Spatio-Temporal Dynamics and Environmental Drivers of Surface Chlorophyll-a in the Gulf of Guinea (2003–2022)

Abstract

The mechanistic understanding of biogeochemical dynamics in the Gulf of Guinea (GoG) has historically been hindered by persistent cloud cover and reliance on static geographic boundaries. In this study, we analysed a 20-year (2003–2022) satellite-derived chlorophyll-a (Chl-a) dataset to overcome these observational limitations through a three-part spatial and machine-learning framework. First, the Data Interpolating Empirical Orthogonal Functions (DINEOF) algorithm reconstructed a gap-free climatology, demonstrating robustness under extreme simulated cloud cover (R2 = 0.884). Second, a Fuzzy C-Means (FCM) clustering algorithm objectively partitioned the basin into three dynamic, physically driven bioregions: an oligotrophic gyre, river plumes, and an upwelling mega-cluster. Third, we applied an explainable Random Forest framework, supported by SHapley Additive exPlanations (SHAP), to identify the main physical and biogeochemical predictors associated with coastal Chl-a variability using hindcast nutrients and a strict chronological split (training: 2003–2018; test: 2019–2022). The models produced conservative but meaningful independent test-period performance across coastal zones, with R2log values from 0.437 to 0.595. Rather than revealing a new ecological paradox, the framework provides a basin-specific interpretation of a globally documented pattern: offshore oligotrophication alongside localized coastal enrichment. The open ocean and transition/upwelling sectors show negative Chl-a tendencies consistent with sea surface warming, enhanced stratification, and reduced upward nutrient supply. Conversely, coastal ecosystems are structured by local hydrological and wind-driven forcings that modulate the regional climate signal. In the Congo plume, Chl-a variability is primarily structured by haline plume dynamics and secondary nutrient constraints, whereas the Niger plume reflects coupled mixed-layer and terrigenous nutrient controls. These findings establish a spatially objective typology of the GoG, providing a regional reference framework for future high-resolution missions, monitoring, and coupled physical–biogeochemical modelling.