We explore to what extent data-driven prediction models have skills in forecasting daily sea-surface temperature (SST), which are comparable to or perform better than current physics-based operational systems over long-range forecast horizons. Three hybrid deep learning-based models are developed within the South China Sea (SCS) basin by integrating deep neural networks (back propagation, long short-term memory, and gated recurrent unit) with traditional empirical orthogonal function analysis and empirical mode decomposition. Utilizing a 40-year (1982–2021) satellite-based daily SST time series on a 0.25° grid, we train these models on the first 32 years (1982–2013) of detrended SST anomaly (SSTA) data. Their predictive accuracies are then validated using data from 2014 and tested over the subsequent seven years (2015–2021). The models’ forecast skills are assessed using spatial anomaly correlation coefficient (ACC) and root-mean-square error (RMSE), with ACC proving to be a stricter metric. A forecast skill horizon, defined as the lead time before ACC drops below 0.6, is determined to be 50 days. The models are equally capable of achieving a basin-wide average ACC of ~0.62 and an RMSE of ~0.48 °C at this horizon, indicating a 36% improvement in RMSE over climatology. This implies that on average the forecast skill horizon for these models is beyond the available forecast length. Analysis of one model, the BP neural network, reveals a variable forecast skill horizon (5 to 50 days) for each individual day, showing that it can adapt to different time scales. This adaptability seems to be influenced by a number of mechanisms arising from the evident regional and global atmosphere–ocean coupling variations on time scales ranging from intraseasonal to decadal in the SSTA of the SCS basin.

On ensemble‐based coupled data assimilation, cross‐component parameter estimation (CPE), has not been as extensively developed and applied as weakly coupled state and parameter estimation along with cross‐component state estimation. This discrepancy is partially attributed to the lack of emphasis on the instantaneous response of coupled model states with respect to parameters across different components. We define so‐called response as the instantaneous parameter sensitivity (IPS). Under the framework of sequential assimilation, the prior information heavily relies on the IPS of coupled states with different time scales. Based on the IPS analysis for an intermediate coupled model, a series of twin experiments of state and parameter estimation are conducted, in which an IPS‐inspired adaptive inflation scheme for parameter ensemble is introduced. Results show that the success of a parameter estimation strategy is closely tied to the significant IPS of the observed state to the parameter targeted for optimization, as it maintains a high signal‐to‐noise ratio in the error covariance between parameter and prior state, thereby enhancing parameter estimation. An interesting finding in the context of IPS‐based CPE is: an atmospheric parameter can be successfully estimated by assimilating observations from slow‐varying oceanic component, but not vice versa. In comparison with cross‐component state estimation, successful CPE significantly enhances the estimation accuracy of coupled states by mitigating model bias.

This study introduces two distinct post-processing strategies to address systematic biases in sea surface temperature (SST) numerical forecasts, thereby enhancing SST predictive accuracy. The first strategy implements a spatiotemporal four-dimensional multi-grid analysis (4D-MGA) scheme within a three-dimensional variational (3D-Var) data assimilation framework. The second strategy establishes a hybrid deep learning architecture integrating empirical orthogonal function (EOF) analysis, empirical mode decomposition (EMD), and a backpropagation (BP) neural network (designated as EE–BP). The 4D-MGA strategy dynamically corrects systematic biases through a temporally coherent extrapolation of analysis increments, leveraging its inherent capability to characterize intrinsic temporal correlations in model error evolution. In contrast, the EE–BP strategy develops a bias correction model by learning the systematic biases of the SST numerical forecasts. Utilizing a satellite fusion SST dataset, this study conducted bias correction experiments that specifically addressed the daily SST numerical forecasts with 7-day lead times in the Kuroshio region south of Japan during 2017, systematically quantifying the respective error reduction potentials of both strategies. Quantitative verification reveals that EE–BP delivers enhanced predictive skill across all forecast horizons, achieving 18.1–22.7% root–mean–square error reduction compared to 1.2–9.1% attained by 4D-MGA. This demonstrates deep learning’s unique advantage in capturing nonlinear bias evolution patterns.

This study developed the MaCOM Operational Hybrid Data Assimilation System (MOHDAS), which integrates 2 independently developed core technologies. The first is the regional high-resolution Mass Conservation Ocean Model (MaCOM), China’s first self-developed, mass-conserving operational ocean circulation model. The second is an innovative hybrid data assimilation framework that combines the analytical 4-dimensional ensemble variational (A-4DEnVar) method with the 4-dimensional multigrid analysis method. The 4-dimensional multigrid analysis method separates multiscale signals from observational data, thereby providing A-4DEnVar with ensemble members of high signal-to-noise ratio. In addition, a dimensionality-reduction algorithm for A-4DEnVar (DRA-4DEnVar), a preprocessing-based efficient observation projection module, and adaptive workflow control and monitoring scripts are also incorporated into the system. Operational evaluations conducted from 2025 July 1 to September 30 demonstrate that MOHDAS outperforms the state-of-the-art GLO12v4 product from Mercator Ocean in the analysis of temperature/salinity profiles as well as sea surface temperature and successfully captures signals from extreme weather events, such as typhoons, and monthly to subseasonal climate anomalies. MOHDAS provides reliable 1- to 7-d forecasts of temperature/salinity profiles, with higher sea-surface-temperature accuracy than forecasts initialized from the GLO12v4 analysis of Mercator Ocean. The system effectively corrects and maintains systematic biases at a low level throughout the forecast period. Furthermore, the DRA-4DEnVar-based hybrid data assimilation strategy demonstrates promising nonlinear assimilation capability, highlighting the robustness of MOHDAS and its potential for broader applications.

High-resolution temperature and salinity ( T & S ) gridded datasets are essential for exploring large and mesoscale ocean phenomena. In this study, a global, weekly T & S gridded dataset with a horizontal resolution of 1/4° and a depth range of 0–1500 m from 2005 to 2023 is reconstructed using a four-dimensional multigrid analysis (4D-MGA) framework, based on the T & S profiles in WOD23. With minimal prior statistical assumptions, the 4D-MGA efficiently extracts information from in-situ T & S profiles and satellite-observed sea level anomalies by integrating multiscale spatiotemporal correlation and physical constraints. The results show that the 4D-MGA product successfully delivers a credible, high-resolution analysis that combines robustness with reliable mesoscale information. Specifically, our product exhibits a lower root mean square error and excellent unbiased performance on a global scale when compared with the ARMOR3D product and the GLORYS reanalysis. Furthermore, the clear recirculation in the western boundary currents (WBCs) can be well captured by the 4D-MGA product. The product is then applied to investigate the linear trends of geostrophic transport within five key sections in WBCs. The Kuroshio and Agulhas Current transports exhibit significant decreasing trends, with rates of 0.34 ± 0.07/0.26 ± 0.06, 0.23 ± 0.05/0.10 ± 0.06, and 0.30 ± 0.06/0.12 ± 0.05 Sv yr ⁻¹ in 4D-MGA, ARMOR3D, and GLORYS, respectively. The weekly reconstructed dataset (4D-MGA) is freely available at https://doi.org/10.5281/zenodo.19378150 (Wu et al., 2026a).