The SNOWED Dataset and Its Application to Po River Monitoring Through Satellite Images

This paper presents an approach for the creation of a water/land segmentation dataset using a combination of satellite imagery and certified shoreline measurements. The dataset is created by selecting Sentinel-2 Level 1C satellite images that align with certified shoreline measurements obtained from the NOAA Continually Updated Shoreline Product (CUSP) program. The effectiveness of the proposed dataset is demonstrated through its application in a water monitoring scenario, specifically in assessing water quantity fluctuations in a region of Po river in Italy. Given the very good results obtained in this application, the dataset proves to be effective in training neural networks for water/land segmentation tasks. This preliminary research contributes to address the increasing environmental challenges, particularly in hydrogeologically vulnerable areas, by providing a reliable dataset for accurate shoreline segmentation and long-term monitoring applications.