Comparison between shorelines derived from radar and multispectral satellite data

Accurately mapping the shoreline using satellite data plays a key role in coastal monitoring and hazard management. Yet, the effectiveness of Multispectral Satellite Imagery (MSI) can be limited by cloud cover and lighting conditions, in contrast to Synthetic Aperture Radar (SAR) data that can therefore be explored as a valuable alternative. This study compares the performance of Sentinel-1 (S1) SAR images and Sentinel-2 (S2) MSI images for shoreline extraction at two Mediterranean sandy beaches. For the first time, the same semiautomatic workflow was applied to both SAR and MSI datasets. The methodology comprises image preprocessing, including an additional step for SAR speckle noise reduction, unsupervised classification using Otsu thresholding, Gaussian Mixture Models, and K-means clustering, as well as a contour extraction algorithm.The extracted shorelines were validated against high-resolution orthomosaics (25 cm), using Mean Absolute Deviation (MAD) and bias. Results show that S2 generally offers higher spatial accuracy, achieving sub-pixel MAD values. Regarding the bias, S1 consistently showed a negative bias with greater variability across all contexts, whereas S2 exhibited a slight positive and more constant bias. These results suggest a different ability to capture the boundaries present on the beach (land-water, dry sand-wet sand) and a different sensitivity to the morphological context of the shoreline, other than suggesting possible strengths and limitations of each sensor.