The European Centre for Medium-Range Weather Forecasts (ECMWF) is an organization run by many European countries. It has three locations: Shinfield Park in Reading, United Kingdom; Bologna, Italy; and Bonn, Germany. The ECMWF uses one of the largest supercomputer centers in Europe and holds the world's largest collection of weather data used for predictions.

History

ECMWF was started in 1975 to combine the scientific and technical resources of Europe's weather services and institutions. This was done to create weather forecasts for periods up to about two weeks and to use the economic and social benefits of these forecasts. The Centre has about 350 employees, most of whom come from different member countries and cooperating countries.

In 2017, the Centre's member states agreed to move ECMWF's data centre to Bologna, Italy. The new location, a former tobacco factory, would be redesigned by the architecture firm gmp.

During 2020, the Centre moved its Copernicus operations from Reading to European Union territory. After considering bids from Toulouse, Italy, Austria, Germany, Spain, and Ireland, Bonn, Germany, was chosen. This move happened because of Brexit.

In early 2024, it was announced that ECMWF's headquarters, currently at Shinfield Road in Reading, England, would move to a new building on the Whiteknights Park campus of the University of Reading. Construction of this building began in early 2025, and it is expected to be completed in early 2027.

Objectives

ECMWF works to create accurate weather forecasts for up to 15 days and seasonal forecasts for up to 12 months. Its forecasts are shared with national weather services in member and cooperating countries to support their short-term and long-term weather activities. These countries use ECMWF's forecasts to warn the public about severe weather that could cause harm.

ECMWF's main goals are to:

• Create numerical weather forecasts and track systems that affect weather
• Improve forecasting accuracy through scientific and technical research
• Keep a record of weather-related data

To achieve these goals, ECMWF provides:

• Global weather forecasts twice each day
• Analysis of air quality
• Monitoring of atmospheric composition
• Climate monitoring
• Study of ocean currents
• Predictions about water systems

ECMWF develops and runs global models of Earth's atmosphere and systems that interact with it. It uses numerical methods to create forecasts and monitor Earth's systems.

ECMWF is currently responsible for managing two services in the EU's Copernicus Programme: the Copernicus Atmosphere Monitoring Service (CAMS) and the Copernicus Climate Change Service (C3S).

In 2020, ECMWF moved its Copernicus operations from Reading to Bonn, Germany. This move was directly linked to Brexit.

Work and projects

Numerical weather prediction (NWP) uses weather data collected by satellites, weather stations, planes, ships, and weather balloons. This data is combined to create a starting point for computer models that simulate the atmosphere. These models are then used to predict weather conditions. Common types of forecasts include:

• Medium-range forecasts, predicting weather up to 15 days ahead
• Monthly forecasts, predicting weather on a weekly basis for 30 days ahead
• Seasonal forecasts, predicting weather up to 12 months ahead

Over the past 30 years, the European Centre for Medium-Range Weather Forecasts (ECMWF) has worked on research to improve how weather data is combined and used in models. This has made weather forecasts more accurate by about one day every 10 years. For example, a seven-day forecast in 2015 was as accurate as a three-day forecast in 1975.

ECMWF’s monthly and seasonal forecasts help predict events like heat waves, cold spells, and droughts. These predictions also help sectors like agriculture, energy, and health prepare for possible impacts. Because ECMWF runs a wave model, it can also predict coastal waves and storm surges in European waters, which helps provide warnings.

Forecasts of severe weather allow authorities and the public to take action, such as planning for emergencies or evacuations. Better warnings can save lives, for example, by helping people leave areas at risk from storm surges. Governments and businesses can also prepare to keep services running during threats like high winds, floods, or snow.

In October 2012, the ECMWF model predicted Hurricane Sandy would hit the East Coast of the United States seven days before it occurred. It also predicted the path and strength of a nor’easter storm in November 2012, which affected the East Coast a week after Sandy.

ECMWF’s Extreme Forecast Index (EFI) is a tool that compares weather predictions from the Ensemble Prediction System (EPS) to typical weather patterns. It shows when weather conditions are unusual without needing specific rules for different times or places.

ECMWF works with organizations like EUMETSAT, ESA, and the EU to use satellite data for weather and seasonal forecasts. More satellite data and better ways to analyze it have improved the accuracy of weather predictions.

ECMWF also studies climate changes using a method called reanalysis. This involves using weather data from many years to recreate past weather, ocean, and land conditions. Reanalysis helps scientists understand how the climate has changed over time.

ECMWF has completed several major reanalysis projects with support from European weather services and the European Commission. These include:
– ERA-15: weather data from December 1978 to February 1994
– ERA-40: weather data from September 1957 to August 2002
– ERA-Interim: weather data starting in 1979
– ERA5: a more detailed weather dataset released in 2019 as part of the Copernicus Climate Change Service

ECMWF’s daily weather forecasts are created using its “Integrated Forecast System” (called the “European model” in the United States). This system runs every 12 hours and predicts weather up to 10 days ahead.

The system includes two types of forecasts:
– A “deterministic forecast,” which is a single, high-resolution model run that is more detailed but requires more computer power
– An “ensemble,” which uses 51 model runs with slightly different starting conditions to show a range of possible outcomes. This is less detailed and less accurate but provides a broader view of possible weather changes

As of 2021, ECMWF’s weather model is generally considered the most accurate global weather forecasting model.

In 2025, the AIFS became ECMWF’s first fully data-driven weather model. Unlike the Integrated Forecast System, the AIFS does not use detailed physics equations to predict weather. Instead, it uses a coarser 28 km grid compared to the IFS’s 9 km grid.

Member and co-operating states

ECMWF includes 23 European countries:

• The 18 original members from 1975: Austria, Belgium, Denmark, Finland, France, Germany, Greece, Republic of Ireland, Italy, Luxembourg, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom.
• Five countries that joined after 2010: Iceland (joined in April 2011), Slovenia (joined in December 2012), Serbia (joined in January 2015), Croatia (joined in January 2016), Estonia (joined in December 2020).

ECMWF also works with other countries through agreements: Bulgaria, Czech Republic, Georgia, Hungary, Israel, Latvia, Lithuania, North Macedonia, Montenegro, Morocco, Romania, and Slovakia.