Quantification of Nearshore Sandbar Seasonal Evolution Based on Drone Pseudo-Bathymetry Time-Lapse Data

Abstract

Nearshore sandbars are dynamic features that characterize shallow morphobathymetry and vary over a wide range of geometries and temporal lifespans. Nearshore sandbars influence beach geometry by altering the energy of incoming waves; thus, monitoring the evolution of sandbars is a fundamental approach in effective coastal planning. Due to several natural and technical limitations related to shallow seafloor mapping, there is a significant gap in the availability of high-resolution, shallow bathymetric data for monitoring the dynamic behaviour of nearshore sandbars effectively. This study introduces a novel image-processing technique that produces time series of pseudo-bathymetric data by utilizing multi-temporal (monthly) drone imagery, and it provides an assessment of local morphodynamics at a sandy beach in the southeast Mediterranean. The technique is called standardized-ratio bathymetric index (SRBI), and it transforms natural-colour drone imagery to pseudo-bathymetric data by applying an empirical formula used for satellite-derived bathymetry. This technique correlates well with laser altimetry depth measurements; however, it does not require in situ depth data for implementation. The resulting pseudo-bathymetric data allows for extracting cross-shore profiles and delineating the sandbar crest with 4 m horizontal accuracy. Stacking of temporal profiles allowed for the quantification of the sandbar’s crest and trough changes at different alongshore sections. The main findings suggest that the nearshore crescentic sandbar at Episkopi Beach (north Crete) shows strong seasonality regarding net offshore migration that is promoted by enhanced wave action during winter months. In addition, the crescentic sandbar is susceptible to morphology arrestment during prolonged weeks of low wave action. The average migration rate during winter is 10 m.month⁻¹, with some sections exhibiting a maximum of 60 m.month⁻¹. This study (a) offers a novel remote-sensing approach, suitable for nearshore seafloor monitoring with low computational complexity, (b) reveals sandbar geometry and temporal change in superior detail compared to other observational methods, and (c) advances knowledge about nearshore sandbar monitoring in the Mediterranean region.

1. Introduction

1.1. Nearshore Sandbar

The nearshore sandbar is a prominent morpho-bathymetric feature of shallow sandy areas, resulting from hydrodynamic activity on the seafloor [1]. Typically, sandbars are divided into straight and crescentic (rhythmic) ones according to their geometric pattern. However, it has been generally observed that a sandbar may alternate between straight and crescentic forms depending on the changing characteristics of hydrodynamic activity [2,3]. Sandbars are further characterized by strong spatio-temporal variability, which spans from daily to inter-annual and even decadal time scales [2,4]. Sandbars switch from straight to crescentic form when wave action decreases and incidence angles of waves are normal [3,5]. Such hydrodynamic conditions favour the development of rip currents that influence cross-shore sediment transport [3,6]. Crescentic sandbars are mainly found in non-tidal or micro-tidal beaches and usually form a system of two or more sandbars arranged parallel to the shore [6]. The sizes of the sandbars vary significantly as well, with bar amplitudes being reported in the scale of a few meters up to a few hundreds of meters wide [2,5,7]. Sandbars influence beach geometry significantly and are considered a major focus subject in coastal research [3,8]. Particularly, sandbars control the erosion/deposition patterns occurring both nearshore and onshore. Furthermore, the rip-channels that develop along with crescentic sandbars play an important role in the seaward dispersion of suspended matter and other biogeochemical parameters (i.e., larvae, pollutants) and pose a threat to swimmers by intensifying current activity [9]. During the last few decades, there has been a significant number of studies focusing on nearshore sandbars, either from a modelling and/or an observational perspective [2,3,8,10]. Cross-shore sandbar migration has been validated both by coastal morphodynamic models and field observations [2,4,9,11,12]. Although there is a wealth of modelling studies on sandbar behaviour, it is difficult to validate model outputs using appropriate in situ data such as nearshore bathymetry [6]. Traditional hydro-acoustic mapping of nearshore areas is time-consuming, costly, and has a limited spatial coverage over the shallow seafloor. Therefore, optical imaging techniques have been adopted for extracting indirect information about the nearshore seafloor.

