Is the Atlantic surface temperature a good proxy for forecasting the recruitment of European eel in the Guadalquivir estuary?

This study analysed the possibility of using the sea surface temperature (SST) of the Atlantic Ocean to predict the recruitment of European eels in one of the most important estuaries of the south of Europe. For this purpose, two different time series concerning glass eel in the Guadalquivir estuary (the first obtained from a set of fishery-independent experimental samplings in this estuary and the second from an unofficial database on commercial catches provided by one of the main local marketer-buyers) were standardised to obtain a single time series on a monthly scale. This series was correlated with a total of 368 SST time series for 368 sectors of 1.95° × 1.95° of the Atlantic Ocean covering the possible migration routes of adult eels and leptocephalous larvae. The significant sectors were clustered and selected as inputs for artificial neural network models (ANNs) with the objective of obtaining a model to forecast glass eel recruitment. Globally, the best result was given by an ANN with only 12 clusters as input variables and 35 neurons in the hidden layer. For this configuration, the explained variance in the test phase was slightly higher than 79%. These results were significantly better than those obtained with classical methods. The strong correlation between predicted and observed glass eel abundance suggests that: (a) there is a marked non-linear relationship between SST and glass eel recruitment in the Guadalquivir estuary; (b) SST is a good proxy for predicting glass eel recruitment and; (c) one of the main factors responsible for the changes in abundance of this species is changes in the ocean conditions.