Markov model studies of the El Niño-Southern Oscillation
Aspects of the El Nino-Southern Oscillation are investigated using statistical models derived from observational data sets of surface and subsurface temperature in the tropical Pacific Ocean. Topics examined include: the existence of a seasonal cycle in the unforced internal dynamics of the sea surface temperature (SST) anomaly field; useful forecasting information contained in the subsurface temperature anomalies not reflected in SST; Geographical patterns of SST anomaly and subsurface temperature anomaly that lead to highest growth of SST anomaly in a given period of time.Annual variation in the Markov model is sought using various tests. It is determined with a high degree of certainty that the seasonal transition matrices are not season-interchangeable, indicating that the internal dynamics are not annually constant. A numerical model representation of ENSO should therefore include a seasonal cycle in the internal (unforced) dynamics. The forecast skill of a Markov model derived form SST anomalies alone is compared to the skill of a Markov model derived from surface plus subsurface temperature anomalies. The Markov model that includes the subsurface information possesses increased forecast skill at all lead times. The second principal component of heat content anomaly, which appears to contribute information on the transition phase between warm and cold ENSO events, is the primary contributor to the improvement in skill. The skill added by including subsurface information, although robust, is not shown to be statistically significant.The qualitative structure of the empirically derived optimal initial and final patterns of SST and subsurface temperature anomaly are found to be fairly robust. The optimal final pattern of both SST and heat content anomaly is identical to the respective fields during the peak of an ENSO event. Skill of forecasts of Nino indices using only the projection of the SST anomaly state onto the t -month SST optimal pattern nearly matches the skill of forecasts by the full empirically-derived model, which indicates that a sizeable fraction of ENSO forecast skill is captured by resolving the ENSO mode at the forecast start time.
- Atmospheric sciences