Forecasting Monthly Rainfall in the Bowen Basin of Queensland, Australia, Using Neural Networks with Nino Indices

摘要

For three decades there has been a significant global effort to improve El Nino-Southern Oscillation (ENSO) forecasts with the focus on using fully physical ocean-atmospheric coupled general circulation models (GCMs). Despite increasing sophistication of these models and the computational power of the computers that drive them, their predictive skill remains comparable with relatively simple statistical models. In this study, an artificial neural network (ANN) is used to forecast four indices that describe ENSO, namely Nino 1+2, 3, 3.4 and 4. The skill of the forecast for Nino 3.4 is compared with forecasts from GCMs and found to be more accurate particularly for forecasts with longer-lead times, and with no evidence of a Spring Predictability Barrier. The forecast values for Nino 1 + 2, 3, 3.4 and 4 were subsequendy used as input to an ANN to forecast rainfall for Nebo, a locality in the Bowen Basin, a major coal-mining region of Queensland.