Improving seasonal forecasting through tropical ocean bias corrections. (4th October 2016)

• Record Type: Journal Article
• Authors: Mulholland, David P.; Haines, Keith; Balmaseda, Magdalena Alonso
• Is Part Of: Quarterly journal of the Royal Meteorological Society. Volume 142:Number 700(2016)
• Page Start: 2797
• Page End: 2807
• Publication Date: 2016-10-04
• Subjects: seasonal forecasting -- climate prediction -- ENSO -- El Niño -- data assimilation -- ocean dynamics -- shocks; Meteorology -- Periodicals; 551.5
• DOI: 10.1002/qj.2869
• Languages: English
• ISSNs: 0035-9009

Abstract

Abstract : Initialization shock is often discussed in the context of coupled atmosphere–ocean forecasting, but its detection has remained elusive. In this article, the presence of initialization shock in seasonal forecasts is clearly identified in the variability of the tropical thermocline. The specific source of shock studied here is the use of a bias correction procedure to account for errors in equatorial wind stress forcing during ocean initialization. It is shown that the abrupt removal of the bias correction at the beginning of the forecast leads to rapid adjustments in the upper ocean, creating a shock that remains in the system for at least 3 months. By contrast, gradual removal of the correction term, over 20 days, greatly reduces the initialization shock. Evidence is presented of substantial increases in sea surface temperature (SST) seasonal forecast skill, at around 3–7 months' lead time, when the gradual removal approach is used. Gains in skill of up to 0.05, as measured by the anomaly correlation coefficient for SST in the Niño‐4 region, are found, using a modest hindcast set covering four seasonal start dates. The results show that improvements in coupled initialization aimed at reducing shocks may considerably benefit seasonal forecasting.