«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
According to WMO plans, routine global analyses were to be prepared at the World Meteorological Centres (WMCs) of Washington, Moscow and Melbourne. There were additional plans to establish a Global Analysis Centre, in connection with the Global Atmospheric Research Programme (GARP), which might be situated at one of the WMCs.
It was therefore assumed that after 1975 the Centre would in principle be able to obtain suitable global analyses from one of the WMO Centres for its forecasting activities, and not have to devote scientific efforts and its valuable computing resources to making its own. Some, especially in Germany and the UK, felt rather strongly that the Centre should not develop its own analysis system.
The Project Study 39 It was however also envisaged that the Centre might at some stage perform its own analyses in real time. This would become a necessity if the Centre were to receive and process observations from European satellites.
Such an extension of its responsibilities would influence the planning for telecommunications and the composition of the scientific personnel.
However, the study did not consider these problems further.
In any event, if suitable, good quality, analyses could not be obtained from the WMO Centres, it was foreseen that a major effort would be required to prepare global analyses at the Centre. These analyses would involve extensive use of satellite data with the development of appropriate techniques for assimilating the new data into the models.
We have noted that the medium-range had been characterized as a forecast period of 4 to 10 days. It was assumed (wrongly, as it turned out!) that models with rather crude estimates of energy production and dissipation and using hemispheric integration areas could successfully cover the short-range period of up to four days.
Though the models were expected to produce full sets of forecast charts from analysis to the end of the medium-range, it was evident that the geographic scale of predictable phenomena would increase with the forecast period. Short-range forecasts should be able to predict the location and intensity of rather small-scale, well-developed pressure centres, and major temperature changes and precipitation amounts over small areas. Mediumrange predictions were expected to indicate the significant changes of the weather over fairly large areas. It would be the Centre’s responsibility to investigate possible long-range prediction methods following a satisfactory solution of the medium-range forecast problems.
With the state of knowledge in 1970, a detailed description of an atmospheric model for medium-range forecasting was not possible without some rather arbitrary assumptions. To make a reasonable estimate of the computer requirements, it was necessary to consider the structure of an unfiltered dynamic model, without implying a recommendation for the characteristics of the actual model to be developed by the Centre. The model corresponded
roughly to the models used in the USA for weather and climate simulation:
at the National Center for Atmospheric Research, the Geophysical Fluid Dynamics Laboratory, and the University of California.
It was agreed that the Centre initially should not develop a very advanced and complicated model, but rather try to set up a first version on the basis of general circulation models already available and proven, and later to produce more advanced versions.
40 Chapter 4Since it was planned to perform operational forecasts for 4 to 10 days, a fairly complete description of the non-adiabatic processes, including the complete hydrological cycle, and also of the dissipative forces in the atmosphere, was deemed necessary.
The computer requirements of the forecasting model had to be thoroughly considered in the Project Study. The estimates made were based on a representation of grid-points for the medium-range model with a grid distance of 150 km from the North Pole to about 20°S, increasing to about 300 km south of 20°S.
Such a grid would consist of somewhat more than 15,000 grid-points horizontally. The corresponding vertical resolution would be about 100 hPa to 150 hPa in the troposphere. If some additional levels in the surface boundary layer and in the stratosphere were added, the model would have at least ten levels. There would thus be about 200,000 grid-points in the computer model. Sub-grid scale phenomena, such as cumulus convection, would be taken into account by describing their effects statistically on the parameters of the large-scale flow, that is to say, they would be “parameterized”. Since only very limited knowledge was available on the effects of the oceans, the sea surface would be represented rather crudely in the first version of the operational model.
Over the continents, the coupling between the atmosphere and the underlying earth depends mainly on the available ground moisture and the snow cover. These time-dependent properties had also to be included in a model for medium-range forecasting.
Both the routine computation of medium-range dynamic forecasts and the corresponding research in atmospheric modelling would determine the main computer requirements for the Centre. All other operational activities, including possible preparation of global analyses and processing of satellite measurements for these purposes, were considered to be smaller by an order of magnitude and did not affect the main requirements for computing speed and capacity of fast internal memory. [With the benefit of hindsight, we can see that this severely underestimated the benefit of a good analysis for a medium-range forecast. We will see later just how important the Centre’s analysis system, and its research into use of satellite data, would become.] For operational weather prediction, a practical ratio of computing time to real time was taken to be about 1 to 20, which corresponds to about one hour computing time for a one-day forecast or half a day for a forecast to ten
days. When these model characteristics:
• 200,000 grid-points, • 3,000 operations per grid-point per time step, The Project Study 41 • 5 minute time-step, and • 1 to 20 ratio of computing time to real time were combined, a total of 600x106 operations per time step would be executed within 15 seconds. This meant a required computing speed of 40x106 instructions per second or a 40 Million Instructions Per Second (MIPS) main-frame computer.
The model assumptions represented a rather conservative estimate based on limited experimental experience. Hence a speed of about 50 MIPS was considered appropriate; further substantial improvements in the model were foreseen to call for speeds of 100 MIPS or even higher. If there was a computer that could be upgraded to at least 50 MIPS without major reprogramming, it was deemed to be economic to equip the Centre in the beginning with a computer system of 10 to 20 MIPS. The fully operational phase, however, could not start before a computing speed of about 50 MIPS became available.
