Optimization of Groundwater Long-Term Monitoring Network with Ant Colony Optimizatio

Groundwater remediation is conducted in polluted sites to remove contaminants and to restore ground water quality. After remediation goals are achieved, long-term groundwater monitoring (LTM) that can span decades is required to assess the concentration of residual contaminants and to avoid the risk of human health and environment. On large remediation sites, the cost for maintaining a LTM network, collecting samples, conducting water quality lab analysis can be a significant, persistent and growing financial burden for the private entities and government agencies who are responsible for environmental remediation projects. LTM network optimization offers an opportunity to improve the cost-effectiveness of the LTM effort while meeting data accuracy requirements. The optimization includes identifying the redundancy in the monitoring network, and recommending changes to protect against potential impacts to the public and the environment.;This study develops a variant ant colony optimization (VACO) method, using ordinary kriging (OK) or inverse distance weighting (IDW) for data interpolation, to identify optimal LTM networks that minimize the cost of LTM by reducing the number of monitoring locations with minimum overall data loss. ACO is a global stochastic search method inspired by the collective problem-solving ability of a colony of ants as they search for the most efficient routes from their nests to food sources.;The performance of ACO variant (VACO) developed in this study is evaluated separately in two test cases. In the first case, VACO is used to solve a simplified traveling sales person problem. In the second case, both enumeration method and VACO are employed for optimization of a synthetic long term monitoring network of 73 wells generated from a groundwater transport simulation model. The two sets of test show that the VACO performs well for optimization problems. The VACO is finally adopted for the optimization of a long term monitoring network of 30 wells in Logistic Center, Washington, with the data interpolation methods of inverse distance weighing, ordinary kriging, and modified inverse distance weighing which is developed in this study. The optimization results are analyzed and group of ideal redundant wells identified. The conclusion of this study is summarized at the end, and future work is suggested.