利用PARAFAC及SOM研究不同来源及粒径胶体的三维荧光光谱特征

Fluorescence characterization of fractionated colloids in different sources of waters is investigated using excitation emission matrix fluorescence spectroscopy (3D-EEM) based on parallel factor analysis (PARAFAC) and self-organizing map (SOM). Meanwhile, the results in the present study are compared to those in the previous study which were analyzed based on the peak-picking method. This study aims to evaluate the different approaches including PARAFAC, SOM and combined PARAFAC and SOM to investigate the size-dependent fluorescence properties of colloids from different sources. PARAFAC identified two humic-like fluorophores (i.e. C1 and C3) and three protein-like fluorophores (i.e. C2, C4 and C5). Thereinto, the fluorescence intensity of C1, C2, and C3 increases with the increased colloidal size; for C4 and C5, the fluorescence intensity decreases. The highest fluorescence intensity is generally present in the 300 kDa~1 μm fraction. The fluorescence intensity of bulk waters from the different sources including domestic wastewaters (Influent and Effluent), agricultural wastewaters (Daying and Tianenqiao) and the pristine natural river (Wusongkou) is decreased as the following order: Wusongkou>Influent>Daying>Tianenqiao>Effluent. The results in SOM model are agreed with those in PARAFAC model, but with a higher visualization. Furthermore, EEM-SOM model needs more input variables and has the same problem as those of peak-picking method. PARAFAC-SOM model not only has the advantages of PARAFAC and EEM-SOM model, but also has advantages on fewer input variables, shorter running time, and higher reliability. In addition, PARAFAC-SOM model was successfully applied to analyze the physicochemical properties of colloids (i.e. Parameters-SOM model). Parameters-SOM model makes the results in the previous study more systematic and visual. Therefore, PARAFAC-SOM model is the best approach to visualize the relationship between colloidal size and the fluorescence properties from different sources.