Reliability assessment of water quality index based on guidelines of national sanitation foundation in natural streams: integration of remote sensing and data-driven models

摘要

Rivers, as one of the freshwater resources, are generally put in the state of jeopardy in terms of quantity and quality due to the development in industry, agriculture, and urbanization. Management of water quality is inextricably bound up with a reliable prediction of the Water Quality Index (WQI) for various purposes. In this way, an accurate estimation of WQI is one of the most challenging issues in the water quality studies of surface water resources. There is a board range of traditional methodologies for the WQI evaluation. Due to the intrinsic limitations of conventional models, Data-Driven Models (DDMs) have been frequently employed to assess the WQI for natural streams. In the present research, WQI values and their typical classifications were obtained by guidelines of the National Sanitation Foundation (NSF). Hence, four well-known DDMs such as Evolutionary Polynomial Regression (EPR), M5 Model Tree (MT), Gene-Expression Programming (GEP), and Multivariate Adaptive Regression Spline (MARS) are employed to predict WQI in Karun River. In this way, 12 Water Quality Parameters (i.e., Dissolved Oxygen, Chemical Oxygen Demand, Biochemical Oxygen Demand, Electrical Conductivity, Nitrate, Nitrite, Phosphate, Turbidity, pH, Calcium, Magnesium, and Sodium) were accumulated from nine hydrometry stations and additionally missing values of water temperature were extracted from images analysis of Landsat-7 ETM+. Furthermore, the Gamma Test (GT), Forward Selection (FS), Polynomial Chaotic Expression (PCE), and Principle Component Analysis (PCA) were used to reduce the volume of DDMs-feeding-input variables. Results of DDMs demonstrated that FS-M5 MT had the best performance for the estimation of WQI classification. WQI values for Karun River were assessed in the reliability-based probabilistic framework to consider the effect of any uncertainty and randomness in the input parameters. To this end, the Monte-Carlo scenario sampling technique was conducted to evaluate the limit state function from the DDMs-based-WQI formulation. Based on the qualitative description of the WQI, it was observed that the WQI of Karun River is classified into “Relatively Bad” quality. Moreover, based on the reliability analysis, there is only a 19% chance exists for a specimen from Karun River to have a better quality index.