Efficient

A New Framework for Adaptive Sampling and Analysis During Long-Term Monitoring and Remedial Action Management
B. S. Minsker, B. Bailey, A. Valocchi, R. Johnson, D. Tcheng, M. Welge
Sponsor: Department of Energy Environmental Management and Science Program

Introduction and Objectives: DOE and other Federal agencies are making a significant investment in the development of field analytical techniques, nonintrusive technologies, and sensor technologies that will have a profound impact on the way environmental monitoring is conducted. Monitoring and performance evaluation networks will likely be based on suites of in situ sensors, with physical sampling playing a much more limited role. Designing and using these types of networks effectively will require development of a new adaptive para-digm for sampling and analysis of remedial actions, which is the overall goal of this project. Specifically, the ob-jectives of this project are to: (1) enable effective interpretation of non-intrusive monitoring data, (2) improve predictions and assessment of remediation performance, (3) develop decision rules for on-site adaptive sampling and analysis, and (4) enable more informed decision making and risk analysis of long-term monitoring systems.

Research Plan and Methodology: These objectives will be accomplished through development of a new framework for adaptive sampling and analysis, decision making, and risk assessment. The framework assumes that existing data have been used to develop some type of conceptual model for the site, be it an analytical, nu-merical, or statistical model. This project will test the effectiveness of new hierarchical models that integrate all sources of process knowledge at the site, including geological data collected during site characterization, meas-urements of parameters collected during routine groundwater monitoring (including surrogate data and data at different spatial and temporal resolutions), and scientific understanding of biochemical transformation processes. An interactive model and parameter identification system will also be developed for creating the new models effi-ciently. The models would then be used to predict future measurements, to identify initial sampling plans, and to create decision rules for adaptive, on-site sampling and analysis. During a particular sampling episode, initial measurements would be compared with predictions and statistically significant deviations would be identified. The decision rules would then be used to identify further sampling needs. This process would be repeated until sufficient data have been collected to verify any anomalies observed and reduce uncertainties to acceptable levels. At the end of the sampling period, or during periodic reviews, the monitoring objectives, conceptual model, initial sampling plans, and decision rules would be re-evaluated and updated to incorporate knowledge obtained from the new sampling data and to accommodate changing site needs. Methods will be developed for automated updating of these components of the framework (with or without user interaction) to ensure efficiency and cost-effectiveness.

Our framework will be tested using data from the 317/319 area at Argonne National Laboratory, where a traditional pump-and-treat and novel phytoremediation system are in place to control off-site migration of a trit-ium and VOC plume. To enable rigorous testing and comparison of methods, the framework will initially be tested using synthetic data generated based upon historical trends observed at the site. Once the initial testing and methods development is completed, the framework will be validated using the most recent available data from the 317/319 area and used to identify: (1) long-term mass measurement trends, (2) future data needs for assessing the performance of the pump-and-treat and phytoremediation systems, and (3) key measurements for model valida-tion.

Relevance to DOE: In a report to Congress, DOE estimated that it would spend approximately five and a half billion dollars on long-term site management (“stewardship”) over the six years between 2000 and 2006 (DOE, 2001). The re-port also states that long-term stewardship costs will be more than 100 million dollars per year over the next 70 years. With these significant costs, it is critical that long-term monitoring and management be as efficient and effective as possible. The framework proposed in this project is expected to contribute substantially to reducing costs and increasing effectiveness of data collection and analysis. Specifically, this project will address the fol-lowing subsurface contamination research needs stated in this notice inviting grant applications:

· Conceptual modeling: Innovative methods will be developed for efficiently integrating all process knowl-edge at all scales into conceptual models. The methods will be tested and validated at a DOE site, ensuring relevance to the kinds of subsurface environments and contamination problems encountered at DOE sites.

· Monitoring and validation: The methods developed in this project will be created to: (1) design monitoring systems to detect both current conditions and changes in system behavior, (2) allow validation and updating of models using present-day measurements, (3) identify key measurements for model validation, including surrogate data and data at different spatial and temporal resolutions, and (4) identify trends in measurements over long time periods.