A hybrid artificial neural network: An optimization-based framework for smart groundwater governance

Given the growing scarcity and strong demand for water resources, the sustainability of water resource management requires an urgent policy of measures to ensure the rational use of these resources. The heterogeneous properties of groundwater systems are related to the dynamic temporal-spatial patterns that cause great difficulty in quantifying their complex processes, while good regional groundwater level forecasts are completely required for managing water resources to guarantee suitable support of water demands within any area. Water managers and farmers need intelligent groundwater and irrigation planning systems and effective mechanisms to benefit from the scientific and technological revolution, particularly the artificial intelligence engines, to enhance the water support in their water use planning practices. Therefore, this work aims to improve the groundwater level prediction based on the previous measures for better planning of hydraulic resource use. For this concern, the suggested method starts with data-preprocessing using the Principal Component Analysis method. Next, we validated the effectiveness of the hybrid artificial neural network, combined with an extended genetic algorithm for the hyperparameters and weight optimization, in predicting the groundwater levels in a selected monitoring well in California. The evaluation results have demonstrated the performance of the optimized ANN-GA model.