Drinking water quality evaluation in supply systems in Wuhan, China: application of entropy weight water quality index and multivariate statistical analysis.

Abstract

In this study, source water, finished water, and tap water were sampled monthly from two large drinking water treatment plants in Wuhan city, China for 12 months where physicochemical and microbiological parameters were measured, and the complex monitoring data was analyzed using single-factor assessment method, entropy weight water quality index (EWQI), and multivariate statistical techniques (i.e., cluster analysis (CA), discriminant analysis, and correlation analysis). The results of the single-factor assessment method showed that the total nitrogen pollution was the main problem in the source water quality, and the finished and tap water met the required quality standards. The EWQI values indicated that the overall quality of the source, finished, and tap water samples was "Excellent." In addition, strengthening monitoring of parameters with high entropy weights, including Pb, Hg, sulfide, Cr in surface water and Hg, aerobic bateria count, and As in drinking water, were suggested, as they were prone to drastic changes. Spatial CA grouped the finished and tap water samples from the same plant into a cluster. Temporal CA grouped 12 sampling times of source water into Cluster 1 (June), Cluster 2 (April-May, and July-November), and Cluster 3 (December-March). Concerning finished and tap water, except the October was regrouped, the result of temporal CA was consistent to that of the source water. Based on similar characteristics of water samples, monitoring sites and frequency can be optimized. Moreover, stepwise discriminant analysis indicated that the spatiotemporal variations in water quality among CA-groups were enough to be explained by four or five parameters, which provided a basis for the selection of monitoring parameters. The results of correlation analysis showed that few pairwise correlations were both significant (P < 0.05) and stable across sampling sites, suggesting that the number of monitoring parameters was difficult to reduce through substitution. In summary, this study illustrates the usefulness of EWQI and the multivariate statistical techniques in the water quality assessment and monitoring strategy optimization.