ECMWF ECMWF

A Supercomputer Centre
ECMWF has operated a world-class HPCF for weather
forecasting since the installation of its first CRAY-1
supercomputer in 1978. ECMWF’s HPC has always been
amongst the most powerful supercomputers in Europe
and provides users with leading-edge HPC technologies.
Over time, various supercomputer architectures have
been used, including vector shared-memory systems,
vector distributed-memory systems and clusters of
scalar shared-memory processing systems. ECMWF
has always ensured that its codes are portable and
it has invested considerable resources in ensuring
that they remain portable for most prevailing
high-performance computing architectures.
The available HPCF computing resources are allocated
to operational activities (25%), ECMWF’s research
programme (50%) and workload from ECMWF’s
Member States (25%). Emphasis is placed on
delivering the operational forecast production to a strict
schedule, providing a good service to users, whilst using
the HPCF resources effectively and maintaining a very
high level of utilisation. The HPCF currently processes
about 370,000 parallel jobs per week, which account
for about 96% of the HPCF resources, and over a million
serial jobs per week, which perform ancillary tasks for
the parallel jobs, such as archiving results to the data
handling system.
European members
the TIGGE-LAM archive. Launched by the World
Weather Research Programme as a new tool to facilitate
research designed to improve global (and more recently
regional) ensemble forecasts of high-impact weather
and so strengthen early warning and disaster prevention,
TIGGE-LAM is a single web portal hosted by ECMWF.
The new regional archive groups 10 European ensemble
predictions in a standard data format produced on grids
between 10 and 2 km resolution and provides detailed
information for the short-range, up to a few days ahead.
This complements the larger-scale information provided
by the global data in the established TIGGE archive and
allows the meteorological community easy access to all
of these ensembles to study their performance.
The TIGGE-LAM archive has been developed as
part of the EU-funded GEOWOW project to improve
Earth observation data discovery, accessibility and
exploitability. It is part of the weather contribution to
the GEO System-of-Systems (GEOSS) and is accessible
through the GEO Common Infrastructure (GCI).
ECMWF’s 20 Member States are Austria, Belgium,
Denmark, Finland, France, Germany, Greece,
Iceland, Ireland, Italy, Luxembourg, the Netherlands,
Norway, Portugal, Slovenia, Spain, Sweden,
Switzerland, Turkey, and the United Kingdom.
ECMWF has concluded co-operation agreements
with the following 14 States: Bulgaria, Croatia,
Czech Republic, Estonia, the Former Yugoslav
Republic of Macedonia, Hungary, Israel, Latvia,
Lithuania, Montenegro, Morocco, Romania, Serbia
and Slovakia.
ECMWF has been working in collaboration with
national meteorological and hydrological services
(NMHSs) and research institutions from many
of these 34 States, to develop its modelling
capabilities, to design new products and to evaluate
and diagnose forecast quality. The relationship with
the Member and Co-operating States is a formal one
governed by the ECMWF Convention, which has the
status of an international governmental treaty.
Advancing weather science to improve
global numerical weather prediction
Member States
Co-operating States
The largest meteorological
archive in the world
Aside from its operational and research usage,
ECMWF’s HPCF provides easy access to the whole
archive for scientists and other users. An important
example of such archive capability is the hosting of
Satellite data
Another reason why weather centres need supercomputers is the amount of observations needed to be processed
every day to produce reliable predictions. ECMWF routinely processes data from 75 different instruments on 30
satellites as part of its operational daily data assimilation and monitoring activities. A total of 40 million observations
are processed and used daily; the vast majority of these are satellite measurements, but ECMWF also processes all
available observations from non-satellite sources, including surface-based and aircraft reports. Information on the
availability, data counts and quality of the different observing systems is updated continuously for the purposes of
monitoring the inputs to the forecasting system. The observation quality statistics are provided as important feedback
to satellite agencies and other observation providers on a regular basis.
ECMWF
European Centre for Medium-Range Weather Forecasts
ECMWF Shinfield Park,
Reading RG2 9AX, UK
Tel: +44 118 949 9000
www.ecmwf.int
August 2014
ECMWF
European Centre for Medium-Range Weather Forecasts
ECMWF is an intergovernmental organisation established by a Convention that
came into force on 1 November 1975. The Centre has 20 Member States and has
co-operation agreements with another 14 States (Co-operating States). It is both
a research institute and a 24/7 operational service producing and disseminating
global numerical weather predictions and other data to Member and Co-operating
States. The Centre also offers a catalogue of forecast data that can be purchased
by businesses worldwide and other commercial customers. Other strategic
activities include maintaining a data archive, assistance in advanced education
and assistance to the World Meteorological Organization (WMO) in implementing
its programmes. The supercomputer facility (and associated data archive)
at ECMWF is one of the largest of its type in Europe.
