Model Calibration and Predictive Uncertainty Analysis using PEST

Kursusindhold

 

PLEASE NOTE         

The PhD course database is under construction. If you want to sign up for this course, please click on the link in order to be re-directed. Link: https://phdcourses.ku.dk/nat.aspx

 

Topics covered on each day are listed below. It is anticipated that the days will be divided between workshops and practical sessions. However this can be varied according to participants’ desires.

Practical exercises will be provided so that attendees can gain experience in the use of PEST. However if course participants would prefer to work on their own models, and discuss these models with the presenter, that will be fine. In fact half of the last day will be set aside for just this purpose.

The course will be as informal as possible, with plenty of time provided for discussions on topics of interest, and for repeating parts of the course material that may not have been well understood on the first time through.

DAY 1: Introduction and Parameter Estimation Basics

  • introductions
  • outline for next five 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
  • 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 to the current inversion problem
  • PEST, Parallel PEST and BEOPEST

 

DAY 2: Ill-Posed Problems and Highly Parameterized Inversion

  • 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
  • examples

 

DAY 3: A: Practical Groundwater Model Calibration

  • 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
  • groundwater modeling utility support software available with PEST
  • PEST and MODFLOW-USG
  • the PLPROC utility

 

DAY 3: B: Practical Surface Water and Land Use Model Calibration

  • lumped-parameter and distributed-parameter models
  • decomposition of a flow time series into its components
  • formulation of a multi-component objective function
  • high-pass, low-pass and baseflow filtering
  • regionalization of model parameters
  • the role of expert knowledge
  • regularization strategies
  • simultaneous calibration of multiple models
  • multiple optima and parameter nonuniqueness
  • the TSPROC model postprocessor and pest preprocessor

 

 

DAY 4: Uncertainty and Sensitivity 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

 

 

DAY 5: Working with Defective Models – and Conclusions

  • 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

Many sessions during the course will be devoted to workshops through which participants can gain experience in using PEST. Files and printed notes for one of these workshops will be installed on participants’ laptops during the first day of the course. Participants can then chose what workshop they would like to do next; workshop files and notes can then be transferred from the memory sticks provided at the beginning of the course.

 

Alternatively, participants may wish to download the following workshop from the PEST web pages at:

 

http://www.pesthomepage.org/

 

If so, it is suggested that documentation for this workshop be printed after downloading and brought to the course.

 

Another alternative for those who use Groundwater Vistas is to do the workshops that are provided with this software. If so, workshops notes should be printed out ahead of time and brought to the course.

 

Yet another alternative is to bring along your own model calibration dataset and work with that during the practical sessions, asking questions as required.

 

Memory Sticks

Memory sticks provided to course participants also include the following.

  • latest versions of PEST and BEOPEST
  • literature on PEST
  • copies of all slideshows used in the course

over 14 PEST workshops (with all files and associated workshop documentation

Engelsk titel

Model Calibration and Predictive Uncertainty Analysis using PEST

Lectures, computer exercises, presentations

ECTS
5 ECTS
Prøveform
Andet
  • Kategori
  • Timer
  • Forberedelse
  • 45
  • Forelæsninger
  • 16
  • Øvelser
  • 30
  • Praksishold
  • 8
  • Undervisningsforberedelse
  • 10
  • Studiegrupper
  • 10
  • Total
  • 119