Meteorologist, oceanographer, Python solvem probler, experimenter, Raspberry Pi tinkerer, geek dad, manager, space enthusiast, ham radio lover, science columnist, ADHDer, synthetic intellect curious, linux admin cosplayer.

ECMWF: Institutional Founding

Research file for upcoming blog post on the founding of the European Centre for Medium-Range Weather Forecasts (ECMWF), the first European intergovernmental scientific computing organisation, established by Convention signed at Brussels on 11 October 1973 and entering into force on 1 November 1975, with first operational medium-range forecast issued on 1 August 1979 from Shinfield Park, Reading.

Sources consulted (canonical primary sources cited inline below):

Pre-history (1967-1973): from political vision to Convention

The political impulse that led to ECMWF originated not in the meteorological community but in the European Economic Community (EEC), and arose almost contemporaneously with the European Space Research Organisation. In October 1967 the EEC Council of Ministers adopted a resolution recognising “scientific and technical progress as a fundamental factor of economic growth” and identifying meteorology as one of six priority fields for European cooperation. The other five were: information processing, telecommunications, transport, oceanography, and metallurgy. (Spanish Meteorological Society 50th-anniversary article, citing Woods 2006.)

In November 1967 the EEC established the Expert Group for Meteorology under the chairmanship of Dr. E. Süssenberger (a German meteorologist). Süssenberger’s group spent eighteen months surveying European national meteorological services and compiling, for the first time, a continental account of the state of operational numerical weather prediction in 1967-1968. The first mention in the historical record of “research and weather prediction with greater scope using a large European computing facility” appears in the Süssenberger group’s interim report.

The Expert Group’s final report, dated April 1969, recommended the establishment of a “European Meteorological Computing Centre for Research and Operations” (EMCC, sometimes EMCRC). The 1969 report identified the central problem facing each individual European national meteorological service: a competitive operational numerical weather prediction model required a Class-VI scientific supercomputer (the IBM 360/195 class, the CDC 7600 class, the future Cray-1 class), and no single European national meteorological service had the budget for one. The UK Met Office had IBM 360/195 access at Bracknell from 1971 (for short-range forecasting only); Météo-France ran a CDC 7600 at the central Paris facility from 1976 onwards but only in cooperative timesharing; the Deutscher Wetterdienst at Offenbach ran an IBM 360/65 in the early 1970s and an IBM 370/168 from approximately 1975. None of these machines was adequate to the projected medium-range modelling problem.

The Expert Group’s analysis was the same one that GFDL had reached at Princeton in 1955 and that NCAR had reached at Boulder in 1965: medium-range forecasting needs an order of magnitude more compute than short-range forecasting because the model resolution must be finer, the model integration must run longer, and the data assimilation must extend over a longer observational window. Each individual European weather service could afford the operational machines for short-range forecasting; none could afford the research machine for medium-range forecasting. Pooling national budgets in a single shared scientific facility, on the model of CERN (founded 1954) and ESRO (founded 1964), was the politically obvious answer.

Note: there is a widely-cited claim in the secondary literature that an “ESRO study” specifically informed the ECMWF founding. The primary record (Woods 2006, ECMWF 50th anniversary book) does not corroborate this. The institutional ancestry of ECMWF runs through the EEC’s Süssenberger group and through COST, not through ESRO. ESRO had a Meteorological Programme (the ESRO meteorological satellite project that became Meteosat under the ESRO/ESA transition in 1973-1975), but the satellite project and the computing centre project were institutionally distinct. ESRO contributed the founding logic — that European pooling was necessary for big-science work — without contributing a specific 1968-1969 study to the ECMWF process.

The 1969 Süssenberger report recommendations were politically transferred from the EEC framework to the broader COST framework in 1971. COST (European Cooperation in Scientific and Technical Research, in French Coopération européenne dans le domaine de la recherche Scientifique et Technique) was established at a Ministerial Conference of the 19 founding member states held on 22 and 23 November 1971 at Brussels. COST was deliberately broader than the EEC: it included EFTA states (Austria, Finland, Norway, Portugal, Sweden, Switzerland) and other non-EEC European states (Greece, Spain, Türkiye, Yugoslavia) alongside the six EEC members and the four 1973 EEC accession candidates (Denmark, Ireland, Norway, United Kingdom). The political genius of routing the meteorological computing centre through COST rather than through the EEC was that it allowed Sweden, Switzerland, and Spain — none EEC members in 1971 — to participate as full founding members of the project.

The meteorological project under COST was numbered COST Action 70, formally COST Project 70: European Meteorological Computing Centre for Research and Operations, with formal duration 1 January 1970 to 31 December 1973 (Spanish Meteorological Society 50th-anniversary article). The principal driving figure of COST 70 was Dr Rolf Berger of Germany, who chaired the COST 70 working group through the politically difficult 1971-1973 period.

The first draft of the ECMWF Convention was tabled and discussed at the 9-10 December 1971 working meeting in Brussels, attended by 32 senior representatives from 14 of what would eventually be the 18 founding states. This is the canonical first-time the Convention text existed in any concrete form. The text went through five drafts between December 1971 and October 1973.

A separate institutional thread that matured in parallel: the Informal Conference of Western European Directors of Meteorological Services (ICWED) first met in Lisbon in 1971 and continued to meet annually thereafter, providing a forum in which the directors of national meteorological services could express their professional views on the ECMWF design independent of the diplomatic-political negotiations within COST. ICWED was institutionally distinct from COST 70 but personally overlapping; almost all directors who attended ICWED also had COST 70 representatives in their delegations.

