Groundwater Modelling Decision Support Initiative (GMDSI)
Lead collaborating organisations: BHP, Rio Tinto, and Flinders University
About GMDSI
GMDSI Website - https://gmdsi.org
GMDSI is an industry-funded and industry-aligned project that is focussed on improving the role that groundwater modelling plays in supporting environmental management and decision-making.
Environmental modelling technology is improving all the time however, putting this technology to work in contexts where data are scarce and simulator forecasts cannot be made with accuracy is a challenging undertaking. Nevertheless, decisions must be made, and risks must be assessed. Hence all available data must be assimilated into computer models so that the potential for management to go wrong can be exposed before a project begins and can be accommodated through strategic adaptive management as a project evolves.
GMDSI recognizes the important role played by model-value-adding software in decision support. These are computer programs that work with simulators (i.e. models) to quantify uncertainties in their predictions, reduce these uncertainties through assimilation of pertinent data, and optimize management outcomes subject to management constraints.
GMDSI’s mission is to promote, facilitate and support the improved use of models in groundwater management, regulation and decision-making. This will be achieved through the following means:
- Promulgating increased industry proficiency in decision-support modelling;
- Promoting industry-wide discussions leading to improved perspectives on the role of modelling in groundwater protection and management;
- Providing the means to achieve these aims through promoting and facilitating the use of model-value-adding software.
GMDSI project activities follow five themes. These are as follows.
Education
GMDSI will develop downloadable, on-line educational videos on topics that deepen understanding and enable implementation of decision-support modelling. Our target audience will span the breadth of modelling skill and experience. It will include those who are new to modelling, those who must regulate modelling or are stakeholders in model-based decisions, and high-end modellers who wish to implement the latest state-of-the-art data assimilation and uncertainty analysis techniques.
GMDSI will also host courses on a range of topics, from technical to philosophical, that fall under the umbrella of decision-support modelling.
Outreach
GMDSI will host an active website that informs the public of its activities, and that provides a forum for discussion of important modelling issues. Issue papers, reports on GMDSI activities, educational videos and news items will also be downloadable from this site.
From time to time GMDSI will host an open workshop on a topic that is vital to model-based decision-support. Our initial workshop was held in Canberra in July 2019; its topic was model uncertainty. This was very well attended, and feedback was positive. For more details click here
GMDSI’s next workshop (planned in the first half of 2020) will focus on appropriate model complexity.
Some GMDSI outputs will be submitted for publication in peer-reviewed scientific journals. Others will appear as articles in popular publications and news websites that serve the water industry.
Worked Examples
In collaboration with government and industry bodies, GMDSI will build models and use them to support decisions. In doing so, we will implement decision-support modelling strategies that assess and reduce management risks. We will demonstrate the use of data assimilation and uncertainty analysis methodologies and software, and record details as we go so that others can follow. We will “think aloud” when deciding on an appropriate level of model complexity.
Research
GMDSI will sponsor three Ph.D. students. These students will come from industry and work on industry-salient topics, possibly in conjunction with at least one of GMDSI’s worked examples. The aims of this research will be aligned with the GMDSI mission.
Steering Committee
Though industry-funded, GMDSI is managed by the National Centre for Groundwater Research and Training (NCGRT). It is advised by a steering committee composed of the following individuals:
Chairperson: Blair Douglas, BHP
Members: Keith Brown - Rio Tinto, Craig Simmons - NCGRT/Flinders University, Wendy Timms - Deakin University, Liz Webb - EMM,Consulting Pty Limited, John Doherty - Watermark Numerical, Peter Baker, Ty Ferre – University of Arizona, Randy Hunt, and Fiona Adamson NCGRT/Flinders University.
National and International Collaborators
GMDSI project personnel work with government, university and industry personnel from Australia and overseas. Existing and potential international collaborators are affiliated with government, university and industry organizations in the USA, UK, Chile, Italy, Switzerland, and France. It is expected that this list will grow over time.
Funding Bodies
The GMDSI project is jointly funded by BHP and Rio Tinto as a service to the water industry.
Keep in Touch
Engagement & Outreach
Webinar Series: GMDSI Decision-Support Modelling: Principles and Practice
Four part webinar series over four weeks. Commencing on Wednesday 20th May. Register Here for the Q & A Session
Webinar 1 - Groundwater Modelling for Decision Support: Concepts and Fundamentals
View webinar recording - Click here. Power point slides - Click here
Webinar 2 - Repercussions for Model Construction, Deployment and Reporting
View webinar recording - Click here. Power point slides - Click here
Webinar 3 - Overview of Data Assimilation Technology – Part 1: Calibration and Linear Analysis
View webinar recording - Click here. Power point slides - Click here
Webinar 4 - Overview of Data Assimilation Technology – Part 2: Uncertainty Analysis and Optimization
View webinar recording - Click here. Power point slides - Click here
Q & A Webinar - GMDSI Decision-Support Modelling: Principles and Practice - View Webinar Recording - Click here
Groundwater Modelling Uncertainty Workshop: Groundwater Modelling Uncertainty Workshop - Canberra, 22nd July 2019.
