Inversion, Uncertainty Analysis and Problem Decomposition in Decision-Support Modelling

  • Dates: 02-06 September, 2024 (just before the IAH Davos conference)
  • Speaker: Dr. John Doherty
  • Location: University of Neuchâtel, Campus Unimail (Room: tbd), rue Émile-Argand 11, Neuchâtel, Switzerland.
  • Cost
    • PhD student members of the Water-Earth Systems PhD School: no fee
    • External PhD students, other academic members, and those in government and industry: please contact School.Earth-Water@unine.ch
  • Registration: please fill in the REGISTRATION FORM
  • Questions: please contact School.Earth-Water@unine.ch

Course Description

While this course is similar to previous Neuchatel courses on model calibration and uncertainty analysis, it differs in some important respects.

The course concentrates on some newer concepts and technologies that are now available through the PEST and PEST++ suites. At the same time, it attempts to meet participants where they are at in their modelling journeys by discussing their concerns, and how these new technologies can meet these concerns.

Decision support environmental modelling is as much of an art as it is a science. The course addresses both of these aspects of environmental modelling through both formal and informal exchanges through which all of us can learn from each other.

Parameter estimation and uncertainty analysis are the vantage points through which the efficacy of environmental modelling can be judged. But first “efficacy” must be defined. This can have different meanings in different modelling contexts. Nevertheless, for decision-support modelling, the metrics are generally clear. These metrics will be defined. The course will then explore ways in which these metrics can be served in everyday modelling practice.

Formal discussions will describe the following technologies.

  • Model calibration through highly parameterised, regularised inversion
  • Relationships between real and estimated hydraulic properties
  • Uncertainty analysis using ensemble methods
  • How to define useable prior parameter probability distributions
  • Flexible, adjustable, non-stationary geostatistics and their use in model predictive uncertainty analysis
  • Data space inversion
  • Complementary use of multiple history-matching and uncertainty analysis methods
  • Predictive uncertainty analysis where models are large and complex
  • Appropriate model simplification

Hands-on workshops will complement the discussions.

More informal discussions will attempt to address the individual needs of participants. Each modelling context is different. While certain rules apply to all of them, strategic application of those rules is context-specific. Meanwhile, community expectations on modelling are high. We explore whether all of those expectations are realistic, and how to engage in discussions with modelling stakeholders.


The class will last for five days. However discussions that begin in the course can continue for weeks or months after the course if this helps. An outline of the course content can be downloaded here or directly from the PEST website.


Bring a laptop (with Windows), an open mind and a willingness to exchange new ideas.