«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
There was one Member State that continued to have strong reservations on scientific and technical as well as formal grounds: the UK. In line with its preference to achieve where possible consensus on issues relating to the Convention, Council decided that wave forecasting would continue as an Optional Project.
In February 1995, Janssen returned to the Centre, now as Head of the Ocean Wave Project. By this time, advantage was being taken of the fact that the wave model predictions depended strongly on the quality of wind forecasts; the wave model was used to validate planned changes to the atmospheric model. Changes to the atmospheric model in April 1995 led to a marked reduction in the errors of the wave forecasts, and therefore in the wind forecasts, in the Southern Hemisphere winter.
In June the resolution of the Mediterranean model was increased to 0.25°.
Software was developed to extract monthly mean wave forecasts, and using data collected by the Portuguese Meteorological Service a study of the interannual variations of the wave field in the North Atlantic began.
On 21 April 1995, ESA launched the ERS-2 satellite. Now work was
begun to cross-compare the data from the two Earth Resource Satellites:
wind, wave, altimeter and more. Software was developed in collaboration with MPI in Hamburg.
Larger computers, scientific advances including improved numerical schemes that used a grid similar to that of the atmospheric model’s “reduced Gaussian grid”, and improved satellite data were now pointing to the desirability of another increase in model resolution. Consequently a feasibility study was made in late 1995 of having a 0.5° global model. The study laid the groundwork for such a model to be introduced in 1996.
154 Chapter 12Also in 1995, Janssen, with Pedro Viterbo from ECMWF, and in collaboration with the scientists at KNMI, began work on a major development: the “coupling” of the wave model with the atmospheric model. Ocean waves play an important role in transferring momentum and heat between air and sea, and vice versa, at the ocean surface. The steeper waves created by locally strong winds increase the drag on the wind by some 50%, thus slowing it.
In essence, in a coupled model, the atmospheric model runs for one time step. The ocean wave model is then run for one time step, using the winds predicted by the atmospheric model. The slowing of the wind from the wave-induced stress is now determined. Thus, this two-way interaction gives quantitative information on the slowing of the airflow. A study of the impact of the modelled rough ocean surface on the predicted development of Atlantic storms showed significant differences between the experimental coupled model and the original uncoupled model; the central pressures of the storms were not so deep in the coupled model.
The performance of the model continued to improve. An assessment of the performance of the wave model during 1995 compared the wave heights and periods with buoy data. Wave heights were underestimated by about 10%; this was associated with the assimilation of ERS-1 data, which were known to underestimate wave height. However the results showed a reduction of 25% in the errors of predicted wave height since 1988 — a real improvement in the quality of wind and wave forecasts. In fact now wave forecasts were being used for quality control of buoys. Prompted by large differences between the observed and predicted waves in the northeast Atlantic, the buoy operator replaced the wave sensors on the buoys. The differences were much reduced. All in all, the scores suggested that Northern Hemisphere wave forecasts were now useful to about five days ahead.
In April 1996, ERS-2 wave height data replaced those from ERS-1.
Analysis and forecast data were being routinely exchanged between the Centre and the UK Met Office, Fleet Numerical Meteorology and Oceanography Center (FNMOC) in Monterey California, Atmospheric Environment Centre Canada and the National Centers for Environmental Prediction (NCEP) Washington, to compare model performance. Operational runs of a 0.5° global model began in parallel with the operational 1.5° model.
In areas of intense storms, the higher resolution could give a better representation of the more intense wave systems. Also, close to the coastline, a dramatic improvement in the quality of wave prediction was noted. The 0.5° global model became the operational model in December 1996. The model achieved the best scores ever achieved to that time in February 1997, with a one-day forecast error close to the known accuracy of the buoy data.
Wave prediction 155 Nevertheless the ERS-2 under-estimation of the wave height in “young wind seas” with steep waves continued to be a problem.
Development of the coupled model was almost completed by end 1997.
A systematic study into the benefits of coupling using 12 forecast cases showed that the hoped-for reduction in model errors was achieved. In the coupled forecasts, near-surface winds were slowed considerably by the rough water surfaces. Even the 500 hPa height scores for the atmospheric model were improved. At the December Council session the UK delegation stated that the UK wished to join the Project if it was added to the core programme of the Centre. All delegations welcomed this.
In June 1998, meeting in Tromsø, Norway, the Council agreed to incorporate the Project in the 1999 budget of the Centre, thus making it a core programme. The Director was now able to give staff contracts to the three scientists working on the Project from 1 January 1999. As a “late joiner” the UK paid £45,000 to contribute to the costs already incurred in setting up the Project.
The coupled version of the model became the operational model, with a resolution of 55 km, in June 1998, after being extensively tested. In the following years, its performance was closely monitored. A large reduction in the errors of the predicted waves was recorded, especially in the tropics and Southern Hemisphere. Forecasts of surface wind forecasts also showed improvements. Comparisons of the forecast errors with those of other centres showed the consistently superior performance of the ECMWF system.
The Mediterranean model had already been extended to cover the Baltic Sea and the Black Sea. It was now extended further, to cover the North Sea, the Norwegian Sea and the North Atlantic north of 10°N. Its resolution was now 28 km. Its forecasts were run to five days, as compared to the ten-day forecasts of the global model.
In 1999, research was under way on use of the ensemble prediction technique, as discussed in Chapter 10, for wave forecasting. In particular an experimental Ensemble Prediction System for ship routeing was developed and tested. Initial results were promising; in half the cases, the lowest cost route was found, compared to one-fifth of the time using the operational system.
