<p>Lignin, a macromolecule found in all vascular plants, can be used as a biomarker for terrestrial dissolved organic matter in the ocean. Measuring lignin in the ocean can help us quantify the supply to and fate of terrestrial carbon in the ocean. Lignin analyses in aquatic samples quantify phenolic products after cupric oxidation using gas or liquid chromatography, with detection either by mass spectrometry or UV-Vis spectroscopy. Mass spectrometry yields low detection limits and high specificity, but requires specialized and expensive instrumentation. In contrast, liquid chromatography coupled with UV-Vis spectroscopy is more readily available and cheaper to operate, but traditionally suffers from lower specificity due to overlapping signals of the bulk organic matter background.</p><p>This study demonstrates a new approach of UV-Vis spectroscopic detection coupled to high-performance liquid chromatography (HPLC) that circumvents common issues and improves the detection limit by a factor of 10. This improvement is accomplished by using the second derivative of the chromatogram and applying a modified parallel factor analysis (PARAFAC2). PARAFAC2 tolerates subtle remaining chromatogram shifts in retention time between samples and successfully separates spectra of co-eluting signals. The isolation of spectra based on this machine learning approach improves both lignin phenol identification and the accuracy of their quantification. The approach developed automates the analysis of chromatograms and considerably reduces the water volumes required, improving the applicability of HPLC-UV-Vis for lignin characterization, which may increase the feasibility for widespread use.</p>