Inscription (closed March 10)
Five Day Short Course
Model Calibration and Predictive Uncertainty Analysis
Dates: 3rd - 7th April 2023
The course will be run by John Doherty (author of PEST) and will cover the broad spectrum of topics (philosophical, theoretical and practical) that embody a scientific response to the imperatives of modern-day decision-support modelling. We have tried to design this week so that the most people can get the most out of it. The atmosphere will be relaxed, with plenty of time for discussion and questions. Social events will be organized on day 1 on the campus and day 4 in Bordeaux city.
Program
Day 1: Introduction and theoretical background
Morning:
Introductory session covering the philosophy of decision-support modelling. It will discuss how decision-support modelling is different from explanatory modelling, and why/how simulation should serve harvesting of information from data, and delivery of that information to decision-makers.
Basic linear algebra and statistics (do not be afraid; it is easy)
Afternoon:
- Solution of well-posed inverse problems
- Basics of PEST ; Interfacing your model with PEST (GUIs, Python, …)
Day 2: The basics of PEST
Morning:
- Regularisation and the quest for uniqueness (manual regularisation, Tikhonov regularisation and SVD)
- Workshop: use of PEST/PEST++ to solve a simple inverse problem with the GLMA
Afternoon:
- Linear analysis, sensitivity analysis, data worth and related topics
- Workshop: highly parameterized regularized inversion with the GLMA
Wine tasting
Day 3: Uncertainty analysis
Morning
- Linear analysis, sensitivity analysis, data worth and related topics
- Workshop: linear uncertainty analysis (with PyEMU)
Afternoon
- Conference : Nonlinear Decision support modeling needed - Conference with local stakeholders and practicioners
- Workshop: using the PESTPP-IES ensemble smoother (with PyEMU)
Day 4: Optimization and data assimilation
Morning
- Decision optimisation and optimisation under uncertainty
- Debate : Dealing with model defects ; art of weighing observations ; fit-for-purpose modeling
Afternoon
- Data space inversion (DSI)
- Workshop: using DSI for predictive uncertainty analysis
Dinner in Bordeaux
Day 5: customized support and group therapy
The content of day 5 will be adjusted to satisfy the expectations of the participants.
Morning
- Model-specific interfaces (FePEST GUI, PyMarthe interface)
- Workshop :A script-based approach for PEST with Python.
Afternoon
- This is where you get to tell us your problems, and we suggest ways in which lessons learned in the previous few days may help you. Or perhaps you can help us (and the rest of the class) by telling us your experiences. Either way, the theme is group therapy!
Practical and administrative information
Course will be held at : ENSEGID - Bordeaux INP (France). [ Campus map - research Bordeaux INP - ENSEGID ]
The course will be led by John Doherty, with support from Alexandre Pryet for practical aspects and workshops.
Course attendance is flexible, Attendees can join us for 1, 2, 3, 4 or 5 days (recommended!)
Full course (5 days): 1400€ (regular participant) - 1000€ (PhD students)
Flexible attendance: 400€ per day
Costs include lunch + 2 social events (ice breaking with a wine tasting session, dinner in Bordeaux).
Contact : alexandre.pryet@bordeaux-inp.fr
Inscription (closed March 10)
Who is John Doherty ?
John Doherty is the author of PEST, a software package that is widely used for groundwater model calibration and uncertainty analysis. He has worked for over 35 years in the water industry, first as an exploration geophysicist and then as a modeller. Over this time, he has been employed by government and industry; he has also worked at a number of universities where he undertook research, and supervised postgraduate students.
These days John works for his own company, Watermark Numerical Computing, undertaking consulting, research, programming and education, mainly on issues related to model deployment in support of environmental management and impact assessment. He also leads the GMDSI initiative – an industry-funded project that assists modellers to undertake scientifically-based decision-support modelling, and non-modellers to understand the contributions that modelling can (and cannot) make to environmental management.
