A Nonlinear Cause for the Seasonal Predictability Barrier of SST Anomaly in the Tropical Pacific

Abstract

AbstractThe seasonal Predictability Barrier (PB) of Sea Surface Temperature Anomaly (SSTA) is characterized by a rapid loss of prediction skills at a particular season in dynamic models. Here, the connection between seasonal PB and the inherent nonlinearity of SSTA was investigated using a statistical method known as Sample Entropy (SamEn). When the SamEn value is large, the chaotic degree of SSTA is high. In the Niño 3 and Niño 3.4 regions, the chaotic degree of SSTA was high in the spring; in the Niño 4 region, it was high in the summer. This result was consistent with the known PB occurrence season in these regions. The month when the chaotic degree of SSTA peaked moved westward longitudinally from March to June. This spatial‐temporal variation of the chaotic degree was consistent with that of the low SSTA variance and PB occurrence timing. Specifically, when the variance of SSTA was low, the low‐amplitude background SSTA signal was more dominant than the El Niño/Southern Oscillation (ENSO)‐related signal, the chaotic degree of SSTA was high, and the PB phenomenon occurred. Hence, the results indicated that the background SSTA signal was more chaotic than the ENSO‐related signal and the seasonal PB may result from the inherent low predictability of the chaotic background SSTA signal. Furthermore, the correlation coefficient between SSTA and shortwave flux anomaly showed similar variation along the longitude compared with the chaotic degree, which suggested that the seasonal connection between SSTA and atmospheric forcing may be responsible for the spatial‐temporal variation of the chaotic degree.