Stochastic Cost Optimization of DNAPL Remediation – Field Application

Document Type


Publication Date


Publication Title

Environmental Modelling and Software


A stochastic remediation design optimization methodology implemented in the program Stochastic Cost Optimization Toolkit (SCOToolkit) was successfully applied to evaluate remediation options at the East Gate Disposal Yard (EGDY) at the former Fort Lewis, now Joint Base Lewis-McChord (JBLM), Washington. Non-optimized forward simulations based on calibrated parameters and their uncertainty inferred from data prior to actual thermal source remediation system implementation at the site indicated a low probability of the actual thermal system design meeting remediation criteria in a reasonable time frame. Calibration using additional data collected during thermal treatment reduced prediction uncertainty, but still predicted a high probability of taking more than 100 years to reach compliance criteria using the actual thermal treatment design with no additional remedial action. Stochastic optimization of the thermal treatment design indicated larger treatment areas were needed to capture source mass due to uncertainty in source delineation. The expected cost for the enlarged thermal treatment system was estimated to be $22M, which is nearly twice that of the actual system, suggesting that additional characterization to reduce source delineation uncertainty or consideration of an alternative strategy that is less sensitive to delineation uncertainty may be warranted. Stochastic optimization of whey injection was investigated to accelerate source zone dense nonaqueous phase liquid (DNAPL) dissolution and enhance dissolved plume biodecay. The optimized design indicated a 93% probability of meeting compliance criteria by 2100 with an expected net present value cost of $4.7M. Whey injection substantially shortened the remediation time compared to no whey injection. The results indicate that the proposed stochastic cost optimization approach is able to reduce expected remediation costs, increase the probability of achieving remediation objectives, and identify data characterization needs. © 2012 Elsevier Ltd.