Search for a command to run...
This study examines the hydro-sedimentological–radioecological controls governing the distribution of natural (K-40, Ra-226, Th-232) and anthropogenic (Cs-137) radionuclides in surface sediments of the western Black Sea shelf. Activity concentrations were determined by high-resolution gamma spectrometry, and radiological indices—including radium equivalent activity (Ra_eq), external hazard index (Hex), and annual effective dose (AED)—were calculated to evaluate environmental safety. All indices remained well below internationally accepted thresholds, confirming the absence of radiological hazard in both coastal and offshore settings. Strong correlations between Ra-226 and Th-232 indicate dominant lithogenic control of natural radionuclides, whereas Cs-137 exhibits geochemical decoupling consistent with its behavior. A significant relationship between the fine-grained sediment fraction (<63 µm) and Cs-137 activity highlights the grain size effect, with offshore depositional zones acting as sediment-focusing areas where Cs-137 and excess Pb-210 co-accumulate under low-energy hydrodynamic conditions. Despite localized offshore enrichment, dose contribution analysis shows that natural radionuclides dominate the absorbed-dose budget, while Cs-137 contributes only marginally. Spatial predictive modeling using Artificial Neural Networks, validated under a Spatial Leave-One-Group-Out framework, yielded moderate generalization capacity (R2 = 0.61 for Ra-226; R2 = 0.41 for Cs-137), reflecting smoother spatial gradients of lithogenic radionuclides than heterogeneous radiocesium deposition. Furthermore, Machine Learning algorithms provided significant analytical value: a Random Forest (RF) model successfully classified environments (nearshore/shelf/depositional basin) based on distinct radionuclide signatures. At the same time, an optimized Artificial Neural Network (ANN-GA) enabled the nonlinear reconstruction of radiometric–granulometric patterns to identify local anomalies. The results show that radionuclide distributions are primarily structured by sediment provenance, grain size sorting, and hydrodynamic energy gradients rather than ongoing anthropogenic inputs.