Quality Prediction of Sustainable Groundwater Resources, a Falaj in Oman

Aflaj are the most important engineering technology in Oman for the abstraction of water under the ground. Aflaj’s water is used for domestic and agricultural purposes in this country. Therefore, the quality of Aflaj’s water is crucial for both domestic users and farmers for agricultural and irrigation purposes. Water decision-makers consistently recognize the need for a reliable model that can predict quality. This study employs four distinct models to achieve this objective. Artificial neural network, adaptive neuro-fuzzy inference system (ANFIS), K -nearest neighbor, and support vector machine are the engaged models. They have been used for the forecasting of electrical conductivity (EC) as a quality parameter in falaj Al-Hamra, Oman. To this end, five reasonable scenarios are defined (S1, S2, S3, S4, and S5). Input data such as precipitation, flowrate of falaj, temperature of water, and EC values with lag time is the only difference among these scenarios. The data collection spans from 1982 to 2021. We implement these models using the MATLAB programming software. We also use four evaluation criteria, namely, mean absolute error, root-mean-square error, Nash–Sutcliff error, and R, to assess the performance. Results showed that, among all the models, ANFIS has the highest accuracy in all stages, including training, testing, and validation. All evaluation criteria indicate this. Also, findings were presented that S4 is closer to the real condition of falaj Al-Hamra, as the errors achieved from this scenario are less than the others. It means that there is a relationship between the contributing parameters in scenario #4 and the quality of water in falaj Al-Hamra. The funding also revealed that changes in flowrate have a greater impact on the water's EC than precipitation. This study assists water decision-makers in developing a well-functioning model for quality forecasting, which can then be enforced in falaj Al-Hamra.