Source Prediction and Distribution of Organic Matter Across a Land to Ocean Sediment Transect: An Integrative Biogeochemical and Multivariate Machine Learning Approach

Identifying organic matter (OM) distribution and origins in sediments is vital for distinguishing archival records, depositional dynamics, and extent of terrigenous inputs to marine systems. However, degradation and modification of OM before deposition can camouflage and mask its source signals in sedimentary sinks. Also, limited number of samples, matrix complexity, and multiple available tracers for single‐proxy mixing models can all challenge source apportionment of sedimentary OM. To overcome these limitations and deconvolute the origins of deposited OM along a land‐ocean transect, this study integrates machine learning and multivariate analyses of elemental contents (C and N), biomarker abundances (C17‐C27 n ‐alkanes), and bulk and compound specific stable carbon isotope values (δ 13 C) of sediments across the St. Lawrence Estuary and Gulf (SLEG). With pronounced provenance and terrestrial‐marine regimes, the most proximal and distal sedimentary endmembers of the SLEG (60 and 780 km away from outlet) retain distinct geochemical information for training Principal Component Analysis and Partial Least Squares Regression model, predicting OM distribution across in‐between stations. Results reveal a gradual shift in sedimented OM composition along the continuum due to OM inputs and source‐to‐sink processes; with a gradient of 2% decrease in terrestrial character of sedimentary OM per each 10 km distance from river outlet, findings reflect that the inputs and signals of terrigenous OM are progressively diluted and transformed offshore. Also, isotopic and molecular signatures exert stronger predicting controls than conventional tracers such as C/N ratios, highlighting that the diagnostic strength of proxies is system‐specific, particularly under dynamic conditions. Integrative multivariate approaches can hence capture the complexity and covariances of multiple biogeochemical proxies to better decode camouflaged sedimentary archives and provide more holistic estimations of OM transportation and transformation across systems with strong geo‐spatial and input gradients.