Multi-Sensor UAV Validation of Sentinel-2 NDVI as an Operational Tool for Environmental Monitoring and Situational Awareness in a Heterogeneous Riparian System (Fusaro Lake, Italy)

UAV imagery resolves fine-scale vegetation structure that satellite sensors cannot; satellites provide the spatial coverage that UAVs lack. Reconciling observations across these scales requires cross-resolution validation and rigorous uncertainty quantification. This study fused UAV multispectral and thermal data with Sentinel-2 imagery to characterize vegetation patterns around Fusaro Lake, Naples, Italy. The UAV NDVI orthomosaic was resampled to Sentinel-2 resolution for direct comparison, and a pixel-level error map was produced to spatialize residual discrepancies. Results indicated strong agreement (r = 0.885, R² = 0.73), alongside a positive bias (MBE = +0.102) attributable to mixed-pixel dilution at sub-pixel land-cover transitions. An RMSE of 0.296 confirms that random variability from spectral response differences and atmospheric correction residuals dominates over systematic error. Vegetation edges registered 43.4% high-error pixels compared to 28.1% in spectrally homogeneous zones, a 1.4× amplification attributable to mixed-pixel effects. Thermal data identified a 4–5 °C land–water gradient spatially coincident with peak NDVI discrepancies, establishing ecotonal heterogeneity as the primary driver of cross-scale observation uncertainty. This multi-sensor approach quantifies satellite uncertainties and enhances situational awareness, providing a transferable framework for sanitary-environmental engineering in complex ecosystems.