We present the China Ocean ReAnalysis version 2 (CORA2) in this paper. We compare CORA2 with its predecessor, CORA1, and with other ocean reanalysis products created between 2004 and 2019 [GLORYS12v1 (Global Ocean reanalysis and Simulation), HYCOM (HYbrid Coordinate Ocean Model), GREP (Global ocean Reanalysis Ensemble Product), SODA3 (Simple Ocean Data Assimilation, version 3), and ECCO4 (Estimating the Circulation and Climate of the Ocean, version 4)], to demonstrate its improvements and reliability. In addition to providing tide and sea ice signals, the accuracy and eddy kinetic energy (EKE) of CORA2 are also improved owing to an enhanced resolution of 9 km and updated data assimilation scheme compared with CORA1. Error analysis shows that the root-mean-square error (RMSE) of CORA2 sea-surface temperature (SST) remains around 0.3°C, which is comparable to that of GREP and smaller than those of the other products studied. The subsurface temperature (salinity) RMSE of CORA2, at 0.87°C (0.15 psu), is comparable to that of SODA3, smaller than that of ECCO4, and larger than those of GLORYS12v1, HYCOM, and GREP. CORA2 and GLORYS12v1 can better represent sub-monthly-scale variations in subsurface temperature and salinity than the other products. Although the correlation coefficient of sea-level anomaly (SLA) in CORA2 does not exceed 0.8 in the whole region, as those of GREP and GLORYS12v1 do, it is more effective than ECCO4 and SODA3 in the Indian Ocean and Pacific Ocean. CORA2 can reproduce the variations in steric sea level and ocean heat content (OHC) on the multiple timescales as the other products. The linear trend of the steric sea level of CORA2 is closer to that of GREP than that of the other products, and the long-term warming trends of global OHC in the high-resolution CORA2 and GLORYS12v1 are greater than those in the low-resolution EN4 and GREP. Although CORA2 shows overall poorer performance in the Atlantic Ocean, it still achieves good results from 2009 onward. We plan to further improve CORA2 by assimilating the best available observation data using the incremental analysis update (IAU) procedure and improving the SLA assimilation method.

To effectively reduce model bias and improve assimilation quality, we adopt a hybrid adaptive approach of ensemble adjustment Kalman filter (EAKF) and multigrid analysis (MGA), called EAKF-MGA, to implement parameter optimization as follows. For each assimilation cycle, observations are used to adjust the prior ensembles of both state variables and parameters using the EAKF without inflation. Then, the MGA is adaptively triggered to extract multiscale information from the observational residual to innovate the ensemble mean of the state once again. Results of biased twin experiments consisting of a barotropic spectral model and idealized observation systems show that the proposed EAKF-MGA is insensitive to state variance inflation and localization during the parameter optimization process, compared with the EAKF with adaptive inflation. We also find that computational efficiency is another important advantage of the EAKF-MGA for both state estimation and parameter estimation since extremely small ensemble size is allowed, while the EAKF with adaptive inflation does not work anymore. In essence, the EAKF-MGA is designed to estimate and correct systematic errors jointly with model’s state variables. Through alleviating biases, including the model bias caused by the biased parameter and the analysis bias resulting from the sampling noise given the limited ensemble size, it can be guaranteed that the analysis in the EAKF-MGA will be proceeded onward with the standard assumption of the unbiased model background field in modern data assimilation theory to be met.

Ocean forecasting is critical for various applications and is essential for understanding air–sea interactions, which contribute to mitigating the impacts of extreme events. While data-driven forecasting models have demonstrated considerable potential and speed, they often primarily focus on spatial variations while neglecting temporal dynamics. This paper presents the TSformer, a novel non-autoregressive spatio-temporal transformer designed for medium-range ocean eddy-resolving forecasting, enabling forecasts of up to 30 days in advance. We introduce an innovative hierarchical U-Net encoder–decoder architecture based on 3D Swin Transformer blocks, which extends the scope of local attention computation from spatial to spatio-temporal contexts to reduce accumulation errors. The TSformer is trained on 28 years of homogeneous, high-dimensional 3D ocean reanalysis datasets, supplemented by three 2D remote sensing datasets for surface forcing. Based on the near-real-time operational forecast results from 2023, comparative performance assessments against in situ profiles and satellite observation data indicate that the TSformer exhibits forecast performance comparable to leading numerical ocean forecasting models while being orders of magnitude faster. Unlike autoregressive models, the TSformer maintains 3D consistency in physical motion, ensuring long-term coherence and stability. Furthermore, the TSformer model, which incorporates surface auxiliary observational data, effectively simulates the vertical cooling and mixing effects induced by Super Typhoon Saola.