Developing a Framework for Quantifying and Effectively Communicating Ground Water Model Uncertainty
* Is it possible to develop a framework, used by all Reclamation ground water modelers, that defines and quantifies ground water model output uncertainty and that can be effectively communicated with Reclamation managers and clients?
Ground water models are imperfect representations of reality and have some level of uncertainty inherent in the solution. Increasingly, Reclamation is being asked to investigate ground water and surface water interactions and impacts to ground water due to additional storage facilities, changes in management practices, and climate change using models that have some degree of uncertainty associated with them. Communicating the significance of model output uncertainty to management and clients is a difficult task. This project proposes to develop a methodology Reclamation ground water modelers to use to quantify model uncertainty and a framework for communicating model results and uncertainty to management and clients.
Need and Benefit
As budgets are reduced, Reclamation management and clients will favor the less costly alternative of developing ground water models to investing in more expensive field explorations in order to make decisions regarding reservoir seepage, ground water/ surface water interactions, and climate change effects. However, this places a greater burden on the ground water model. In turn, decisions are made with these tools that have varying levels of confidence (or uncertainty), based on data availability and the complexity of the issue. Therefore, a methodology is proposed that can not only establish a mechanism for quantifying uncertainty in output results, but that can effectively communicate the ramifications of what this uncertainty means to decisions. An additional benefit of this methodology is an opportunity to cost effectively structure and focus field investigations to reduce model uncertainty based on initial results.
Currently, it is possible to quantify the level of uncertainty in model results by various statistical methods and techniques. Model uncertainty can result from the level of confidence in model input data and/or a lack of input data. However, a lack of input data does not necessarily produce a model solution with a low level of confidence. At the current time, Reclamation does not recommend a standard for quantifying uncertainty in its ground water models for the purposes of interpreting output.
In a recent reservoir seepage study, ground water modelers were asked to develop a model to determine the amount of potential seepage from the reservoir and the seepage path. The lack of input data resulted in multiple unique solutions to determine how much seepage would be directed toward a nearby contaminated site. It was important for the modelers to explain to Reclamation management that, given the lack of input data, there was not a single seepage value that could answer the question. To better explain the uncertainty, the modelers ran a Monte Carlo simulation of a specific area within the model to determine the range of possible solutions.
A traditional sensitivity analysis was performed on another model developed by Reclamation. This analysis allowed the modelers to communicate to the client to what degree they could utilize results from the model. The modelers felt that, with the type of uncertainty analysis that was performed, the model could only be used to look at differences of head conditions from the base case and that absolute head conditions calculated were within 20 feet of actual head conditions.
In both models, Reclamation management and/or clients would have benefited from a more detailed uncertainty analysis. Since Reclamation does not have a standardized method to calculate uncertainty nor a requirement that uncertainty be calculated, uncertainty is not always calculated or fully understood. Consequently, management and clients receive results with confidences that vary widely. Since the models are used for decisionmaking purposes, it is important that the decisionmakers understand the context of the results and how they can be used.
Contact the Principal Investigator for information about partners.
Bureau of Reclamation Review
The following documents were reviewed by experts in fields relating to this project's study and findings. The results were determined to be achieved using valid means.
Framework for Understanding Ground Water Model Uncertainty (final, PDF,
By Jennifer Johnson
Publication completed on September 30, 2010
A methodology that would be used by all Reclamation groundwater modelers to quantify the uncertainty in groundwater model outputs. (final, PDF,
By Jennifer Johnson
Report completed on June 28, 2011