MULTIPLE LINEAR REGRESSION MODEL TO PREDICT THE WATER QUALITY PARAMETERS OF MUSI RIVER NEAR HYDERABAD, INDIA

The river Musi is one of the tributaries of river Krishna. The water quality of any river deteriorates due to the discharge of untreated municipal and industrial wastewater. Assessment of river water quality is necessary to know the extent of contamination of the river Musi. Four multiple linear regression (MLR) equations are developed to predict sodium adsorption ratio (SAR), electrical conductivity (EC), biological oxygen demand (BOD) and dissolved oxygen (DO) using the monthly water quality data for the year 2021 of river Musi at four measuring stations near Hyderabad. The open access online database of Telangana State Pollution Control Board (TSPCB) is used. Multiple linear regression (MLR) models are developed using regression tool available in Microsoft Excel. F-tests ant t-tests are used to measure the goodness of fit of MLR equations developed. The MLR model predicting EC has the highest multiple correlation coefficient (multiple R) value of 0.77 and it is followed by MLR model predicting SAR with multiple R value of 0.76. The values of coefficient of multiple determination (R square) of four MLR models developed indicate that the explained variance is 0.58, 0.59, 0.48 and 0.3 respectively. From this we can conclude that the unexplained variance is 0.42, 0.41, 0.52 and 0.7 respectively. The F observed values are higher than the F critical values for single-tailed test obtained from the standard tables at 5% level of significance. This shows that all the independent variables are statistically related to the dependent variables used for prediction at 95% confidence level. t- statistic observed values of all the independent variables are higher than the tcritical values obtained from the standard tables for single-tailed test at 5% level of significance except in two cases.