1.2. Coastal Seafloor Monitoring

In the last decade, more bathymetric mapping studies were based on airborne hydro-optical data [13,14]. Bathymetric light detection and ranging (LiDAR) is considered the most recognized optical technique for shallow seafloor mapping. However, LiDAR sensors’ cost and the logistic effort of acquiring bathymetric data are often limiting factors [15]. Thus, observational studies on nearshore seafloor mainly rely on low-cost equipment, including video monitoring tools such as the Argus camera network and surfcams, as well as satellite imagery [16,17,18,19]. Video monitoring coupled with long-exposure photography techniques (i.e., timex) has been employed extensively in Argus, as well as surfcam image analyses for deriving the spatiotemporal variability of sandbar geometry (2D) [19,20,21,22]. Though video monitoring has proved an effective approach in sandbar monitoring studies, its application may become problematic as it is influenced by image noise and low spatial resolution due to sensor size, viewing geometry, and recording style [21]. Satellite imagery has proved a valuable dataset in nearshore morpho-bathymetric monitoring by applying satellite-derived bathymetry (SDB) algorithms on temporal imagery [15,23,24,25]. Apart from SDB, recent studies developed colour-based image-enhancement techniques for deriving the position of the bar crest from satellite images [4,17,26]. Despite its merits, the applicability of satellite imagery for nearshore sandbar monitoring may be affected by cloud cover, the spatial/temporal resolution of satellite sensors, and the increased cost of commercial datasets [27,28]. A number of recent studies have focused on shallow bathymetric mapping with drone sensors. The transferability of SDB algorithms in drone imagery offers the possibility for landscape-scale, very-high-resolution mapping of the nearshore seafloor, showing good results with up to 40 cm vertical accuracy [27,28,29,30,31,32,33]. Apart from bathymetric studies, drone-based multi-temporal imaging has been deployed for monitoring coastal morphodynamics onshore [34,35,36,37]. In recent years, there have been a few studies utilising UAS topo-bathymetric LiDAR sensors for high-resolution bathymetric mapping [38,39,40]. Furthermore, other studies have applied structure-from-motion (SfM) techniques on drone imagery for bathymetry retrieval [41,42]. SfM produces acceptable results only in cases where the seafloor surface is texture-rich so that the SfM algorithm can identify corresponding matching points on the images. Thus, the SfM method suits bathymetry extraction over rocky or rugged seafloor areas. However, significant morpho-bathymetric changes occur mainly in smooth (low-texture) seafloor, comprising soft sediment types, which are more susceptible to hydrodynamic activity than rocky areas. Accordingly, Ref. [43] applied a multispectral data-fusion approach for short-term monitoring of nearshore bathymetric change on a beach in the Chania area (Crete, Greece).

This study aims to examine the morphodynamic behaviour of a crescentic sandbar system located in northern Crete by analysing a unique set of drone-based, pseudo-bathymetric data collected on a quasi-monthly interval during an one-year monitoring survey. Drone data consist of standard RGB orthomosaics from which a pseudo-bathymetric surface is constructed by applying a simple, empirical transformation that does not require in situ bathymetric measurements. This transformation is based on the spectral band ratio from [44,45]; it is linearly correlated with actual bathymetry and allows for a comparison between data obtained at different dates. This approach resulted in a set of temporal pseudo-bathymetry models (PBMs) of the nearshore seafloor, from which several temporal bathymetric profiles are extracted and analysed. The novelty of this study is that it allows for effective monitoring of the nearshore seafloor in superior detail (25 cm spatial resolution) without the need for in situ measurements, thus revealing fine-scale characteristics of sandbar evolution. Taking into account the latest literature about sandbar observations in the Mediterranean Sea [46,47], this is the first study assessing the temporal variability of nearshore sandbars in the SE Mediterranean region using remote sensing technology.