Media Release: Global Miners Back Groundwater Modelling Project
Educational Videos presented by Dr. John Doherty
The videos that are presently accessible through this page were prepared by John Doherty. John is the author of PEST and a participant in the GMDSI project. While these videos will be of interest to users of PEST and PEST++, their subject areas are much broader than this. In fact, most of these videos are of general interest, for they describe the theoretical and conceptual outcomes of history-matching, however, it is achieved.
It is our intention to add to these videos over time, with contributions by different modellers on different topics. Some of our planned videos will continue to target users of model-value-adding packages such as PEST, PEST++, and their supporting suites. Others will be of a more general nature. All will address, in one way or another, the way that models are used in decision support.
This video provides an overview of PEST and the three families of utility software that accompany it. It also provides a brief discussion of the demands of decision-support environmental modelling, for this is the context in which PEST must operate.
This short video discusses what it means to calibrate a groundwater (or other) environmental model. Calibration implies uniqueness. The quest for uniqueness is not a quest for truth. Uniqueness of solution of an ill-posed inverse problem requires regularization. If properly applied, regularization yields a solution to that problem which is of minimized error variance. However the potential for error in a calibrated parameter field, and in many predictions made by a calibrated model, can still be high.
This video provides a short reminder of some aspects of matrix and vector algebra that we all learned at school but have since forgotten. It does the same for a few basic statistical concepts. It may be useful to watch this video before watching some of the videos listed below.
This video shows how parameters can be estimated when model calibration constitutes a well-posed inverse problem. Unfortunately, in groundwater modelling, well-posedness does not occur unless preceded by manual regularization. This is not recommended practice, for reasons discussed in the video. Nevertheless, the discussion provides a solid foundation for other videos in this series which focus on solution of ill-posed inverse problems. Of particular interest is the discussion on how to extend linear inverse theory to calibration of nonlinear models.
Problems with Manual Regularisation
This video extends the discussion of the preceding video, while laying foundations for ensuing videos. It explains how calibration based on manual regularization may fail to find a parameter field that minimizes the error variance of important model predictions. It also shows why calibration achieved through manual regularization does not provide a good foundation for post-calibration uncertainty analysis, and does not resolve the relationships between estimated and true properties of an environmental system. The “cost of uniqueness” can therefore not be assessed.
Singular value decomposition (SVD) is explained. Also explained is the important role that SVD can play in solving ill-posed inverse problems, and the insights that it can provide into what calibration of an environmental model can and cannot achieve. Important concepts such as the null space are introduced. An explanation of over-fitting is provided; the reasons why this should be avoided are explained.
The importance of Tikhonov regularization in solution of ill-posed inverse problems is discussed. PEST’s implementation of Tikhonov regularization in groundwater model calibration is explained. It is shown how appropriate implementation of Tikhonov regularization can facilitate the use of expert knowledge, and of information forthcoming from site characterization studies, in groundwater model calibration.
Pilot points are often employed as a parameterization device for groundwater models. This is because they can provide a sound basis for highly-parameterized inversion, while restricting parameters to a number which is low enough to allow filling of a Jacobian matrix using finite difference derivatives. This video discusses options for their emplacement, and how Tikhonov regularization is best applied in estimation of pilot point parameters. The video finishes with an example which demonstrates how, despite our best efforts to introduce geologically-meaningful Tikhonov regularization to the groundwater model calibration process, important hydrogeologically-significant structures may remain invisible to it.
Getting the Most out of PEST - Part 1
This is the first video of a two-part series whose intention is to provide PEST users with some information on how to use PEST, and some of its supporting software, to best effect. This video covers the PEST control file, the difference between singular value decomposition and SVD-assist, and easy ways to add Tikhonov regularization to a PEST control file.
Getting the Most out of PEST - Part 2
This second video of a two-part series covers calculation of finite-difference derivatives, defences against model output numerical granularity, some aspects of observation weighting, termination criteria, Marquardt lambda settings, and the use of some important PEST support utilities.
Approaches to improve the value of models and to reduce total project costs
This is the first of a two-part series. It discusses correlated random variables. It shows how knowledge of one such variable conditions estimation of the other, and reduces its uncertainty. These principles are then applied to regionalized random variables to demonstrate the concepts behind random parameter field generation and kriging. The semivariogram, and its relationship to the spatial covariance function, are also discussed.
In this continuation of the first video of this series, links between geostatistics and history matching of groundwater models are explored. The use of geostatistical concepts and tools in model calibration, and in calibration-constrained uncertainty analysis are also discussed. Some shortcomings of traditional geostatistics, and how these have been addressed by newer geostatistical concepts such as multiple point geostatistics, are also explained.