Enhancing coastal wave forecasts by improving forcings with deep learning – The Copernicus Marine Service Evolution KAILANI project

Ocean wave forecasting is highly demanded by end-users. There is a pressing need for reliable forecasts, to be applied in emergency services, harbour logistics, search-and-rescue operations, renewable energy or pollutant transport. In addition to this wide variety of uses, the coastal zone represents a modelling challenge due to the joint superposition of physical processes that make it a highly dynamic environment (including wind, waves, circulation and air-sea-land interactions).

In the observational side, remote sensing products such as those derived from Satellite Synthetic Aperture Radar (SAR, e.g. from the Sentinel missions) and High Frequency Radar (HFR, e.g. available at the Copernicus Marine Service - In Situ TAC) offers vast quantities of high-resolution spatio-temporal fields. However, their applicability within the operational ocean forecasts services is not straightforward.

The Copernicus Marine Service Evolution KAILANI project (2022 - 2024) aims to enhance the Copernicus Marine regional wave forecasts by improving the forcings required by spectral wave models: i.e. wind forcings and surface current fields. This enhancement comes from blending remote sensing observations with wind and surface currents forecasts. Artificial Intelligence Neural Networks (ANNs) has been proposed as the basis for this blending, as they allow to extract complex spatio-temporal features from remote-sensing data.

The impact on bias and error reduction would be assessed by testing these blended fields under a preoperational environment. The Iberia-Biscay-Ireland (IBI) area has been selected for this Proof of Concept, due to the good coverage of HFR along its coastline. Selection of pilot study sites in areas at the Cantabrian Sea (macrotidal), the Canary Islands (mesotidal), the NW Mediterranean (microtidal), and in the hot spot that is the Gibraltar Strait will ensure that KAILANI applicability ranges different environments.

This methodology focuses on the post-processing of the forcings. Then, it could be a complement for Data Assimilation algorithms. If successful, the proposed KAILANI methodology could be exportable to different Copernicus Marine Monitoring and Forecasting Centers (MFCs); without significant changes in their numerical codes and operation chain. Finally, the expected enhancement of the delivered coastal wave spectra and their integrated parameters (i.e. wave height, period and direction) will be key to foster downstream nearshore applications.