Performance of backpropagation artificial neural network to predict el nino southern oscillation using several indexes as onset indicators

Abstract

El Nino Southern Oscillation (ENSO) is a world’s climate anomaly that occurs repeatedly, unavoidable, has significant natural disaster impact for countries around the Pacific Ocean include Indonesia. ENSO has a time series of predictors, so it can be predicted the Artificial Neural Network (ANN). ANN has several important advantages over the more statistical models traditionally used. ANNs can accommodate non-linear relationship and the flexibility in testing multiple inputs. This research aims to predict the onset of ENSO using the ANN-backpropagation method of learning rate and momentum variation. The prediction is based on several indexes during 1979-2018, i.e., Sea Surface Temperature (Nino 1.2, Nino 3, Nino 3.4, and Nino 4), Southern Oscillation Index (SOI), Multivariate ENSO and then verified with prediction data from the International Research Institute (IRI). The results of this prediction stated that in the period JAS (July-August-September) and DJF (December-January-February) 2020/2021 world climate conditions are in normal ENSO in which there are no El Nino and La Nina phenomena. Thus, the ANN-backpropagation method is an appropriate method to predict ENSO.