1.3. Study Area

The study area is located at Episkopi Beach (Figure 1) on the northern coastline of Crete (Greece, SE Mediterranean). The wider coastline stretches for several kilometres, consists of consecutive beaches, and is interrupted by three small rivers which provide sediments during winter rainfalls [48]. Episkopi Beach comprises mainly sandy sediments, and it is completely exposed to northern winds and waves that typically prevail [48]. In general, the waters in the area show good visibility when the sea state is calm as a result of the oligotrophic character of the South Aegean, and the tidal range in the region is approximately 10–20 cm at maximum. According to [48], the wider area of Episkopi beach comprises a low-relief dune system in front of a Holocene-age alluvial platform. Particularly, the Episkopi beach is bounded by two minor rivers, one in the west and one on the east side, which also provide material for maintaining the beach in the long-term. According to [20], the role of tides in altering beach geometry at microtidal environments (such as at the Episkopi beach) is negligible since the wave-breaking pattern is not regulated by tide. The study area belongs to Almyros Bay, which is mainly influenced by winds of northern directions (NW to NE). However, the presence of Cape Drapano in the west of the bay restricts the distance over which NW winds can generate waves, resulting in relatively small offshore waves, which typically remain under 1 m in height and have short periods. Although less frequent, north winds have a greater fetch, leading to waves that can reach considerable heights and longer periods, creating more dynamic sea conditions [48]. Meanwhile, the NE winds, despite having the greatest fetch, occur infrequently. As a result, while these waves can reach moderate significant heights (<2 m) and longer wave periods, their impact remains limited due to their rarity in the area [49].

2. Methods

2.1. Acquisition and Processing of Drone Imagery

Drone imagery was acquired quasi-monthly using a DJI Mavic Air 2 equipped with a 1/2”CMOS sensor along a 2 km segment of the Episkopi Beach (

Table 1. Summary of the drone survey metadata.

Drone Survey Date		Flight Altitude (m.asl)	Number of Images	Sun Elevation (Degrees)
1.	03/11/2022	254	36	19
2.	07/12/2022	287	45	21
3.	04/01/2023	300	38	24
4.	17/02/2023	305	44	34
5.	14/04/2023	327	42	18
6.	09/05/2023	319	42	29
7.	20/06/2023	330	38	32
8.	13/07/2023	356	47	29
9.	16/08/2023	351	34	19
10.	15/09/2023	360	41	23
11.	23/10/2023	353	43	18
12.	08/11/2023	354	46	20

, Figure 1). Image capture occurred manually, hovering above each waypoint with the camera sensor facing the nadir. The flight altitude was 254–360 m above sea level for capturing a large area of the surf zone and part of the onshore area, thus assisting in producing large-scale orthomosaics from raw drone imagery. The camera sensor records in three broad-band channels (Red, Green, and Blue; JPEG format) that are similar to the band wavelengths of the DJI Phantom 4 Pro RGB camera: Red = 590 ± 25 nm, Green = 525 ± 50 nm, and Blue = 460 ± 40 nm. Still images with 12 Mp resolution were collected, resulting in a ground-sample distance (GSD) between 15–25 cm.

