Optimizing the Spatial-Temporal Extent of Environmental Factors in Forecasting El Nino and La Nina Using Recurrent Neural Network

The periodic fluctuations in sea surface temperature (SST) and overlying air pressure across the Equatorial Pacific region causes El Nino-Southern Oscillation (ENSO). ENSO influences weather patterns around the globe and could cause severe rainfalls, heat waves, floods, and droughts, affecting regions far from the tropics. Hence, forecasting these events for longer lead times is of great importance. This study applies LSTM for forecasting ENSO events in the next year, based on environmental variables from previous years, including SST, sea level pressure, zonal wind, meridional wind, and zonal wind flux. The grid search method derives spatial and temporal features of ENSO by optimizing the extent of spatial and temporal windows of the environmental variables impacting ENSO. The results underscored the sensitivity of the ENSO forecast accuracy to the extent of spatial and temporal windows of predictors