«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
Following the recommendation of the group, Bengtsson decided to invite three scientists to the Centre for the two months of January and February 1986: Peter Janssen, Anne Guillaume from France and Luciana Bertotti from Italy. The work was done as a “special project” under the auspices of KNMI — actually a COST Project headed by Gerbrand Komen. Large computing resources were made available, but there were no free offices at the time — the group worked from tables in the Centre’s Classroom and a small meeting room!
By the end of the two months, a working analysis and forecasting system had been set up. The wave analysis was derived from the analysed winds, and the predicted waves were generated from the forecast winds using a third-generation model. In March, the researchers returned to their home institutes; while the group as such disbanded, work continued and the team met again from time to time as part of the COST Project.
At the same time, three other research scientists working at the Centre, two from the USA, V. J. Cardone and J. A. Greenwood, and Magnar Reistad from the Norwegian Meteorological Institute, implemented a high-resolution (25 km grid spacing) model for simulation of the sea state during hurricanes in the Gulf of Mexico. The results were encouraging, in that the predicted sea state agreed well with the data observed from buoys in the Gulf.
ESA concluded a 12-month contract with the Centre to run from March 1986 to study the use of information from the ERS-1 satellite. This satellite had scatterometer instruments designed to measure the sea state all over the globe. Wind speed and direction, wave height, height of the sea by an altimeter, and sea temperature by a radiometer were all measured.
In September 1986, Janssen returned to work at the Centre under a oneyear contract with ESA. During the course of the work, it became increasingly clear that the Centre would benefit from a global wave model for its work in assimilating data, and in particular in processing the data from the ERS-1 satellite. Further, for accurate near-surface wind prediction, the parameterization of the momentum transfer from the air to the ocean, Wave prediction 149 which results in a slowing of the wind, required knowledge of the wave spectrum, such as was produced by the wave model.
In December 1986, a new flexible version of the model and the required pre-processing software was implemented on the ECMWF computing system, from generation of the required initial spectral fields to archiving of the output and plotting of the output forecast fields.
There were important developments elsewhere too. In 1987, a major international field experiment in the Labrador Sea was begun, to increase understanding of wind-generated ocean waves, and to assess the relative superiority of the recently developed third-generation wave models. Dutch and Canadian research ships, Canadian and American research aircraft and the American Geodetic Satellite (GEOSAT) spacecraft all took part.
GEOSAT was a US Navy satellite designed to measure sea surface heights to within 5 cm.
Janssen decided that the ECMWF model should be in at least a quasioperational state as soon as possible. At 15.00 on Saturday 7 March 1987 a button was pushed to start the Centre’s first quasi-operational global wave forecast. From then on a 24-hour analysis and a five-day forecast were run daily. Verification of the forecast quality by comparison with any available buoy data and with the measurements from the Labrador field experiment was begun. Results were promising.
After spending some months working with Piero Lionello from Italy developing the first data assimilation scheme for waves, Janssen returned to KNMI on 1 October 1987. Work continued at the Centre by a series of visiting scientists: Lionello who developed the data assimilation system further, Liana Zambreski from the USA who stayed until October 1989 and collaborated with MPI in the work, Heinz Günter who worked on the numerical scheme and the efficiency of the wave model, Bjorn Hansen, working for ESA on ERS-1 altimeter and scatterometer data, and others.
Up to now the work was formally to further the Centre’s research. In June 1987, the Council discussed the Director’s report on the experiments in
wave modelling. Comments now were generally favourable. For example:
“The Netherlands was fully in favour”, “Sweden strongly supported”, Germany “welcomed progress”, Finland “supported the proposal that wave forecasts be carried out operationally” and the UK “welcomed the research in wave modelling”. There was some caution: Germany, France and the UK suggested that operational implementation should be clarified with respect to the Convention.
Council asked the Director to prepare a paper on global wave modelling.
An assessment on the quality of the wave forecasts, the resources required and the formal aspects of the Centre running an operational wave model,
150 Chapter 12were all to be considered. The Director’s document noted that a wealth of sea state data would become available from several satellites to be launched in the following two to three years. Use of these data would require a data assimilation system such as that at the Centre to provide analyses of the global wave and low-level wind fields. There was a strong coupling between the winds over the oceans and the waves generated by those winds.
Successful assimilation of wave data required a good wave prediction model, such as the WAM model.
Operational wave prediction at the Centre would require only three scientific staff. At this stage it was estimated that less than 10% of the “number crunching” computing power needed for the atmospheric model would be needed, and 10% or less of the Centre’s archive and telecommunications resources.
The Scientific and Technical Advisory Committees (SAC and TAC) considered the Director’s paper that autumn. The SAC noted that the wave model “appeared to be well based”, and because of the “possible impact on both the representation of the oceanic boundary layer and the optimal use of future satellite data it was scientifically important for the Centre to be involved with global wave modelling”. The TAC recommended that research into wave modelling continue at the Centre.
At the Council discussion in December 1987, Germany “noted that improved medium-range forecasts could be expected as an outcome of global wave modelling”, and “it was important that the European countries be in a position to take advantage of [European remote-sensing satellites] when they were launched”. The UK agreed, “otherwise the very large investment in this satellite would be partly wasted”. France “was in favour of a continuation of the research programme at approximately the same level of resources as before”. This was what Council agreed.
Why was Council somewhat reluctant to allow the Centre to become more deeply involved in wave prediction at the Centre? Partly this was because the formal position was not entirely clear. ECMWF is an independent international organisation established by nations to carry out specific objectives that are specified in Article 2 of the Convention. In a sense, the Convention can be compared to the constitution of a State.
