A review of recent hybridized machine learning methodologies for time series forecasting on water-related variables

Time series forecasting tools, particularly machine learning models, are essential for diverse water applications, including infrastructure management, flood warning systems, and water treatment plant optimization. While standalone models have proven effective, hybridizing multiple models can enhance performance and better meet end-user requirements. Given the commonalities in time series forecasting methodologies across industries and water applications, a comprehensive review of recent novel models can facilitate the cross-pollination of methods. To this end, we review recent works on forecasting water demand, water quality, and hydrological variables. It examines key considerations at different hybridization stages and synthesizes methodological architectures, significant findings, and modeling philosophies. Additionally, the growing availability of data has catalyzed the development of specialized time series forecasting models. This paper analyzes data properties across water-related variables and evaluates recent novel models to assess their potential. By highlighting similarities among various forecasting methodologies, this review provides insights to guide future approaches through leveraging collective knowledge.