Parameterization and Model Limitations

The last lecture showed that no matter how fine a model’s grid, some processes are always too small to capture. This lecture, with Emanuela Clementi, is about what a model does with those unresolved processes – and, just as importantly, where the limits of any model lie.

The lecture builds the idea step by step:

• Why parameterization is needed – the ocean’s equations are continuous, but models solve them on a finite grid. Only processes larger than a grid cell can be resolved explicitly; everything smaller – the sub-grid processes – must be represented another way
• What remains unresolved – even “eddy-resolving” models miss turbulent motions, sub-mesoscale eddies, vertical mixing, and small bathymetric features. There is always a scale below the grid
• What parameterization is – representing those unresolved processes through simplified mathematical relationships, based on theory, observations, or empirical formulas. Rather than resolving a small-scale process directly, the model estimates its average effect on the scales it can see
• The key parameterizations – vertical turbulence and mixing (which control stratification, mixed-layer depth, and heat exchange), horizontal eddy mixing and diffusion, and – especially in coastal work – wave-current interactions and sub-grid bathymetry
• Why it’s critical for forecasting – operational forecasts must run fast, so high resolution everywhere isn’t feasible; forecast skill therefore depends strongly on the parameterization choices
• Strengths and weaknesses – parameterizations make complex modelling possible, cut computational cost, and enable operational forecasting. But they rest on empirical assumptions, need tuning and calibration, and may not carry over from one region or condition to another

The message to carry forward is an honest one: parameterizations are not exact physics, but informed approximations – and they are essential to every ocean forecasting system. Improving them matters just as much as increasing resolution. This is also where a model’s limitations come from, and why every forecast carries some uncertainty – the theme that runs into the rest of the course.