Previous tests with the camera suggested that using a neutral density (ND-256) filter and long shutter speeds improves image quality significantly and eliminates image noise due to wave-focusing on the seafloor. This approach is of paramount importance for imaging the shallow seafloor. Therefore, the shutter speed was set to 1.3–2 s, ISO sensitivity was set to 100, and image overlap was determined visually by having a prominent ground feature captured in three consecutive images along the flight path. Image acquisitions occurred when the sea state was calm (<0.2 m significant wave height), sun elevation was <30 degrees, or cloud coverage was low and homogeneous to avoid sun-glint noise. This is also why flying at exact 30-day intervals was not possible. Images were processed in Pix4D v4.5.6 proprietary software for obtaining a set of RGB orthomosaics [50]. The overall workflow is presented in the diagram of Figure 2. Initially, the cameras’ positions were reconstructed using SfM processing, and camera-specific geometric corrections were applied. Following this, bundle adjustment and georeferencing were performed using the navigation metadata of imagery. Since no RTK mode was available onboard the drone, the positional accuracy of the drone is ±1.5 m according to the manufacturer. Camera-specific radiometric corrections, including sensor bias, sensitivity, gain settings, exposure settings, and lens vignette effects, were also applied using Pix4D v4.5.6 software. The output orthomosaics cover a coastal area that is ~2 km long and more than 200 m wide. For data harmonization purposes, each orthomosaic was resampled to 25 cm pixel size. At this point, it has to be noted that although SfM succeeded in creating the orthomosaics, it cannot be applied to producing a comprehensive bathymetric point cloud for the nearshore area. This is due to the limited number of matching points in the submerged part between consecutive images, which is a typical drawback of SfM when applied on featureless surfaces such as smooth, sandy seafloor [43]. This study is based on a spectral method suitable for estimating bathymetry over smooth, featureless seafloor types such as those found in a sandy beach area. Following this, the coastline was digitized manually and the land area was clipped so that each orthomosaic depicts only shallow seafloor. The clipped orthomosaics were transformed according to the method described in Section 2.2 for enhancing the relief of shallow seafloor. By utilizing the transformed mosaics, the bar crest was manually identified for each date, and then its distance to the coastline was calculated. In addition, ten cross-shore profiles (Figure 1) were extracted and then stacked along their temporal dimension to highlight the bar evolution at each profile location.

2.2. Standardized-Ratio Bathymetric Index (SRBI)

Considering the absence of in situ bathymetric data, the type of shallow seafloor, and the temporal character of drone imagery, a novel data-transformation approach is suggested for producing a multi-temporal PBM for analysing the morphodynamic behaviour of nearshore sandbar. Therefore, the SRBI concept is introduced as an extension of the spectral-based logarithmic band ratio from [44]:

$$z=m_1{\displaystyle \frac{ln\left(n{R}_w\left({\lambda }_i\right)\right)}{ln\left(n{R}_w\left({\lambda }_j\right)\right)}}-m_0$$

(1)

where z is relative depth, R_w (λ_i,j) is the water column reflectance of optically shallow water recorded at wavelength λ_i nanometres (with i < j), m₁ is a tuneable constant to scale the ratio to depth, n is a fixed constant for all areas, and m₀ is the offset for a depth of 0 m (e.g., tidal offset). The fixed value of n is chosen arbitrarily in order to assure that the logarithm will be positive under any condition.

This method performs well in areas where, ideally, only a single seafloor type is present and bathymetric changes occur in a gradual pattern. Consequently, Episkopi Beach is a suitable candidate for applying the logarithmic band ratio approach. Although the RGB sensor does not record in narrow-band wavelengths, the resulting pixel values appear highly influenced by exponential light absorption with increasing water depth (see Section 3.1). Therefore, the logarithmic band ratio is applicable for quantifying bathymetry according to the corresponding pixel-value variability between the Green and the Red bands. These bands were selected, as they are known to have good penetration in clear coastal waters [51]. Initially, the coastline is extracted visually and a land mask is applied. Following this, the logarithmic ratio between the Green and the Red bands is calculated using map algebra according to the following equation:

Band_ratio = −(ln(Green)/ln(Red))

(2)

The negative sign in Equation (2) ensures that pixel values are numerically arranged as an actual digital elevation model (DEM) where Z_deep < Z_shallow. Considering the lack of in situ bathymetric measurements for calibrating the band ratio output and that the band ratios from different dates hold different value ranges, the output grid is then standardised by subtracting the mean value of the raster and multiplying it by its standard deviation. The SRBI rasters have the same range (from 0 to −5) for all temporal data, allowing for comparisons between surfaces from different dates. This is considered a major advantage of this approach. The validity of the SRBI concept was tested with a subset of 2D bathymetric data, resulting from a double track of ICESAT-2 LiDAR altimetry sensor (ATL03_20230225093959_10211802_006_01_3R & _3L). The LiDAR sensor has a horizontal resolution of 4–5 m [52] and is able to resolve elevation differences greater than 0.25 m. The ICESAT-2 LiDAR overpass occurred on 25 February 2023 at the study area, just 8 days after collecting the drone imagery, meaning that it has captured bathymetry with the lowest possible difference since the drone survey date. However, some discrepancy is expected between the LiDAR and SRBI data due to the intense wave activity during this period. Here, the LiDAR bathymetry was extracted from geo-located photon-count data (Figure A1, Appendix A) by manual filtering of outliers. Then, a constant value corresponding to sea surface elevation was subtracted for obtaining the actual depth profile. The extracted LiDAR depths were corrected for tidal and refraction effects. A tidal offset of 10 cm (typical tide in the area) was added to the data, and a scale factor of 0.25416 was used as a universal correction for water refraction, as suggested in [53].

