«Medium-Range Weather Prediction Austin Woods Medium-Range Weather Prediction The European Approach The story of the European Centre for Medium-Range ...»
142 Wave prediction 143 Accurate prediction of ocean wave and swell is required, not only for commercial applications, such as avoiding damage to ships, cargo and crew by routeing vessels away from strong head winds and high waves, but also for the protection of lives and property on land.
In Chapter 11 “Seasonal prediction” we noted Lennart Bengtsson’s farseeing “Ten-year Plan 1985–94”, which he presented to Council in November 1984, but which was not adopted. As we saw in that Chapter, the problem was not in the science, but in the presentation of a glossy brochure, without prior consultation with the Council. In Bengtsson’s Plan was a twopage section on “Wave Prediction”. It is worth quoting this in full.
5.2.5 Wave prediction An important application of the Centre’s medium range forecast is that associated with marine activities. Shipping, fisheries and offshore operations, for example, are all strongly dependent on weather and typically require marine weather forecasts extending to the full limit of the medium range deterministic forecasting period.
Surface or near-surface parameters, available directly from the model or derived from model parameters, are routinely made available to the Member States, including for example, wind at the 10-metre level or temperatures at 2 metres. An integral part of marine weather, however, is the sea state, which is not included in the present operational forecast system of the Centre. Wave forecasts are needed both globally and for the medium range (for example for ship routeing) and with a spatial and temporal resolution of the same scale as is required for the prediction of the synoptic scale weather disturbances which generate the waves. The integrations of the numerical wave model such as that which has been developed at the Max Planck Institute for Meteorology at Hamburg are best carried out on the same spatial grid and with the same time step as the atmospheric model used to predict the surface winds driving the model. In practice this can be achieved effectively only by integrating the wave and atmospheric models in tandem.
There have been significant developments in recent years in ocean wave modelling. Our understanding of the dynamics of surface waves has increased significantly as a result of a series of field programmes and experience with a sequence of first- and second-generation wave models. The European wave modelling community is currently in the process of developing a new, third-generation wave model, which may be expected to yield a further significant improvement in wave forecasting skill.
However, the full potential of these advances can be realised only with
144 Chapter 12access to powerful computing facilities and the use of a sophisticated global atmospheric model to drive the wave field.
Present wave prediction models are based on the integration of the radiative transfer equation for the two-dimensional wave spectrum.
Simple empirical wave prediction tables relating wind or sea parameters such as the significant wave height and period to the wind speed, fetch and duration are still sometimes used in engineering applications; these have long been superseded in routine forecasting operations. The transfer equation describes the propagation of the different wave components of the spectrum, with different frequencies and propagation directions, at their appropriate group velocities, and the changes in the energies of the components produced by wind forcing, dissipation and higher order nonlinear wave-wave interaction. The models predict the full two-dimensional spectrum (typically several hundred components) at each time step and grid point. The number of degrees of freedom carried by a wave model is therefore normally higher than that carried by an atmospheric model. However, the physical processes are simpler to compute (after parameterization of the highly complex multi-dimensional Boltzmann integral representing the nonlinear interactions) and it is therefore estimated that the integration time needed for a third generation spectral wave model is of the order of 10% of the integration time of an atmospheric model of the same spatial resolution.1 The operation of a global wave model by the Centre would also be timely in view of the advent of the first European remote sensing Earth Resource Satellite ERS-1. The high rate of surface wind and wave data to be produced by ERS-1 can be effectively exploited in an operational or quasi-operational model only at a large forecasting centre such as ECMWF. However, the Centre will need to cooperate with universities, research centres and weather services in order to develop the necessary models and data assimilation techniques. An active participation in the development of good wave forecasts is also in the interest of the Centre and of the World Meteorological Community for winning the cooperation of ship operators to obtain an improved data coverage over the oceans.
1 Twenty years later, the integration time needed for the third generation spectral wave model running operationally at the Centre was in fact 9% of the integration time of the operational atmospheric model, which had similar spatial resolution.
Klaus Hasslemann wrote this part of the text — what excellent forecasting skill was being demonstrated!
Wave prediction 145 It is therefore proposed that the operations of the Centre be extended to include global, medium range forecasts of the two-dimensional surface wave spectrum as an additional ocean component of the global weather forecast products of the Centre.
In addition to a global wave model, there will also be a need for limited area, high-resolution wave models for the Eastern Atlantic, North Sea, Mediterranean and Baltic, which may require input from the global wave model to provide boundary condition. These can be developed and operated by the national weather services in a similar way to that in which limited area high-resolution atmospheric models are run by the Member States.
The Council discussion in November 1984 on Bengtsson’s Plan showed that some were in favour of wave prediction: Italy “expressed great pleasure that wave prediction was included; this should be done also for the Mediterranean and the North Sea”. Denmark noted that “wave prediction is of interest to the Member States engaged in ship routeing”. For the UK, “testing of wave models as part of the Centre’s research” could be carried out. Some were neutral: for Ireland and France, wave prediction should be considered a special application, and “wave prediction is not important for Austria, but it is not opposed”. The question of resources was raised. None spoke against wave prediction being carried out at the Centre.
At its next session in May 1985, Council considered a document on the “ECMWF Long-term Strategy”. This document still contained a significant section on “Operational medium-range forecasting of ocean waves”. There was now some unease being expressed: Germany “was not ready to agree to operational wave forecasting, it did not believe that this corresponded to the provisions of the Convention”. Council asked its Scientific and Technical Advisory Committees to examine the strategy.
