Machine Learning–Based Uncertainty Analysis of Multi-Temporal GEBCO Bathymetry in the Nigerian EEZ

Reliable bathymetric information is essential for marine spatial planning, offshore engineering, and environmental management, yet large portions of the global ocean remain constrained by sparse acoustic survey coverage. This study presents a machine learning–based uncertainty analysis of multi-temporal GEBCO bathymetry (2019–2025) within the Nigerian Exclusive Economic Zone (EEZ), a data-limited region of high strategic relevance. Rather than generating a new bathymetric surface, the analysis focuses on quantifying and predicting spatial bathymetric uncertainty associated with successive updates of global bathymetric products. Temporal bathymetric differencing (ΔZ) was applied to successive GEBCO releases to characterise systematic depth refinement and to derive proxy uncertainty metrics. Morphometric attributes, temporal refinement indicators, and data-provenance variables distinguishing predicted from acoustic measured bathymetry were used as predictors in ensemble machine-learning models. Model performance was evaluated using 10-fold cross-validation, with accuracy assessed using the coefficient of determination (R²) and root mean square error (RMSE), which quantify prediction error in proxy uncertainty, expressed in metres, rather than seabed depth. Results indicate that the large-scale geomorphological structure of the Nigerian EEZ remains stable across all datasets, while successive GEBCO releases show statistically significant mean depth refinements of 0.8–2.3 m per release and a cumulative adjustment of approximately 4.2 m between 2019 and 2025. Proxy bathymetric uncertainty declined by approximately 70% between 2019 and 2023. The Random Forest model achieved the highest predictive performance (R² = 0.86; RMSE = 7.8 m), enabling spatially explicit uncertainty mapping to support survey prioritisation and marine governance.