Validating Satellite-Derived Shorelines with UAV-SfM: A Multi-Sensor Accuracy Study in NW Türkiye

Accurate shoreline monitoring is critical for coastal zone management in tectonically active regions. However, in micro-tidal environments like the coast of NW Türkiye, the positional accuracy of freely available satellite data remains poorly quantified. The key uncertainty is whether observed errors stem from sensor limitations or short-term environmental noise, hindering reliable sensor selection for operational use.

We conducted a rigorous benchmark analysis using centimeter-precision Unmanned Aerial Vehicle (UAV) Structure-from-Motion (SfM) data as the definitive reference. Shorelines were extracted from Sentinel-2 MSI, Landsat 8 OLI, Sentinel-1 SAR, and Copernicus DEM GLO-30 using Google Earth Engine (GEE) and compared using transect-based metrics (RMSE, MAE) at two contrasting sites: the complex rocky coast of Güneyli (Gulf of Saros) and the urbanized sandy coast of Altınova (Sea of Marmara).

Our analysis reveals a critical finding: all satellite sensors exhibited RMSE values clustered between 8–15 m. This range aligns with the magnitude of expected hydrodynamic noise (e.g., wave run-up) in such micro-tidal settings, suggesting an environmental constraint on practical accuracy. Site-specific patterns emerged: In urbanized Altınova, Sentinel-1 SAR achieved the lowest RMSE (8.79 m), proving robust against spectral confusion from anthropogenic structures. In natural Güneyli, Sentinel-2 demonstrated superior geometric fidelity with the lowest MAE (7.45 m), effectively capturing complex morphology obscured by radar speckle. The Copernicus DEM was consistently unsuitable for precise delineation, with errors exceeding 15 m due to vertical uncertainties amplified by coastal topography.

This study establishes that in micro-tidal, active margins, environmental variability sets a practical accuracy floor (~10 m) for operational satellite monitoring. Therefore, we propose a tiered framework: (1) Sensor choice must be context-driven (SAR for modified coasts, optical for natural settings), (2) Detected changes near this threshold require caution, and (3) High-resolution UAV data is indispensable for validation and for resolving sub-satellite-scale geomorphic features (e.g., submerged beachrock geometry) critical for hazard assessment. Our work thereby provides a calibrated benchmark for coastal scientists and managers.

This work was supported by the Scientific and Technological Research Council of Türkiye (TÜBİTAK) under the Grant Number 119Y567.