Application of multiple linear regression and artificial neural networks in river water quality modelling to predict dissolved oxygen in rivers: A case study of Krishna river in India.

Sustainable River water quality is to be maintained and it is the responsibility of all concerned stakeholders. The prime objective of Hydrology Project of Government of India is to ensure the river water quality and quantity by continuous capacity building and development of relevant both structural infrastructure (such as water quality laboratories) and cyber infrastructure (such as Decision Support Systems). Engineers and scientists working in the area of water quality need to be open minded with the ever uplifting attitude of lifelong learning to upskill with the ever-evolving relevant software and hardware technologies. Dissolved oxygen in rivers is essential for sustainability of aquatic life. Twelve year annual mean values from 2003 to 2014 of Dissolved Oxygen, pH, Electrical Conductivity, Nitrates, Biochemical Oxygen Demand and Temperature at nine stations along Krishna River are considered. Four types of models with varied input variables are developed in Multiple Linear Regression (MLR) using Microsoft Excel and Artificial Neural Networks (ANN) using nonlinear autoregressive model Artificial neural network identification model (NARX-ANN) to predict the dissolved oxygen. Levenberg – Marquardt Method (LMM) algorithm is used. The performances of both methods are compared. This paper takes care of Goal 6 of United Nations Sustainable Development of ensuring availability and sustainable management of water and sanitation for all as river water quality prediction is involved. ANN model with Mean Square Error (MSE) of 0.01 and R value of 0.987 is adopted. MLR model with R value of 0.574 is adopted. [ABSTRACT FROM AUTHOR]