Numerical weather prediction

Categories: Meteorology

Numerical Weather Prediction is the science of predicting the weather using computational techniques. Manipulating the huge datasets and performing the complex calculations necessary to do this on a resolution small enough to make it accurate requires some of the most powerful supercomputers in the world.


Contents

[edit]

Models

There are many different methods of determining the weather in this way; each of these methods is called a Model. Numerical Weather Prediction models may be global or regional, and typically represent meteorological variables in three spatial dimensions as they evolve in time. Global models are generally based on a two-dimensional spectral horizontal coordinate system with a discrete finite-difference vertical coordinate. Regional models are generally finite-difference in all three spatial dimensions. These models are based on the equations of motion as well as equations that govern the conservation of energy and mass. Some of the better known ones are:

  • GFS Global Forecast System (previously AVN) - developed by the NOAA - output is freely available
  • WRF Weather and Research Forecasting Model - developed jointly by the operational and research NWP communities
  • NAM North American Mesoscale (Formerly ETA) an international model originating in Yugoslavia in the 1970s
  • MM5 The Fifth Generation Penn State/NCAR Mesoscale Model
  • UKMET developed by the UK Met Office - output is restricted
  • GEM Global Environmental Multiscale - developed by the Meteorological Service of Canada (MSC)

others include RUCS, NOGAPS, COAMPS. The University Corporation for Atmospheric Research has a page describing the characteristics of these models

[edit]

GFS

The GFS is the global model run by the NOAA. Its output can be downloaded for free from the National Centers for Environmental Prediction website. This model is run 4 times a day and produces 3-hourly forecasts up 16 days in advance, but with decreasing spatial and temporal resolution over time (though it is widely accepted that beyond 7 days the forecast is little better than guesswork). Unlike conventional forecasts the 6 hour gap between runs does not allow actual measurements to be collected in time to be used as the starting conditions, nor is the distribution of weather stations globally of high enough density to give worthwhile data. For this reason the input of the GFS is satellite derived data.

The model is run in two parts: the first part has a higher resolution and goes out to 180 hours in the future, the second part runs from 180 to 384 hours. The resolution of the model varies in each part of the model: Horizontally, it divides the surface of the earth into 35 or 70 kilometre grid squares; Vertically, it divides the atmosphere into 64 layers and Temporally, it produces a forecast for every 3rd hour for the first 180 hours, after that they are produced for every 12th hour .

Incidentally, this is the only global NWP output available for free over the internet (as determined by the US legislature), and as such is the basis for many web based forecast services, eg Weather Underground, AccuWeather, The Weather Channel, MetCheck.

[edit]

WRF

The WRF model is a community-oriented regional finite difference model developed by operational centers including (but not limited to) NCEP and AFWA as well as research organizations like NCAR, FSL. University researchers have made significant contributions to the project.

The project was designed to facilitate cooperation between the research and operational communities and to improve the efficiency of the research to operations cycle.

The NCEP is currently running WRF operationally for two hiresolution domains in the United States and plans to replace the Eta model with the WRF model as its North American mesoscale model in Spring of 2006 with a horizontal resolution of 10 km.

[edit]

Ensemble

Given the chaotic nature of the weather, very small changes in starting conditions can produce a radically different forecast. For this reason, where time allows, many runs of the same model are performed, each with slightly different initial conditions or model configurations, and the results combined to give an Ensemble forecast. If the output of all the models is similar, one can be confident in the predictions. If the results diverge, forecasts produced are likely to be prone to errors. The ensembles are designed to provide a spread of forecasts that fully captures the probability of occurrence of weather phenomena.

[edit]

Organizations

An incomplete list of major government organizations that develop and run NWP models includes:

The National Centers for Environmental Prediction (NCEP), the European Centre for Medium-Range Weather Forecasts (ECMWF), the UK Met Office (UKMO), the US Navy Fleet Numerical Meteorological Oceanographical Center (FNMOC), and the US Air Force Weather Agency (AFWA)

[edit]

See Also

[edit]

External Links