"Which data do I need?" is the first question of any oceanographic project, and the one most often answered badly. Choosing the wrong source means paying for useless resolution or, the other way round, missing the very phenomenon you are trying to describe. Here is how we arbitrate between WaveWatch III, CROCO and Copernicus Marine.
Start from the variable, not the model
Before talking about tools, you must name the physical quantity that drives your decision. Each variable has a natural source:
- Waves (height, period, direction) — WaveWatch III, a spectral sea-state model.
- Currents, water level, temperature, salinity — CROCO in the coastal zone, CMEMS at basin scale.
- Basin conditions and long reanalyses — Copernicus Marine, for context and validation.
WaveWatch III — the sea state
WW3 resolves the wave spectrum at basin scale, typically from a few kilometres to a few tens of kilometres of resolution. It is the reference for offshore swell and its propagation towards the coast. On the other hand, it describes neither currents nor the fine transformation of waves in very shallow water, which belong to a dedicated coastal model.
CROCO — the coast at high resolution
CROCO models coastal ocean circulation: currents, tide, water level, temperature and salinity, down to a resolution of a few tens of metres. It is the right tool as soon as bathymetry and the shoreline govern the phenomenon — estuaries, bays, harbour approaches. Its computational cost grows quickly with resolution, which makes it a model you nest over a targeted area rather than over a large extent.
Copernicus Marine — the basin context
CMEMS provides operational forecasts and reanalyses at regional and global scale, free of charge and homogeneously over time. Two major uses: feeding the boundary conditions of a coastal model, and serving as a reference to validate a modelling chain over several years. Its resolution, on the order of a kilometre to tens of kilometres, remains too coarse to describe littoral dynamics on its own.
The right combination rather than the right model
In practice, the question is almost never "which one to choose" but "how to articulate them". A common pattern: CMEMS and WW3 at the boundaries, CROCO nested over the area of interest, then AI-based downscaling to reach the final resolution at a controlled cost — an approach we detail in our article on AI coastal forecasting.
The right starting point remains a scoping exercise: which variable, which extent, which resolution and which acceptable error. Tell us about your need and we propose the leanest data assembly that answers it.