Home
/ Nature & Environment / Climate Change / Fundamentals of Ocean Forecasting / How Observations Improve Ocean Forecasts
How Observations Improve Ocean Forecasts
Forecasts drift as errors grow. See how a steady stream of observations keeps them sharp, detailed, and skilful.
The last lecture showed what data assimilation is. This one, again with Babette Tchonang, answers the follow-up question: how do observations actually improve a forecast? The answer turns on a fundamental feature of all forecasting – that errors grow with time – and on how observations keep pushing back against that growth.
The lecture builds the argument in stages:
- Observations are the critical link – the whole forecasting cycle depends on them. Observations from satellites, Argo floats, moorings, drifters, gliders, and ships are quality-controlled, assimilated to make the analysis, and turned into forecast products for marine safety, fisheries, climate monitoring, and disaster risk reduction – with a continuous feedback loop that keeps improving both the models and the observing networks
- Why forecast error grows – even starting from the best possible analysis, a forecast is never perfect. Small uncertainties in the model and the observations accumulate, so the forecast gradually diverges from the real ocean. Error rises with lead time – at different rates for temperature, salinity, sea surface temperature, sea level, and currents – which is exactly why fresh observations must be fed in again and again
- How assimilation pushes back – each time new observations arrive, a new analysis is made, and the next forecast cycle restarts from the most accurate estimate available. Without these regular updates, forecast skill would steadily decay
- Seeing the difference – a striking real example: as more observations are assimilated, the ocean comes into focus. A model with no assimilation captures only the broad circulation, with fronts and eddies weak or missing; adding satellite observations sharpens the currents, fronts, and mesoscale eddies; and adding high-resolution SWOT data reveals finer structures still. Each layer of observation makes the analysis more realistic – and the forecast more skilful
The take-home message is simple: observations are what keep a forecast anchored to reality. Left alone, any model drifts; fed a steady stream of good observations, it stays sharp. The more – and the better – the observations, the more realistic the analysis and the better the forecast. The final lecture asks how we measure that improvement: validation and quality control.