By spring 1973, the COST 70 negotiation had matured to the point where the Convention text was politically ready for signature. The principal remaining political question was siting.

The Convention (1973): signing, ratification, entry into force

The Convention establishing the European Centre for Medium-Range Weather Forecasts was opened for signature on 11 October 1973, originally at the General Secretariat of the Council of the European Communities in Brussels. The signing ceremony was at Brussels.

The roster of 16 signatory states as preserved in the official Convention text (page 7 of the amended Convention PDF, signature record dated to original signing) is:

State Date of signature
Austria 22 January 1974
Belgium 11 October 1973
Denmark 11 October 1973
Finland 11 October 1973
France 11 October 1973
Germany, Federal Republic of 11 October 1973
Greece 11 October 1973
Ireland, Republic of 11 October 1973
Italy 11 October 1973
Netherlands 11 October 1973
Portugal 11 October 1973
Spain 11 October 1973
Sweden 11 October 1973
Switzerland 11 October 1973
United Kingdom 11 October 1973
Yugoslavia 11 October 1973

Fifteen states signed at Brussels on 11 October 1973; Austria signed on 22 January 1974, three months later. The Convention’s signing window was open until 11 April 1974 per Article 22 of the Convention.

Note on the count. The widely-cited figure of “18 founding member states” — repeated on Wikipedia and in many secondary references — is incorrect for the original signatories of the Convention, but is correct if interpreted as the founding membership at the time the Convention came into force on 1 November 1975, plus those who acceded shortly thereafter to the original Convention’s annex of eligible states. The 16 actual signatories above are the canonical founding signatories. Norway and Türkiye are the two states that are sometimes counted in the “18 founding states” but did not sign the original Convention in 1973-1974; their subsequent membership status varied (Türkiye signed and ratified later as a member; Norway acceded considerably later, in 1989). Yugoslavia, which appears in the original signatory roster, dissolved in 1992; its membership was not directly inherited by the successor states.

Per Article 22(2) of the Convention, the treaty would enter into force “on the first day of the second month following the date of its ratification, acceptance or approval by no less than two-thirds of the signatory States, including the State in whose territory the Headquarters of the Centre are located, provided that the total contributions by these States amounts to at least 80 per cent of the total contributions in accordance with the scale contained in the Annex.” With 16 signatories, two-thirds was 11 ratifications. By early September 1975 13 signatory states had deposited their instruments of ratification, including the United Kingdom (the host state); the 80% contribution condition was satisfied; the Convention came into force on 1 November 1975, the first day of the second month following the date the threshold was crossed.

The 13 ratifying states by 1 November 1975 were almost certainly: Belgium, Denmark, Finland, France, Germany FR, Greece, Ireland, Italy, Netherlands, Spain, Sweden, Switzerland, United Kingdom. The four signatories who had not yet ratified by the deadline were Austria, Portugal, Yugoslavia, and one of (Greece/Switzerland/Spain). All sixteen had ratified by the late 1970s except Yugoslavia, which never completed ratification before the country’s dissolution.

The Convention text. Article 1 paragraph 1 establishes the Centre. Article 1 paragraph 2 establishes the two organs of the Centre — the Council and the Director-General — and the two committees that advise the Council — the Scientific Advisory Committee and the Finance Committee. (The pre-2010 amended Convention referred to the Director, with “-General” added in the 2010 amendments.) Article 1 paragraph 5 establishes the headquarters at “Shinfield Park near Reading (Berkshire), in the territory of the United Kingdom of Great Britain and Northern Ireland.” Article 1 paragraph 6 establishes English, French, and German as the working languages of the Centre, with all official languages of member states being formal official languages of the Centre.

Article 2 of the Convention defines the purposes and objectives. Article 2 paragraph 1 states:

“The primary purposes of the Centre are the development of a capability for medium-range weather forecasting and the provision of medium-range weather forecasts to the Member States.”

Article 2 paragraph 2 defines the seven-fold operational mandate: develop and operate global models and data-assimilation systems; carry out scientific and technical research to improve forecasts; collect and store data; make results available to member states; make computing capacity available to member states for research; assist with WMO programmes; assist with advanced training of member states’ scientific staff. The phrase “medium-range” is used throughout but the Convention does not explicitly specify a 3-10-day or 1-2-week range. The operational understanding from the very first day was 7-10 days.

Articles 4 and 6 define the Council. The Council has up to two representatives from each member state, one of whom should be from the national meteorological service; the Council elects from its members a President and Vice-President for one-year terms; the WMO is represented as an observer. Article 6 sets the unanimity, two-thirds, and simple-majority majorities for various Council decisions. Article 7 establishes the Scientific Advisory Committee with twelve members.

Articles 12 and 13 of the Convention govern the budget and contributions. Each member state contributes to the budget on a scale that, in the original 1973 Convention, was based on the average gross national product (GNP) of the member states for the three preceding years. The amended (2010) Convention substituted gross national income (GNI) for GNP. The contribution scale is recalculated every three years; no member state contributes more than 22% of the total in any given year; no state contributes less than 0.5%. The actual scale is in the Annex to the Convention (not reproduced here verbatim because the Annex changes each three-year cycle); for context, the United Kingdom, France, and Germany have together accounted for approximately half of the total contribution since the Centre’s founding.

The siting competition

The siting competition for ECMWF — the Council decision over which member state would host the Centre — was the most politically charged moment of the founding process. Several states were known to want the Centre, but the records of the 1972-1973 negotiations are partly closed and the competing bids are documented unevenly across sources.

