Hyperspectral imaging and 3D sediment modeling for organic carbon stock mapping in an estuarine environment

Accurate quantification of organic carbon (OC) stocks in estuarine sediments remains challenging due to strong spatial heterogeneity and limited spatial integration of point-based measurements. This study develops an approach combining hyperspectral imaging (HSI)-based predictive modelling with 3D stratigraphic reconstruction to improve the quantification and spatial mapping of OC stock in heterogeneous sedimentary environments. Ten sediment cores were collected from an estuarine deposit in the upper Seine Estuary (France). A single reference core was analyzed for sediment organic carbon (SOC) using Rock-Eval pyrolysis and for dry bulk density (DBD), then scanned using short-wave infrared imaging to calibrate predictive models subsequently applied to all cores. Stratigraphic correlations derived from visible-near-infrared imaging constrained the 3D reconstruction of deposit geometry, allowing spatial integration of predicted SOC and DBD. HSI-based models showed excellent predictive performance for both SOC (R² = 0.97) and DBD (R² = 0.87), enabling quantitative applications. The resulting OC stock was estimated at 488 Mg C ha -1 , more than twice the estimate obtained using the classical core-averaged extrapolation method over a fixed depth. This difference highlights the limitations of sparse sampling and fixed-depth assumptions, and demonstrates the importance of improved spatial integration for OC stock estimation in heterogeneous sedimentary environments. The proposed workflow is robust, cost-effective, and transferable provided that representative local calibration datasets are available, offering strong potential for spatially explicit OC stock assessments across diverse sedimentary settings and for supporting large-scale carbon budget evaluations.