Formatted Title

Optimization of Groundwater Remediation Design and Performance

Background/Objectives

There are a number of complexities related to the selection, design, and application of groundwater-restoration remedies that can be addressed through remedy optimization. Remedy optimization includes many definitions ranging from (1) any incremental improvement to (2) engineering evaluations of existing systems performance with recommended actions to (3) numerical modeling linked with optimization algorithms to improve remedy performance. For this paper, optimization refers to linking numerical modeling of groundwater flow/contaminant transport models with optimization algorithms to determine the most cost-effective approach for remedy optimization while incorporating site access restrictions and management considerations (e.g., budget, schedule, sustainability, etc.). Optimal remedial designs consider multiple objectives including contaminant mass and water volume removal, hydraulic control, cleanup time, and cost. This approach allows for many thousands of possible remedy configurations to be evaluated to determine which configuration of the remedy achieves cleanup goals, typically for the lowest cost and/or shortest remedy lifecycle. Recommendations are developed for optimal configuration of key elements, including well locations, pumping rates, and operational approaches that lead to timely and cost-effective remedial strategies while still achieving target cleanup goals. The primary objectives of this paper are to illustrate the application of this remedy-optimization approach at groundwater-contamination sites and share lessons learned during efforts to evaluate individual and combined groundwater-restoration remedies.

Approach/Activities

The general approach for applying optimization algorithms to groundwater remedies is to (1) update or develop the conceptual site model (CSM) with recent data, (2) calibrate a numerical groundwater flow and contaminant transport model based on the updated CSM, and (3) link the calibrated model to the optimization algorithm. The optimization algorithm is used to both guide and evaluate the population of individual candidate remedy configurations to see which configuration provides the best approach to groundwater restoration as defined by the constraints and objective functions of the design problem. Each remedy configuration requires a complete flow and transport simulation. Given the large number of simulations required, modeling is executed on an internal computing cluster or on the Cloud. The approach is illustrated through recent optimization efforts at different groundwater-contamination sites with remediation technologies including a pump-and-treat system comprised of up to 74 existing system wells; an in situ treatment system; a combined pump-and-treat and in situ treatment system utilizing up to 119 treatment system wells; and a pump-and-treat system incorporating mass transfer from immobile to mobile domains.

Results/Lessons Learned

Unlike in traditional trial-and-error optimization approaches where only a handful of scenarios are subjectively considered, the results of numerical optimization in this approach include the ability to evaluate the performance of thousands of possible remedy scenarios (e.g., cost versus percent cleanup and time to clean up) across each modeled remedy configuration and to illustrate remedy progress through time, both critical elements of communication to stakeholders to gain approval when tens of millions of dollars will be required for remediation. Integration of subject matter experts (e.g., remediation engineers, geologists, hydrogeologists, risk assessors) during CSM updates and development of potential remedial technologies can be critical to ensure proper remedy representation within the numerical model and for the formulation of the optimal design problem. Often, key site-specific data such as hydraulic conductivity, mass distribution, sorption coefficients, and natural degradation rates are not sufficiently available, which introduces uncertainty into the calibrated model used in remedy optimization. This uncertainty can be addressed through a robust-design approach incorporating equally-likely realizations of site parameters into the optimization process and communicated to stakeholders. Access to cloud computing means no problem is too big, but cloud time and memory can be expensive, which needs to be managed by fully exploiting the parallelization potential of the models involved in remedy-optimization. A process model, including input/output files for the final remedy configuration, can be set up and transmitted to clients for their or their contractor’s implementation over the remedy lifecycle to modify the design as performance data become available. This paper will demonstrate related results and lessons learned through the application of the numerical remedy-optimization approach at various groundwater-contamination sites.