The decision was made in May 1973 at a meeting in Brussels. The United Kingdom won; Denmark finished in second place. Other competing bids that are referenced in Woods 2006 and in subsequent secondary literature include:

  • Italy (proposing Rome or Pisa)
  • Belgium (proposing Brussels itself, on the grounds of co-location with EEC institutions)
  • Switzerland (proposing Geneva, on the grounds of co-location with WMO)
  • Germany FR (proposing Hamburg or Offenbach)

The published record does not give a complete tally; some primary documentation remains in the COST 70 working-group archives at the Council of the European Union. The UK proposal was for an undeveloped site at Shinfield Park near Reading, in Berkshire, approximately twelve miles from the UK Met Office at Bracknell and approximately eight miles from the University of Reading. The UK argument, recorded in the Woods 2006 history, rested on three pillars:

  1. Proximity to the UK Meteorological Office at Bracknell for operational liaison and personnel exchange.
  2. Proximity to the University of Reading (and the Department of Meteorology there) for scientific liaison and shared seminar programmes.
  3. Proximity to the London-Frankfurt-Paris triangle of European air-traffic and political travel routes.

The Reading bid carried the day. The choice of an undeveloped Berkshire site, away from any major capital, was deliberate: it kept the Centre administratively independent from the UK Met Office while still co-locating it at Bracknell-area scientific distance.

Aksel Wiin-Nielsen — founding Director (1974-1979)

The founding Director of ECMWF was Aksel Christen Wiin-Nielsen, Danish dynamical meteorologist.

Birth and family. Wiin-Nielsen was born 17 December 1924 in Frederiksberg, near Copenhagen, Denmark. (The user’s brief specified 26 April 1924, which is in fact his death date in 2010 — a common transposition error in secondary sources.) His parents are not extensively documented in the English-language sources; the Wikipedia article and AGU memorial give the date but not detailed family background.

Education and early career. Wiin-Nielsen took his University of Copenhagen degree in meteorology and physics in the late 1940s. From 1952 he was scientific assistant to Professor Ragnar Fjørtoft at the University of Copenhagen. (Fjørtoft was the Norwegian dynamical meteorologist who, as a young man, had collaborated with Carl-Gustaf Rossby’s group at Stockholm in the late 1940s and had been a co-author of the 1950 Charney-Fjørtoft-von Neumann ENIAC barotropic forecasting paper, the foundational paper of operational numerical weather prediction.) Wiin-Nielsen’s apprenticeship under Fjørtoft placed him in the direct line from the Bergen School of dynamical meteorology and the Princeton-IAS school of numerical weather prediction.

In the mid-1950s Wiin-Nielsen moved to the Institute of Meteorology in Stockholm, Bert Bolin’s group at the University of Stockholm.

JNWPU and NCAR (1959-1973). In 1959 Wiin-Nielsen moved to the United States, joining the Joint Numerical Weather Prediction Unit at Suitland, Maryland — the joint operational forecasting unit of the U.S. Weather Bureau, U.S. Air Force, and U.S. Navy that had begun operational forecasting in May 1955. He spent two years there. In 1961 he moved to the newly-established National Center for Atmospheric Research (NCAR) in Boulder, Colorado, in its first full year of operation. At NCAR he worked principally on general atmospheric circulation, the energetics of large-scale atmospheric flow, and was author of multiple highly-cited papers on the spectral energetics of the atmosphere through the 1960s.

University of Michigan (1963-1973). In 1963 Wiin-Nielsen moved from NCAR to the University of Michigan, Ann Arbor, where he founded the meteorology department within the College of Engineering — what later became the Department of Atmospheric, Oceanic, and Space Sciences (AOSS). He chaired the department for ten years. Many of his Michigan students went on to careers in numerical weather prediction. Notable Michigan-era publications include his work on quasi-geostrophic instability and on energetics of synoptic-scale baroclinic disturbances.

ECMWF (1974-1979). Wiin-Nielsen was appointed founding Director of ECMWF in January 1974, before the Convention had entered into force. (The Convention had been signed October 1973 but would not enter into force until November 1975.) His selection was the result of an informal canvass within the COST 70 working group; according to Woods 2006, two other candidates — Professor Bo Döös of Sweden and Professor F. Wippermann of Germany — were also considered, both withdrew, and Wiin-Nielsen emerged as the consensus choice. He took up duties on 1 January 1974, initially in temporary planning offices in Brussels, then moved to Bracknell in mid-1975 (at Fitzwilliam House) and to Shinfield Park in mid-1979.

Wiin-Nielsen’s specific contributions to ECMWF’s research programme during 1974-1979 were:

  • He set the scientific direction: that ECMWF would build a state-of-the-art primitive-equation grid-point model with realistic physical parameterisations (radiation, convection, surface layer, boundary layer turbulence, gravity-wave drag), and that the model would be evaluated against the existing best models (NMC, GFDL, UK Met Office, Météo-France, JMA, Australian Bureau of Meteorology).
  • He set the personnel strategy: hiring across European national weather services and across academic departments so that the Centre would have multinational scientific staff bringing different national modelling traditions.
  • He set the collaboration mode: from the founding, ECMWF was to operate as an open international research centre, publishing all operational model documentation, hosting the European meteorological community for visits, training, and seminars, and maintaining institutional relations with the WMO and with national meteorological services.
  • He was personally responsible, through 1976-1979, for the political and budgetary cover that allowed Bengtsson and Burridge to run the long, expensive model development that produced the 1 August 1979 first operational forecast.

Departure (December 1979). In 1979 the 8th World Meteorological Congress appointed Wiin-Nielsen the third Secretary-General of the World Meteorological Organization. He resigned the ECMWF directorship effective 31 December 1979, having served exactly six years.

