Water quality assessment of river Beas, India, using multivariate and remote sensing techniques.

Abstract

River Beas originates in the Himalayas and merges into river Sutlej at Harike, a Ramsar wetland. This river is a habitat of the endangered freshwater dolphin, Platanista gangetica minor R. Twenty-five water quality parameters, including eight heavy metals, were studied at four sampling sites over a stretch of 63 km between Beas and Harike towns for pre-monsoon, post-monsoon and winter seasons. Principal component analysis of the data proved to be an effective tool for data reduction as the first three principal components of all the water quality parameters explained 100% variance. Factor analysis delineated three factors underlying the water quality. Factor 1 comprised pollution-related parameters like BOD, COD, DO, PO4(-3) and hardness. Factor 2 was a natural water quality determinant and explained maximum variance in turbidity, alkalinity and TDS. Factor 3 comprised NO3(-1), a fertilizer-related parameter. Reflectance values from bands 2 (green), 3 (red) and 4 (near infra-red) of Landsat (TM) digital data were regressed on PO4(-3), turbidity and TDS using multiple linear regression analysis. PO4(-3) contributed positively to the spectral radiance, whereas TDS contributed negatively. Beta regression analysis revealed that PO4(-3) had a positive relation with BOD, whereas turbidity and TDS were negatively regressed with BOD. Artificial neural network models were fitted to the data. Correlations between the target values from ANN for turbidity, BOD and bands 2 (green), 3 (red) and 4 (near infra-red) were highly significant.