2.3. GIS Analyses

The resulting 12-month time-lapse of pseudo-bathymetry datasets was used for extracting morpho-bathymetric information regarding the nearshore sandbar. Initially, the sandbar crest and coastline position were digitized manually for each temporal dataset. Following this, a set of 10 cross-shore SRBI profiles were used for quantifying the sandbar’s dynamic behaviour over time. Temporal stacking of each batch of cross-shore profiles was applied to illustrate the transitional states of the sandbar. Thus, each SRBI profile set was placed in an arbitrary raster with the X-coordinates representing the temporal dimension, and then each raster was interpolated using the inverse-distance weighted (IDW) algorithm. The SRBI contours were calculated to highlight the morpho-bathymetric changes occurring over time. The distance from the bar crest to the coastline was calculated using a set of 40 points along the bar crest. The original RGB orthomosaics used for estimating the SRBI were georeferenced in QGIS v3.28 using the orthomosaic from May 2023 as a reference, given that it was aligned well with the GoogleEarth© basemap. A set of 7–10 visual control points were selected for georeferencing each orthomosaic. These control points were different for each case, but they always represented stable, onshore structures such as corners of buildings or road marks, for example. The root-mean-square error (RMSE) between points on each orthomosaic varied from 0–1.5 m. However, horizontal differences up to 4 m were found by visually examining independent target points between random orthomosaics.

2.4. Wave Data

Time series of significant wave height and direction were extracted (hourly) from the Copernicus product MEDSEA_ANALYSISFORECAST_WAV_006_017 [54] for a location ~2.5 km offshore from the study area (Latitude: 35.34218; Longitude: 24.35512). The waves forecast component (Med-WAV system) is a wave model based on the WAM Cycle 6. This dataset is used for a qualitative description of the local wave regime and assists in interpreting the dynamic behaviour of nearshore sandbar during the monitoring period.

3. Results

3.1. Correlation with Satellite LiDAR Altimetry Data

In Figure 3A, it is shown that the logarithmic values between the Green and Red bands show a linear relation when sampled over increasing distance from the coastline (assuming increasing water depth, Figure A1, Appendix A). Therefore, it is anticipated that the SRBI values should have some correlation with actual bathymetry data. ICESAT-2 LiDAR data provided a good opportunity for comparing the correlation between the SRBI values and the corresponding LiDAR depth values. Figure 3B shows the relation between the ICESAT-2 bathymetric data and SRBI values along the LiDAR tracks shown in Figure A1 (Appendix A). It is inferred that the SRBI provides a PBM that relates linearly with ICESAT-2 LiDAR bathymetric measurements. Therefore, it can be utilised as a surrogate of seafloor DEM when actual bathymetric data are unavailable.

3.2. Temporal Wave Data and PBMs

Episkopi Beach is north-faced, meaning that incoming waves from general directions between WNW and ENE are the main drivers of nearshore morphodynamics. According to hourly model data 2.5 km offshore from the study area, there were several occasions during the monitoring period when the significant wave height was greater than 2 m, while at the beginning of February 2023, waves of more than 4 m were recorded (Figure 4).

