A new DRP-4DVar-based coupled data assimilation system for decadal predictions using a fast online localization technique

A new coupled data assimilation (CDA) system based on dimension-reduced projection four-dimensional variational data assimilation (DRP-4DVar) for decadal predictions is developed and applied to a fully coupled model FGOALS-g2, which applies a fast online localization technique. The improved CDA system can assimilate more observational information than the previous system, with a larger reduction in the observational cost function and smaller biases and root mean square errors (RMSEs) in global mean ocean temperature and salinity. The assimilated climate variabilities in the Pacific and Atlantic, such as El Nin ~ o-Southern Oscillation (ENSO), Pacific Decadal Oscillation (PDO), North Pacific index (NPI), Pacific-North American teleconnection (PNA) and Atlantic Multi-decadal Oscillation (AMO), are generally better reproduced by the new CDA system compared to the previous system and the uninitialized simulation. The Atlantic meridional overturning circulation (AMOC) is better described than the previous system. After partially restoring the climatology of the initial condition to that of the model to relieve the initial shock, ten-member decadal prediction experiments are started each year from 1961 to 1996. Higher decadal prediction skills of near-surface air temperature anomalies over the North Pacific, the Atlantic, and the Indian Oceans and the Eurasia Continent are achieved by the new system compared to those obtained by the previous CDA system, persistence and the uninitialized simulation.