Project Summary

• This research supports UN SDG 6 (Clean Water and Sanitation) aiding policymakers and environmental agencies in coastal water management.
• The project aims to develop an AI-driven water quality monitoring system using Copernicus Marine Service data and Φsat-2 satellite imagery.
• Initially, the previous study relied on in situ measurements from 76 coastal monitoring stations. However, to enhance scalability and generalization, the research now prioritizes Copernicus Marine Service data, allowing for broader regional analysis.
• The project utilizes AI trained on satellite-derived datasets to predict water quality parameters. The models are optimized for edge processing on Myriad 2, ensuring efficient onboard execution.
• The study is ongoing, with efforts focused on optimizing AI generalization, transferability, and hardware adaptation. Future research will further refine model deployment on Myriad 2, making the system more adaptable for global applications.

Development Tools

• PyTorch for the AI code
• GitHub repositories
• Overleaf

Development Outputs

• First analysis on a new type of dataset
• New exploration of AI optimization model
• Analysis on satellite data and optimization on its features

Project Description

Water quality monitoring is crucial for the protection of aquatic ecosystems and the well-being of public health. Effective monitoring helps identify pollution, eutrophication, and harmful algal blooms, all of which can degrade water quality and pose significant risks to both the environment and human populations. The United Nations recognizes the importance of this issue through Sustainable Development Goal (SDG) 6 – "Clean Water and Sanitation." Specifically, a specific target aims to improve water quality by reducing pollution, minimizing the release of hazardous chemicals, and significantly increasing recycling and safe reuse of water by 2030.

Monitoring key water quality parameters such as turbidity, chlorophyll-a (Chla), and suspended solids helps address these global goals by providing data needed to assess the ecological status of water bodies and ensure the sustainability of water resources.

By tracking parameters like turbidity (which indicates sediment or pollutant concentrations), Chla (a key indicator of phytoplankton biomass and nutrient levels), and suspended solids (which affect water clarity and marine life), water quality monitoring becomes essential for managing aquatic environments. These indicators are critical not only for ecosystem health but also for guiding policies in sectors like fisheries, tourism, and coastal development, all of which rely on clean, healthy water bodies. An efficient, real-time water quality monitoring system is necessary to address both local and global environmental challenges, including climate change, marine degradation, and sustainable water management. In a previous work in fact, an in situ data used come from 76 coastal monitoring stations, from ARPA Liguria, where water samples are collected and analyzed to assess physical and chemical parameters of water quality. These monitoring stations are strategically located at different distances from the coast (500 m, 1000 m, and 3000 m) to provide a comprehensive representation of coastal water conditions. Among the most critical parameters monitored, turbidity and pH play a fundamental role in evaluating marine pollution and water quality. Turbidity, a key indicator of suspended particles in the water, is crucial for understanding sediment transport, algal blooms, and contamination events. pH levels provide insight into the chemical balance of the water, affecting aquatic life, metal solubility, and biological processes. These parameters serve as ground truth for training and validating the AI model, allowing it to learn the correlation between spectral satellite data and real water quality measurements.

This study is still ongoing, with further optimization efforts in progress to make the AI model as generalized and adaptable as possible for a wide range of real-world applications. The research is focused on ensuring that the integration of Copernicus Marine Service data, Φsat-2 satellite imagery, and AI-driven techniques leads to a scalable and transferable solution for coastal water monitoring.

The goal is to refine the workflow to handle greater variability in water quality conditions, increasing the robustness and generalization ability of the predictive models. To improve monitoring accuracy and efficiency, the research also incorporates Φsat-2 satellite imagery. Using the Φsat-2 simulator, satellite data are aligned with water quality parameters under study, allowing for a direct comparison between spectral data and water quality indicators.

Another critical aspect of the study is adapting the AI model for onboard processing. The Myriad 2 processor, used for edge computing in space missions, introduces hardware limitations that must be accounted for. The model is being optimized to ensure efficient deployment on Myriad 2, balancing computational efficiency and prediction accuracy within the available processing power. Future research will focus on more results. By continuing to enhance model scalability and adaptability, this approach contributes to UN Sustainable Development Goal 6 (Clean Water and Sanitation), providing a real-time, reliable, and efficient solution for coastal environmental monitoring and decision-making.