How does ENSO diversity limit the skill of tropical Pacific precipitation forecasts in dynamical seasonal predictions?

Abstract

The prediction skill of tropical Pacific sea surface temperature (SST) and precipitation has been investigated, based on retrospective forecasts from 1983 to 2005 of the APEC Climate Center multimodel ensemble (MME) 6-month climate prediction, with a focus on El Nino-Southern Oscillation (ENSO) and its diversity. It was found that the MME prediction skill is related to both strength and flavor of ENSO. To better analyze the relationship between ENSO diversity and tropical climate predictions on a year-to-year basis, anomalous SST data were stratified into those due to typical ENSO and their residue. The former is defined as the reconstruction of data based on the leading empirical orthogonal function of the Pacific SST itself. It was found that typical ENSO and its impacts on tropical rainfall are well captured by dynamical seasonal prediction systems. Singular vector decomposition (SVD) analyses were further carried out, in order to identify climate modes related to the co-variability between residual SST signals and precipitation. Observational results show that the first SVD mode gives a SST singular vector that greatly resembles the east-west shift from typical El Nino features to El Nino diverse patterns. Dynamical models, on the other hand, have difficulties in reproducing the strength of this mode on the interannual scale. Their skill in predicting the temporal variation of this mode is also much lower than that for typical ENSO. Finally, the role of ENSO diversity on SST and precipitation potential predictability was also examined. Typical ENSO is the major predictability source of tropical Pacific SST. On the other hand, it was demonstrated that residual ENSO variability actually acts as a limit to tropical rainfall predictability. In other words, residual ENSO plays a rather crucial role in seasonal forecasting of tropical Pacific rainfall.