Model Calibration, Uncertainty Analysis and PEST - A Brief Tour
This short course is targeted at both modellers and those who use the results of models to set policy and make decisions.
Its purpose is to provide a broad overview of parameter estimation and uncertainty analysis in groundwater modelling.
While examples will focus on PEST, and an introduction to PEST and its supporting software will be provided,
the intention of the course is to discuss the principles behind PEST rather than its working details.
A number of broad subject areas will be covered, including why there is so much nonuniqueness in estimating parameters for groundwater models,
and why and how numerically stable solutions to the inverse problem of groundwater model calibration problem can nevertheless be obtained.
It will be shown that calibration should be only the first step in the history matching process,
and that calibration-constrained uncertainty analysis should be the second.
The idea that a groundwater model (or any environmental model) can tell us what will happen in the future is self-evidently wrong.
However a great deal of model-based environmental decision-making seems to imply that this is the case. Instead,
a model should be asked to tell us what will NOT happen in the future, and how unwanted outcomes of planned management
strategies can therefore be avoided. When a model is used in this way, implementation of the scientific method and the needs of decision-makers
coincide. The course will show how the partnership of a model with high-end parameter estimation and uncertainty analysis software
such as PEST can provide a much needed basis for scientifically-based decision-making.
Trainer: John Doherty
Training fee: 300 €, with current SMA 270 €