How weather models work

Short answers to the questions people ask most about weather models, each linking to the full explanation.

A computer simulation of the atmosphere that is started from today's observations and stepped forward in time.

Global, regional, ensemble, wave, ocean, AI and nowcasting models: what each is for and how they fit together.

Grid boxes, equations, time steps and the parts of the weather that have to be approximated.

Satellites, balloons, aircraft, ships, buoys and stations, and how data assimilation turns them into a starting point.

Machine-learning models such as AIFS, GraphCast and Pangu: how they differ from physics-based models, and where they are weaker.

Grid spacing, why a model sees features about five times larger than its grid, and what that means near a coast.

Even at the same resolution, models start from different data, use different physics and different settings.

Many slightly different runs instead of one: how to read plumes and spaghetti plots.

Why the first two or three days are reliable, day five shows trends, and day ten is mostly a range of possibilities.

What the wind number in a forecast really is, why gusts are higher, and why the coast is where forecasts go wrong.

Significant wave height, wind sea and swell, and why the period matters as much as the height.

Why track forecasts are better than strength forecasts, and what the cone of uncertainty does and does not show.

A practical way to combine a global model, a local high-resolution model and an ensemble for your area and passage.

Which model an app uses, how it picks a point from the grid, and what it adds on top.

A run is named after its start time, but the files appear hours later. Real delays for the models in the catalog.

What a GRIB file contains, how big one is for your area, and how to keep it small offshore.

How accuracy is measured, and a simple routine to learn how a model behaves in your own waters.

Short definitions of the terms used across the Learn pages.