FY 1991

Comparison among autoregression models for forecasting El Niño events

Shuzhen, P., Y. Huiling, J. Toole, B. Millard, M.J. McPhaden, and L.J. Mangum

In Air-Sea Interaction in Tropical Western Pacific, Proceedings, US-PRC International TOGA Symposium, China Ocean Press, Beijing, 1988, 59–66 (1990)

SST (Sea Surface Temperature) data in a large area is used for spacial averages from which annual variation is subtracted to obtain a time series of SST anomalies of the cold tongue in the Tropical Eastern Pacific Ocean (Jan. 1951–Dec. 1985). An autoregression model, a self-exciting threshold autoregression model and an open loop autoregression model were developed, based on the time series. The interannual variations are simulated by those models. Results of simulations show that all the three models give good hindcasting of the nine El Niño events since 1951. To test reliability of the open loop threshold model, forecasting was extended to the period Jan. 1986–Feb. 1987. It can be seen from the forecasting that the model can forecast the beginning stage and strengthening stage of the recent El Niño event (1986–1987). Correlation coefficients of estimations to observations are respectively 0.84, 0.88 and 0.89, for the three models. All the models obviously work well. The open loop threshold autoregression model is the best and most useful for monitoring the interannual variation of the cold tongue SST in the Eastern Tropical Pacific Ocean and for estimating El Niño strength.