Groundwater Level Modeling Using Multiobjective Optimization with Hybrid Artificial Intelligence Methods

Estimating groundwater level (GWL) fluctuations is essential for integrated water resource management in arid and semiarid regions. In this study, we used hybrid evolutionary algorithms to promote the multilayer perceptron (MLP) learning process. A hybrid metaheuristic algorithm was applied to overcome MLP difficulties in the learning process, including low conversions and local minima. Additionally, the hybrid model benefited from the advantages of two objective function procedures in obtaining MLP parameters resulting in a robust model regardless of over- and underestimation problems. These algorithms include the nondominated sorting genetic algorithm (NSGA II) and the multiobjective particle swarm optimization (MOPSO) algorithm in different combinations, including MLP-NSGA-II, MLP-MOPSO, MLP-MOPSO-NSGA-II, and MLP-2NSGA-II-MOPSO. Temperature, precipitation, and GWL datasets were used as model input candidates in various combinations with different delays. Finally, the best model inputs were selected using the coefficient of determination (R2). In summary, the contribution of the paper is the development of a robust model for estimating groundwater level fluctuations in arid and semiarid regions through the application of hybrid evolutionary algorithms, careful selection of input parameters, and the identification of a superior model that combines the advantages of multiple optimization techniques. The input parameters include temperature and precipitation with delays of 3, 6, and 9 months and GWL with delays of 1 to 12 months. In the next step, the performance of the different combinations of the MLP and hybrid evolutionary algorithms was evaluated using the root mean square error (RMSE), correlation coefficient (R), and mean absolute error (MAE) indices. The evaluation outcomes revealed that the MLP-2NSGA-II-MOPSO model, with RMSE=0.112, R2=0.97, and MAE=0.095, outperforms the other models in estimating GWL fluctuations. The selected model benefited from the advantages of both the MOPSO algorithm and NSGA-II regarding accuracy and speed. The results also indicated the superiority of multiobjective optimization algorithms in promoting the MLP performance. This study's outcomes not only advanced the field of GWL prediction but also offer practical insights crucial for sustainable water resources management in arid and semi-arid regions.