What Is Data Assimilation?

This lecture, with Babette Tchonang, introduces one of the most important ideas in the whole week – and in modern operational oceanography: data assimilation, the process that combines observations with a numerical model to produce the best possible estimate of the ocean state.

The lecture builds the idea step by step:

• Why it’s needed – observations are accurate but sample only part of the ocean; models give continuous coverage but gradually drift away from reality. Data assimilation regularly pulls the model back toward the observations, producing an estimate – the analysis – that stays close to the real ocean
• How it works – the elegant cycle at the heart of assimilation, told through four key terms:
  • The first guess (or background) – the model’s estimate before new observations are added
  • The innovations – the differences between the observations and that first guess: where, and by how much, the model disagrees with reality
  • The increments – the corrections the assimilation system computes from those innovations, and applies to the model’s temperature, salinity, sea level, and currents
  • The analysis – the improved estimate after assimilation, combining the accuracy of observations with the coverage of the model
• Where it fits – the analysis becomes the initial condition for the next forecast, which is why the quality of assimilation directly shapes the quality of every forecast
• What gets assimilated – both satellite (sea surface height, temperature) and in-situ (Argo, moorings) observations
• The methods, benefits, and limits – a conceptual look at the main assimilation approaches and how they’re used in global, regional, and coastal systems, and an honest account of what assimilation can and can’t do

Don’t worry about the mathematics – the lecture keeps it conceptual. The one idea to hold onto is the cycle: observe, compare, correct, forecast – and repeat. That loop, running continuously, is what keeps an operational forecast anchored to the real sea. The next lecture shows how much it actually improves forecast skill.