In this paper, chemometric approaches based on cluster analysis, classical and robust principal component analysis were employed to identify water quality in Daya Bay (DYB), China. The results show that these approaches divided water quality in DYB into two groups: stations S3, S8, S10 and S11 belong to cluster A, which lie in Dapeng Cove, Aotou Harbor and the north-eastern part of DYB, where water quality is related mainly to anthropogenic activities. The other stations belong to cluster B, which lie in the southern, central and eastern parts of DYB, where the quality is related mainly to water exchange with the South China Sea. Cluster analysis yields good results as a first exploratory method for evaluating spatial difference, but it fails to demonstrate the relationship between variables and environmental quality on the one hand and the untreated data on the other. However, with the aid of suitable chemometric approaches, the relationship between samples or variables can be investigated. Classical and robust principal component analysis can provide a visual aid for identifying the water environment in DYB, and then extracting specific information about relationships between variables and spatial variation trends in water quality.
2
Dostęp do pełnego tekstu na zewnętrznej witrynie WWW
The paper presents overview of literature pointing up usefulness of two chemometric methods of data exploration: cluster analysis (CA) and principal component analysis (PCA) for interpretation of multidimensional data obtained during surface water monitoring programme. The exceptional importance of fresh water quality rests on the fact that it determines condition and development of all human beings as well as animals and plants. In sections 1 and 3 a concise description of the two methods of pattern recognition was presented. CA provides visual presentation of clustering mode of samples or variables in the form of tree-like scheme called dendrogram [3, 9]. Samples which exhibit considerable similarity fall into one cluster and at the same time they differ to the greatest extent from samples belonging to other clusters [1, 70-74]. PCA enables significant reduction of data matrix dimension without an excessive loss of information [5, 6, 82]. In addition, PCA provides intelligible graphic visualization of the data structure by scattering samples, described by many variables, on plane formed by the following principal components [74]. In CA methods of calculation the distance between different clusters or objects in a hierarchical dendrogram were also mentioned [75]. They define whether two clusters or objects are sufficiently similar to be joined together into one cluster. It was also denoted that among most commonly applied methods Ward's clustering method turns out to be the best and most effective for surface water monitoring results as it gives the largest number of correctly classified observations. The sections 2 and 4 present several examples of the two multidimensional chemometric techniques application in assessment of surface water quality. It was perceived that CA and PCA allow classification of sampling stations in relation to chemical composition of water and therefore permit to pinpoint areas with similar water quality [5, 82]. CA and PCA also help to indicate areas with distinctive chemical composition of water [6, 73, 90, 105]. They also help to identify factors which determine water quality at different regions [4, 5, 88, 113]. When applied for the results acquired from one watercourse at different sites CA and PCA enable to examine changes of water quality downstream [73, 80, 114]. The two chemometric methods are also useful tools for determination if season of sampling or specific weather conditions can influence variation of physicochemical parameters of water [90, 95, 103]. Other chemometric methods applied for monitoring results evaluation are mentioned in section 5 [74, 75, 116].
JavaScript jest wyłączony w Twojej przeglądarce internetowej. Włącz go, a następnie odśwież stronę, aby móc w pełni z niej korzystać.