A Preliminary Exploration On The Predictability Barrier Of Warm Pool Type Eino Events

As the strongest ocean-air interaction phenomenon with interannual variability inthe tropical Pacific Ocean, El Ni o-Southern Oscillation has significant impacts onglobal climate, ecosystem, economy and so on. In recent decades, the observations,physical mechanism, numerical simulation and prediction of El Ni o have advancedby leap and bounds. However, since1990s, a new type of El Ni o events occur moreand more frequently in the tropical Pacific, which has warm sea surfacetemperature(SST) anomaly confined in the central Pacific Ocean and lead toatmospheric response and oceanic processes different with the "traditional" El Ni o.The new type El Ni o also brings significant impacts on global climate, ecosystemand economy. So far, many different nomenclatures have been used for the new typeEl Ni o events. Herein we collectively refer them as the warm pool El Ni o (WP typeEl Ni o), and the original―traditional‖El Ni o event as cold tongue El Ni o (CT typeEl Ni o)).It is known that when forecast a CT type El Ni o event before and through thespring, the prediction skill may decrease rapidly in numerical models. That’s theso-called "spring predictability barrier"(SPB). In this paper, we focus on whether suchpredictability barrier exists in WP type El Ni o events similarly. Using a complicatedglobal coupled ocean-atmosphere-land-ice model GFDL CM2.1, we try to explore thepossible seasonal predictability barrier and the initial error that correspond to asignificant seasonal predictability barrier in WP type El Ni o events.First, we investigate the model capability of GFDL CM2.1. The results show thatit can faithfully reproduce the spatial distribution of the sea surface temperature, seasurface current and wind field over tropical Pacific Ocean. Their climatology issimilar to the observation. Using a modified Ni o3index and a modified Ni o4index, the model could distinguish two type El Ni o events successfully, with maxwarm sea surface temperature anomalies located in eastern Pacific(CT type) and central-western Pacific(WP type),respectively. During CT type El Ni o events, widewesterly anomalies dominate the central-eastern Pacific Ocean, while during WP typeEl Ni o events, there is a wind convergence center in the central-western PacificOcean. The evolutions of sea surface temperature anomalies and wind anomaliescoincide with the observation as well.On this basis, we choose six typical WP type El Ni o events from the simulationand try to investigate the role of initial error plays in the prediction of WP type ElNi o events. The result shows that for different prediction area, the seasonalprediction error variability is also different. For tropical eastern Pacific Ocean, it isfound that the spring predictability barrier occurs most frequently. However, whentake tropical central-western Pacific Ocean as a prediction area, the prediction errorlikely increase fastest in summer. Besides, by analyzing the mean seasonal predictionerror variability, we found that:1. For the same event with different start months, the seasons of max meanprediction error variability are not consistent.2. For the same start month of different events, the seasons of max meanprediction error variability are not consistent.3. For the same events with the same start month, the seasons of max meanprediction error variability are not consistent in different prediction area.Clustering analysis for the initial error which lead to significant seasonalpredictability barrier shows that, some season predictability barrier doesn’t correspondto some initial error pattern. In other words, there is a kind of initial error which maylead to predictability barrier in all seasons. Therefore, for WP type El Ni o, theseasonal predictability barrier may be influenced by the event itself rather than thespatial pattern of initial error.