Changes in the fluorescence composition of multiple DOM sources over pH gradients assessed by combining parallel factor analysis and self-organizing maps

Abstract

Dissolved organic matter is a ubiquitous constituent of natural waters that plays key roles in several important processes. The fluorescence properties of DOM have been linked to its functionality, but these properties may vary with pH. In this study Kohonen's self-organizing maps (SOMs) were applied to excitation-emission matrices (EEMs) of fresh dissolved organic matter (DOM) from three sources: senescent sugar-maple leaves and white spruce needles, and humified white spruce needles, over a pH range of ~4.5 – 12.5. SOMs were applied to raw EEMs, EEMs reduced in dimensionality by pre-processing using parallel factor analysis (PARAFAC), and PARAFAC loading proportions normalized to values at initial pH. Some separation of EEMs into source-based clusters was achieved in the SOM of raw EEMs, but commingling was apparent and evidence of changes over pH gradients was overshadowed. SOMs of PARAFAC component proportions demonstrated clear source-based clustering, and pH-based gradients were visible for DOM from senescent and humified spruce needles. Changes in optical properties were obvious over pH gradients in the SOM of components normalized to starting condition. Component proportions decreased to values as low as 5% of the initial values for microbial humic-like peak M and increased to as high as 278% for a humic-like component. Tyrosine-like fluorescence increased to 112% of initial over increasing pH in humified spruce leachates but decreased to as low as 45% in the other leachates. The combination of PARAFAC and SOM drastically enhanced visualization and interpretability of pH-induced changes in DOM compared to either method alone.