State-of-the art-on irrigation water quality management using data-driven methods: Practical application, limitations, and prospective directions

The use of brackish water resources in agriculture is a promising alternative to overcome water scarcity issues under global change and to implement Sustainable Development Goal (SDG) target 6.3. Meanwhile, according to the World Bank report in 2020, bad water quality can lead to the worldwide loss of food up to 9.54 trillion kilocalories per year. The rapid development of Artificial Intelligence-based technologies is a promising opportunity to modernize irrigation water quality (IWQ) management. This review endeavors to provide a comprehensive overview of the extent to which Machine Learning (ML) models overcome the limitations of conventional methods. This paper began with an introduction section focusing on the background research, followed by a bibliometric analysis of IWQ. Subsequently, a comprehensive review is presented, including discussions on model performances, data availability, and existing limitations. The review revealed that there is a potential accuracy of the ML models to develop ML-based sensor technologies for monitoring IWQ. However, it highlights the need to improve the applicability of ML models through selecting appropriate input and output variables, as it was approved that the efficiency of ML models not only depends on the prediction accuracy but also on the used variables. Overall, this review presents prospective directions to overcome the current limitations with a particular focus on the practical application and integration of the ML models into innovative technologies to manage IWQ.