Weather forecasting

The practice of weather prediction dates back millennia, with early civilizations using astronomical and phenological signs to anticipate seasonal changes. The modern scientific era began in the 19th century with the invention of the telegraph, which enabled the rapid collection of weather observations over large areas. In 1854, the French astronomer Urbain Le Verrier organized a storm warning system after the Crimean War, and by the 1860s, the first daily weather maps were produced in Britain and the United States.¹

The theoretical foundation was laid by Vilhelm Bjerknes in 1904, who proposed that weather prediction could be treated as an initial-value problem in physics, governed by the equations of fluid dynamics and thermodynamics. This concept led to the first numerical weather prediction (NWP) attempt by Lewis Fry Richardson in 1922, though his manual calculation was impractical. The advent of electronic computers in the 1950s made NWP feasible, with the first successful computer-based forecast produced in 1950 by a team led by Jule Charney.²

Accurate forecasting depends on comprehensive observations of the atmosphere, ocean, and land surface. The global observing system includes surface weather stations, weather balloons (radiosondes), aircraft reports, ships, buoys, and a constellation of satellites. The World Meteorological Organization (WMO) coordinates the exchange of data through the World Weather Watch program, ensuring that forecasters worldwide have access to real-time observations.

Satellites provide crucial data over oceans and sparsely populated regions, measuring cloud cover, temperature profiles, humidity, and wind. The Geostationary Operational Environmental Satellites (GOES) and polar-orbiting satellites like the Joint Polar Satellite System (JPSS) are key components. Additionally, radar networks detect precipitation and severe weather, while lightning detection systems and wind profilers add to the observational mix.

Numerical weather prediction (NWP) is the core of modern forecasting. It involves solving the primitive equations of the atmosphere using supercomputers. The atmosphere is divided into a three-dimensional grid, and the equations of motion, thermodynamics, and moisture are integrated forward in time. The initial conditions are created by data assimilation, which combines observations with a short-range forecast (the background) to produce the best estimate of the current state.³

Global models, such as the European Centre for Medium-Range Weather Forecasts (ECMWF) Integrated Forecasting System (IFS) and the U.S. Global Forecast System (GFS), run at horizontal resolutions of about 9–13 km. Regional models, like the High-Resolution Rapid Refresh (HRRR) in the U.S., provide finer detail (3 km) for short-term forecasts. Ensemble forecasting, which runs multiple simulations with slightly perturbed initial conditions, is used to quantify forecast uncertainty and produce probabilistic forecasts.

Weather forecasts are categorized by lead time: nowcasting (0–6 hours), short-range (1–3 days), medium-range (4–10 days), and long-range (monthly and seasonal). Each type serves different users. Nowcasting relies heavily on radar and satellite imagery to predict severe weather, such as thunderstorms and tornadoes, with lead times of minutes to hours. Short- and medium-range forecasts are used for daily planning, aviation, and marine operations. Long-range forecasts, based on climate patterns like El Niño, are used for agriculture, water resource management, and energy planning.

Specialized forecasts include aviation weather (METARs and TAFs), marine forecasts, air quality forecasts, and fire weather forecasts. The National Weather Service (NWS) issues watches and warnings for hazardous conditions, while private sector companies provide tailored forecasts for industries.

Despite advances, weather forecasting has inherent limits. The atmosphere is a chaotic system, as described by Edward Lorenz's butterfly effect, which makes long-range deterministic prediction impossible beyond about two weeks. Errors in initial conditions grow exponentially, limiting the skill of forecasts. Additionally, small-scale phenomena like thunderstorms are difficult to predict precisely due to their short lifetimes and localized nature.⁴

Data gaps exist over oceans and developing regions, and the increasing frequency of extreme weather events due to climate change poses new challenges. However, improvements in observations, model resolution, and data assimilation continue to extend forecast skill.

Beyond the mainstream, weather forecasting has several niche and surprising dimensions. One is the use of paleoclimate proxies to reconstruct past weather patterns, which helps validate long-range forecast models. Another is the role of amateur observers in the Cooperative Observer Program (COOP), who have provided daily temperature and precipitation data for over a century, contributing to climate records.⁵

In the 19th century, Francis Galton coined the term 'anticyclone' and pioneered the use of weather maps for understanding pressure systems. The Bureau of Meteorology in Australia was one of the first to issue daily forecasts via radio in the 1920s. More recently, crowdsourced weather data from personal weather stations and smartphone barometers are being integrated into models, improving local forecasts. Additionally, the World Weather Attribution initiative uses forecast models to assess the role of climate change in extreme events, a field known as 'extreme event attribution'.

Numerical weather prediction (NWP)

A method of forecasting that uses mathematical models of the atmosphere and oceans to predict the weather based on current conditions.

Data assimilation

The process of combining observational data with a short-range forecast to produce the best estimate of the current atmospheric state.

Ensemble forecasting

A technique that runs multiple forecasts with slightly different initial conditions to quantify uncertainty and produce probabilistic predictions.

Nowcasting

Weather forecasting for the very short term, typically up to 6 hours ahead, using real-time observations and extrapolation.

Radiosonde

An instrument carried by a weather balloon that measures atmospheric parameters and transmits them to a ground station.

Primitive equations

The fundamental equations of fluid motion and thermodynamics that describe the behavior of the atmosphere.