Numerical weather prediction

Numerical weather prediction
Name	Numerical weather prediction
Established	1922
Discipline	Atmospheric science
Type	Computational forecasting
Methods	Finite difference, spectral, ensemble, data assimilation

Numerical weather prediction is the use of mathematical models of the atmosphere and ocean to forecast weather by computing future states from initial conditions. Developed through collaborations among mathematicians, physicists, and meteorologists, it integrates observations, dynamical equations, and computational algorithms to produce operational forecasts for aviation, agriculture, and emergency management. Major contributors include early pioneers and institutions that advanced atmospheric dynamics, computational methods, and observing systems.

History

The origins trace to the work of Lewis Fry Richardson, who attempted a manual forecast and inspired later efforts at University of Washington, Palmerston North, and University of Cambridge. Progress accelerated with contributions from Vilhelm Bjerknes, whose conceptualization influenced the Norwegian School of Meteorology and researchers at Geophysical Fluid Dynamics Laboratory and Massachusetts Institute of Technology. The advent of electronic computers such as the ENIAC and development at National Meteorological Center and European Centre for Medium-Range Weather Forecasts transformed practice. Cold War era investments by organizations like National Aeronautics and Space Administration and Defense Advanced Research Projects Agency enabled global observing systems including TIROS and Global Atmospheric Research Program. Key milestones include the establishment of operational global models at UK Met Office, the development of spectral methods at Met Éireann and Institut Pierre-Simon Laplace, and ensemble forecasting advances at US National Oceanic and Atmospheric Administration and Canadian Meteorological Centre.

Fundamentals and governing equations

Models solve the primitive equations derived from conservation laws first formalized by Carl-Gustaf Rossby, incorporating the Navier–Stokes equations on a rotating sphere and thermodynamic relations influenced by studies at Scripps Institution of Oceanography and Max Planck Institute for Meteorology. Governing equations include momentum, mass continuity, thermodynamic energy, and moisture continuity with parameterized subgrid processes shaped by work at National Center for Atmospheric Research and Woods Hole Oceanographic Institution. Boundary conditions use information from observing networks like Global Telecommunication System and satellite missions from European Space Agency and Japan Aerospace Exploration Agency. Coupled atmosphere–ocean–land surface systems integrate components developed at Hadley Centre and Institute Pierre Simon Laplace.

Numerical methods and models

Discretization methods—finite difference, finite element, finite volume, and spectral techniques—were refined at institutions including Princeton University, California Institute of Technology, and ETH Zurich. Time integration schemes trace back to algorithms used in Los Alamos National Laboratory computational projects. Model families include global spectral models from UK Met Office and ECMWF, gridpoint models from NOAA GFS and Canadian GEM, and nonhydrostatic convection-permitting models from COSMO and Weather Research and Forecasting Model. Ensemble systems pioneered by Tim Palmer and groups at ECMWF and NCEP address uncertainty. Parameterizations for convection, radiation, and boundary layers were advanced at Monash University, Ruhr University Bochum, and Columbia University.

Data assimilation and observations

Data assimilation integrates observations into model initial states using methods such as three-dimensional variational (3D-Var), four-dimensional variational (4D-Var), and ensemble Kalman filters developed at ECMWF, NCEP, and JMA. Observing systems include radiosondes launched by Royal Netherlands Meteorological Institute, satellite sounders from NOAA and EUMETSAT, aircraft reports coordinated via International Civil Aviation Organization, and radar networks maintained by Deutsche Wetterdienst and Australian Bureau of Meteorology. Advances in remote sensing from missions like MetOp, GOES, and Sentinel series improved assimilation of moisture and temperature profiles. Reanalysis projects led by NCAR, ECMWF, and JMA provide long-term datasets for model development.

Model evaluation and verification

Verification metrics and methodologies evolved through collaborations among World Meteorological Organization working groups, with operational practices at UK Met Office, ECMWF, and NOAA comparing forecasts to observations from Surface Synoptic Observations and reanalyses by ECMWF. Studies at University of Reading, University of Oklahoma, and University of Hamburg examine skill scores, reliability diagrams, and probabilistic calibration for ensembles. Model intercomparisons in projects like Coupled Model Intercomparison Project and regional evaluation initiatives sponsored by Intergovernmental Panel on Climate Change facilitate systematic assessment of biases, resolution dependency, and extreme-event representation.

Operational forecasting and applications

Operational centers such as ECMWF, NOAA National Weather Service, UK Met Office, Météo-France, and Japan Meteorological Agency run suites of deterministic and ensemble forecasts to support aviation regulated by International Civil Aviation Organization, energy markets linked to International Energy Agency, and disaster response coordinated with United Nations Office for Disaster Risk Reduction. Sectoral applications include flood forecasting with hydrological models from USGS, air quality prediction tied to European Environment Agency networks, and marine forecasting for port authorities like Port of Rotterdam. Commercial weather services from groups including The Weather Company and AccuWeather adapt model outputs for decision support.

Challenges and future directions

Challenges include improving representation of convection studied at University of Melbourne and ETH Zurich, reducing systematic model biases identified by ECMWF and NCAR, and assimilating new observing systems from CubeSat constellations and novel sensors developed by NASA Jet Propulsion Laboratory. Computational limits drive research in exascale computing initiatives at Oak Ridge National Laboratory and algorithmic improvements from Argonne National Laboratory. Future directions emphasize machine learning contributions from Google DeepMind, hybrid physics–ML parameterizations investigated at MIT, and expanded coupled Earth system modeling coordinated via IPCC and World Climate Research Programme initiatives. Continued international collaboration among centers such as ECMWF, NCEP, JMA, and UK Met Office will shape forecast skill improvements and societal applications.

Category:Meteorology