Author(s): Pedro Magana; Pedro Otinar; Marcus Santana; Manuel Cobos; Asuncion Baquerizo

Linked Author(s): Pedro Magaña, Pedro Otiñar, Manuel Cobos

Keywords: Satellite Derived Shoreline; Remote Sensing; Google Earth Engine; Sub-pixel co-registration

Abstract: Coastal zones are areas of special interest for their natural and economic values. However, the modeling of their morphodynamics is rather complex due to the great variety of physical processes involved, that interact at different time and spatial scales. Human activities and future expected changes in mean sea levels and wave climate severity and, in particular, in the incoming direction may also affect sedimentary processes, exacerbating morphological changes. In the last decades, the increasing computational capability has favored the development of tools capable to model many of these processes. However, field data is needed for their implementation with appropriate boundary conditions and posterior calibration and validation. In this framework, techniques that allow to observe the evolution of the shoreline in an approachable and global manner and without the realization of expensive field campaigns is very appealing. In this respect, the use of remote sensing techniques known as Satellite Derived Shoreline (SDS) has been expanded significantly in the last few years. In addition, the emergence of the Google Earth Engine tool has allowed to make massive analysis with spatial and temporal resolutions that, even depending on the satellite used, go further away from the available ones only a few years earlier. This has facilitated the use of large collections of satellite images that in the recent past would have required enormous computational resources. Several toolkits have been developed by using Google Earth Engine as a base within the SDS domain. Although these tools give satisfactory results in most sandy beaches, depending on the method applied this may not be the case when the beach has unconventional characteristics or when the beach selected is too large and comprises different features. Also, due to the offsets between the different satellite passes, images are not accurately geo-referenced, and shorelines can be detected with errors that can be of several meters. Thus, geospatial co-registration is a mandatory prerequisite when dealing with remote sensing data within SDS field and is not generally integrated within these toolkits. In this work we present the case study of the shoreline detection from images of a mixed sand and gravel beach. The method of shoreline detection consists fundamentally of the following steps: (a) co-registration of satellite images; (b) calculation of water index; (c) computation of edges using a Canny edge detector; (d) setting of an adaptative threshold value using Otsu method; (e) estimation of threshold index values; and finally (f) extraction of shoreline. The validation of the proposed methodology has been carried out by comparing results with shorelines obtained at field campaigns, as well as with shorelines manually digitized by experts. The obtained accuracy is fully satisfactory for many applications.

DOI: https://doi.org/10.3850/IAHR-39WC2521711920221628

Year: 2022