Model upgrade improves ocean wave forecasts

Jean-Raymond Bidlot


Changes in the ocean wave model used in ECMWF’s Integrated Forecasting System (IFS) are set to improve forecasts of some of the most common ocean wave variables, including significant wave height. The changes are part of IFS Cycle 46r1, which is scheduled to be implemented in June 2019. They include new parametrizations for wind input and deep-water dissipation of waves as previously implemented by Météo-France, based on work by Fabrice Ardhuin (Ifremer, France) and collaborators.

Wave modelling is an important part of numerical weather prediction systems: wind-generated surface waves modulate momentum, heat, and mass fluxes between the atmosphere and the oceans. Ocean waves in turn influence the upper ocean circulation and mixing. As part of ECMWF’s Earth system approach, the wave model component of the IFS is actively coupled to both the atmosphere and the ocean modelling subsystems.

The evolution of the wave field over the deep ocean is modelled by the wave energy balance equation. This equation makes use of different parametrizations to represent how waves are generated by wind and how they are dissipated. It also includes a nonlinear interaction term that allows for wave energy redistributions among the different wave components. Once locally generated, waves can propagate across whole ocean basins as swell. Their interaction with surface currents is also represented in the model.

Better forecasts

The Météo-France changes to the wave model code have been adapted and optimised to run efficiently as part of the coupled Earth system models used in the IFS. It was important to ensure that the new parametrizations return a similar level of feedback from the modelled sea state to the atmosphere model. The impact of modelled ocean waves on the modelled ocean circulation when using the new parametrizations was also assessed. Finally, as part of the upgrade, when the ocean and ocean waves are coupled in the forecast model, surface currents have an impact on the wave propagation.

The main impact of these changes is increased accuracy and realism of the most frequently used ocean wave parameters, such as significant wave height (roughly the average height of the highest one third of waves). The new formulation reduces the overprediction of long period swell energy and the small wave height under-estimation in the storm tracks. Forecasts are generally improved up to 10 days ahead. One effect of the new parametrizations is that the activity level of significant wave height forecasts is increased. The figure shows that the new formulation reduces negative wave height forecast biases and increases the activity level, as measured by the standard deviation of the model simulations, in better agreement with the observation standard deviation (symmetric slope closer to 1). There is also a small reduction in the symmetrically normalised root-mean-square error throughout the 10‑day forecast range.

Improved scores. The charts show reduced bias (left), increased activity as shown by a symmetric slope closer to 1 (middle) and reduced symmetrically normalised root-mean-square error (SNRMSE – right) in forecasts of significant wave height using the new parametrizations. Bias is determined by comparison against in-situ wave height observations and, for the northern hemisphere (NH), southern hemisphere (SH) and the tropics, wave height estimates from space-borne altimeters. In-situ observations are mostly located in near-coastal areas in the northern western hemisphere. The altimeter data were assimilated by the wave model while the in-situ data are completely independent. The results are based on experiments at TCo399 resolution (25 km grid spacing) for the period June to August 2017 and December 2017 to February 2018.

Freak wave modelling update

Output wave parameters available from the IFS comprise a set of parameters to describe the mean sea state as well as the different major wind sea and swell wave components. There is also a set of variables to describe the single largest wave in a record, which, when substantially larger than the mean state, is commonly referred to as a freak wave. Based on work by Peter Janssen (ECMWF Technical Memorandum No. 813) subsequently developed by Peter Janssen and Augustus Janssen, the freak wave parameter calculation has been updated. The main impact is a more realistic increased probability of larger waves in shallow water compared to the old version.