A new remote sensing-based algorithm for estimating dissolved organic carbon in an urban eutrophic reservoir

Inland waters play a pivotal role in the global carbon cycle, acting not only as conduits but also as active processors of carbon. Monitoring dissolved organic carbon (DOC) in these systems is essential for understanding biogeochemical transformations, particularly in eutrophic reservoirs where anthropogenic pressures intensify carbon cycling. This study parameterized a DOC retrieval algorithm, optimizing the Quasi-Analytical Algorithm (QAA-M13) for application to Sentinel-2 imagery of the Billings Reservoir (Brazil). The QAA-M13 was recalibrated using in situ remote sensing reflectance (Rrsn = 73) to retrieve total absorption Rrs, achieving high correlation with field measurements (R between 0.74 to 0.81 for visible bands). Subsequently, the study implemented a semi-analytical proportionality model to estimate Rrs(440), achieving strong agreement with in situ data (R = 0.95, ε = 13.14%). A regression-based approach retrieved DOC from Rrs(440), and produced accurate estimates (R = 0.77, ε = 8.12%). When applied to Sentinel-2 imagery under matchup conditions, the model successfully captured the spatial gradients of DOC across the reservoir. However, the quantitative agreement with field measurements was limited, showing a low correlation (R = 0.32), although presenting low errors (ε = 8.62%). Despite this limitation, the model effectively captured areas of elevated DOC, accurately representing the spatial distribution near tributary inflows and anthropogenically influenced zones. Overall, the proposed methodology provides a robust framework for DOC monitoring in optically complex inland waters, offering a cost-effective tool for large-scale and long-term water quality assessments.