Case Study
GenAI for Innovative Coastal Monitoring
We developed an advanced solution based on generative Artificial Intelligence for the coastal area monitoring and the analysis of marine ecosystems with innovative environmental data processing approaches.
Where: Italy
Where: Italy
Challenge
ISPRA, the Italian National Institute for Environmental Protection and Research, needed to enhance its marine monitoring activities with new, more effective models capable of operating in complex coastal areas and managing heterogeneous data and ambiguity in observed phenomena.
This project is part of ISPRA's commitment within the broader MER* (Marine Ecosystem Restoration) program.
Approach
To overcome computational limits and improve modeling resolution, a hybrid approach based on the integration of generative AI and physical modeling was adopted. The custom models realized leverage cutting-edge technologies such as Retrieval-Augmented Generation (RAG), Diffusion-Transformers, and Physics-Informed Neural Networks (PINN). In this solution, AI is used to emulate complex mathematical simulations with fewer computational resources, process satellite data, and refine and validate the predictions of existing models. The project involved scientific collaboration with UNIVPM (Università Politecnica delle Marche).
Digital Ecosystem
Solution
Digital Ecosystem
Using MLOps methodologies, the project led to the development of an advanced AI platform composed of several interconnected modules:
- Hybrid predictive models based on approaches integrating GenAI and phenomenological (“phase-resolver”) models, capable of predicting and simulating coastal scenarios, even in critical areas, with greater speed and availability, reducing computational costs and increasing operability.
- AI-driven spatial smoothing module, designed with Transformer autoencoder architectures, to harmonize simulated data on a national scale, improving consistency between adjacent areas and data from different sources.
- Automatic satellite validation module, based on advanced computer vision techniques to analyze Sentinel-2 images, identify river plumes (e.g., related to E. coli), and automatically validate the results of numerical models.
Results
Consistent national biogeochemical datasets
Reduced computational load for nearshore wave forecasting
Accurate and fast forecasts, analyses, and simulations with ensured operational continuity
Timely and reliable insights to support decision-making
*Marine Ecosystem Restoration
*PNRR MER – National Recovery and Resilience Plan – Mission 2 “Green Revolution and Ecological Transition” – Component 4 “Protection of the Territory and Water Resources” – Investment Line 3.5 “Restoration and Protection of Seabeds and Marine Habitats” – Funded by the European Union – NextGenerationEU – CIG A05B735AD5 – Procedure Number: MEPA order 7884683 – Accounting Code PNRR Z0PNRR03 – CRA Code 04
The project carried out with ISPRA focuses on interventions B31-32-25 within investment area 3.5 “Restoration and Protection of Seabeds and Marine Habitats” of the PNRR MER. It involves the development of an integrated monitoring system based on modeling tools, field data, and satellite data to unify existing information and derive consistent data and insights.
This project falls under Mission 2 “Green Revolution and Ecological Transition” of the PNRR, Component 4 “Protection of the Territory and Water Resources,” which aims to implement actions to make the country more resilient to inevitable climate changes, protect nature and biodiversity, and ensure the safety and efficiency of the water system.
Impacts
To know more
Recommended for you
Explore additional content associated with the topic
