A self-organizing map approach to the analysis of lake DOM fluorescence for differentiation of organic matter sources.

Abstract

The sources and properties of dissolved organic matter (DOM) in two lakes with different non-point source inputs were investigated by combining conventional three-dimensional fluorescence spectroscopy methods with a self-organizing map (SOM). To assess the DOM humification level, the representative neurons 1, 11, 25, and 36 were assessed. The SOM model showed that the DOM humification level of the Gaotang Lake (GT) which has a mainly agricultural non-point source input was significantly higher than that of the Yaogao Reservoir (YG) which has a mainly terrestrial source input (P < 0.01). The GT DOM mainly came from factors such as agricultural-related farm compost and decaying plants, while the YG DOM originated from human activities around the lake. The source characteristics of the YG DOM are obvious, with a high level of biological activity. Five representative areas in the fluorescence regional integral (FRI) were compared. The comparison showed that during the flat water period, the GT water column showed more terrestrial characteristics, even though the humus-like fractions in the DOM of both lakes were derived from microbial decomposition. Principal component analysis (PCA) showed that the agricultural lake water DOM (GT) was dominated by humus components, while the urban lake water DOM (YG) was dominated by authigenic sources.