El Niño and Solar Activity: Granger Causality on a Neural Network

El Niño (ENSO), a consequence of changes in ocean circulation patterns, has a significant impact on the global climate and associated economic activity. According to our hypothesis, in addition to internal climatic factors, the ocean circulation regime can be controlled by small changes in total solar irradiation (TSI) occurring in the 11-year solar activity cycle. In this case, positive feedback with a gain of about 10 is possible in near-equatorial regions. In this paper, we attempt to predict monthly averages of an index describing ENSO using TSI as an additional predictor. For prediction, we train a recurrent neural network with a long- and short-term memory (LSTM) unit on ENSO alone and with the addition of TSI. As a result, we find that the ENSO training error is reduced when TSI is added as a predictor. Our result indicates the possibility of using TSI as one of the predictors in constructing modern nonlinear predictive global climate models.