2007 GSA Denver Annual Meeting (28–31 October 2007)

Paper No. 2
Presentation Time: 1:55 PM

PREDICTIVE MODELING AND PROJECT PERMITTING: USING INVERSE METHODS TO HELP PROVIDE CONTEXT FOR DECISION MAKING


CARLSON, Christopher P., Watershed and Minerals & Geology Staffs, USDA Forest Service Washington Office, 1400 Independence Avenue SW, Washington, DC 20250, ccarlson@fs.fed.us

Ground water flow and transport models are increasingly used as inputs to public agency decision-making processes for large-scale and often highly controversial projects with potentially substantial environmental impacts. Given the complex systems that can be involved, such modeling is often the only reasonable tool available for estimating potential project impacts and the efficacy of potential mitigation measures. Though most conscientious practitioners are well aware of the inherent limitations of their work, many decision makers view models as providing them with the answer. In their minds, the model takes them “off the hook” politically (and potentially legally), regardless of the whether the final decision is for or against the proposed project. This misconception, combined with a lack of focus on the part of modelers to fully disclose the assumptions, limitations, and ensuing uncertainty of their results, has often led to inadequate public disclosure of the range of potential impacts and, almost certainly, to many poorly informed decisions.

This has to change. Full disclosure of limitations and uncertainty, including a range of reasonable potential results where appropriate, should be the way ground water modeling supporting public agency permitting processes is conducted and reported.

Recent developments in the application of inverse methods to ground water modeling has the potential to begin to change the status quo. By automating some of this work, incorporation of inverse methods in both calibration and prediction can make the presentation of uncertainty and ranges of results the standard approach in this field. Why should modelers provide their clients the one number the client thinks they want (the proverbial "answer"), when we all know it is or will be wrong and the tools at their disposal readily allow them to put that number into appropriate context? Modelers should be erring on the side of disclosure, and inverse methods can be the tool that moves us a long way toward this goal.