Drone-based temporal orthomosaics were utilized for producing twelve SRBI grids, representing time-lapse PBMs with 25 cm spatial resolution (Figure 5). These datasets highlight the complex crescentic bar system at Episkopi Beach and other morpho-bathymetric features of nearshore seafloor, such as rip-channels and sand-waves (Figure 6b,d). According to the SRBI grids, the crescentic bar system appears to consist of three components: (a) the inner bar or platform (towards the coastline), (b) the outer bar (towards offshore), and (c) the intermediate bar (in-between the previous features). All bars show a degree of connectivity at one or multiple locations, while the inner bar is more intermittent and periodically transforms into a shallow platform. The outer bar is less prominent, as it appears mainly in the offshore boundaries of the mosaics, and it has a relatively straight geometry. The intermediate bar is a typical example of a crescentic sandbar with a more consistent morphology, including rip-channels (Figure 6c,d) and horns. It is often connected to the inner bar. The high-resolution SRBI data allowed for manual digitising of the crescentic bar crest (Figure 5) of the intermediate crescentic bar. However, it occasionally merged with the inner bar’s crest (Figure 6e,f).

A set of ten cross-shore beach profiles were extracted from temporal PBMs (C1–C10, Figure 1). The spacing between the profiles, as well as their length, is ~200 m. At the beginning of the monitoring period (November 2022), profiles C1, C2, and C5 (Figure 7) exhibited a low-height bar formation ~100 m from the coastline, which remained relatively stable until January 2023. Since then, the bar has been smoothed out and transformed into a generally homogeneous shallow platform attached to the coastline until the end of the monitoring period. Profiles C3 and C4 show a rather steep, narrow bar formation slightly around 120 m from shore in November 2022. In contrast to C1, C2, and C5, the bars in C3 and C4 were smoothed out during December 2022. Since January 2023, a new, gently sloping bar developed ~140 m from the shore, moving gradually towards 100 m from the shore between January and November 2023. Continuing further to the east, stacked profiles C6 and C7 (Figure 7) showed similar behaviour to profiles C1, C2 and C5, with a bar feature at ~100 m from shore, the morphology of which is smoothed out until February 2023 and, since April 2023, becomes part of a wider shallow coastal platform. However, the temporal stack of profile C6 indicates the development of a bar with a low slope at 150–180 m from shore between April and August 2023. In addition, the temporal stack of profile C7 shows the development of a bar formation ~100 m from shore during the last three months of the monitoring period (September–November 2023). On the other hand, (and similarly to stacks of C3 and C4), profile stacks of C8–C10 show the formation of a bar ~150 m from shore in January 2023, which then moved towards the shore until the end of the monitoring period. Interestingly, the temporal stack of C10 depicted the periodic transition of the bar, starting at ~150 m from shore (November 2022), migrating to a maximum of ~180 m from shore (April 2023), and moving back to ~150 m from shore in November 2023. It has to be noted that the trough separating the intermediate bar from the shallow platform in C8 was much shallower during most of the monitoring time (January–November 2023) compared to the troughs from C9 and C10. Thus, the bar in C8 had a wider amplitude and smoother slope than the bars in C9 and C10. Additionally, the bar in C8 fluctuated within 120–140 m from shore compared to the bars in C9 and C10, where they migrated further offshore.

A set of 40 point measurements (~50 m apart) along the bar crest was used for extracting the differential distance between the bar and the coastline over consecutive month pairs (Figure 8). The diagram in Figure 8 provides a detailed overview of the seasonal variations of the sandbars and cross-shore migration. In the first few months of the monitoring survey (November 2022–January 2023), there was a relative trend for offshore migration, while some eastern bar locations showed a strong migration towards the coastline. The average migration rate was 6 m.month⁻¹. From January 2023 until April 2023, the average offshore migration rate reached 9 m.month⁻¹ (with eastern bar sections reaching maximum rates of 40–60 m.month⁻¹), though some locations showed a clear nearshore direction of migration. For a brief period from April 2023 to May 2023, the bar was at a tipping point, as it entirely moved towards the coastline with an average rate of 11 m.month⁻¹. From May 2023 to July 2023, the bar reached an equilibrium state, having the lowest migration rate of 0.4 m.month⁻¹ towards the coast. From July 2023 to October 2023, there was an increasing trend for nearshore migration with an average rate of 8 m.month⁻¹ (with some parts located in the eastern part of the bar exceeding 50 m.month⁻¹). From October to November 2023, the bar was found again in a relatively balanced state with a pace of 2.6 m.month⁻¹ towards the coast. Given the horizontal uncertainty of these datasets, it is suggested that bar movements greater than 4 m can be accurately resolved.

