Sentinel-2-Derived Water Surface Mapping and Bathymetric Analysis of Reservoirs for Floating PV Deployment: A Case Study of the Basilicata Region, Southern Italy

Abstract

Floating photovoltaics (FPV) offer a sustainable approach to renewable energy production while enhancing water resource management by reducing land use, improving module efficiency through natural cooling, and limiting water evaporation. However, large-scale deployment requires careful site selection, operational feasibility assessment, and integration with existing water management practices. This study presents a fully cloud-native Sentinel-2-based framework to map recent water surface extents of reservoirs and reconstruct their bathymetric profiles. Developed within the Google Earth Engine (GEE) environment, the approach employs harmonized time series, automated cloud filtering, and median composites combined with adaptive NDWI/MNDWI thresholds to generate high-resolution (10 m) water masks. These dynamic surface extents are integrated with monitored water levels to estimate bathymetry through an annual multitemporal log-ratio band-switching configuration. The method has been tested across the nine main strategic reservoirs of the Basilicata region (Southern Italy). The water-surface extraction approach achieved an Overall Accuracy of 95.33% and a Kappa coefficient of 0.907 in the internal thematic assessment. The bathymetric reconstruction results led to relative volume errors for medium-to-large water bodies ranging between 13.8% and 35.1%, while larger percentage discrepancies in smaller impoundments were driven by scale-dependent normalization effects on low storage volumes. Overall, the proposed method offers a scalable, cost-effective, and Earth Observation (EO)-driven screening tool for the initial site selection, capacity assessment, and planning of FPV systems, particularly in data-scarce regions lacking updated bathymetric surveys.