Abstract
The Gulf of Guinea represents a wide tract of the African coast with complex and rich coastal ecosystems undergoing various pressures. The seasonal variations of chlorophyll-a concentration (Chla) along the Gulf of Guinea (GoG) and their relations with physical oceanographic variables were analyzed using satellite observations covering the period 2002-2012. The effects of sea surface temperature (SST), sea level anomalies (SLA), winds, geostrophic currents, eddy kinetic energy (EKE), mesoscale eddies and fronts were considered on a monthly time scale. The analysis for each unit area was carried out on a chlorophyll index (IChla) computed as the product of the mean distance from the coast to the eutrophic threshold (1 mg m(-3) isoline) and the average Chla in the eutrophic area. The study, based on satellite-derived Chla, was allowed by the unprecedented coverage given by the products distributed by the ESA Ocean Colour Climate Change Initiative (OC_CCI) resulting from the merging of data from several satellite missions. The physical variables served as potential predictors in a statistical Boosted Regression Tree (BRT) model. To account for the heterogeneous nature of the GoG, the analysis was conducted on eight systems that made up a partition of the whole region defined on the basis of the BRT model results and climatological properties. The western-most domain, from Guinea-Bissau to Sierra Leone, was associated with upwelling properties in boreal winter and appeared to share some characteristics with the overall Northwest African upwelling system. The region of Ivory Coast and Ghana also had upwelling properties but the main upwelling season was in boreal summer. In general upwelling conditions with cold SST, negative SLA, fairly strong frontal activity, and moderate winds, appeared as the environmental window most favorable to high IChla values. For these systems, the BRT model fitted the IChla data well with a percentage of explained total deviance [Formula: see text] between 70% and 91% when using only physical oceanographic variables. Finally, the systems associated with the coasts of Nigeria to Gabon showed some mixed properties, with [Formula: see text] values of 54-60%. Among these systems, a common feature seemed to be the importance of river discharge to explain IChla variations. Where possible (for the Niger River in the Nigeria system), the addition of river data as predictor in the BRT model resulted in a significant increase of [Formula: see text] to 75%. Further progress is needed to understand the observed relationships and to predict how they can evolve in the face of climate change.