In reporting to Council in November 1985, the Scientific Advisory Committee “considered that, in terms of European science and the European remote sensing programme in particular, it would be very desirable for the Centre to become involved with the data to come from Earth Resource Satellite ERS-1 and to provide a central focus for the Ocean Wave Modelling Programme”. [The ERS-1 satellite, launched in July 1991, produced a large volume of surface wind and wave data, which required powerful computing resources with sophisticated software for its exploitation.] Neither Germany nor France could agree with this opinion of the Committee.
In the “ECMWF Long-term Strategy 1987–1996” adopted unanimously by Council in May 1986, the only mention of wave prediction was under “Operational Aspects”: “The forecasting scheme and the range of
146 Chapter 12dissemination products will be enhanced to include... should Council so decide, forecasts... of ocean waves”. As with seasonal prediction, however, Bengtsson was determined that the Centre would not stand aside from developments in this important area, in spite of the less than warm reception of the proposal by Council. And at the least, the ice had been broken, and some Member States had expressed support for the Centre’s involvement in prediction of ocean waves.
It is of course the wind that makes the waves — the so-called wind-driven sea. The transfer of momentum downwards from the rapidly moving air forces the formation of waves, which are the visible manifestation of this downward transfer of horizontal momentum from the air to the water. Swell is different — this is the result of distant storms. Swell from the North Atlantic beating against the west coast of Ireland may very well have been caused by hurricanes some days ago in the Caribbean.
Beginning in the late 1950s, numerical wave models were being formulated in terms of the so-called energy balance equation for the two-dimensional wave spectrum. These “first-generation” models developed through the 1960s assumed that the waves suddenly stopped growing when they reached some prescribed empirical saturation level. They greatly underestimated the effect of interactions between waves. In mathematical terms, these interactions are non-linear, and not easy to treat or to model.
Klaus Hasselmann, who was Director of the Max-Planck-Institut für Meteorologie (MPI), Hamburg from February 1975 until November 1999, had developed the theory of the general structure of the source function of the deep-water transport equation in 1960. However none of the wave models developed to the mid-1980s were able to compute the wave spectrum from first principles. Klaus Hasselmann and his wife Susanne had begun significant research at the Institute to parameterize better these non-linear interactions. They developed the theory of non-linear transfers of energy and momentum between waves in the 1960s — a theory that could be introduced into the numerical models only in the 1980s.
Through the 1970s measurements of wind effects on waves led to the development of second-generation models, which attempted to model better the wave-wave interactions. Although an improvement, the models were still unable to handle the complex seas generated for example by hurricanes or intense small cyclones — the very situations for which wave forecasts were most required. Also they had difficulty in treating the transition from the sea waves, which are locally generated, to the swell.
A study in 1984 compared the success of first- and second-generation models. Severe weaknesses were identified in the models.
Wave prediction 147 Knowing that computing power would continue to increase quickly, and knowing from contacts with the European Space Agency (ESA) the nature of the global sets of observational data of wind and waves that would become available in the coming years, Klaus and Susanne Hasselmann decided to speed up the pace of research by increasing collaboration with other groups. They contacted the Royal Netherlands Meteorological Institute (KNMI).
Peter Janssen, who was later to become Head of Ocean Waves Section at the Centre, had joined KNMI in 1979 from the University of Eindhoven, where he had completed his doctorate on plasma physics. At KNMI, a simple numerical wave prediction model had already been introduced to complement the manual techniques for wave and swell prediction, based on wind forecasts, which had been developed at KNMI during the 1950s and 1960s. The second-generation model was based on sound if simple physical principles.
While there was some representation of wind, sea state and swell in the models of the time, there was full awareness that much work was required to improve the parameterisation. Janssen’s work in plasma physics allowed him quickly to involve himself in wave modelling. At the beginning of October 1979, he started working on the theory of ocean waves, including the interaction of wind and waves. By 15 December 1979, Janssen had developed the theory of the two-way interaction between the wind and waves. This was far ahead of model development at the time — in fact it was not until June 1998 that the theory was satisfactorily introduced into operational wave prediction.
The establishment of the international WAM — acronym for “Wave Modelling” — Group in 1984 stimulated European research into numerical wave prediction, in particular by collaborating on the development of a thirdgeneration model. The Group included Klaus and Susanne Hasselmann — whose work laid the foundations for the model — Gerbrand Komen, Luigi Cavaleri from Italy, and Peter Janssen. The WAM Group had grown by 1990 to include about 40 scientists, mainly but not exclusively European.
The necessary good, efficient algorithm for computing non-linear transfers, and a reliable parameterization of the dissipation of energy, had been developed. The third-generation model would predict not merely wave height at a point but the full spectrum of waves, without a separation between wind-driven sea and swell.
Bengtsson had known Klaus Hasselmann for some time, and was familiar with his work — he had in fact done some of his work at the Centre as a scientist working on a special project. Bengtsson invited a group to meet
148 Chapter 12at the Centre on 12 December 1985, six months before Council would adopt the Strategy, to discuss the elements that would be required for an operational system of wave prediction. The group consisted of Klaus and Susanne Hasselmann, Janssen, Komen, Cavaleri, and Dorethea von Berg from MPI.
There was a little understandable resistance from some — but not all — senior staff in the Centre’s Research Department. Some were not in favour of the Centre being diverted from its main task of medium-range prediction of the atmosphere. Bengtsson was convinced that he was making a correct and important decision, and he pressed ahead.