ECMWF was established as a major initiative in
European scientific and technical co-operation
in meteorology, based on a high-performance
computing facility, a scientific and technical
workforce, the production of medium-range
weather forecasts, and related research and
development. The collaborative aspect of ECMWF
remains to this date a key to its success; our staff
of 270 are from over 30 countries, and developing
effective partnerships with meteorological
services, universities and other organisations
that help ECMWF to achieve its targets is a key
priority. Establishing closer and more effective
collaborations with leading institutions is helping
the Centre to continue to develop its models and
satisfy its users’ increasing requirements.
ECMWF operates with an annual budget of £53m,
of which Member States contribute 78%, sales
of data 8%, and external research funding 13%
(ECMWF Annual Report 2013). ECMWF is based
in Reading, UK.
Copernicus
The Copernicus programme, a joint initiative
of the European Union and the European
Space Agency, aims to ensure operational
monitoring of the atmosphere, oceans, and
continental surfaces, and to provide reliable,
validated information services for a range of
environmental and security applications.
Based on the exploitation of space-based and
in situ observations and models, Copernicus will
provide information services for land, marine,
atmospheric and climate monitoring, as well
as emergency management and security. The
full operational phase of Copernicus is starting
in 2014 and is funded by the EU, under its
Multiannual Financial Framework 2014-2020.
ECMWF has had a strong association with the
development of the Copernicus programme
particularly via its MACC, ERA and wave
forecasting activities. It is expected that this will
be continued and strengthened as Copernicus
services enter their fully operational phase.
Photo: Sentinel-1A liftoff
Copyright ESA–S. Corvaja, 2014
Global numerical weather
prediction at ECMWF:
Accurately forecasting the
most likely future weather and
the associated uncertainties
ECMWF’s key scientific objectives are to improve
the data assimilation and forecast models and develop
fundamental research in numerical weather prediction.
This includes ocean waves, currents and sea-ice,
atmospheric composition aspects, and coupled
ocean-land-atmosphere extended-range predictability.
This in turn helps ECMWF meet its strategic goal of
providing its Member and Co-operating States with
global analyses and predictions that enable national
meteorological services to provide forecast products
to their users.
Model components
Medium-range and monthly predictions utilise a common
atmospheric model that is typically upgraded twice a
year. The seasonal forecasts, reanalyses, atmospheric
composition predictions, ocean model, and analyses
are also upgraded regularly but on a less frequent basis.
The latest model cycle to date, 40r1 was introduced on
19 November 2013. Scientific developments include the
vertical resolution upgrade for ensemble predictions, the
coupling to the ocean from the initial time, the enhancement
of the ensemble of data assimilations to 25 ensemble
members producing background error co-variances, the
improved diurnal cycle of convective precipitation, and
the upgrades to vertical diffusion and orographic drag,
producing significant medium-range predictive
skill enhancements.
An Integrated Forecasting System
Our Integrated Forecasting System (IFS) produces
forecasts for multiple time ranges to address different
user requirements. These provide key aspects of the
forecast evolution and the associated uncertainty. Specific
products designed to highlight potential severe weather
events include the Extreme Forecast Index and tropical
cyclone activity.
Our predictions cover the medium-range (up to two weeks
ahead), extended range (up to a month ahead), and long
range (up to a year ahead).
and an Open IFS!
ECMWF launched in 2014 an open version of its IFS,
in order to facilitate its use by the academic and scientific
community worldwide, but also to help improve it through
external research collaboration. OpenIFS provides research
institutions with an easy-to-use version of the ECMWF IFS,
offering the forecast capability of IFS (no data assimilation),
a Single Column Model (SCM) version and supporting
software and documentation.
OpenIFS is free but requires a license from ECMWF.
