New oceanic data assimilation system improves prediction accuracy

Over 70% of the Earth is covered by oceans. Altogether, the Earth’s oceans contain approximately 1.338 billion cubic kilometers of water that greatly influence the world’s weather and climate.

Collecting oceanic data across such a vast area is no simple task, and remote sensors, including buoys, satellite data and autonomous vehicles, are limited. Despite this, researchers can combine relatively sparse oceanic observations with numerical models to produce an estimate of ocean conditions at any given time, improving the accuracy of weather forecasts and climate predictions.

The data oceanographers collect from the oceans has improved over time to include real-time temperature and salinity data. However, improved forecasting also relies heavily on numerical modeling that can accurately reflect ocean conditions based on the data provided. Over many decades, researchers have enhanced these numerical models to produce the global ocean data assimilation and forecast systems we use today, such as the GLO12V4, GIOPS 3.5.0, OceanMAPS v4.0i, FOAM GOSI9, GOFS 3.5, and NMEFC-NEMO platforms.

Recently, a group of researchers from National University of Defense Technology in Changsha, China has improved the numerical modeling for Earth’s oceans. Their platform, called the Yin-He Global Ocean Data Assimilation and Forecast System (YHGO), provides more accurate estimations of oceanic conditions than older platforms.

The team published the study in the November 2025 issue of the journal Ocean-Land-Atmosphere Research.

“We aimed to overcome the limitations of traditional ocean data assimilation systems, which often rely on linear or weakly nonlinear assumptions and Gaussian error distributions and are built on volume-conserving ocean models that do not fully represent real ocean physics,” said Senliang Bao, assistant researcher in the College of Meteorology and Oceanography at the National University of Defense Technology and author of the research paper.

Specifically, traditional systems utilize data assimilation and numerical models that don’t accurately reflect true ocean dynamics. Older platforms, for example, may rely on linear assumptions where the output is a simple proportional function of the input. Gaussian error distributions assume that random measurement errors will follow a bell-shaped curve. True ocean physics may not always follow these assumptions, limiting modeling accuracy.

“YHGO successfully integrates a fully nonlinear and non-Gaussian data assimilation method with a mass-conserving ocean model, leading to improved accuracy in sea-surface temperature and sea-level anomaly forecasts compared to the widely used GLO12V4 system,” said Yan Chen, assistant researcher from the College of Meteorology and Oceanography at the National University of Defense Technology.

The Mass Conservation Ocean Model (MaCOM) assumes that mass is neither created nor destroyed and that the total amount of water in the model remains constant. Unlike traditional volume-based models, mass conservation models can more directly predict actual sea surface height and salinity changes by accurately accounting for global changes from factors like ice melt, which adds mass to the oceans. The model was developed independently by the National Marine Environmental Forecasting Center at the Ministry of Natural Resources, China and includes integrated global–regional numerical simulation capabilities and sea-ice simulation features.

To test the accuracy of YHGO versus traditional ocean data assimilation systems, the team  generated one- to 10-day forecasts between August 15, 2024 to November 30, 2024 using the YHGO and GLO12V4 platforms. The results demonstrated that YHGO achieves higher accuracy in both sea-surface temperature and sea-level anomaly predictions compared to the popular GLO12V4 system. The team found systematic biases in YHGO forecasts, however, which need to be resolved to enhance the overall performance of the platform.

“Our next step is to identify and reduce the systematic biases in temperature and salinity profile forecasts. The ultimate goal is to further optimize the YHGO system to provide more reliable and comprehensive global ocean predictions for both scientific research and operational applications,” said Bao.

Yu Cao, Huizan Wan and Weimin Zhang from the College of Meteorology and Oceanography at the National University of Defense Technology in Changsha, China  also contributed to this research.

This work was supported by the National Key R&D Program of China (2021YFC3101500), the National Natural Science Foundation of China (NSFC, 42306039, 42406195, and 42305176) and the Youth Independent Innovation Science Foundation (ZK24-54).