Characterizing dissolved organic matter in Taihu Lake with PARAFAC and SOM method.

Abstract

The three-dimensional fluorescence spectrum has a significantly greater amount of information than the single-stage scanning fluorescence spectrum. At the same time, the parallel factor (PARAFAC) analysis and neural network method can help explore the fluorescence characteristics further, thus could be used to analyse multiple sets of three-dimensional matrix data. In this study, the PARAFAC analysis and the self-organizing mapping (SOM) neural network method are firstly introduced comprehensively. They are then adopted to extract information of the three-dimensional fluorescence spectrum data set for fluorescence characteristics analysis of dissolved organic matter (DOM) in Taihu Lake water. Forty water samples with DOM species were taken from different seasons with the fluorescence information obtained through three-dimensional fluorescence spectrum analysis, PARAFAC analysis and SOM analysis. The PARAFAC analysis results indicated that the main fluorescence components of dissolved organic matter in Taihu Lake water were aromatic proteins, fulvic acids, and dissolved microorganisms. The SOM analysis results showed that the fluorescence characteristics of the dissolved organics in Taihu Lake varied seasonally. Therefore, the combined method of three-dimensional fluorescence spectrum analysis, PARAFAC and SOM analysis can provide important information for characterization of the fluorescence properties of dissolved organic matter in surface water bodies.