ECMWF

In this final section we briefly describe how wave forecasts are validated and the way we apply wave forecasting at ECMWF .

Figure 7.1 Standard deviation of wave height error of day 1, 3, 5, 7 and day 10 wave forecast. Here, the wave forecast is validated against the wave analysis, and the period is August 1995 until January 2001.

An important task of any weather centre is the verification of the forecast. We routinely verify analysed and forecast wave products against buoy data, we verify forecast wave height against the analysis. An example of forecast validation over the Northern and the Southern Hemisphere is given in Fig. 7.1 . A more detailed discussion of the quality of the ECMWF wave forecast is given by Janssen et al. (1997a, 2000). Furthermore, we have been validating extensively the quality of wave products from Altimeter and SAR from ERS1/2 and will do so from ENVISAT which will be launched in the middle of 2001. An account of this work may be found in Hansen and Günther (1992), Janssen et al. (1997a) and Janssen (2000).

Following the work of Janssen (1982, 1989, 1991) on the feedback of ocean waves on the airflow we have made a dedicated effort towards an integrated forecasting system for our geosphere. Ultimately, it is expected to have a model consisting of the atmosphere and the oceans where the ocean waves are the agent that transfer energy and momentum oss the interface in accordance with the energy balance equation. This role of the ocean waves is illustrated in Figs. 3.3 and 3.4 . On the one hand ocean waves receive momentum and energy from the atmosphere through wind input (hence they control to a large extent the drag of airflow over the oceans), while on the other hand, through wave breaking, the ocean waves transfer energy and momentum to the ocean, thereby feeding the turbulent and large scale motions of the oceans. Ocean waves are in general not in an equilibrium state determined by a balance of the three source functions, because advection and unsteadiness are important as well. Typically, of the amount of energy gained by wind about 90% is lost locally to the ocean by wave breaking, while the remaining 10% is either advected away or is spent in local growth.

Presently, we have taken the first step by coupling the IFS atmospheric model with the WAM model in a two-way interaction mode. This coupled model provides the 10 day weather and wave forecast since the 29th of June 1998. Here, every coupling time step surface winds are passed through the WAVEMDL interface towards the wave model, while the Charnock parameter as determined by the sea state (cf. Eq. (3.11) ) is given to the atmospheric model and is used to estimate the slowing down of the surface winds during the next coupling time step. An overview of results is given by Janssen et al. (2001) who show that the introduction of two-way interaction gives improvements in the prediction of surface winds and waves. As a next step ECMWF is developing a coupled atmosphere, ocean-wave, ocean-circulation model. This coupled model will be used in seasonal forecasting and monthly forecasting in the near future.

Since December 1992 ECMWF is providing estimates of forecast error by running an ensemble prediction system. With the introduction of the coupled wind-wave prediction scheme in June 1998, ensemble wave products became available as well. This new product may provide useful information on the uncertainty in the sea state prediction (Saetra and Bidlot 2002, 2004), which could be applied among other things to ship routes (Hoffschildt et al. 1999, Saetra 2004).

As described in Chapter 3, the coupling between the IFS and WAM was extended to include air density and gustiness effects on the wave growth. This change became operational in April 2002 (CY25R1). Finally, the flexibility of the coupling interface between the two models was further extended to provide WAM with 10m neutral winds instead of the usual 10m winds since WAM cycle 4 was derived in term of friction velocity with the neutral logarithmic profile. This latest change was part of CY28R1, implemented in operations in March 2004.

This concludes our discussion of the software aspects of the ECMWF version of the third generation WAM model. Although the software description only comprises a small part of this document, it should be realised that the greater part of the efforts of the WAM group was devoted to the development of the WAM model code. One can imagine how the strong involvement of a number of WAM people in the wave model development has led to heated debates on aspects of the model design during the yearly WAM meetings. However, all this has paid off. The present cycle 4 version of the WAM model is a beautiful looking fortran code. It combines efficiency with flexibility. It has been installed at over 200 institutes world wide and is used for research and operational applications. Furthermore, at ECMWF we have seen a steady evolution of the WAM software towards a better integration in the ECMWF software system and towards a tight coupling with the atmospheric model.

Part VII: ECMWF Wave-model documentation

CHAPTER 7 Wind-wave interaction at ECMWF