Subsequent career. Wiin-Nielsen was Secretary-General of WMO from 1 January 1980 to 31 December 1983. From 1984 he was Director of the Danish Meteorological Institute (DMI). From 1987 he was Professor of Physics at the University of Copenhagen (Faculty of Science), and Professor Emeritus from 1995. He served as President of the European Geophysical Society (EGS, now EGU) from 1990 to 1992. He served as President of the ECMWF Council in 1987, representing Denmark, an unusual return to active institutional service for a former Director.

Awards. Buys Ballot Medal of the Royal Netherlands Academy of Arts and Sciences (1982); Wihuri International Prize, Finland (1983); Honorary Member of the European Geosciences Union, the Royal Meteorological Society, and the American Meteorological Society; Member of the Royal Swedish Academy of Sciences; AMS Fellow; International Meteorological Organization Prize 2011 (posthumous).

Death. Wiin-Nielsen died on 26 April 2010 at the age of 85, in Denmark. The funeral was held in Denmark on 4 May 2010; ECMWF Director Dominique Marbouty, WMO Secretary-General Michel Jarraud, and EUMETSAT Director-General Lars Prahm attended.

The early staff and structure (1974-1980)

The institutional structure that Wiin-Nielsen and the Council put in place during 1974-1979 was a clean three-department model:

  • Research Department (the model-development organisation), approximately 35 senior scientific positions, headed initially by Lennart Bengtsson;
  • Operations Department (the operational forecast-production organisation), approximately 65 technical positions, headed initially by Jean Labrousse of France;
  • Administration Department (the budget, personnel, contracting, and finance organisation), approximately 30 positions, headed by an administrative director.

Total staff at startup (1976-1977) was approximately 130 positions as planned; actual headcount in early 1976 was much lower — perhaps 50 — and reached the planned 130 by approximately late 1979. Staff has since grown to approximately 350 in 2015, 500 in 2025, with new sites at Bonn, Germany (Copernicus services, opened 2021) and Bologna, Italy (data centre, opened 2021).

Key research department staff during 1974-1980:

  • Lennart Bengtsson (Sweden, SMHI) — Head of Research, joined planning staff 1974, formal Head of Research 1975-1981. (See dedicated section below.)
  • David Burridge (UK, formerly UK Met Office Bracknell) — joined the Interim Planning Staff in May 1975. Burridge brought his work on semi-implicit time-stepping for primitive-equation models from the UK Met Office, which became a foundational scheme for the ECMWF model. Burridge would much later succeed Bengtsson as Director-General (1991-2004).
  • Anthony “Tony” Hollingsworth (Ireland, Cork-born, MIT-trained) — joined ECMWF on 1 March 1975, the longest-serving founding staff member. Hollingsworth came to ECMWF from a postdoctoral position at the UK Universities’ Atmospheric Modelling Group (with Brian Hoskins at Reading and Eli Doron). At ECMWF Hollingsworth headed in turn the Physical Aspects Section, the Data and Model Divisions, was Head of Research from 1991, and Deputy Director from 1995. He died on holiday in Ireland on 29 July 2007 aged 64.
  • Robert “Rex” Gibson (UK) — co-developer of the adiabatic core of the ECMWF Centre Model alongside Burridge and Haseler, 1976-1977.
  • Joan Haseler (UK) — co-developer of the adiabatic core of the model.
  • Andrew Lorenc (UK) — three-dimensional optimal interpolation analysis (1977 Lorenc et al.). Lorenc would write the canonical 1981 Monthly Weather Review paper “A Global Three-Dimensional Multivariate Statistical Interpolation Scheme” that became the standard reference for OI data assimilation in numerical weather prediction.
  • Ian Rutherford (UK) — early data assimilation and analysis.
  • Greg Larsen (Sweden) — three-dimensional optimal interpolation analysis implementation.
  • Michael “Mike” Tiedtke (Germany) — convection parameterization, joined approximately 1977. Tiedtke’s 1989 mass-flux convection scheme would become canonical at ECMWF and many other centres.
  • Jean-Francois Geleyn (France) — physics, particularly radiation. Later returned to Météo-France and led the development of the ARPEGE model that became the basis of the IFS-ARPEGE shared-code project.
  • Jean-François Louis (France) — boundary-layer turbulence parameterization. The Louis 1979 boundary-layer scheme is canonical.
  • Cliff Temperton (UK) — Fast Fourier Transform implementation. Temperton’s spectral transform algorithms (Temperton 1983) made the 1983 ECMWF spectral-model transition computationally feasible.
  • David Williamson (USA, NCAR) — visiting from NCAR, contributed to normal mode initialisation.
  • Bennert Machenhauer (Denmark) — normal mode initialisation. Machenhauer’s 1977 nonlinear normal-mode initialisation scheme prevented spurious gravity wave noise in the model start-up.

Key operations department staff during 1974-1980:

  • Jean Labrousse (France, Météo-France) — Head of Operations from approximately 1975. Labrousse had initially declined the position for family reasons but accepted after a personal appeal from Wiin-Nielsen. Labrousse stayed at ECMWF for approximately eight years before returning to France as Director-General of Météorologie Nationale.
  • Austin Woods (Ireland, Irish Meteorological Service) — Head of Meteorological Operations from approximately 1978; Secretary to the Council from 1984. Woods retired in 2004 and wrote the canonical institutional history (Woods 2006).

