All the services cover the sea areas only up to 10 km of distance from the coast.
The current scenario for sea state analysis and forecast is based on physical models and the two main operational services over the Mediterranean Sea are run by the Copernicus Marine Environment Monitoring Service (CMEMS) and the Italian National Agency for New Technologies, Energy and Sustainable Economic Development (ENEA).
- Wave Historical Analysis;
- Wave-Circulation Forecast based on wave and circulation physical models:
- Sea wave daily forecast over the next day and sea wave daily forecast over the next 5 days;
- Sea circulation forecast daily service over the next day;
- Sea circulation daily forecast over the next 5 days
The expected DAMAS scenario will provide:
- Much faster data processing and generation of results,
- Improved accuracy by ensembling various results (from different initialization parameters),
- Improved geo-spatial resolution (1x1km) and coastal coverage (up to 1 km from the coast) also by using Earth Observation (EO) satellite data (e.g., Sentinel-1 and Sentinel-3) and meteorological data (Mistral from CINECA),
- Wave-Circulation Nowcasting over the next 12 hours, with 30 minutes temporal resolution and 1x1Km geo-spatial resolution,
- Easy evolution of the model and tuning to different geographic areas.
The main result of the experiment is a fast and accurate sea-state forecast for off-shore industry operations security and Cities/Regions coastal area planning and monitoring. An AI based model for sea-state analysis and forecast has been implemented, providing results aligned with the physical model currently used at national level.
Main Results of the 1st stage of the programme: The DAMAS model requires much less time for processing the sea-state forecast and its performance is going to be improved with the assimilation of Copernicus EO satellite data.
The objective of DAMAS is the development of an AI based model for sea state analysis and forecast with improved geospatial and temporal resolution as well as extended close to coast coverage, with respect to currently available services.
The cooperation between GMATICs and the two selected DIHs will provide:
- Better debugging capabilities due to a deeper understanding of the model;
- Improved reliability and robustness of the models;
- Better capability of detecting biases in the training dataset;
The DAMAS AI models will be implemented on a cloud/HPC infrastructure to speed-up the runtime during both training and application phases. The services from the selected DIHs will enable the definition and development of a container-orchestration framework for automating the application deployment, scaling, and management. These steps will make it possible to have faster AI model applications, allowing to reduce the forecast horizons from the current hourly steps to 30 minutes steps (for nowcasting), and to change the initial conditions of the AI model for evaluating several alternative scenarios by running more solutions in parallel and by ensembling their outputs.
The DAMAS experiment will contribute to the Green Deal data space, by supporting offshore renewable energy production and improving efficiency and safety of the shipping sector. As a matter of fact, offshore renewable energy has the greatest potential as a contributor to the achievement of the decarbonization of the power sector, which is one of the key objectives of the European Green Deal. Both offshore renewable energy and shipping challenges are linked to business opportunity with the related industry sectors, exploiting the new DAMAS services based on Data Analytics and Big Data. DAMAS will also contribute to assess the impact of Climate Change on sea state evolution trends, will support coastal management (with sea state information close to the coast) and will support marine oil spill countering (through improved pollution drift prediction).
- The Key Exploitable Result of this experiment are:
- AI models engineered on CINECA cloud/HPC infrastructure as a core element of the DAMAS services and for their future geographic extension and evolution. KER (1) also includes the developed software framework, and the experience acquired in Data Analytics and Big Data that will be exploited also in other monitoring services being currently developed by GMATICS (air pollution analysis/forecast and land-use/cover change detection).
- Historical analysis chain for sea waves, with 1kmx1km geo-spatial resolution and hourly temporal resolution hindcast;
- Wave and circulation forecast chains for daily forecasts over the next 5 days, with hourly temporal resolution and 1kmx1km geo-spatial resolution, up to 1 km from the coast;
- Wave and circulation nowcast chains for nowcasting over the next 12 hours, with 30 minutes temporal resolution and 1kmx1km geo-spatial resolution, up to 1 km from the coast
KPI1: AI model runtime. The runtime for producing the results over the whole Mediterranean Sea will be improved to be collected in less than 30mins for each run. KPI2: The geo-spatial resolution of the Wave and Circulation AI models will be improved to became 1km x 1km. KPI3: The Root Mean Squared Difference between the Wave and Circulation AI and the ENEA models will be improved by a 10% up to 1km from the coast. KPI4: The geo-spatial correlation coefficient between the wave and circulation AI and ENEA models will be improved by more than 90% up to 1km from the coast.