GRSG 35th Conference 2024 Abstract

Title: Synthetic Aperture Radar Shoreline processor: methodology and results.

Author: Salvatore Savastano

Organisation: isardSAT

Across the globe, the transformation of coastlines is one of the most immediate consequences of climate change, driving natural processes like erosion, sediment deposition, and rising sea levels. Satellite remote sensing serves as a powerful and precise tool for mapping these dynamic changes and monitoring the ongoing shifts in coastal landscapes over time. Using advanced imaging technology, satellite systems allow for the accurate observation and analysis of coastal evolution, providing valuable data for environmental management and coastal preservation efforts.

Currently, many satellites (both optical and radar) are in operation around the Earth, acquiring large amounts of data. Thanks to programs like Copernicus, high-quality satellite images are freely available, enabling the monitoring of coastal evolution with reasonable revisit times.
In recent years, many tools and software have been developed that utilize optical data, providing the geology community with valuable instruments for coastal studies. These tools are often preferred over radar images, which are considered less reliable for this purpose. However, when the study area is frequently affected by persistent cloud cover or extreme sea conditions, it becomes difficult to monitor coastal evolution over long time periods using only optical images.

This limitation is overcome by Synthetic Aperture Radar (SAR) imagery, which is unaffected by factors like darkness, clouds, or rain. SAR offers the added advantage of potentially providing more frequent updates for shoreline mapping and can also detect features that optical technology may miss.
isardSAT has developed an innovative Shoreline (SL) processor (Savastano et al., https://doi.org/10.3390/jmse12010163) that generates coastal change products using Sentinel-1 (C-band) data, though it can be extended to other SAR missions with different frequencies as well. This processing chain can extract a shoreline from each available SAR image, enabling the creation of long time series that accurately track coastline variations over time.
The SAR Shoreline (SL) processor consists of three stages. First, in the Georeferencing stage, georeferenced images are generated for each available Sentinel-1 dataset using the SNAP toolbox, preparing the data for analysis.

The second stage, SL Extraction, involves four steps: Enhancement, Segmentation, Healing, and Vectorization. This process reduces errors due to the speckle effect, producing a shoreline vector from each SAR image, based on intensity thresholds calculated using the Kittler method. The resulting shoreline is a sequence of points with geographic coordinates, incidence angles, and elevations.

In the third stage, SL Filtering, a new method refines the extracted shorelines to capture coastal changes. This involves:
• Heatmapping to identify SL concentration patterns.
• Polygon Creation along a Reference Line (RL) to divide the scene into segments for detailed analysis.
• Distance Assignment, where SL points are measured from the RL and given a sign based on location (seaward or landward).
• Filtering with Gaussian Mixture Distribution (GMD) to refine SL points by calculating the mean and standard deviation of distances.
• Mean Distance Calculation for each polygon, generating a time series to detect shoreline changes and determine change rates (CR) via linear regression, with positive or negative slopes indicating accretion or erosion.

Incidence angles are also analysed for each acquisition (ascending/descending paths) to assess the impact of satellite geometry and topography on SL positions.
The research findings consistently show that SAR-derived SLs align with positions above the high-water mark across all the studied sites. Depending on the topography, the SLs acquired in ascending and descending tracks show overlapping or displacement, providing the end users with a combination or separate time series and change rates derived from the SLs acquired in different geometric acquisitions.

Several examples of such occurrences will be presented at the conferences, in addition to validating the results from three different end users. This methodology offers the coastal scientist and stakeholder community a unique tool to detect coastal variations in various coastal types (such as flat beaches, cliffs, dunes, etc.), addressing the challenges in SAR-SL detection posed by factors such as coastline orientation, coastal topography, radar signal incidence angle, backshore type, and soil moisture.