Risk-based in situ bioremediation design using a noisy genetic algorithm

Risk-based corrective action (RBCA) is rapidly becoming the method of choice for remediating contaminated groundwater. In this paper, a management model is presented that simultaneously predicts risk and proposes cost-effective options for reducing risk to acceptable levels under conditions of uncertainty. The model combines a noisy genetic algorithm with a numerical fate and transport model and an exposure and risk assessment model. The noisy genetic algorithm uses sampling from parameter distributions to assess the performance of candidate designs. Results from an application to a site from the literature show that the noisy genetic algorithm is capable of identifying highly reliable designs from a small number of samples, a significant advantage for computationally intensive groundwater management models. For the site considered, time-dependent costs associated with monitoring and the remedial system were significant, illustrating the potential importance of allowing variable cleanup lengths and a realistic cost function.