Editing Numerical weather prediction (section)

===Weather forecasts===

Because weather drifts across the world, producing forecasts a week or more in advance typically involves running a numerical prediction model for the entire planet. Agencies use various software to do this, including:
* [[North American Ensemble Forecast System]], which combines results from:
** [[Global Forecast System]] from the US [[National Weather Service]]
** [[Global Environmental Multiscale Model]] from the [[Canadian Meteorological Centre]]
* [[Integrated Forecast System]] from the [[European Centre for Medium-Range Weather Forecasts]] and [[Météo-France]]
* [[Unified Model]], produced by a partnership of:
** UK [[Met Office]]
** Australia [[Bureau of Meteorology]]
** (South) [[Korea Meteorological Administration]]
** India [[National Centre for Medium Range Weather Forecasting]]<ref>[https://ncmrwf.gov.in/ncmrwf/NCUMG-Writeup-for-WEB-June2020.pdf Global NCMRWF Unified Model (NCUM-G) System]</ref>
** New Zealand [[National Institute of Water and Atmospheric Research]]
* [[Icosahedral Nonhydrostatic]] (ICON) from [[Deutscher Wetterdienst]], the German Meteorological Service
* [[Navy Global Environmental Model]] from the US Navy [[Fleet Numerical Meteorology and Oceanography Center]]
* Global Spectral Model and Global Ensemble Prediction System from the [[Japan Meteorological Agency]]<ref name="JMA">[https://www.jma.go.jp/jma/en/Activities/nwp.html Numerical Weather Prediction Activities]</ref>
* [[China Meteorological Administration]] Global Assimilation Forecasting System<ref name="CMA">[https://www.cma.gov.cn/en/forecast/highlight/202311/t20231117_5892086.html Numerical Weather Prediction]</ref>
* Brazilian Global Atmospheric Model (BAM) from [[Centro de Previsão do Tempo e Estudos Climáticos]] (CPTEC)

The global models can be used to supply [[boundary conditions]] to higher-resolution models that provide more accurate forecasts for an area of interest, such as the country served by a government agency, or an area where military action or rescue efforts are planned.
* Users of the Unified Model re-run the same system (hence the name) for a specific country or crisis zone at a higher horizontal resolution, feeding it the output of the global Unified Model run. This is given a different name, such as the UKV model or the New Zealand Limited Area Model.<ref>[https://www.nesi.org.nz/case-studies/improving-new-zealands-weather-forecasting-ability A 36 hour forecast by NZCSM takes 130 minutes to complete using 810 processors spread across 13 nodes of FitzRoy]</ref>
* The US National Weather Service runs the [[Weather Research and Forecasting Model]] with different parameters to create:
** [[North American Mesoscale Model]] (NAM) every six hours (with an ensemble called Short Range Ensemble Forecast, SREF)
** [[Rapid Refresh (weather prediction)|Rapid Refresh]] (RAP) and High Resolution Rapid Refresh (HRRR), every hour<ref>[https://rapidrefresh.noaa.gov/ Rapid Refresh (RAP)]</ref><ref>[https://rapidrefresh.noaa.gov/hrrr/ The High-Resolution Rapid Refresh (HRRR)]</ref>
* The Japan Meteorological Agency runs:<ref name="JMA" />
** Meso-Scale Model (MSM) every 3 hours, looking 39 and 78 hours ahead
** Meso-scale Ensemble Prediction System every 6 hours, looking 39 hours ahead (providing uncertainty estimation)
** Local Forecast Model every hour, looking 10-18 hours ahead
* The China Meteorological Administration runs the Regional Numerical Forecasting Model (CMA-MESO)<ref name="CMA" />
* CPTEC runs the Brazilian Regional Atmospheric Modelling System (BRAMS) and ETA Regional Model (ETA) for South America

The output of higher-resolution models may be further modified by [[model output statistics]] to take into quirky local phenomena that general models do not handle well, such as [[mountain waves]].