Exploratory analysis of excitation-emission matrix fluorescence spectra with self-organizing maps as a basis for determination of organic matter removal efficiency at water treatment works

In the paper, the self-organizing map (SOM) was employed for the exploratory analysis of fluorescence excitation-emission data characterizing organic matter removal efficiency at 16 water treatment works in the UK. Fluorescence spectroscopy was used to assess organic matter removal efficiency between raw and partially treated (clarified) water to provide an indication of the potential for disinfection by-products formation. Fluorescence spectroscopy was utilized to evaluate quantitative and qualitative properties of organic matter removal. However, the substantial amount of fluorescence data generated impeded the interpretation process. Therefore a robust SOM technique was used to examine the fluorescence data and to reveal patterns in data distribution and correlations between organic matter properties and fluorescence variables. It was found that the SOM provided a good discrimination between water treatment sites on the base of spectral properties of organic matter. The distances between the units of the SOM map were indicative of the similarity of the fluorescence samples and thus demonstrated the relative changes in organic matter content between raw and clarified water. The higher efficiency of organic matter removal was demonstrated for the larger distances between raw and clarified samples on the map. It was also shown that organic matter removal was highly dependent on the raw water fluorescence properties, with higher efficiencies for higher emission wavelengths in visible and UV humic-like fluorescence centers.