Glossary
Short definitions of the terms used across the Learn pages.
Short definitions of terms used in the Learn pages, in alphabetical order.
Terms
- Analysis
- The starting picture of the atmosphere for a model run, built by blending observations with the previous short forecast.
- Background
- The previous short forecast from the same model, used as the first guess when making a new analysis.
- Bias
- The average of a forecast’s errors, with sign. A negative wind bias means the model tends to forecast less wind than occurs.
- Boundary conditions
- The values a regional model takes from a larger model at its edges, updated through the run.
- Control run
- The unperturbed run in an ensemble, started from the best estimate of the atmosphere.
- Convection-permitting
- A model with a grid fine enough (roughly 4 km or less) to simulate showers and thunderstorms directly instead of approximating them.
- Cycle (run)
- One start of a model. Global models run at 00, 06, 12 and 18 UTC; some regional models run every three hours or hourly.
- Data assimilation
- The method that combines observations with a forecast to produce the analysis.
- Deterministic forecast
- A single forecast run, as opposed to an ensemble of many.
- Downscaling
- Adding local detail to a coarse forecast, with a finer model or a statistical method.
- Effective resolution
- The smallest size of feature a model shows well, typically about five to seven grid spacings.
- Ensemble
- A set of forecasts from one model started from slightly different conditions, used to measure uncertainty.
- GRIB
- GRIdded Binary: the standard compact file format for gridded weather data.
- Grid spacing
- The distance between neighbouring points in a model; the usual measure of resolution. 1 NM = 1.852 km.
- Gust
- A brief peak in wind speed, usually averaged over about three seconds, above the mean wind.
- Initial conditions
- The state of the atmosphere at the start of a run; the analysis.
- Knot (kn)
- One nautical mile per hour: 0.514 m/s, 1.852 km/h.
- Lead time
- How far ahead a forecast is, counted from the run time.
- Limited-area (regional) model
- A model that covers only part of the globe, at finer detail, with edges taken from a global model.
- Mean wind
- Wind speed averaged over a period, typically ten minutes in observations; the value forecasts show.
- Member
- One run within an ensemble.
- Nest
- A finer model grid placed inside a coarser one.
- Nowcasting
- Very short-range forecasting, for the next few hours, by extending the latest radar and satellite observations.
- NWP
- Numerical weather prediction: forecasting by solving the equations of the atmosphere on a computer.
- Parametrisation
- A simplified rule used for a process too small for the grid, such as convection or turbulence.
- Peak period (Tp)
- The time in seconds between crests of the dominant waves.
- Post-processing
- Correcting or combining model output after the run, for example against past observations.
- Reanalysis
- A re-run of the data assimilation over past decades with one fixed system, giving a consistent record of past weather.
- RMSE
- Root-mean-square error: a measure of the typical size of forecast errors, where large errors count more.
- Significant wave height (Hs)
- The average height of the highest third of the waves; the largest single wave is typically about twice that.
- Skill
- How much better a forecast is than a simple reference, such as assuming that tomorrow is like today.
- Spread
- The difference between the members of an ensemble; a measure of uncertainty.
- Step (forecast hour)
- The time after the run start that a forecast value is for. Valid time = run time + step.
- Swell
- Waves that have travelled out of the area where the wind made them: long, regular and smooth.
- UTC (Z, Zulu)
- Coordinated Universal Time, the time standard of forecasting and navigation. Model runs are named in UTC.
- Wind sea
- Short, steep waves made by the local wind, which change as the wind does.