PEST - Model-Independent Parameter Estimation and Uncertainty Analysis

Training

Tutorial PEST Courses


A Tutorial

Download a comprehensive tutorial that shows you how to use PEST to undertake state-of-the-art model parameter and predictive uncertainty analysis. Two complete worked examples are provided, one involving a surface water model (HSPF) and the other involving a groundwater model (MODFLOW). Also provided is the background theory (explained in an easy-to-read manner) and suggestions on how modern methods of data processing and uncertainty analysis such as are provided by PEST can be used as critical components of the environmental decision-making process.

Topics covered include the following:

  • Monte-Carlo analysis;
  • Linear parameter and predictive uncertainty analysis;
  • Predictive error and predictive uncertainty;
  • Parameter contributions to pre- and post-calibration predictive uncertainty;
  • Assessment of data worth through its ability to reduce predictive uncertainty;
  • Uncertainty analysis for over-determined and under-determined systems;
  • Null space and solution space components of parameter and predictive uncertainty;
  • Nonlinear predictive calibration-constrained maximization/minimization;
  • Assessment of parameter identifiability;
  • Null space Monte Carlo analysis;
  • Scientific hypothesis-testing using models;
  • Using Pareto methods to calculate predictive confidence intervals.

Many thanks to the South Florida Water Management District for making development of this tutorial possible.


PEST Courses

Two to three professional PEST courses are normally held each year.

A two-day course on parameter estimation and uncertainty analysis using PEST and Groundwater Vistas will be held following the MODFLOW conference in Golden, Colorado. This will take place over 8th/9th June and will be conducted by John Doherty and Jim Rumbaugh. This will follow a two day course conducted by Sorab Panday, Jim Rumbaugh and John Doherty featuring use and support for MODFLOW-USG.

A five-day course will be held in Belgrade, Serbia starting on Monday 23rd September with heavy emphasis on practical aspects of PEST usage. This will be run by John Doherty with help from Dragan Kaludjerovic.  

Beginning on the 6th October, a five day course will be held in Neuchatel, Switzerland. This course will be both practical and theoretical and will cover parameter estimation and uncertainty analysis. It will be run by John Doherty with help from Philip Brunner.

In mid to late October (date has yet to be decided) a course will be held in Bethesda, Maryland. This will feature a two-day beginners course run by Matt Tonkin of SSPA and a two day advanced course run by John Doherty.

Contact us for further details on any of these courses.

In-house training is also offered. This can be tailored to your needs. Contact us if you are interested.

Topics covered in a typical PEST course include the following:
  • Traditional parameter estimation
  • Regularized inversion
  • Tikhonov and SVD regularization
  • Model calibration using SVD-assist
  • Techniques for calibration of ground water models
  • Use of pilot points as a parameterization device
  • Techniques for calibration of surface water models
  • Parameter and predictive uncertainty analysis
  • Calibration-constrained Monte Carlo analysis (including null space MC)
  • Optimization of data aquisition
  • Dealing with model imperfections
  • Handling of "real" measurement noise and structural noise
  • Use of models in decision-support
  • Model-based hypothesis-testing 
Courses include theory and workshops. Participants are provided with a memory stick whose contents include the following.
  • The latest version of PEST
  • Copies of all slides shown during lectures
  • Files and documentation for all PEST workshops
  • Copies of papers and other literature on PEST
Workshops include the following.
  • Using PEST with a simple storage model
  • Using PEST with MODFLOW (three workshops)
  • Using PEST with HSPF
  • Using PEST with SEAWAT (two workshops)
  • Using pilot points
  • Linear and nonlinear uncertainty analysis

C4SF Login | Web Site Managed via C4Site Factory,™ a trademark of Echo Valley Graphics, Inc.
©2013 Copyright Echo Valley Graphics, Inc. All Rights Reserved.