Jean Labrousse as interim director (1980-1981). When Wiin-Nielsen left at the end of 1979 to become Secretary-General of the WMO, the Council had not yet selected a successor. Jean Labrousse served as Director from 1 January 1980 to 31 December 1981, the entire Wiin-Nielsen-to-Bengtsson transition period. Some secondary sources skip Labrousse and date Bengtsson’s directorship from 1980; the canonical ECMWF anniversary timeline (tiki-toki 2025) lists Labrousse explicitly. The “Bengtsson 1981-1990” claim in some references (Wikipedia and others) is incorrect; Bengtsson’s directorship began 1 January 1982.

Site at Shinfield Park, Reading

The ECMWF site at Shinfield Park is a former Royal Air Force administrative complex on the southern outskirts of Reading, Berkshire, immediately south of Junction 11 of the M4 motorway and approximately twelve miles east of Bracknell. The site was developed during World War II as an RAF training and administrative facility, and continued in RAF use into the post-war era.

Site acquisition. The Centre’s site at Shinfield Park was made available by the United Kingdom government in 1975-1976 as part of the UK’s host-state commitment under the Convention. The site was leased rather than purchased; the lease has been extended periodically. The exact lease terms are not in the public record; the Convention is silent on the precise UK contribution but Article 22 of the Protocol on Privileges and Immunities (annexed to the Convention) treats the Centre’s premises as inviolable territory under international law, in the manner of an embassy.

Original buildings. The Centre’s first headquarters building at Shinfield Park was a purpose-built, three-storey office and computer-hall complex designed in the modernist British civic-architecture style of the late 1970s, with brick exterior, expanses of bronze-tinted glass, and an attached single-storey computer hall designed to accommodate the Cray-1 (and successors) and the operational tape and disk archive. The building was opened officially by HRH Prince Charles (subsequently King Charles III) on 15 June 1979, accompanied by the Lord Lieutenant of Berkshire. The opening ceremony preceded the start of operational forecasting by six weeks and three days.

The computer hall. The Cray-1A serial number 9 was installed at Shinfield Park on 24 October 1978, in the dedicated computer hall designed for it. The hall had specialised power conditioning (the Cray-1 drew approximately 115 kilowatts), Freon-coolant plumbing for the chilled-water heat exchangers, and EMI shielding. The Cray-1A was the first Cray supercomputer in Europe.

Earlier transitional facilities. Before the Shinfield Park site was ready, ECMWF operated from temporary facilities. The interim location was Fitzwilliam House at Bracknell, on the UK Met Office complex. Computing during this period (August 1975 to October 1978) was on a CDC 6600 rented from Control Data Limited under a Service Agreement (later converted to a Lease Agreement in December 1975), supplemented by purchased time on the UK Met Office’s IBM 360/195. The CDC 6600 was inadequate to the projected operational workload — Wiin-Nielsen had reported that the machine took 12 days to produce a 10-day forecast, an obviously useless ratio — and was always understood as a development machine, not an operational machine. The transition from the CDC 6600 to the Cray-1A in late 1978 was the moment ECMWF acquired the operational compute capacity it needed.

Subsequent expansion. The Shinfield Park site has been expanded periodically: a second building was added in 1990; a third building in approximately 2000; a “Pink Cottage” (formally, Keeper’s Cottage, a 5000-square-metre adjacent property) was acquired in 2004 by Director Marbouty under exceptional procurement powers when it came up for auction at short notice; a new headquarters complex on the University of Reading’s Whiteknights Park campus is under construction (started 2025, scheduled completion 2027). The Centre will move from Shinfield Park to Whiteknights at approximately the same time as it is opening its 50th-anniversary commemorative book (October 2025).

The organisational principles

The institutional architecture of ECMWF, as fixed in the Convention and as it has operated since 1975, is a particular blend of intergovernmental research-centre conventions:

Council of Member States (Convention Article 4-6). The Council is the highest organ of the Centre. It meets at least once a year (in practice twice). Each member state has up to two representatives, one of whom is normally the Director-General of the national meteorological service. The Council elects from its members a President and Vice-President for one-year terms (renewable up to three times). The Council adopts the annual budget, the long-term strategy, the scale of contributions, the programme of activities, and is responsible for hiring the Director-General. The WMO is represented as an observer.

Director-General (Convention Article 8). The Director-General is the chief executive officer of the Centre, appointed by the Council on a fixed-term contract (originally five years, currently five years renewable once). The Director-General is the legal representative of the Centre, manages all departments, and reports to the Council.

Scientific Advisory Committee (Convention Article 7). Twelve members appointed by the Council in their personal capacity for four-year terms (renewable once). The SAC advises the Council on the scientific programme, on the Director-General’s scientific proposals, and on the implementation of the programme. The SAC is structurally analogous to the scientific advisory committees of CERN, ESO, and other intergovernmental research centres.

Finance Committee (Convention Article 9). Advises the Council on the budget, financial regulations, and accounts.

Technical Advisory Committee (added by the 2010 amendments). Advises the Council on operational and technical aspects of the programme, particularly communications, computing systems, and the four-year activity programme. Before 2010 this function was performed by the Technical Sub-Committee of the Council.

Policy Advisory Committee, Advisory Committee of Co-operating States, Advisory Committee for Data Policy (later additions). Specialised committees added during the institution’s evolution.

Member-state contributions. Article 13 of the Convention establishes that member states contribute to the budget on a scale based on their average gross national product (GNP, in the original Convention; gross national income, GNI, in the 2010 amended Convention) for the three preceding years. The scale is recalculated every three years. The Convention sets a 22% maximum contribution by any single state and a 0.5% minimum. In practice the United Kingdom, France, and Germany account for approximately half the total budget; Italy and Spain together account for another approximately 12-15%; the remaining 35-40% is distributed across the other member states in roughly proportional shares.

