Search

Contaminants in groundwater may enter through various sources which are required to be identified for informed decision-making regarding remediation. In early stages of contamination of an aquifer, the sources of contamination are generally unknown. To estimate the unknown release histories at potential sources, the governing equations of contaminant transport are solved backwards in time which renders the inverse problem as ill-posed. Furthermore, due to measurement or operational errors, observed breakthrough curves can be noisy which can induce uncertainty. Simulation-optimization (SO) techniques are generally used for solving the inverse model. Nevertheless, swarm population-based optimization algorithms may get trapped in local optima in the nonlinear search space related to the problem. In this study, SO model based on meshless radial point collocation method (RPCM) and multiverse optimizer (MVO) is proposed. MVO implements gradual transformation from global exploration to local exploitation phase which helps in better estimation of unknowns. The performance of RPCM-MVO-SO model is compared with two other SO models developed by combining RPCM with grey wolf optimizer and particle swarm optimization in two case studies. The proposed model performs better compared to the other two SO models considered which suggests the efficacy of the model for release history estimation in groundwater. HIGHLIGHTS A novel RPCM-MVO-SO model is proposed for contaminant source identification.; The RPCM-MVO model performs better compared to RPCM-PSO and RPCM-GWO for the two case studies considered.; RPCM-MVO model estimated release intensities have low variations for noisy observations.;