Workshop Summary

The Groundwater Resources Association (GRA) is conducting a three-day workshop "Model Calibration and Predictive Uncertainty Analysis Using PEST" on September 13-15, 2006. GRA is pleased to announce that the principal instructor for the workshop is Dr. John Doherty, author of PEST and President of Watermark Numerical Computing. He will be assisted by Matt Tonkin of SS Papadopulos & Associates, Inc. The workshop will provide attendees with the foundations of parameter estimation theory, an understanding of the sources of uncertainty in predictions made by models, and experience in using PEST to calibrate groundwater, surface water and other environmental models and explore their predictive uncertainty.

By Request - Advanced Model Predictive Error Analysis Morning Session

Current attendees and those who previously completed a PEST course are invited to join us for a special additional 4-hour session on Saturday September 16th (8am to midday) when model predictive error analysis will receive detailed attention. Participants will gain hands-on experience in predictive error analysis through a comprehensive ‘soup-to-nuts’ model calibration, regularized inversion and predictive error analysis workshop. Topics covered will include:

•  Sources of error in a calibrated model - the model null space and solution space
•  Linear and nonlinear analysis of model predictive error
•  Evaluating the contributions of various sources of model predictive error
•  Optimization of data acquisition to reduce predictive error
•  Predictive error analysis as an adjunct to regularized inversion
•  Appropriate model complexity

Wine-and-Cheese Reception

These PEST courses are hard work – what better way to get to know your classmates and the instructors than during a wine-and-cheese social, the night before the hard work begins!

About PEST

PEST (Parameter ESTimation) is a general-purpose, model-independent, parameter estimation and model predictive error analysis package developed by Dr. John Doherty. PEST is the most advanced software readily available for calibration and predictive error analysis of groundwater, surface water, and other environmental models. Using PEST you can:

•  apply advanced and efficient regularization techniques in calibrating your models to extract    maximum information content from your data
•  undertake linear and nonlinear predictive error analysis of model outputs
•  simultaneously parameterize several models using multiple datasets
•  accommodate heterogeneity using advanced spatial parameterization
•  combine PEST with stochastic field generation to explore calibration non-uniqueness
•  conduct parallel model optimization runs across PC or UNIX networks
•  compare the worth of different proposed data acquisition strategies in reducing model     predictive error thereby optimizing resources allocated to such tasks
•  quantify the contributions to model predictive error made by different parameter types
•  establish the irreducible uncertainty of a model prior to calibrating that model
•  quantify the reduction in predictive uncertainty accrued through model calibration

Traditional methods of parameter estimation and uncertainty analysis based on a handful of zones do not provide the flexibility to extract vital information from expensive data. Some of the advances made over the last few years in PEST to overcome this include:

•  The combination of regularized inversion with the use of pilot points as a spatial    parameterization device
•  The unique and extremely efficient “SVD-assist” regularized inversion methodology which    combines subspace and Tikhonov methods. Inversion can be carried out with model-run    efficiencies compatible with traditional approaches - but with hundreds of parameters
•  The combination of regularized inversion with stochastic field generation to produce    calibration-constrained parameter fields

The development of PEST and its ancillary software in 2005 focused on model predictive error analysis. This will be discussed in the 3-day class and focused upon in the four-hour ‘soup-to-nuts’ workshop – including linear and nonlinear predictive error analysis as an adjunct to regularized inversion; how regularized inversion develops mathematical parsimony on the basis of the data rather than preemptive best-guess parsimony; extracts maximum information from the available data; enables the rigorous investigation of model predictive error variance; and in the groundwater context can support inference on the whereabouts of heterogeneity based on the calibration data.

Course Attendee Information

Practitioners from a wide range of backgrounds will benefit from this course, whether new to PEST or with previous PEST experience. To get the most out of the course, attendees should have modeling experience, preferably in the groundwater or surface water disciplines. Hands-on labs are GUI-independent and cover a variety of modeling disciplines, so that anyone interested in model calibration, parameter optimization, or the analysis of numerical model uncertainty could attend. Experience in working at the command line level is an advantage in doing the workshops. However, whether or not you are a modeler you can still benefit from the course because the concepts explained in this course are vital to an understanding of the role of modeling in environmental management, from which modelers and non-modelers alike will benefit.

Program Agenda - Complete Course Outline >>

Tuesday, September 12, 2006
6pm to 8pm	Free wine and cheese reception - a first for the PEST course! Meet instructors and attendees in a relaxed atmosphere before the grind starts!
Wednesday, Setember 13, 2006
Day 1: Morning	Lecture - Introduction to non-linear parameter estimation: mathematical theory, development, and integration with models
	Lab - Hands-on exercise calibrating a basic unsaturated zone/storage model
Day 1: Afternoon	Lecture - Fundamentals of PEST - the implementation of nonlinear parameter estimation theory within PEST
Thursday, September 14, 2006
Day 2: Morning	Lecture - The use of PEST in the groundwater modeling context - flow and transport
	Lab - Hands-on exercise - choice of prepared case studies for groundwater and surface modeling, or attendees' own studies
Day 2: Afternoon	Lecture - Regularized inversion - methodologies and benefits.
	Lab - Choice of hands-on problems including surface and groundwater, forestry, unsaturated zone, etc.
Friday, September 15, 2006
Day 3: Morning	Lecture - The use of PEST in surface water modeling, with particular emphasis on HSPF.
	Lab - Hands-on exercise - choice of prepared case studies for groundwater and surface water modeling, or attendees' own studies
Day 3: Afternoon	Lecture - Sources and analysis of uncertainty in environmental modeling.
	Lab - Choice of hands-on problems including surface and groundwater, forestry, unsaturated zone, etc.
Saturday, September 16, 2006 - Optional Advanced Model Predictive Error Analysis Morning Session
	Current and former course attendees can join a special additional 4-hour session (8am to midday) when model predictive error analysis will receive detailed attention.
Optional Evening Sessions
	Optional sessions each evening. Exercises and instructor-attendee interaction typically as long as attendees want.

Course Location

The Savoy Hotel
580 Geary Street
San Francisco
California 94102
Tel: 1.866.445.8102

Travel & Leisure Magazine says the Savoy Hotel is: "One of the ten best small hotels in the city that invented them."

Accommodation: $139 per night
(please indicate “ PEST Course Attendee” to obtain block rate)

Continental breakfast is included in the course fee, together with mid-morning and mid-afternoon coffee and cookies.

Registration Fees (prices assume laptop provided by attendee)

Early-bird Discount: receive a 10% discount by signing up before 30 th June, 2006!
$1295 Private Industry/ non-GRA
$1125 GRA member
$1210 Registration plus GRA Membership
$825 Full time student or state employee.
$200 Laptop rental (please indicate to GRA at registration time)
Predictive Analysis Soup-to-Nuts Morning Session
$150 for current PEST course attendees
$245 for previous PEST course attendees

> Register For This Event

Additional Information

Go to http://www.sspa.com/pest for more information on PEST , together with the latest version available for download. Contact Mathew Tonkin at pest@sspa.com for information about the course content or if any difficulties are encountered booking accommodations, etc. For all other information, contact Mary Megarry at mmegarry@nossaman.com or 916-446-3626.