Forecasting air temperatures using time series models and neural-based algorithms

Recently, neural network models have been popular and found useful for forecasting a wide variety of time series data in many disciplines. This is due to their favorable modeling properties of simplicity, fault and noise tolerance and their capability to adapt to process changes. Nevertheless, applications in climatology have been less widespread than other disciplines such as economics. In this study, feed-forward neural-network (FFNN) and autoregression (AR) time series models are used in forecasting the annual air temperature time series data in Jordan. The performance of the two predictors was compared using out-of-sample forecasts. The test period was shifted through the whole available time. As demonstrated by the forecasting experiments, the FFNN models gave better forecasts and were able to identify the dynamics of the temperature time series and gave more realistic forecasts. Both predictors showed a cooling trend in annual air temperatures for the coming 10 years.