The models in practice: a tour of operational and research systems
The video surveyed the models used around the world. This reading turns that survey into a reference you can return to – a curated tour of the major model families, what each is for, and how they fit together. The key idea: there is no single “best” model, only the right one for a given job, scale, and process.
Operational versus research models
Before the models themselves, one distinction shapes everything. Operational models run routinely – daily or more often – to deliver real-time forecasts, so above all they must be robust, efficient, and reliable. Research models are built to understand processes, and can be more complex or experimental, run for a specific study rather than on a fixed schedule. Many of the models below serve both worlds.
A hierarchy of scales
The other organising idea is scale. Models form a hierarchy, each handing information to the next:
- Global models (~1/12°, ~8-10 km) simulate the whole ocean. They provide the large-scale context – gyres, overturning circulation – and the boundary conditions that regional models need
- Regional models (~1-10 km) refine a specific basin or sea, forced at their edges by global output, adding local detail and local observations
- Coastal models (kilometres down to tens of metres) resolve ports, estuaries, storm surge, and flooding, handling complex bathymetry and wetting-and-drying
Downscaling is the strategy that links them: a global model’s output becomes the boundary condition for a finer regional or coastal model.
The major ocean circulation models
These are the “engines” that simulate ocean physics – the ones named through this week:
| Model | Key characteristic | Best known for |
|---|---|---|
| NEMO | Ocean, sea ice, and biogeochemistry components | Backbone of many European operational services (Copernicus Marine) – global and regional |
| HYCOM | Hybrid vertical coordinates (isopycnal / z-level / sigma) | Global forecasting; the US Navy’s operational system |
| ROMS | Terrain-following (sigma) coordinates | Coastal and regional forecasting – upwelling, estuaries, shelves |
| CROCO | High resolution, nesting, sediment and biogeochemistry modules | High-resolution coastal and multi-scale work |
| MITgcm | Highly flexible; can run non-hydrostatic | Ocean, atmosphere, and climate research; process studies |
| FVCOM | Unstructured grids | Complex coastal geometries – estuaries, tides, storm surge |
Beyond circulation: the wider model family
Ocean circulation is only part of the picture. A full forecasting system draws on specialised models for other components, many of which couple to the circulation models above:
- Biogeochemical models (PISCES, BFM, ERSEM) simulate the “Green Ocean” – carbon, nutrients, and plankton – often embedded in NEMO
- Wave models (WAVEWATCH III, WAM, SWAN) forecast the sea state, from global swell to nearshore breaking
- Sea-ice models (CICE and LIM, coupled to NEMO or HYCOM) simulate the “White Ocean” – ice growth, melt, and drift
- Storm-surge models (SLOSH, HAMSOM) predict coastal water levels for hazard warning
- Sediment-transport and coastal-process models (Delft3D, TELEMAC, MIKE) handle erosion, deposition, and engineering in estuaries and harbours – TELEMAC, for instance, has been used for tidal studies in the Wouri estuary, Cameroon
- Coupled Earth-system models (the CMIP6 models behind IPCC reports) join ocean, atmosphere, land, ice, and biogeochemistry for long-term climate projection
A full catalogue of these models and their applications is provided as a downloadable reference at the end of this step.
The challenges that remain
Even with all these tools, operational ocean modelling faces real limits:
- The high computational cost of high-resolution models
- Limited observations to constrain the small scales a fine model resolves
- The continuing need for accurate parameterizations
- The complexity of multi-scale interactions
And a recurring lesson from the whole week: more resolution does not always mean a better forecast – if there aren’t enough observations to constrain the extra detail, the gain may be small. Good modelling is a balance of accuracy and efficiency, not resolution for its own sake.
Key takeaways
- Models divide into operational (routine, robust, real-time) and research (scientific, flexible) – many serve both
- They form a hierarchy of scales – global, regional, coastal – linked by downscaling
- The major circulation models each have a niche: NEMO and HYCOM for global/operational, ROMS, CROCO, and FVCOM for coastal and regional, MITgcm for research
- A full system also needs wave, biogeochemical, sea-ice, surge, and sediment models, often coupled together
- There is no single best model – the right choice depends on the scale, the processes, the application, and the computing available
