Advanced Residuals Analysis for Determining the Number of PARAFAC Components in Dissolved Organic Matter

Area of Science:

• Environmental Chemistry
• Analytical Chemistry
• Biogeochemistry

Background:

• Parallel Factor Analysis (PARAFAC) is crucial for linking dissolved organic matter (DOM) fluorescence to its functions and biogeochemical cycles.
• Traditional split-half analysis for PARAFAC model validation requires a large number of excitation-emission matrices (EEMs) and may not adequately capture DOM differences.
• Underfitting or overfitting PARAFAC models can lead to inaccurate descriptions of DOM fluorescence.

Purpose of the Study:

• To develop and validate an enhanced method for estimating the optimal number of PARAFAC components for dissolved organic matter (DOM) fluorescence data.
• To improve the robustness of PARAFAC models by supplementing split-half analysis with residuals analysis.
• To enhance the accuracy of DOM characterization and classification using fluorescence spectroscopy.

Main Methods:

• Employed self-organizing maps (SOM) for clustering and analyzing residuals from PARAFAC models.
• Compared changes in model residuals with the addition of PARAFAC components to determine the optimal number.
• Assessed model overfitting by examining correlations between components and their distribution across samples.

Main Results:

• Self-organizing maps (SOM) revealed persistent peaks in model residuals, indicating limitations of split-half validation alone.
• Visualizing residuals effectively demonstrated the removal of fluorophore-like fluorescence with increased PARAFAC components.
• Advanced residuals analysis significantly improved the correct identification of size-fractionated leaf leachates (56.0±0.8% to 75.2±0.9%) and whole leachates (51.7±1.4% to 92.9±0.0%).

Conclusions:

• Advanced residuals analysis offers a superior method for validating PARAFAC models of DOM fluorescence, complementing split-half analysis.
• This enhanced approach improves the resolution of variations in DOM fluorescence, leading to more accurate biogeochemical interpretations.
• The proposed method provides a more reliable assessment of the number of PARAFAC components, crucial for understanding DOM cycling.