Groundwater parameter estimation using multiquadric-based meshfree simulation with covariance matrix adaptation evolution strategy optimization for a regional aquifer system

This study proposed a novel combination of a meshfree simulator with the covariance matrix adaptation evolution strategy (Mfree-CMA-ES) to construct a simulation-optimization (SO) model for accurate estimation of aquifer parameters. In most regional aquifer systems, minimal temporal fluctuations are observed throughout the year. Therefore, the widely used approaches of least squares error minimization with metaheuristics (MH) optimization, such as differential evolution (DE), particle swarm optimization (PSO), and their hybrid versions, are often prone to premature convergence and lead to incorrect estimations of aquifer parameters. In the proposed model, the Mfree simulator was used, which produces accurate values of groundwater head, particularly when compared to the popularly used grid or mesh-based finite difference method (FDM) or finite element method (FEM). Furthermore, CMA-ES, with its auto-updated-strategy parameters, provided a competitive resultant population in each generation, leading to fast convergence and excellent matches with observed head data. The developed SO model was applied to both a synthetic regional aquifer and a real field case, i.e., the Mahi Right Bank Canal aquifer in India. Results showed that Mfree-CMA-ES requires less computational time and has a greater accuracy of estimated parameter values in comparison to various other models, i.e., FEM-DE, Mfree-DE, FEM-PSO, Mfree-PSO, FEM-DE-PSO, Mfree-DE-PSO, and FEM-CMA-ES. A sensitivity analysis showed the relative composite scaled sensitivity (RCSS) values of all the hydraulic conductivity zones that are within the range of 1–0.1. These obtained RCSS values validated the ability of the proposed model to estimate zonal hydraulic conductivity values using the Mfree-CMA-ES model.