Member-state computer time allocations. A consequence of the Convention’s Article 2 paragraph 2(e) is that ECMWF must “make available to the Member States for their research, priority being given to the field of numerical weather forecasting, a sufficient proportion of its computing capacity, such proportion being determined by the Council.” In practice, approximately 25% of ECMWF’s supercomputer time is allocated to member states for research uses; the remaining 75% goes to ECMWF’s own operational and research programme. This 25:75 allocation was the basic design parameter from 1979 onwards and has held with minor variation through every supercomputer generation.

Lennart Bengtsson — second Director (1982-1990)

The Centre’s second long-tenure Director was Lennart Bengtsson, Swedish meteorologist.

Birth and family. Bengtsson was born 5 July 1935 in Trollhättan, Sweden, an industrial town on the Göta älv river in Västergötland province (about 80 km north of Gothenburg). He celebrated his 90th birthday on 5 July 2025 and was, as of late 2025, still alive, scientifically active, and an Honorary Senior Research Fellow at the University of Reading.

Education. Bengtsson studied at Uppsala University, where he took his Bachelor of Science in 1959. He then moved to Stockholm University to study under Bert Bolin (the same Bert Bolin who chaired the IPCC at its founding in 1988), where he took his Filosofie Licentiat (the Swedish pre-doctoral degree, intermediate between MSc and PhD) in meteorology in 1964. His Docent (post-doctoral teaching qualification) was awarded in 1978.

Pre-ECMWF career: SMHI (1961-1974). Bengtsson joined the Swedish Meteorological and Hydrological Institute (SMHI) at Norrköping in 1961, eventually becoming Head of Aerological Forecasting and Numerical Prediction. During his SMHI years he developed numerical weather prediction models for operational use in Sweden; some of these were also adopted operationally by the Royal Netherlands Meteorological Institute (KNMI) and by other smaller European weather services in the 1960s. By 1969 he was sufficiently senior in the European NWP community to be involved with the COST project EMCC (European Meteorological Computing Centre for Research and Operations) — the working group that drafted the ECMWF Convention.

ECMWF (1974-1990). Bengtsson moved from SMHI to ECMWF in 1974, joining the planning staff at Bracknell. He became formal Head of Research from 1975 to 1981. As Head of Research during 1975-1981 he was responsible for the model development that produced the 1 August 1979 first operational forecast. He served as Director-General of ECMWF from 1 January 1982 to 31 December 1990, nine years, the longest single tenure of any ECMWF director.

The 1979-1981 transition. Wiin-Nielsen left the directorship on 31 December 1979 to take up the WMO post; Jean Labrousse was Director from 1 January 1980 to 31 December 1981; Bengtsson was Director from 1 January 1982. Bengtsson’s own retrospective, in the 50-years anniversary book (October 2025), describes his Bracknell-era beginnings:

“I had the great opportunity to be involved with the COST project EMCC (European Meteorological Computing Centre for Research and Operations) in 1969 and from 1974 as member of the planning staff in Bracknell. The period prior to January 1982, when I became Director of ECMWF, was a highly active scientific and technical period including the development of a fully operational system for global medium-range weather forecasts.”

Bengtsson’s principal contributions as Director (1982-1990) were:

  • The transition from grid-point to spectral model (1983), with truncation T63, 16 sigma levels.
  • Subsequent transitions to T106 (1985) and continued increase in resolution.
  • The decision to enter into the IFS-ARPEGE collaboration with Météo-France (formally launched 1987), shared codebase between the two centres.
  • The design of the Meteorological Archival and Retrieval System (MARS) in 1985, the model output archive that is still in use today (and that reached one petabyte in 2003 and one exabyte in 2024).
  • The procurement of the Cray X-MP/22 (1984), Cray X-MP/48 (1985), and Cray Y-MP 8/8-64 (1990) supercomputers in succession, the lineage of vector supercomputers that powered the Centre through the 1980s.
  • The decision to begin the First ECMWF Re-Analysis (ERA-15) activity, which was politically launched in 1989-1990 (against initial Council scepticism, as Burridge later described in his anniversary essay) and which produced the first multi-decade reanalysis dataset of any forecasting centre.
  • The negotiation of full ECMWF membership in the Coordinated Organizations system (NATO, ESA, OECD, Council of Europe, EUMETSAT), facilitated by Lord Carrington’s appointment as NATO Secretary-General. This gave ECMWF staff full pension and health rights on the same scale as NATO and ESA staff.

Subsequent career: Max Planck Institute (1991-2000). Bengtsson left ECMWF in 1990 and from 1991 to 2000 was Director of the Max Planck Institute for Meteorology in Hamburg, where he established a third departmental section (“Theoretical Climate Modeling”) and converted the ECMWF model into the climate-modelling tool ECHAM that became Germany’s principal climate model. Jochem Marotzke, current managing director of MPI-M, said in the 2025 90th-birthday announcement: “Bengtsson was instrumental in advancing climate modeling at the institute in the 1990s.”

University of Reading (2000 onwards). From 2000 Bengtsson was Senior Research Fellow at the Environmental Systems Science Centre (ESSC) at the University of Reading; he transitioned to Honorary Senior Research Fellow on retirement in approximately 2015. From 2008 to 2013 he was Director of Earth Sciences at the International Space Science Institute (ISSI) in Bern, Switzerland.

Awards. Milutin Milanković Medal of the European Geophysical Society (1996); Alfred Wegener Medal of the European Geophysical Union (2009); René Descartes Prize for Collaborative Research from the European Union (2005); International Meteorological Organization Prize from WMO (2006); Rossby Prize from the Swedish Geophysical Society (2007); Honorary Member of the Royal Meteorological Society, the American Meteorological Society, and the European Geosciences Union; German Environmental Reward.

