Inside the end-to-end ocean forecasting system

After exploring the general concepts of ocean forecasting and the historical timeline from ancient mariners to today’s digital opportunities, it is time to understand the main components that make up the ocean value chain (Alvarez Fanjul et al., 2022). This value chain combines ocean observations, data processing, and numerical models to build the services that end-users rely on.

Data processing and pre-processing

Before observations can be used by a forecasting system, they need to be processed. Data processing and pre-processing covers the set of procedures involved in accessing observations — and upstream data more broadly — and applying numerical procedures to perform quality control, standardise data formats, and deliver the data to a data store for final access by the marine core service (Novellino et al., 2025).

This step is essential but often invisible to end-users. Without rigorous quality control and standardisation, the raw data flowing in from satellites, buoys, and other observation platforms would be inconsistent and unreliable for use in forecasting.

Data assimilation

Data assimilation is the science of combining different sources of information to predict the possible states of a system as it evolves over time. The term covers the entire chain of operations — from the observations of a system, through the necessary constraints needed to build a process, to ultimately providing an estimate of that system’s state.

In ocean science, data assimilation is a fundamental process that integrates models and observations into comprehensive and reliable estimates of the ocean’s state. It serves two key purposes (Martin et al., 2025):

• Producing near-real-time initial conditions, known as analyses, from which ocean forecasts are generated
• Generating reconstructions of the ocean’s past state, known as reanalyses

Without data assimilation, numerical models would gradually drift away from reality — data assimilation is what keeps them anchored to what is actually happening in the ocean.

Numerical models

Numerical models are a digital representation of the ocean, where physical processes are solved on a three-dimensional grid (Bell et al., 2025). To run, a numerical model requires several types of upstream data:

• Bathymetric datasets, describing the shape of the ocean floor
• Atmospheric forcing, capturing the influence of wind, pressure, and heat exchange with the atmosphere
• River forcing data, accounting for freshwater input from rivers
• Other ocean model data, used to set initial and lateral boundary conditions

Once a numerical model has been run, it produces a set of numerical results. These results are then post-processed, applying standardisation procedures to ensure the final delivery of Essential Ocean Variables (EOVs) to users in a consistent, usable format.

How the components fit together

Each of these three components — data processing, data assimilation, and numerical modelling — depends on the others. Observations alone cannot give us a complete, dynamic picture of the ocean. Numerical models alone, without real-world observations feeding into them, would gradually lose touch with reality. It is only when these elements are combined, through careful data processing and rigorous data assimilation, that an end-to-end ocean forecasting system can deliver products that are both accurate and trustworthy (Alvarez Fanjul et al., 2022).

This is the system you will continue to explore throughout the rest of this course — each subsequent week will take you deeper into one or more of these components.

Key takeaways

• Data processing and pre-processing ensure that raw observations are quality-controlled, standardised, and made accessible to the marine core service
• Data assimilation combines observations with model output to produce accurate estimates of the ocean’s current and past states — generating analyses and reanalyses respectively
• Numerical models represent the ocean digitally on a 3D grid, requiring upstream data such as bathymetry, atmospheric forcing, and river forcing to run
• These three components work together, not in isolation, to produce the reliable ocean forecasts and products that end-users depend on

References

• Alvarez Fanjul, E., Ciliberti, S., and Bahurel, P. (Eds.) (2022). Implementing Operational Ocean Monitoring and Forecasting Systems. IOC-UNESCO, GOOS Report No. 275. Available at: https://www.unoceanprediction.org/en/resources/etoofs-guide
• Novellino, A., et al. (2025). Merging and serving ocean observations: a description of marine data aggregators. State of the Planet, 5-opsr, 8. https://doi.org/10.5194/sp-5-opsr-8-2025
• Martin, M. J., et al. (2025). Data assimilation schemes for ocean forecasting: state of the art. State of the Planet, 5-opsr, 9. https://doi.org/10.5194/sp-5-opsr-9-2025
• Bell, M. J., et al. (2025). Numerical models for simulating ocean physics. State of the Planet, 5-opsr, 10. https://doi.org/10.5194/sp-5-opsr-10-2025