Risk-Based Decision Making A Framework and Automation Tool for Guiding Site Characterization
Project Description and Significance
The Sandia Environmental Decision Support System (SEDSS) project is developing a new probabilistic methodology for site safety or remedial options and incorporating this methodology and existing ones into a decision framework to accommodate the effects of uncertain site-specific information and to facilitate consistent decision making.
To accomplish this objective, the SEDSS project has set the following goals:
Create a robust decision analysis framework (referred to here as the decision framework) and framework components that allow continued improvement of the decision process.
Automate the decision framework into a user-friendly software tool that will result in a consistent application of the methodology and a documented trail of assumptions that can be the basis for negotiations. The use of the tool will allow the decision-making process to become faster, more consistent, better documented, and less costly.
Identify and establish working agreements and coalitions with critical parties (e.g., laboratories, universities, companies) developing significant research that may contribute to the SEDSS decision framework.
Sandia's Contribution
SEDSS utilizes an iterative approach which incorporates probabilistic risk analysis, cost-benefit analysis, and site sampling optimization techniques. Quantitative estimates of risk to human health and the environment are based on probabilistic analysis of site conditions that provides an explicit measure of the uncertainty associated with the decision to be made. Results of the risk analysis can be used to direct additional site characterization and monitoring activities and to provide a consistent framework for comparing concepts for alternative systems conceptualizations. With the iterative approach, new data are used to repeatedly update the analysis and reduce uncertainty. To support the decision analysis, the SEDSS is designed to:
Capture and store user input data and link to GIS site information
Guide the user through the process of developing a site conceptual model based on available information
Evaluate the conceptual model with physical process models
Allow explicit comparison of results with a performance objective
Identify additional data needs
The SEDSS methodology can be applied to a variety of environmental and exposure pathways for both radionuclide and hazardous contaminants, and can be used with any form of site characterization and monitoring. With the capabilities provided, decision makers can more effectively address the following types of environmental decisions:
What are the environmental and human health risks posed by a contaminated site?
What is the best approach for remediating the site or establishing containment?
How can it be determined and demonstrated to the regulator that remediation is complete or containment is effective?
Is a newly proposed waste disposal facility likely to comply with applicable regulatory criteria?
Where should monitoring wells be placed to optimize contaminant detection?
Where should resources be applied to effectively address critical problems?
The first two components of the proposed SEDSS framework (shown in the figure) consist of (1) defining the regulatory drivers or performance measures, and (2) assessing existing knowledge about a site. After this preliminary work, (3) a conceptual model of the site is constructed and translated to models that are physics- and chemistry-based. Using initial estimates of uncertainty in model parameters garnered from either field data or expert opinion, (4) the physical models are run in a Monte Carlo simulation. (Monte Carlo simulation consists of running a computer code multiple times, using input parameters that vary between runs. The parameters are usually randomly selected from a predefined statistical distribution. This method propagates uncertainty in the values for model input parameters to provide a distribution of possible model outputs that represent the range of output uncertainty.) (5) Then the question is posed, "Given the uncertainty (spread) in the output response, do I have enough information (confidence) to say whether the site meets the performance measures?" If so, the site is compared with the performance measures, and a regulatory decision is made; if not, then more data must be collected. If more data are required, the next step involves (6) a sensitivity analysis to determine which changes in model parameters (associated with field data) would have the greatest effect in changing model results. Given a list of important parameters, (7) a data-worth calculation is used to indicate how these data would be necessary to dictate a change in the model output and estimate the cost of collecting these data. The decision maker must then ask if collection of that data is cost effective. Assuming it is, statistical optimization is used to select data-collection sites and (8) how data are collected. Finally, the site conceptual and numerical models are revisited and adjusted based on the newly acquired data. This repetition of step (3) begins a new iteration of the decision framework.
This framework can be applied to many different applications by changing the measures used to assess performance. For example, the performance measure for the question, "Is the site safe?" might be phrased as: "Am I 95% confident that the cumulative human health risk for a given site is less than 10-6 excess cancers," or "Am I 95% confident that the concentration of a given contaminant is below its regulation-defined, maximum concentration limit (MCL)?" For the question, "What remediation scheme should be used?" the performance measure may be phrased as, "Do I have a certain level of confidence that a specific design for a remediation scheme (i.e., pump-and-treat) will reduce the human health risk below the defined regulatory limit, and is this the most cost-effective remedial approach that can be applied to my problem?" The final question has a performance measure much like the first inquiry: "Do I have a certain level of confidence (can I defend) that the remediation scheme applied to the site has reduced risk or contaminant concentration below its regulatory limits?" This includes the interim question, "Is my current monitoring well network adequate to capture a contaminant plume if a release occurs?"
The SEDSS program has been developed as a strong object-oriented software architecture that will allow future expansion. It was designed using CASE tools, and the entire development process uses industry standard software quality assurance and configuration management techniques.
Release 1 of SEDSS will begin its beta test in late 1996. At that time the software will have automated the decision framework for setting performance objectives, assimilating existing information, defining the conceptual model, running the appropriate codes in a Monte Carlo process, and displaying the uncertainty decision point that will allow the user to determine if a decision can be made or if it is necessary to continue through the rest of the decision framework steps.
Future Work
Potential work for the next fiscal year includes:
Completing the remainder of the decision framework loop (both the scientific methodology and the automation for sensitivity analysis, data worth/cost analysis, and decision point as to whether additional information is needed)
Adding more complex numerical analytic codes and more conceptual models
Submitted October 1996