Abstract
Reliable Earth system prediction requires ocean reanalyses that accurately represent both slowly varying circulation and high‑frequency surface variability. ORAS6, the successor to ORAS5, delivers major advances in ECMWF’s ocean and sea‑ice data assimilation, notably through an ensemble-based framework that enables variational assimilation of sea-surface temperature. These developments substantially improve the representation of upper-ocean thermal structure, ocean transports, the Gulf Stream pathway, and the SST diurnal cycle. Coupled forecasts initialized from ORAS6 show robust improvements from the medium to the seasonal range. By providing a more physically consistent ocean state, ORAS6 addresses several limitations identified in its predecessor ORAS5. One such limitation that will be discussed in more detail is a deficient initialisation of the Atlantic Meridional Overturning Circulation prior to the mid-1990s, which results in seasonal forecasts with strong biases and an apparent loss of predictive skill in North Atlantic SST. ORAS6 will underpin ERA6 and the next-generation seasonal forecasting system SEAS6 and is scheduled for publication in 2026.