Performance of Satellite-Derived Shoreline algorithms on a beach with banquettes of Posidonia oceanica

In recent years, advances in remote sensing sensor resolution and data processing techniques have increased our capability to monitor coastal environments, facilitating a more complete understanding of their dynamics and temporal changes. The development of algorithms able to (semi)automatically detect the shoreline on high spatial and spectral resolution images is particularly noteworthy. However, most of these algorithms have been tested on typical sandy beaches, and the indices used to classify the images often overlook the presence of substantial dark-colored organic material on beaches. This oversight can significantly impact the performance of the algorithms. The presence of casts of dead leaves and rhizomes of Posidonia oceanica (the so-called banquettes) in the swash/intertidal zones, indeed, poses a challenge for the shoreline detection, increasing the inaccuracy of the identified sand-water interface. The present study includes a comparison of the results obtained using different algorithms for Satellite-Derived Shoreline (SDS) extraction on a beach where large quantities of Posidonia oceanica banquettes are located in the swash area. Specifically, the capability of three available algorithms, developed for multispectral (CoastSat; Vos et al., 2019 - https://doi.org/10.1016/j.envsoft.2019.104528, and SAET; Palomar-Vázquez et al., 2023 - https://doi.org/10.3390/rs15123198) and hyperspectral (HyperSho; Souto-Ceccon et al., 2023 - https://doi.org/10.3390/rs15082117) imagery was tested, using Sentinel-2 and PRISMA satellites. In order to validate the SDS extracted with the different algorithms and to carry out an accuracy analysis, three RTK-GNSS surveys were performed in September 2022, May and October 2023. The algorithms were tested on the Arborea beach in the Gulf of Oristano, on the western coast of Sardinia (Italy). Arborea beach is characterised by a shallow sloping seafloor that contributes to the accumulation of banks of Posidonia oceanica. Normally, the presence of Posidonia on the beach plays a crucial role in regulating erosional and sedimentological processes and contributes to beach regeneration. Error analysis revealed that the accuracy obtained is around the spatial resolution of the satellite. Focusing on the multispectral algorithms, we find that SAET obtains better results, considering that the performance of the CoastSat algorithm returns an RMSE of about 25 m, on the three dates studied. The error observed in CoastSat could arise from the misclassification or non-recognition of various classes, particularly the sand class, which is frequently mistaken for buildings or remains unrecognised due to its dark appearance. The results obtained from this study demonstrate the advantages of adopting various algorithms, each employing different image processing modes and spectral indices. This approach allows us to assess the accuracy levels of the analysis and emphasise the diverse limitations inherent in each methodology. Additionally, the analysis conducted, underscores the significance of testing shoreline extraction algorithms on beaches characterised by non-standard or uncommon features. The field data were acquired during fieldwork activities of the OVERSEE project, financed by ASI under contract 2022-14-U.0.