Statistical analysis of water quality data of river Yamuna: A research report by National Institute of Hydrology

The study models the variations in water quality parameters of Yamuna river by means of stochastic approach using Central Water Commission’s monthly data for the period 1990-95 and makes projections for the year 1996. The results obtained for different water quality variables have been compared with the observed values and were found to be satisfactory. 

Stochastic approach has been used to enable the identification of trends and periodic phenomenon present in the natural data series. These have been decomposed and subsequently synthesized for data generation and forecasting. It is necessary to probe some of characteristics viz., non-normal distributions, seasonality, missing values, values below the limit of detection, external interventions and serial correlation, which complicate the analysis of water quality time series along the river.

Data of twenty-two important water quality variables were obtained from sixteen sampling stations of river Yamuna and utilized for trend analysis, cross-correlation, statistical estimations and ARIMA modeling. The data has been reproduced showing the longitudinal and spatial variation of water quality variables, the basic statistics, cross-correlation of different water quality variables and statistical modeling.

The study draws the following conclusions –

• The water quality variables are non-normally distributed in some of the cases as the coefficient of skewness is not equal to zero in them.
• Most of the time series plots did not show any definite trend. However, seasonal effect has been observed in all the cases for the data sets. This pattern in water quality may be due to the influence of annual cyclic pattern of the hydrologic inputs to the river water environment.
• The cross-correlation developed between the stations is not significant in all the cases. The regression equations developed can be used for approximate estimation of the water quality variables.
• Analysis of the data using ARIMA model framework such as correlogram structure and minimization for sum of squares of the residuals indicated that the models having both non-seasonal and seasonal components were, in general, appropriate for modeling the water quality time series at all the stations.

Download the report here: