«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
Physicists, for example, had already created a number of international and highly regarded institutions in which problems exceeding the capacity of individual countries were tackled. Meteorology with its long tradition of successful co-operation was seen as being particularly appropriate for such a joint European venture. On 19 October 1970 the Working Party on the
EMCC was instructed by a Committee of COST Senior Officials to continue its work, and to state its views on whether the Centre should:
• have scientific and/or public service roles, and how these roles might be combined, and
• be a centralized body, or a network between national centres, or a combination of the two.
The Working Party was also asked to specify the cost and arrangements envisaged for each alternative.
The International Meteorological Institute in Stockholm was seen as a model. By 1970 there had been enormous progress in data handling, atmospheric modelling and computer technology. This had made feasible, in principle, the production in reasonable time of useful weather forecasts beyond the period of up to three days or so, which was at the time the Meteorological developments 1967 to 1971 33 absolute limit in operational NWP. Indeed in operational forecast offices of the time, while 24- to 48-hour numerical predictions were routinely used operationally, forecasts to 72 hours or more were at best treated with caution. Experimental prediction with an advanced general circulation model in the United States had shown considerable promise for forecasts up to 4 to 10 days ahead; see the figure.
There was agreement among scientists that no fundamentally new principles would be encountered in developing dynamically derived medium-range forecasts, and that these forecasts would prove to be superior to those produced by then-current methods.
70 60 50 2004/05 40 30 20 1979/80 1964/69 10
Scores showing the average accuracy of a series of predictions to 10 days ahead:
• for 12 experimental forecasts made during the period 1964 to 1969,
• for ECMWF forecasts for the winter of 1979–80, the first year of operations, and
• for the winter forecasts of 2004–05.
A score above 0.6 is generally accepted as indicating that the forecasts are still on the average useful. The score remained above that level until about 31⁄2 days in the experimental medium-range forecasts during the late 1960s.
For comparison, the score remained above 0.6 until 51⁄2 days in the December to February forecasts made during the Centre’s first year of operations, and until 81⁄2 days in the forecasts made 25 years later.
The experiment made at GFDL at Princeton New Jersey in the late 1960s was the first indication that medium-range numerical forecasts would be feasible.
Score: Anomaly correlation, 500 hPa, Northern Hemisphere. The 1964/69 score is adapted from Miyakoda et al. (1972) Cumulative results of extended forecast experiments. Monthly Weather Review 100: 836–855.
34 Chapter 3However, at that time, it was clear from experience both with operational short-range prediction and with general circulation models that the future application of dynamic methods to medium-range forecasting would involve more than a simple extension of short-range models.
While the latter’s success depended mainly on their ability merely to redistribute the kinetic and available potential energy within the atmosphere, medium-range models would have to be able to describe energy production and dissipation. They would have to include the hydrological cycle. The models would have to allow extra-tropical cyclones to form, develop and decay. Also, interactions with tropical phenomena implied that the circulation of the Southern Hemisphere had to be taken into account to make Northern Hemisphere forecasts for periods longer than about a week. Further it was anticipated that treatment of the interactions between atmospheric and oceanic circulations would be required for good-quality medium-range predictions.
In contrast to the situation in medium-range forecasting, there was at the time no promising approach to long-range prediction for a period of a month or a season by dynamical methods. Therefore, the logical decision was made to concentrate research and development capacities on the construction of atmospheric models suitable for dynamic medium-range predictions of increasing quality, thereby extending the range of useful deterministic forecasts as far as possible. Monthly and seasonal predictions were assigned lower priority for the initial work of the Centre.
In addition, numerical experiments had shown that further progress in short-range forecasting was likely to be achieved by studying the dynamics of small-scale phenomena and developing appropriate fine-mesh models. This research would be undertaken in parallel with that done in atmospheric modelling for medium-range predictions. It was foreseen that continuous interaction would benefit both.
It was clear that the development and routine application of atmospheric models for medium-range forecasting required tremendous computing power. Even the fastest computers operational at the time would not suffice. Establishing a meteorological computing centre devoted largely to the development of routine medium-range forecasting, therefore, was foreseen to be a costly and challenging project. It was likely to be beyond the financial resources and the research capabilities of most European National Meteorological Services. A combined effort was called for.
This conclusion was supported by the anticipation that the future development in short-range weather forecasting, including the need for finer resolution and quantitative precipitation forecasts, would considerably Meteorological developments 1967 to 1971 35 increase the computer requirements of national centres and would take up much of the computing capacity available to them.
Although the preparation of routine medium-range forecasts together with the associated research activities provided the main arguments for the projected European Meteorological Computing Centre, it was clearly reasonable to expect that this Centre would provide advanced training to post-graduate scientists in NWP and related disciplines.
It was foreseen that the Centre would make available its advanced computing facilities to National Meteorological Services for activities beyond their computing resources, for example research into the dynamics of small-scale systems. The computing facilities of the Centre could be accessed by national institutions via Remote Job Entry, using the same telecommunication network as would be required for rapid dissemination of the medium-range predictions to the computer systems of the services.
Furthermore, the Centre could support related national research activities, for example numerical studies of local phenomena, by offering suitable working facilities to visiting scientists from national centres. The Centre should also serve as a European meteorological data bank.
The Working Party on the EMCC met on 9 November 1970 and 15 January 1971 under its Chairman Dr Süssenberger and Vice-Chairman Mr Schneider. At a meeting of an Expert Group on 19–20 November 1970, Study Groups were set up to prepare a Report to Dr Süssenberger’s Working Party.
• Project programme, especially the cost of the work (Reiser, Hipp).
• Forecasting model and its effect on the computing power required (Bengtsson, Lavalle).
• Requirements and production of the data of the Centre (Palmieri, van Isaker).
• Time schedule for the achievement of the Centre.
At its meeting on 15 February 1971, the Expert Group set up another Study Group “EMCC — Telecommunication aspects” with Chairman Jean Labrousse, which produced a detailed Report on 3 April 1971.
Lennart Bengtsson and Lodovico La Valle, head of the Meteorological Computer Centre of the Italian Meteorological Service in Rome, visited laboratories and factories in the USA between 8 and 21 March 1971: NCAR, GFDL, NMC in Suitland, IBM in Poughkeepsie, Burroughs and UNIVAC in Pennsylvania and Washington, Control Data in Minneapolis, and Texas Instruments in Austin. On 29 April 1971, they prepared a report on “Present activities, organisation and plans for the future of some advanced laboratories for dynamical meteorology and numerical forecasting”. The other Study Groups also prepared input for the Project Study.
The Project Study
The Study Group chaired by Dr Heinz Reiser had the task of preparing the Report: “Project Study on European Centre for medium-range weather
forecasts”. The Group had representatives from a wide range of nationalities, all experts in their fields:
H. Reiser Germany Chairman J. Van Isacker Belgium J. Labrousse France R. Pone France L. La Valle Italy S. Palmieri Italy D. J. Bouman Netherlands K. Cehak Austria D. Söderman Finland L. Bengtsson Sweden E. Knighting United Kingdom The impressive and important 76-page Report, with Annexes totalling 130 pages, was presented to Dr Süssenberger, Chairman of the Working Party on the European Meteorological Computing Centre, on 5 August
1971. The Report incorporated the results of all the Study Groups mentioned at the end of the last Chapter.
With hindsight, the work of Reiser’s Group is remarkable. The basic ideas on the organisation, implementation and performance of the Centre as prepared by the Group in 1971 and summarised below bear a striking resemblance to the Centre 35 years later. All the important aspects, organisational, administrative, scientific and technical, were covered. These are now described.
36 The Project Study 37 The operations of the Centre should effectively supplement the present activities of the national centres; duplication should be avoided as far as possible.
Numerical Weather Prediction (NWP) requires the use of the most powerful available computers; the expected development in medium-range forecasting would call for even more computing power. Accordingly, the Centre should be in a position to take advantage of new developments in computer systems.
The development and continuous improvement of operational mediumand long-range predictions at the Centre would be the main responsibility of the research section. In order to extend this research capacity, to promote co-operation and contacts with national institutions and to facilitate the exchange of views and knowledge, working facilities should be provided for temporary groups working on associated research problems. These groups would consist of visiting scientists from national groups and members of the permanent research staff.
The problems of routine forecasting and associated research in atmospheric modelling are closely related. Experience at National Meteorological Services, however, suggested the need for a separate “research group” which was to be independent of the routine operations. The latter would be the responsibility of an “operational group”; there would be continuous interactions between these groups. The results obtained from operational forecasts would influence the development of more advanced atmospheric models; new developments by the research group would be included in updated versions of the routine model.
The main responsibilities of the research group were to be the development and intensive testing of dynamic models for medium-range predictions of increasing quality. The operational group would be responsible for all applications outside the research sector, including operating the computing system. Their tasks would include the preparation, dissemination and verification of dynamic medium-range forecasts as well as special services to national centres, telecommunications problems and the creation and maintenance of a European data bank.
A clear organisational separation between the research staff and the staff needed for the operation of the Centre was desirable, to protect research and development activities from the increasing operational requirements.
For the implementation of the Centre three main phases were considered:
initial phase, transition phase and fully operational phase. The initial phase would start well before the installation of the computing system and involve construction and testing of a first model. The transition phase would be
38 Chapter 4characterized by a tentative routine application of this model with initial analyses obtained from other centres and a steady approach towards the fully operational phase. For this final phase a fixed ratio of the computing time available for (a) routine operations, (b) model-orientated research and (c) the requirements of National Meteorological Services was recommended: “at a first guess, nearly equal parts should be assigned” to each of these.
It was calculated that the daily routine forecast would normally take about one hour per forecast day, so about 10-12 hours for the ten-day prediction.
It was suggested that about one-third of the computing resources would be available for Member State use — not leaving much time for research when operations got under way!
If better but more time-consuming models became operational their computational requirements should be satisfied by extension of the computing capacity rather than by reduction of the computing times allotted to research and other services.
High-speed data links between the Centre and associated National Meteorological Services were indispensable for the dissemination of the medium-range predictions. Some of these data links were also necessary for rapid input of digital data in the form of grid-point values or pre-processed data originating from European and other centres. The satisfactory incorporation of all these data requirements into existing and projected WMO telecommunication channels appeared unlikely; a separate data net for the envisaged computing centre was needed. All these high-speed data links should be capable of operation in full or half duplex mode and hence would provide an ideal basis for teleprocessing of data.
For routine medium-range forecasts, analyses of the current global atmosphere from which the predictions would begin would of course be needed.
The supply of analyses was to be the responsibility of the operational group.