About the project

State-of-the Art and Objectives: The main objective of the NEURAL project is to research and to build an efficient and reliable prototype of ocean surface current forecasting system, based on neural network algorithms.

Project Abstract: An operational oceanography forecasting service still does not exist in Croatia, an unacceptable omission for a country which has been investing major resources into its infrastructure in support to its development as a leading tourist destination, built on its natural beauty, climate and geographical position. The aim of this project is to investigate and to develop a hybrid ocean forecasting system for the eastern coastal regions of the Adriatic Sea based on the neural network approach. The project will use surface current fields measured by high-frequency oceanographic radars and mesoscale surface winds simulated by the high-resolution numerical weather prediction (NWP) models. A state-of-the-art atmospheric hydrostatic model Aladin/HR, which is used for the operational NWP at the Croatian national weather service (Meteorological and Hydrological Service, DHMZ, http://meteo.hr), will be used in the project. In addition, a high-resolution version of the non-hydrostatic research WRF-ARW model nested into the ALADIN model and operating in real time in a research mode will provide a complementary input wind dataset for validation and intercomparison. The models’ outputs and the HF radar data will be introduced to the neural network and self-organizing maps algorithms to learn about the wind effects on the ocean and to create characteristic circulation patterns in the Adriatic. Once created through the learning process, the ocean current patterns will be forecasted by using outputs from the meteorological models only. The skill of the forecast will be estimated, and the models will be tuned to reach the best score. The forecast process will be in real-time and automatized, with forecasts published online and thus made available to numerous potential users.

The advantages of the final neural-network forecasting operational system versus classical oceanographic models are numerous: (i) their results are based on real data and therefore highly reliable, (ii) they need several orders of magnitude less computational time and resources, and (iii) forecasts can be made available to final users in very short time. Societal benefits include various applications in disastrous events in the Adriatic. For example, our products can significantly decrease the search and rescue time and area at the sea, allow for better forecasting of oil spill and pollution spreading and can provide invaluable benefits to shipping, fishing, the tourist industry, and the community in general. Once established and validated, the new procedures and the hybrid ocean forecasting system can become a basis for an operational oceanography service.