Characteristic region extraction from time-series three-dimensional fluorescence spectroscopy

Abstract

Three-dimensional fluorescence spectroscopy of the online monitoring data of water pollution, i.e., as time-series three-dimensional fluorescence spectroscopy plots, provides a great deal of information about the constituents of water, including not only the information about the usual organic matter as monitored by spectroscopy but also about the unusual organic matter as outliers. In order to assist analyzing the organic matter and the outliers, we propose a method to extract the characteristic region from time-series three-dimensional fluorescence spectroscopy, utilizing the time mode and the spectral mode. From the time mode, some time series changes can clearly be seen because of the presence of outliers. Hence, the characteristic region of a three-dimensional fluorescence spectroscopy plot, i.e., the one related to unexpected events, such as pollution incidents, can be characterized by the outliers region. From the spectral mode, there are significant differences between the fluorescence intensities of the noise and the organic matter, and therefore the characteristic region of organic matter can be extracted by a cluster analysis. To test the proposal, the time-series three-dimensional fluorescence spectroscopy of tryptophan in water samples with catechol as the outlier was carried out. Numerical experiments show that the proposed method can effectively extract the organic matter fluorescence region of tryptophan and the outlier fluorescence region of catechol.