Integrating Machine Learning and Geospatial Modelling for Seafloor Geomorphological Classification in the Nigerian EEZ Using Multi-Temporal Bathymetry (2019–2025)

Abstract Accurate seafloor bathymetry is vital for marine spatial planning, offshore engineering, and environmental management. However, much of the Gulf of Guinea remains under-surveyed. This study integrates machine learning (ML) and geospatial modelling to improve seafloor geomorphological classification and uncertainty assessment within the Nigerian Exclusive Economic Zone (EEZ) using multi-temporal General Bathymetric Chart of the Oceans (GEBCO) datasets (2019–2025). Four successive GEBCO grids were analysed through temporal differencing (ΔZ), morphometric analysis, and provenance mapping. Ensemble algorithms (Random Forest, Gradient Boosting) were applied to predict spatial uncertainty using morphometric and temporal predictors. Statistical analyses, including paired-sample t-tests and Taylor-diagram validation, quantified the significance and accuracy of observed changes. The Nigerian EEZ exhibited a systematic mean deepening of 0.8–2.3 m per release and a cumulative refinement of ~ 4.2 m, reflecting the transition from gravity-derived to multibeam-based bathymetry (p < 0.01). The seafloor structure remained stable, comprising ~ 9–10% continental shelf, 20% slope, 34% rise, and 38% abyssal plain. Proxy uncertainty declined by ≈ 70% between 2019 and 2023. The Random Forest model achieved the best performance (R² = 0.86; RMSE = 7.8 m; r = 0.93), outperforming Gradient Boosting and traditional interpolation methods. Integrating ML with multi-temporal bathymetry enables spatially explicit uncertainty mapping, supporting targeted survey planning, offshore infrastructure design, and environmental monitoring. The framework enhances Nigeria’s contribution to the Seabed 2030 initiative, promoting a data-driven foundation for sustainable blue-economy development and improved marine governance across West Africa.