Wave prediction was not mentioned, so it was necessary that Council could convince itself that wave prediction was somehow an integral part of the objectives. Partly, there was a strong commercial interest in wave prediction; this was a profitable and increasing source of revenue for some Member States. It has always been desirable that the work of the Centre and that of its Member States should not overlap. The United Kingdom in Wave prediction 151 particular had a well-established programme of wave prediction of its own, with which it was commercially successful.
Research advanced at an increasing pace through 1988 and 1989. Italian teams visited the Centre for weeks at a time, developing a high-resolution wave prediction model for the Mediterranean and Adriatic Sea, and carrying out research into the effects of severe storms. Verification of the Centre’s WAM global model against data from buoys in the Atlantic near the east coast of the USA showed generally low prediction errors but some biases, and in two cases the model failed to increase waves in response to increasing wind speeds. At KNMI, the third generation WAM model with 75 km resolution was implemented on the Convex computer. Optimisation of this KNMI version of the model continued. Janssen developed a theory for wind and wave coupling. He spent three months May to July 1990 at the Centre.
Research at MPI progressed on modelling and data assimilation. The WAM model was implemented at IBM’s Bergen Scientific Centre in Norway for research by staff of the Norwegian Meteorological Institute. Data of the sea state under Hurricane Josephine obtained by the crew of the Challenger space shuttle were compared to the waves predicted by different models. By early 1990, the WAM model had been implemented for operational testing at the US Navy’s Fleet Numerical Oceanography Center in Monterey California. It was implemented also at Tsinghua University in China.
It was now time to place the activities at the Centre under a more formal umbrella. The Council in May 1989 had adopted a procedure for “Optional Projects”. These were to be Projects from which individual Member States could opt out, so to speak, by declaring that they were not interested in participating. A year later, in May 1990, the Council considered a proposal for a Project for “prediction of ocean waves (associated with the validation of ERS-1 data)”. A proposal for the Project had to come from a Member State, and the document was presented by the Netherlands. In fact David Burridge helped significantly in its preparation.
Council discussion was generally in favour; considerable support was expressed. However an ad-hoc working group was formed to clarify difficulties which were identified during the discussion: for example the UK “would have to be convinced of the value of the third generation wave forecasting model in relation to the resources required before it would be prepared to join”. Italy, France and Germany also expressed the need for clarification relating to commercial interests, funding of the project and more.
At its following session in December 1990, the Netherlands presented an enhanced proposal. The UK noted that “great difficulties could be foreseen for funding meteorology in Europe in the coming years.... It was concerned that
152 Chapter 12all possible alternative sources for the prediction of ocean waves had not been addressed.... the UK had offered to make available to other national services the wave forecast which it produced”. However other delegations supported the proposal including Germany: “this would be a potentially useful application of resources of the Centre”, Italy: “at a time when funding was difficult nationally, funds for Optional Projects should be encouraged”, and France, which: “saw a link to the medium-range atmospheric model”. The position of the UK was softening, partly perhaps because of the strong interest its Director-General John Houghton had in satellite meteorology and in the ERS satellites in particular. Dr Houghton later became renowned also for his work as Co-chairman of Working Group 1 (Science) of the Intergovernmental Panel on Climate Change.
Council approved in principle operational wave forecasting as an ECMWF Optional Project, with all except the UK (which abstained) voting in favour. Council gave a two-year deadline for the provision of the necessary computing, manpower and financial resources; otherwise its approval would lapse. In fact the Member States involved were able to get their act together quickly, and at its session in June 1991, the Council approved the implementation of the Project. A “Reduced Council” was set up to oversee the Project, consisting of representatives of the 14 participating States.
Greece, Austria, Switzerland, Turkey and the UK opted out, but Iceland, a Co-operating State, participated from the beginning.
The scientists working on the Project were not ECMWF staff. They were employed as consultants, and normally there had been a two-year limit on consultants’ contracts. Council therefore had to waive this limit for the staff to be employed on the Project. Janssen returned to the Centre in early 1992, working on a project funded by the Dutch Remote Sensing Board (BCRS).
On 1 July 1992, operational wave forecasting formally began with a 3° global model forecasting to ten days, and a 0.5° model covering the Mediterranean forecasting to five days. Forecasts were made once per day.
Operational verification of the forecast quality was given high priority, and implemented within a year. Research continued with the implementation of ERS-1 altimeter data in the model, and installation of software from MPI to allow regular comparison of the model waves with the ERS-1 data. In August 1993, the sea-ice boundary in the model was improved.
Spain had been active in wave modelling for some time, and there was by now growing collaboration between the scientists at the Centre and those in Spain and Italy to compare the Mediterranean model with buoy data, and also with researchers in France.
In 1993, the UK was invited to join the Project. The UK delegation noted that Europe benefited from the additional work done by the UK on wave Wave prediction 153 forecasting, but suggested that the UK should be entitled to the Project’s software. Greece and Germany suggested that wave forecasting be done not as an Optional Project with the existing 14 States participating, but as a “core activity” to be covered by the Centre’s normal budget. In June 1994, Council agreed that the Project’s software would be available to non-participating States on an exchange basis.
In early 1994, the main technical work to increase the resolution to 1.5° was completed. This model was run in parallel with the operational 3° model for some months to validate it scientifically before stopping the run of the 3° model.
The value of the Centre’s work to ESA was quickly demonstrated: partly in response to feedback from the Centre’s monitoring of the ERS-1 data, ESA changed the software used to calculate significant wave height.
In December 1994 Council considered at length having wave prediction as a core activity. The cost of wave prediction was about £200,000 per year.