Prediction of the upwelling phenomenon at the northwest African Atlantic coast-a connectionist approach

This article presents a connectionist approach based on an artificial neural network technique for the prediction of coastal upwelling in a study window off the northwest African coast. The prediction approach is based on the correlation of the coastal wind and the sea surface temperature anomalies during upwelling events. This correlation is first investigated by theoretical studies with a 3-dimensional Atlantic circulation model to obtain more information about the process. Then the correlation is approximated with an artificial neural network adapted with time series of the local wind and the upwelling sea surface temperature index. The ANN prediction result varies with the seasons. It is within a range of ±1 K for seasons with little coastal upwelling dynamics (winter, spring) and within a range of ±4 K for dynamic seasons (summer and autumn)