Draft Program for PEST Course in Association with XXVII Nordic
Hydrological Conference
Model Calibration and Predictive Uncertainty Analysis
http://www.pesthomepage.org/Home.php
General
This intensive short course will instruct participants on the automated calibration of
environmental models, and on the analysis of the predictive uncertainty associated with such
models. The principal instructor is the developer of PEST, the industry standard for modelindependent, automated calibration and predictive uncertainty analysis of environmental models.
What you will learn
While the course will include a thorough coverage of the theory and applications of nonlinear
parameter estimation techniques in the calibration of different types of models, there will also be
a strong practical aspect of the course. Participants will gain hands-on experience in the use of
PEST, including its advanced regularization and predictive analysis functionality in the
calibration of groundwater flow and transport models, surface water quality and quantity models,
as well as other types of models.
Topics covered will include:















theory of nonlinear parameter estimation
application of nonlinear parameter estimation to model calibration
“nuts and bolts” of using PEST
the need for regularization
parameter identifiability
Tikhonov and subspace regularization
“SVD-Assist” as a mechanism for model calibration
pilot points as a spatial parameterization device
linear model predictive uncertainty analysis
nonlinear model predictive uncertainty analysis
null-space Monte Carlo as a mechanism for exploring predictive uncertainty
optimization of data acquisition to reduce uncertainty
so called “global optimization”
use of PEST in groundwater model calibration
use of PEST in surface water model calibration
In the practical sessions, participants will gain hands-on experience in using PEST with a
number of different models, including MODFLOW, MT3DMS, SEAWAT, SWIM (a Richardequation-based, unsaturated zone, water-movement model), 3PG (a forest production model)
and HSPF (a much-used USEPA/USGS model for simulation of non-urban, non-point pollution
of surface water systems). Participants will also be introduced to a suite of utility programs that
automate PEST set-up for these (and similar) models, and that expedite implementation of
model predictive uncertainty analysis.
Who should attend?
Those across the whole spectrum of modeling experience will benefit from this course. There
are many insights to be gained into the use of models, and what can be expected of models,
1
from an understanding of the role that parameterization plays in providing integrity to model
predictions, and in analyzing the extent to which those predictions may be in error. Hence
managers, those who use models as a basis for environmental policy formation, as well as
seasoned modelers, will all benefit from this course.
What you will receive
Participants will receive a CD containing the following:
 the latest version of PEST
 the latest version of all PEST support utilities (over 200 programs)
 copies of files and documentation for over 10 PEST workshops
 literature (mainly published papers) on the use of PEST
About the Instructor
Dr. John Doherty is the author of PEST. John has worked for over 35 years in the water
industry, first as a groundwater exploration geophysicist, then as a modeler. He has worked in
the public, private and education sectors. He now directs his own company, Watermark
Numerical Computing, which undertakes software development and advanced modeling for
mining, environmental, agricultural, water supply and remediation applications. He also works as
a senior research scientist for the National Centre for Groundwater Research and Training,
Flinders University, Australia where he supervises a number of post-graduate students who are
pursuing research into issues related to model parameterization and predictive uncertainty
analysis.
John has had over fifteen years of experience in presenting short courses all over the world.
Course material is presented clearly and descriptively with many practical examples and
illustrations. He attempts to create a learning environment that is both educational and
enjoyable.
COURSE PROGRAMME
DAY 1 – 16 August
Excursion to Rokua esker with discussion about groundwater modeling and monitoring (as you
remember, we agreed to include this as part of the course)
The fee for excursion is 85 euros
DAY 2: Introduction and Parameter Estimation Basics – 17
August










introductions
outline for next four days
basic statistics
what is “calibration”?
well-posed problems and ill-posed problems
linear theory of well-posed inverse problems
inferring parameter uncertainty in the well-posed context
extension of theory to nonlinear models
observation weighting
prior information
2









parameter nonuniqueness
use of parameter bounds
the Marquardt lambda
analysis of residuals
PEST and model-independence
template and instruction files
the PEST control file
tuning PEST performance
PEST, Parallel PEST and BEOPEST
DAY 3: Ill-Posed Problems and Highly Parameterized Inversion
– 18 August














the nature of expert knowledge
the need for regularization
metrics for uniqueness
brief discussion of geostatistics
kriging as regularized inversion
Tikhonov regularization
use of pilot points as a device for spatial parameterization
combining pilot points and regularization
utility software to implement regularized inversion
truncated singular value decomposition as a regularization device
information transfer expressed through singular value decomposition
model simplification as a regularization device
PEST’s “SVD-assist” methodology
the resolution matrix
DAY 4: Calibration in Different Contexts – 20 August
Groundwater modelling
 use of PEST with MODFLOW, MT3D and SEAWAT
 coping with cell drying and re-wetting in MODFLOW
 pilot point emplacement guidelines
 strategies for steady state model calibration
 strategies for transient model calibration
 strategies for multi-layer model calibration
 handling uncertain boundary conditions
 calibration and hypothesis testing
 PEST groundwater modeling utility software
Surface water modeling
 problems encountered when calibrating surface water and land use models
 overcoming these problems
 multiple objective function minima and how to handle them
 global methods
 the SCE and CMA methods
 multi-objective parameter estimation
 simultaneous calibration of multiple models with/without inter-model regularization
 incorporation of exceedence probabilities and other “secondary observations” into the
parameter estimation process
3




digital filtering for baseflow separation and incorporation of this into the calibration
process
description of utility software supplied with PEST
use of constituent and sediment data in model calibration
incorporation of relationships between model parameters and catchment properties into
the calibration process
DAY 5: Uncertainty Analysis and Model Simplification – 21
August
Uncertainty analysis
 sensitivity analysis
 loss of detail incurred through model calibration
 the difference between “uncertainty” and “potential for error”
 linear propagation of uncertainty and error
 nonlinear predictive uncertainty and error variance analysis
 stochastic field generation
 calibration-constrained stochastic field generation
 calibration constrained Monte Carlo analysis
 Markov chain Monte Carlo
 null space Monte Carlo
 data worth analysis
 parameter contributions to predictive uncertainty
 examples
Working with defective models
 expressing model defects mathematically
 the nature of structural noise
 ramifications for predictive uncertainty analysis
 surrogate roles taken by parameters and repercussions for model calibration
 prediction-specific calibration
 uncertainty quantification through hypothesis-testing
 model-based decision-making
Practical Sessions
As time permits, a number of sessions during the course will be devoted to workshops through
which participants can gain experience in using PEST. Fourteen such exercises are provided –
many more than can be done during the time allotted. All participants will be given a memory
stick with these exercises so that they can do them later, in their own time. These memory sticks
will also include copies of all slides used during the course, the latest version of PEST, and a
variety of PEST-related literature.
4