Identification of temporal and spatial patterns of river water quality parameters using NLPCA and multivariate statistical techniques

• Published in: International journal of environmental science and technology (Tehran) Vol. 17; no. 5; pp. 2977 - 2994
• Main Authors: Rezaali, M., Karimi, A., Moghadam Yekta, N., Fouladi Fard, R.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-05-2020
• Subjects: Aquatic Pollution; Earth and Environmental Science; Ecotoxicology; Environment; Environmental Chemistry; Environmental Science and Engineering; Original Paper; Soil Science & Conservation; Waste Water Technology; Water Management; Water Pollution Control; River water quality assessment; Discriminant analysis; Nonlinear principal component analysis; Principal component analysis; Auto-associative artificial neural network
• ISSN: 1735-1472; 1735-2630
• DOI: 10.1007/s13762-019-02572-4

River water quality is assessed by collecting samples from rivers. During this process, a significant amount of data is generated, which often results in challenges in interpreting the dataset. In this study, 14 water quality parameters of the Gadarchay River basin in Iran, collected monthly, were analyzed to identify pollution sources and patterns. Nonlinear principal component analysis was compared with frequently used multivariate statistical techniques. Results suggested that spatial and temporal nonlinear principal component analysis outperformed the other multivariate techniques by explaining 80.34% and 80.78% of the total variances, respectively. Cluster analysis categorized 20 sampling stations into three groups: less polluted, moderately polluted and highly polluted. Discriminant analysis was carried out both spatially and temporally for each of the three groups. The performance of the spatial discriminant analysis for less polluted, moderately polluted, highly polluted and overall was observed to be 95.83%, 70.14%, 64.58% and 76.85%, respectively. Temporal discriminant analysis was also done for each season to find the most significant variables. The performance of temporal discriminant analysis for summer, winter, autumn and spring was 85%, 85%, 40% and 61.67%, respectively. For source identification, principal component analysis was implemented on raw data. The results of spatial and temporal discriminant analysis were used to better interpret the results of principal component analysis for the less polluted, moderately polluted and highly polluted groups; five principal components covered 76% of the variance, four principal components covered 75% of the variance, and four principal components covered 77% of the variance, respectively.