A quantitative comparison of ENSO prediction methods

The El Niño Southern Oscillation (ENSO) is one of the most important inter-annual climate phenomena with worldwide impacts. It can influence daily temperature and rainfall, as well as cause extreme weather events and natural disasters. Therefore, early and reliable prediction of the onset and magnitude of ENSO is crucial for different stakeholders. In order to overcome the “spring predictability barrier” in ENSO prediction, recent studies have developed some analysis tools and put forward some forecasting indices based on climate network, claiming they have achieved the long-lead-time (over 6 months) forecasts. However, there are few kinds of research to quantitatively compare the predictive power of these methods. Thus developing a method to measure the quality of these forecasts and compare their predictive power is necessary and meaningful for the improvement of ENSO prediction skills. In these existing researches, in order to set the threshold or estimate the accuracy of the prediction, the standard El Niño indices such as Oceanic Index (ONI), Niño 3.4 Index and etc., are often used to be compared with the invented indices series. In this research, we look into these comparisons and results, and use the receiver operating characteristic curve (ROC) to quantitatively compare these recent analysis tools. Additionally, for demonstrating that the results are not accidental, randomized series obtained by reshuffling the temperature records are analyzed. In this paper, we use the method of surrogate data instead of using shuffle data in the evaluation procedure of the prediction to further improve the evaluation method of the El Nino prediction.

 

(This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 813844.)