Water Quality Zoning Using Probabilistic Support Vector Machines and Self-Organizing Maps

Water quality zoning is an important step for studying and evaluating surface and groundwater quality variations with time and space. It can also provide important information for developing efficient water quality management strategies. Most common methods for water quality zoning do not consider the uncertainties associated with water resources systems. Also, these methods only categorize the water quality monitoring stations into some classes based on existing water quality but do not provide any information about the quality of water which is categorized in a class. In this paper, a new methodology for probabilistic water quality zoning is developed which utilizes the capabilities of Probabilistic Support Vector Machines (PSVMs) and Fuzzy Inference System (FIS). The required data for training the PSVM is generated by using the FIS in a Monte Carlo Analysis. The trained PSVM-based water quality index provides the probabilities of belonging quality of water to different water quality classes. The applicability of the proposed methodology is investigated by applying it to two surface and groundwater resources systems in Iran. Also, for more evaluating the efficiency of the methodology, the results are compared with those obtained from two clustering techniques, namely Fuzzy Clustering Technique (FCT) and Self-Organizing Map (SOM). The results of surface and ground water quality zoning are depicted in maps by utilizing the Geographic Information System (GIS).