2.3.3. Shoreline Classification

The coastline was classified into the eight categories that were discernible on aerial images: sandy beaches, mangroves, trees (tall vegetation on a muddy substrate), grass (or reeds), natural rocky shores, road embankments (and seawalls, necessary for urbanisation purposes), private embankments (to consolidate lands), and quays (cf. Figures 3 and 4; "method similar to [3]). The coastal classification for 2019 was performed first using the ground-based survey of the different coastal typologies to identify the bird-eye aspect of each category (Figure 3) and obtain a baseline from which to work backward in time and classify previous images through photointerpretation. The classification was performed by splitting the shoreline (Editor tool on ArcGIS). The length of each segment was calculated, and the percentage of the shoreline belonging to each category was extracted (cf. Figure 3 for examples of each category on aerial images).

**Figure 4.** Examples of aerial view of coastal classes. (**A**) road embankment (red) and private embankment (pink); (**B**) quay; (**C**) sandy beach; (**D**) mangrove (dark green), grass (clear green), private embankment (pink); (**E**) trees.

#### 2.3.4. Shoreline Position

The Digital Shoreline Analysis System (DSAS [19]) plug-in on ArcGIS was used to study erosion and accretion phenomena along the shoreline. The baseline was placed 25 m inland with respect to the innermost shoreline at any given position. Transects were cast every 5 m, with a maximum search distance from the baseline of 250 m and a smoothing distance of 200 m. The shorelines were set to have a default data uncertainty of 5 m (linked to the uncertainty of the determination of dark reef flat features vs. coastal bushes). Two parameters were calculated between each date and over the whole period (1955 to 2019) with the DSAS plug-in on ArcGIS, with a confidence interval of 95.5 (2σ): the Net Shoreline Movement (NSM)—total distance between the earliest and most recent shorelines for each transect, in meters—and the End Point Rate (EPR)—NSM divided by the number of years between the earliest and the most recent shorelines, in meters per year.

These statistics were studied in relation to the categories obtained beforehand to identify which categories were responsible for the most change in shoreline position around the island since the mid-1950s. For selected parts of the island, the average EPR was calculated for each 10◦ change in azimuth (i.e., average from 0 to 10◦, 10 to 20◦) and time interval. To identify temporal changes in sedimentary dynamics, linear correlation coefficients were calculated by comparing the average EPR at each 10◦ azimuth change between each pair of year intervals. For instance, a hypothetical net erosion of −1 m between 0 and 10◦ and accretion of 1 m between 10◦ and 20◦ for a given time interval, and net erosion of −0.5 m at 0–10◦ and accretion of 0.5m at 10–20◦ for another time interval would yield a linear correlation coefficient r of 1 between those two time intervals.