For more information, please email
[email protected]
ECMWF’s Numerical Weather Prediction
•Uses an ensemble made of 52 members
•One member (HRES) is higher resolution with the most accurate initial conditions and model (16 km horizontal resolution, 137 vertical levels)
•One member (CNTL) has a lower resolution (32 km, 91 vertical levels) but otherwise has the most accurate
initial conditions and model
•50 members are perturbed relative to the CNTL
to explore the uncertainties
Assessing the uncertainty: ensembles
•Perturbations are made to the observations and to the
model physics in an ensemble of 4DVAR data assimilations
•Ensemble spread is a measure of the uncertainties
associated with possible future weather
•Error of the ensemble mean should be equal
to the ensemble spread for a balanced prediction
of uncertainties
•Reliability assesses the frequency of predicted
events compared to actual frequency of occurrence
Most accurate individual forecast
•HRES is on average the most accurate prediction
of the larger scale aspects of the weather up to
about 10 days ahead
•Its skill has been advancing at approximately
one day per decade
•Comparing with re-forecasts (using a fixed model version)
allows the variations in atmospheric predictability to be taken into account in assessing the skill improvements from new science
Key scientific advances
•Reduced numerical errors by increasing resolution, enabled by increasing supercomputer capacity
•Improved quality of initial conditions using data assimilation to combine increasing number and
variety of observations with prior information
•Improved representation of physical processes
using meteorological research
•Design of reliable ensemble predictions
Looking to the future:
Over the next decades the science of NWP is expected
to advance significantly and so the following capabilities
have the potential to be realised:
•Global NWP at kilometre-scale horizontal resolution
(towards “forecasts on the human scale”)
•Predict high-impact weather out to 2 weeks ahead: accurately and reliably
•Predict large-scale weather patterns and regime transitions out to a month or more ahead
•Predict global circulation anomalies out to a year ahead
•Predict aspects of natural environment in addition
to the weather
Building effective partnerships
ECMWF has a history of successfully integrating externally-funded projects with its core activities to produce
results that consistently meet the expected outcomes for which the funding was granted. Over the last five
years we have managed or participated in 70 different projects with a combined total value of over €75
million. Two major programmes ECMWF has delivered with EU funding have been in the fields of atmospheric
composition and climate reanalysis. Both programmes owe their success to extensive scientific and technical
collaboration with the national meteorological services in the Member States, but also with the worldwide
meteorological community.
Atmospheric composition
Starting in the early 2000’s, atmospheric composition has gradually
become an increasingly important (optional) component of the
Integrated Forecasting System (IFS). We provide pre-operational
daily global forecasts of aerosols and atmospheric species such
as ozone, carbon monoxide, and methane.
As part of the European Union’s Copernicus programme on
environmental monitoring, greenhouse gases, aerosols, and
chemical species have been introduced in the ECMWF model
allowing assimilation and forecasting of atmospheric composition.
At the same time, the added atmospheric composition variables
are being used to improve the Numerical Weather Prediction
(NWP) system itself, most notably through the interaction with the
radiation scheme and the use in observation operators for satellite
radiance assimilation.
The atmospheric composition programme, operating under
the acronym MACC (Monitoring Atmospheric Composition and
Climate), is led by ECMWF with contributions from a consortium
of 36 partners from 13 countries. It offers daily services of analyses
and forecasts ranging from air quality to climate forcing, solar
radiation and ozone, and a range of products including aerosol,
fire, greenhouse gasses forecast. Its users include National
agencies responsible for air quality forecasting and assessment,
companies providing tailored information to users, such as solar
radiation information for solar power generation, and the science
community. The data generated by MACC-II have proven to be very
useful to study atmospheric composition and climate interactions.
Climate reanalysis
ECMWF periodically uses its forecast models and data assimilation
systems to ‘reanalyse’ archived observations, creating global
data sets describing the recent history of the atmosphere, land
surface, and oceans. Reanalysis data are used for monitoring
climate change, for research and education, and for commercial
applications. Another major use of reanalysis data at ECMWF
is for verification of the daily forecasts.
Current research in reanalysis at ECMWF focuses on the
development of consistent reanalyses of the coupled climate
system, including atmosphere, land surface, ocean, sea ice,
and the carbon cycle, extending back as far as a century or
more. The work involves collection, preparation and assessment of
climate observations, ranging from early in-situ surface observations
made by meteorological observers to modern high-resolution
satellite data sets. Special developments in data assimilation are
needed to ensure the best possible temporal consistency of the
reanalyses, which can be adversely affected by biases in models
and observations, and by the ever-changing observing system.
Current datasets available from ECMWF cover the reanalysis
of over 100 years of weather.