An Integrated Hydrologic Bayesian Multi-Model Framework for Water Quality and Nutrient Analysis

The point and non-point sources of pollution have disturbed the sustained natural quality of surface water of the Brahmani river basin beyond the self-cleaning capabilities. Deteriorating of surface water quality has become a serious issue. To improve the quality of water, enhanced techniques for water quality monitoring, analysis and modeling are required. The multivariate statistical methods, fuzzy entropy methods, Bayesian approximation and MCMC algorithms are used for the study. A total of 20 water quality indicators (Cl, Na, SO42-,PO43-, K, B, F, Cd, Cu, Pb, Ni, Zn, Fe, Cr, NO3, NO2, Dissolved Oxygen (DO), Permanganate Index (CODMn), Biological Oxygen Demand (BOD), Total Alkali as CaCO3 (TA as CaCO3)) are collected for 15 years from 2002 to 2016 for two seasons (dry and wet season) from 9 sampling stations along the river course. Multivariate Statistical methods are used for the spatiotemporal variations of the water quality. The Hierarchical Agglomerative Cluster Analysis (HACA) grouped the nine gauging stations into four groups such as Considerable (Co), Low Pollution (LP), Moderate Pollution (MP), and Extreme Pollution (EP), whereas the water quality indicators are discriminated spatially and temporally based on the similarities in water quality characteristics. The Functional Water Quality Index (FWQI) are calculated by fuzzy entropy weight method, which identifies the principal contributing factors affecting the quality of water. The Bayesian inference along with MCMC algorithms are used for the water quality modeling. The Bayesian models are developed on the basis of prior information on trends of water quality and sources of pollution and the models are used as the basic model required for the evaluation of water quality using both informative and non-informative priors. The convention Water Quality Index (WQI) method along with Bayes’ rule and Shannon’s entropy weight method are applied for the estimation of likelihood and indicator weights. The likelihood estimates are used for the estimation of the posterior distribution. Finally, the influence of heterogeneous stochastic uncertain parameters is investigated using Advection Dispersion Equation (ADE) and the models developed by Bayes’ rule and MCMC simulations. The impact of uncertain factors is analyzed considering the ADE parameters and water quality indicators. The collaboration of ADE and Bayesian methods encourage the water quality management and environmental modeling techniques