calibrating numerical models against field measurements. It is not tied to one program: any model that reads input and writes output as text files can work with it. In practice it is especially common for calibrating MODFLOW and hydrological models.
PEST++
USGS
Parameter estimation, uncertainty analysis and optimisation for any numerical model — common for groundwater and surface water.
Get it from the vendorWhat you get
- Deterministic calibration (glm) — adjusting parameters so the model reproduces the measurements.
- Ensemble smoother (ies) — calibration and uncertainty analysis with an ensemble of realisations.
- Global sensitivity (sen) — which parameters really drive the result.
- Optimisation (opt, mou) — under uncertainty and with multiple objectives.
- Parallel runs — spreading scenarios over several cores or computers.
What to know
You need to prepare control files and templates in PEST format, and that is the slowest part at the start. Ready-to-run binaries exist for Windows and Linux; on macOS you build from source. The code is public domain (USGS) and free for any use, including commercial projects.
Related tools
To build a MODFLOW model in a script — FloPy. For the groundwater engine itself — MODFLOW 6. For general uncertainty analysis — OpenTURNS.
FloPy
USGS and FloPy developers
Build, run and post-process MODFLOW groundwater models from a Python script.
Get it from the vendor
MODFLOW 6
U.S. Geological Survey (USGS)
The world standard for groundwater flow and solute transport modelling, from the USGS.
Get it from the vendorModelMuse
U.S. Geological Survey (USGS)
The free graphical interface to MODFLOW — build the model with a mouse instead of text files.
Get it from the vendor
OpenTURNS
OpenTURNS consortium
A Python and C++ library for uncertainty treatment — probabilistic modelling, reliability, sensitivity and metamodels for engineering risk.
Get it from the vendorThe bottom line
An uncalibrated model is a sophisticated guess. PEST++ turns calibration and uncertainty analysis into a systematic process — free, and with any text-based model.