One worrying problem remained. There was concern at a systematic under-prediction of wave height when large waves were observed; underprediction of about 1 m was found when waves of about 10–15 m were observed. An extensive study showed that under-estimation of wind speed in the analysis was the cause. This was related to the 5 m height of the anemometers on buoys; the analysis assumed that they were at the standard
156 Chapter 1210 m level. Since the wind is slower at 5 m than at 10 m, the completed analysis ended up with slow winds. A fix was introduced in November 2000, at the same time as a further increase in model resolution to 40 km. Forecast error was immediately reduced; compared to buoy data, both wind and wave forecasts were improved. Nevertheless, persistent low model wind speeds continue to be a problem. However increasing the horizontal resolution helps alleviate this.
In recent years a dedicated effort has lead to an increase in the number and type of observations in the wave analysis scheme. In January 2003 assimilation of low-frequency spectra from the satellite-borne Synthetic Aperture Radar (SAR) started. Following the successful ERS missions launched in 2002 the Environment Satellite (ENVISAT) with ten remote sensors, including a dual frequency radar altimeter, built according to new design specifications, was launched. The altimeter gave significantly better measurements of the wave height. Use of the ENVISAT altimeter data in the analysis scheme from October 2003 improved the quality of the wave forecasts.
Following the considerable progress in wave forecasting made during the past 20 years, what need is there for further development? Let us consider some of the applications in which the wave spectrum plays an important role.
Recently there has been rapid progress in the understanding of the generation of extreme sea states such as freak waves. Prediction of the likelihood of events like these would be of clear benefit to the marine world.
To achieve this, accurate predictions of the detailed “low-frequency” part of the wave spectrum, that is to say the long waves, are required. The windwave forecasting systems developed up to now cannot provide such predictions. More work is needed to investigate the relationship between spectral shape and the occurrence of these extreme states.
Remote sensing applications require knowledge of how the sea surface reflects and emits radiation. This includes instruments like the Advanced TIROS Operational Vertical Sounder (ATOVS), altimeters, scatterometers and Special Sensor Microwave Imager (SSM/I) that are carried on satellites. The reflection and emmision of radiation from the ocean surface depends in a straightforward manner on the range and distribution of wave slopes — the “slope spectrum”. We need to know about the “high-frequency” part of the wave spectrum, the small choppy waves, for this.
We have seen that knowledge of the high-frequency spectrum is important if we want to determine the air-sea momentum exchange. This is the case also for the exchange of carbon dioxide between atmosphere and Wave prediction 157 oceans. In the Centre’s current wave model the parametrization of the highfrequency spectrum is a good first guess. The actual spectral shape is not well understood. Much experimental and theoretical work is needed to obtain a convincing and working model for these high frequencies.
And work is just beginning on the impact of the ocean waves on the largescale ocean circulation.
Exciting times in the field of ocean waves lie ahead.
Data from on high
Satellites are very expensive — but vitally important — sources of data for weather prediction. At the time of writing, the Centre is using data from about 30 instruments on 17 satellites, instruments that are probing and measuring the earth’s atmosphere, and its oceans and land. A single instrument on a weather satellite can provide many thousands of bits of data each second.
Proper exploitation of the vast flow of global data streaming from satellites requires the most powerful computers and the most sophisticated data-handling and analysis software. The Centre has a long history of fruitful relations with the European Organisation for the Exploitation of Meteorological Satellites (EUMETSAT), as well as with the European Space Agency (ESA). There has also been good co-operation with the National Oceanographic and Atmospheric Administration (NOAA) and the National Aeronautics and Space Administration (NASA) in the USA. The Centre has had quick access to many new kinds of satellite data, and has often been the first operational user.
The first European operational geostationary satellite Meteosat was launched in 1977. Initially, ESA was responsible for the satellite, and for its observation processing and so on. Mr R. Mittner, Director of Météorologie Nationale of France, was appointed Chairman of a Meteosat Operational Programme Working Group at an international meeting in Paris in 1981.
Plans were in hand to establish an organisation, later called EUMETSAT, that was to be given the task of carrying responsibility for Meteosat. Serious consideration was given to attaching EUMETSAT to the Centre. There were good practical reasons for this. The Centre’s system would be using the satellite data heavily, and the Centre’s requirements could be expected to be influential in the design of future satellites. The proposal would depend on the willingness of the 17 Member States of ECMWF to modify the Convention establishing the Centre. However, it would have been awkward 158 Data from on high 159 to bring this proposal to fruition. Amending the Convention would take a great deal of time, and the groups of States supporting the two organisations would not necessarily be identical.
Later, an alternative solution, with a co-operation agreement between the Centre and EUMETSAT, was considered, with EUMETSAT having its own legal personality, operating so to speak under the wing of the Centre. The Centre would administer EUMETSAT staff and make its equipment available to EUMETSAT.
In the end, after much discussion, EUMETSAT was established in June 1986 as an independent organisation with headquarters in Darmstadt, Germany.
EUMETSAT inherited the Meteosat satellite programme from ESA in January 1987. Today, EUMETSAT establishes, maintains and exploits European systems of operational meteorological satellites. As well as being responsible for the launch and operation of the satellites, EUMETSAT delivers satellite data for monitoring the climate and for detecting global climate change as well as for operational weather prediction.
In May 1988, the Centre and EUMETSAT concluded a co-operation agreement, formally agreeing to keep each other informed of activities “in which there may be mutual interest”. By this time, the Centre was already using a vast quantity of information from satellites.