The “Royal Astronomical Society Gold Medal 2014” claim. The user’s brief asked about a Royal Astronomical Society Gold Medal awarded to Bengtsson in 2014. I find no primary documentation of this award. The 2014 Royal Astronomical Society Gold Medals (in Astronomy, Geophysics) for that year went to other recipients; Bengtsson is not on the historical list of RAS Gold Medal recipients. The claim should be flagged as probably incorrect; what is documented in 2014 is Bengtsson’s controversial board membership of the Global Warming Policy Foundation (a UK climate-policy think tank) and the public dispute over the rejection of his and colleagues’ Environmental Research Letters submission on climate sensitivity. No award.

Status in 2025. Living, age 90, residing in Switzerland. Honorary Senior Research Fellow at the University of Reading.

The first operational forecast — 1 August 1979

The first ECMWF operational medium-range weather forecast was produced and disseminated on 1 August 1979.

The forecast. The forecast was a global atmospheric prediction extending to 10 days ahead, of which 7 days were disseminated to member states. The forecast was issued from the 00 UTC analysis on 1 August 1979 (or possibly the 12 UTC, sources differ marginally; the canonical citation is “the 12:00 Z forecast” per the Spanish 50th anniversary article); the canonical operational milestone was the dissemination at the end of that day’s compute cycle, several hours later.

The model. The first ECMWF operational model was a finite-difference grid-point primitive-equation model of the global atmosphere, with:

  • Horizontal resolution: 1.875 degrees (approximately 210 km), on a regular latitude-longitude grid with 48 computed latitudes at the equator (the model’s “N48” grid in the modern ECMWF notation).
  • 15 vertical levels in sigma coordinates (terrain-following pressure-based vertical coordinates); the user’s specification of “15 sigma levels” is correct.
  • 6-hour data assimilation cycles using three-dimensional optimal interpolation analysis (Lorenc-Rutherford-Larsen 1977).
  • Atmosphere-only model, with sea surface temperature held constant at each grid-point through the forecast (no ocean model coupling).
  • Five operational forecasts per week through the first year (1979-1980).

The user’s specification of “T63 spectral, 15 sigma levels” is incorrect for 1979. The first ECMWF operational model was a grid-point model. The T63 spectral model was introduced operationally in April 1983, four years after the start of operations, as the first major model upgrade under Bengtsson’s directorship. By April 1983 the model was T63 with 16 vertical levels. The grid-point model of 1979-1983 was not a spectral model.

The computer. The model ran on the Cray-1A serial number 9 that had been installed at Shinfield Park on 24 October 1978. CPU time: approximately 5 hours per 10-day forecast, on the single-processor Cray-1A. The Cray-1A was the second machine of the Cray-1 family (after Serial 1 at Los Alamos); it was the first Cray-1 in Europe; it had 1 million 64-bit words of bipolar memory (8 megabytes), the maximum Cray-1A configuration.

Launch day. As recounted in the ECMWF technical newsletter and in the Spanish 50th anniversary article: 1 August 1979 was tense. The forecast began but the Cray-1 experienced operating-system problems during the evening run. Staff had to “dead-start the Cray-1 with a different version of the operating system.” The forecast was eventually completed and disseminated to member states with a four-hour delay relative to schedule.

The transition from research to operations. Operational forecasting in 1979-1980 ran five days per week (Monday-Friday). On 1 August 1980, after one full year of operational experience, the schedule was extended to seven days per week — the full operational service. Member-state national meteorological services subscribed to the ECMWF medium-range output and incorporated it into their own forecast products from August 1980 onwards. The improvement in forecast skill that ECMWF demonstrated over national-service forecasts in the 7-10 day range was, by 1981-1982, sufficiently dramatic that the Centre’s operational role was institutionally established beyond debate; the political question of whether a single European medium-range forecast centre could deliver on the original 1969 Süssenberger vision was answered.

A note on forecast quality. According to ECMWF’s 1979 Annual Report (cited in the Newsletter 161 article), the early forecasts varied seasonally: “some forecasts proved ‘remarkably good’ through day 10, while others showed practical predictability limits under a week.” David Burridge’s 50-years-anniversary essay (October 2025) summarises the 1979 launch in his characteristic understatement: “The goal to deliver global forecasts operationally to the Member States in 1979 was met on 1 August 1979 — an outstanding achievement, the more so as they were the best available and they have remained so ever since.”

Cross-references for the post

The ECMWF founding post links to several earlier posts in the NWP history series:

  • The Süssenberger 1969 report comes out of the same European-supercomputing institutional pressure that produced the ESRO/ESA satellite programme, the IBM 360/91 deployments at major European universities, and the broader 1960s realisation that European national-level scientific computing was inadequate to international competition. (See post on IBM 360/91 and the European mainframe market, and post on the CDC 6600.)
  • The principal scientific challenge ECMWF was set up to solve — global primitive-equation models with realistic physics, run at high resolution — comes out of the Mintz-Arakawa lineage at UCLA (see post on Arakawa) and the Smagorinsky-Manabe lineage at GFDL Princeton.
  • The Cray-1A at Shinfield Park (October 1978) was Cray Research Inc.’s first European installation; the Cray-1 architecture and Seymour Cray’s company are the direct successors to the CDC 7600 lineage covered in post 32. The Cray-1 will be a future post in this series.
  • The grid-point primitive-equation ECMWF model of 1979 is intellectual descendant of the JNWPU operational grid-point models that began at Suitland in May 1955 and that Wiin-Nielsen worked on briefly in 1959-1961.
  • The COST 70 framework that produced ECMWF was contemporaneous with ESRO’s COST-equivalent activity that produced ESA in 1975. ECMWF, ESA, and EUMETSAT (founded 1986) became the three principal European intergovernmental scientific organisations of the late twentieth century.

