资源环境科技发展态势分析平台

The El Nino-Southern Oscillation (ENSO) has been shown to manifest as primarily two types, the eastern Pacific (EP) type and central Pacific (CP) type, in terms of the zonal positions of the sea surface temperature (SST) anomalies. This study focuses on examining the predictability of the two types of ENSO by developing statistical models for their corresponding Nino indices, which have their own distinct key precursors. The results show that the statistical predictability of the Nino indices representing the two types of ENSO primarily originates from the preceding variations in the equatorial Pacific upper-ocean heat content and the surface zonal wind stress, which intrinsically reflect the zonally uniform and contrasted thermocline patterns, respectively. The traditional Nino3 and Nino4 indices are more predictive than the Nino indices of the EP and CP ENSO types; however, all the indices are subject to predictability barriers with different timings and intensities, which might be weakened by introducing additional external precursors. The EP ENSO indices have overall higher skills than the CP indices, in which the statistical model has much higher skill scores than persistence forecast for the EP ones while it does less for the CP ones. We demonstrate that the precursors outside the tropical Pacific, e.g., the Indian Ocean Dipole, North Pacific oscillation, North American dipole, and Southern Hemispheric SST modes, except the northern tropical Atlantic SST, as suggested in previous studies, only make limited contributions to improving the prediction skills of the two ENSO types at specific initial months and leads compared to a benchmark model built using the equatorial Pacific heat content and zonal wind stress indices. This is primarily because these precursors have already transferred most of their signals into the variation of the two indices in the benchmark model. We further show that conditionally adding the northern tropical Atlantic SST precursor to the benchmark could provide considerable additional prediction skill scores for both types of ENSO and weaken the intensity of the ENSO predictability barriers that occur during boreal springsummer.