Characterization of Dissolved Organic Matter (DOM) in a coastal lagoon utilizing PARAFAC decomposition of fluorescence Excitation-Emission Matrices

Dissolved organic matter (DOM) is a complex mixture of organic compounds whose composition and transformation strongly influence biogeochemical processes in coastal aquatic ecosystems. Fluorescence excitation-emission matrices (EEMs), combined with parallel factor analysis (PARAFAC), provide a useful approach for resolving distinct fluorescent components of DOM and assessing their spatial variability. This study aimed to characterize the fluorescent fraction of DOM (FDOM) in Rodrigo de Freitas Lagoon, a costal lagoon in Rio de Janeiro, Brazil, using EEM-PARAFAC and to evaluate the spatial distribution of the resolved fluorescent components. Water samples were collected at ten sampling points and filtered in a 0.45 µm PES filter. EEMs were acquired over excitation wavelengths of 200–450 nm and emission wavelengths of 300–600 nm. The fluorescence matrices underwent blank Milli-Q correction and spectral preprocessing, including Raman normalization and removal of Rayleigh scattering. PARAFAC models containing three to seven components were evaluated using multiple random initializations and convergence diagnostics. A four-component model was selected based on model convergence, residual structure, and spectral interpretability. It explained 83.3% of the variance in the EEM dataset. Two components exhibited spectrally plausible fluorescence maxima and were retained as the principal fluorescent DOM components. Component 1 (C1) showed a maximum at E_x/E_m = 242/495 nm, consistent with a humic-like fluorescence signature. Its F_max ranged from 0.549 to 0.579 RU, with a coefficient of variation of only 1.58%, indicating a highly homogeneous spatial distribution. Component 2 (C2) exhibited a maximum at E_x/E_m = 221/360 nm and showed substantially greater spatial variability, with F_max ranging from 0.444 to 0.670 RU and a coefficient of variation of 13.45%. C2 reached its highest fluorescence at sampling point P3, where it accounted for 54.6% of the combined F_max of C1 and C2, compared with 47.3% on average across the sampling sites. The remaining two model components (C3 and C4) presented emission and excitation maxima separated by approximately 1 nm, indicating that they were strongly influenced by residual Rayleigh/Tyndall scattering rather than representing independent fluorescent DOM pools. Consequently, interpretation of the chemically interpretable DOM fluorescence was focused on C1 and C2. The contrasting spatial variability of these components suggests that the fluorescent DOM pool comprises both a relatively homogeneous humic-like fraction and a second fraction exhibiting stronger spatial heterogeneity. Overall, EEM-PARAFAC successfully resolved spatial patterns in the fluorescent DOM of the lagoon and demonstrated its potential for distinguishing relatively conservative and spatially variable fluorescent fractions in coastal environments.