Open research questions

A few questions remain incompletely answered from this round of research and might benefit from primary archive work:

  1. The exact list of competing siting bids (1973). The Brussels, Geneva, Hamburg, Rome competitor list is asserted in some sources but the primary record of the May 1973 siting decision is not in the public-facing ECMWF or COST archives; Woods 2006 references it but does not enumerate. Definitive answer would require the Council of the European Union’s COST 70 archive.

  2. The exact list of 13 ratifying states by 1 November 1975. The primary record is the ratification deposits in the Council of the European Union archive in Brussels; the consolidated Convention’s signature page lists signers but not ratification dates. The deduction of “all 16 except Austria, Portugal, Yugoslavia, and one other” is by elimination from secondary sources, not from primary record.

  3. The “ESRO 1968 study” claim. Several secondary sources reference an ESRO study on European meteorological computing as part of the pre-history. I have not found primary documentation of such a study. The institutional ancestry as recorded in primary sources runs through the EEC’s Süssenberger group (1967-1969) and through COST 70 (1970-1973). The “ESRO study” claim should be flagged as not documented in primary sources until or unless such a study can be located.

  4. The Royal Astronomical Society Gold Medal 2014 for Bengtsson. Not documented; probably incorrect. RAS Gold Medal recipients are well-documented and Bengtsson does not appear in the canonical list.

  5. The exact 1976-1980 staff numbers. Sources broadly indicate “approximately 50 in 1976 growing to approximately 130 by 1980,” but precise headcount by year is only in Woods 2006 and the ECMWF annual reports of the period.

Summary of corrected facts (vs the user’s brief)

  • Wiin-Nielsen birth date: 17 December 1924, not 26 April 1924. (26 April is his 2010 death date.)
  • Wiin-Nielsen ECMWF tenure: 1 January 1974 to 31 December 1979. (The user’s “1974-1979” is correct; the brief was unsure of exact start.)
  • Bengtsson birth: 5 July 1935 in Trollhättan, Sweden. Confirmed.
  • Bengtsson ECMWF tenure: Head of Research 1975-1981; Director-General 1 January 1982 to 31 December 1990. (The user’s “1981-1990” for the directorship is incorrect. Jean Labrousse was Director from 1 January 1980 to 31 December 1981 in the Wiin-Nielsen-to-Bengtsson transition.)
  • Bengtsson is alive as of 2025-2026, age 90. Confirmed.
  • Royal Astronomical Society Gold Medal 2014: not documented. Probably incorrect.
  • First operational forecast: 1 August 1979 grid-point model, 1.875°×1.875° (approx 210 km), 15 sigma levels, 5 hours Cray-1A CPU per 10-day forecast, 7-day disseminated forecast range. (The user’s “T63 spectral 15 sigma levels” specification is incorrect; T63 spectral was the 1983 model, not the 1979 model. The 15-level vertical structure is correct.)
  • Convention signing: 11 October 1973 at Brussels, by 15 of 16 founding signatory states (Austria signed 22 January 1974). Convention came into force 1 November 1975.
  • 16 signatory states, not 17 or 18: Austria, Belgium, Denmark, Finland, France, FR Germany, Greece, Republic of Ireland, Italy, Netherlands, Portugal, Spain, Sweden, Switzerland, UK, Yugoslavia.
  • Norway is widely cited as a “founding state” but did not sign the Convention in 1973-1974; Norway’s accession was much later (1989).
  • Reading was chosen in May 1973; Denmark finished second. Other competitor cities probably included Brussels, Geneva, Hamburg, Rome, but the full bidder list is not in the public record.

Voice notes for the post

The previous two posts in the series (Post 32 on the CDC 7600 NCAR and Post 33 on ILLIAC IV at NASA Ames) are both American-supercomputing-and-weather-modelling stories. The ECMWF post is the European pivot. The voice should:

  • Open with a specific scene — perhaps the 1 August 1979 first forecast, or the May 1973 Brussels siting decision, or Wiin-Nielsen taking up duties on 1 January 1974 in temporary Brussels offices.
  • Establish that ECMWF is the European response to the same 1960s realisation that NCAR was the American response to: that medium-range modelling needs more compute than any single national-level meteorological service can buy. NCAR (founded 1960, federal funding through NSF) is the obvious comparison; CERN (founded 1954, intergovernmental) is the institutional template.
  • Bring in Wiin-Nielsen as the founding director and trace his line back through Fjørtoft, Stockholm, JNWPU, NCAR, Michigan to ECMWF. Wiin-Nielsen’s career embodies the trans-Atlantic Bergen-Princeton lineage.
  • Show the institutional politics — the COST 70 framework, the 11 October 1973 Brussels signing of 15 states, the 1 November 1975 entry into force, the Reading siting decision.
  • Show the Cray-1A arriving at Shinfield Park on 24 October 1978 — first Cray in Europe — and the 1 August 1979 first operational forecast running on it.
  • Show Bengtsson taking over the directorship in January 1982 and the spectral-model transition of 1983 setting up the long Bengtsson era.
  • Avoid forward-looking promises about future posts (per project memory). Avoid Co-Authored-By Claude in commits (per project memory). Cross-link to previous posts using full attribution sentences (per project writing-style memory).

Word count target was approximately 5000; this research file is approximately 6800 words. Suitable length for the source material, allowing the post itself to draw 3000-4000 words from this background.