The Working Group on the required telecommunication links made an extensive study under the guidance of its Chairman Jean Labrousse of France. The data volume would be considerable, so that a speed of 2400 bits/sec for the telecommunication lines was considered necessary.
With respect to personnel it was foreseen that the Centre would need a Director, and a Research Department with six senior scientists. These would have experience in NWP and atmospheric modelling and special qualifications in one of the following fields.
• Atmospheric physics
• Boundary layer physics
• Small-scale phenomena
• Initialization procedures
• Numerical methods
• Statistical diagnoses Two junior scientists — capable of original research — would assist each of these six senior scientists, programming model codes and carrying out related research and development activities under supervision of the senior scientists. Eight assistants for auxiliary work, for example lower-level programming, were considered to be necessary within the research staff. Research work was to be co-ordinated and inspired by its Deputy Director.
Besides this permanent research staff, financial provision was planned for at least five additional posts reserved for visiting scientists from other research groups. These facilities would not only reinforce the potential of
42 Chapter 4the Centre but also offer excellent opportunities for European scientists to work on special problems in NWP and associated fields.
The Operations Department under a Deputy Director would be subdivided into two sections. One would be responsible for the technical operations of the computing system, with its size dependent on the requirements of the eventual computing installation. The other section would be more scientifically orientated. It would be responsible for the meteorological aspects of routine applications and for the contacts with the National Meteorological Services and WMO. Under a computer manager there would be five scientists and five system analysts. In addition there would be five programmers and seven assistants as well as thirty-two operators and eight additional auxiliary personnel. This would mean a staff of 64 persons in the Operations Department. Together with a second Deputy Director and 26 staff in the Research Department, and a further 21 in the Administration Department under a third Deputy Director, a total staff of about 110 persons was expected to be required. About 40 would have a university education or equivalent qualifications.
The study recommended that the Centre should be used also for training.
With the rapid development of NWP and its growing influence on the daily routine work of National Meteorological Services, there was an increasing need for adequate training facilities in NWP for postgraduate meteorologists. Since the successful application of dynamic methods in NWP required a broad operational basis, universities were normally not in a position to provide the training required.
Some of the National Meteorological Services with experience in NWP had already organised regular training courses, some in co-operation with universities. These courses were normally intended as an introduction to NWP, and were designed for meteorologists without specific experience in this field and thus emphasized the basics. High-level training facilities for scientists actively engaged in research and development work on advanced NWP were provided at the time either in a fairly unsystematic way by their temporary assignment to an established research group, or by special seminars, symposia and similar arranged by interested organisations or societies.
In particular, it was noted that more ambitious seminars or training courses in applied dynamic meteorology for postgraduate participants were best organised on a basis of international co-operation. Thus the establishment of the Centre offered an excellent opportunity to create central training facilities for NWP and related disciplines in Europe. Such an extension of the Centre’s activities would not only serve directly the National Meteorological Services involved but would also help to build up the Centre’s scientific image.
The Project Study 43 It was noted that the Centre’s operations should effectively supplement the activities of National Meteorological Services with a minimum of duplication. Furthermore the Centre should co-operate with the existing international organisations, and in particular with WMO. For this reason a representative of the Secretary-General of WMO was invited to attend the more important sessions of the Expert Group for Meteorology.
We have seen that at this time WMO had well-developed plans for the advance of meteorology, especially the World Weather Watch Plan and the joint WMO-ICSU Global Atmospheric Research Programme (GARP). This Programme was aimed at improving understanding of the physical basis of the general circulation of the atmosphere and at increasing forecast accuracy for extended periods. As well as large observational experiments, GARP planning called for tremendous efforts in atmospheric modelling and numerical experimentation.
The World Weather Watch (WWW) is an impressive worldwide weather observing system. Its origin lies in the 1961 UN General Assembly Resolution on the Peaceful Uses of Outer Space, which owed much to the address made by American President J. F. Kennedy to the General Assembly.
It is designed to make up-to-the-minute meteorological and related information available to all countries. The WWW is a truly remarkable example of international co-operation. It is composed of the Global Observing System (GOS), the Global Telecommunication System (GTS), and the Global DataProcessing System (GDPS). The WWW has supplementary programmes dealing with Satellite Activities, Instruments and Methods of Observation, Tropical Cyclones, and Emergency Response Activities.
One of the very important purposes of the WWW is to stimulate and facilitate the research work necessary to improve the accuracy and extend the useful range of weather forecasts. The Centre would be developing methods of medium-range forecasting as its primary task, and subsequently providing routine operational forecasts. Its proposed objectives were thus closely related to those of WMO. Indeed the work of the Centre would have a considerable impact on the development planned by WMO.
The creation of the Centre, with its aim of developing advanced models for extended forecasts and with a considerable potential for numerical experimentation, coincided very well with the plans of GARP and would contribute to its implementation. In turn, the Centre would profit considerably from the scientific progress expected from GARP.
The Centre would contribute to the Global Data-Processing System by storing data and making them available.
The economic benefits of meteorological activities were well known.