Apart from focusing on the behaviour of each profile individually, an overview map showing the extent of sandbar mobility and SRBI variability in the entire study area was examined (Figure 9). The map in Figure 9A depicts the relative cross-shore movement of the entire sandbar at each monitoring interval. As suggested by the analysis of individual profiles, there is an initial phase (during the first four months of the monitoring) during which the crescentic sandbar moves rapidly seawards, where it decreases its sinuosity due to extreme wave action. Following this, the sandbar moves steadily backwards to the shore during spring–summer 2023. The envelope within which the entire cross-shore movement occurs is about 40–60 m wide on average for most of the sandbar crests (Figure 9A). Regarding the long-shore movement of the bar, it is observed that, in general, the bar remains relatively stable, with only the western sector showing 80–90 m displacement during the 1-year monitoring. Temporal statistical analysis of all the SRBI mosaics reveals the variability of SRBI values over the monitoring period. Areas within the troughs (e.g., p1, p4, in Figure 9A) and the rip-channels appear to hold greater SRBI differences throughout the year. This suggests that the depth at these areas has undergone significant changes. Apparently, areas near the sandbar crest and close to the coastline show little SRBI variability. Locations p1, p3, and p4 in Figure 9B show that initially, there is a gradual increase in the SRBI values (getting shallower) followed by a relatively stable period before they start decreasing slowly (getting deeper). In contrast, location p2 demonstrated more stable SRBI values, meaning that it experienced small bathymetric changes.

4. Discussion

According to temporal PBM data analysis, the intermediate crescentic sandbar at Episkopi Beach exhibited a wide range of cross-shore variability patterns during the 12-month monitoring period. In addition, it was observed that these varied for the different sections of the bar (eastern, middle, and western). Given the absence of significant tides in the area, it is suggested that wave action is the main hydrodynamic cause of morpho-bathymetric change. A general characteristic of sandbar morphology at Episkopi Beach is that it does not change abruptly (e.g., hours, days), but rather, over weekly or monthly intervals. This occurs because low wave conditions (i.e.,: H_s < 1 m) may last for a few weeks (Figure 4) in the area, resulting in arrested bar morphology [55]. Bar arrestment is typical in micro-tidal areas with low wave energy, and it has been observed in other coastal areas of the western Mediterranean Sea [46,47]. In this study, sandbar variability was mainly driven by occasional increases in wave energy, occurring roughly once per two months during the summer period and once per month during the winter period (Figure 4). It is noteworthy that not all bar sections responded in the same way to wave forcing (Figure 7 and Figure 9). Through temporal stacking of cross-shore profiles, it was observed that some areas (profiles C1, C2, C5, C6, C7) showed a smooth morphological transition throughout the monitoring year, while others (profiles C3, C4, C8, C9, C10) showed a more dynamic behaviour. Particularly, the latter profiles depicted the development of a bar ~120 m from the coastline in January 2023 which then migrated further offshore, briefly, before it started migrating shoreward gradually (Figure 7). A similar pattern was further observed in Figure 9A that depicts the temporal movement of the entire sandbar. This behaviour is typical of the net offshore migration (NOM) cycle (see next paragraph) which has been described in earlier studies about several sandbar locations worldwide [2,11,56]. The different responses of the bar segments suggest that their morphology creates a unique seafloor pattern that influences hydrodynamic activity differently at each bar location [20]. The SRBI variability shown in Figure 9 shows that seafloor areas with relatively deep bathymetry, such as rip-channels and troughs, experienced substantial bathymetric change over the monitoring period (Figure 9B), which is probably due to sediment redistribution caused by the circulation of rip-currents.

4.1. Bar Migration

Although not all bar sections exhibited the same migration trend each month of the monitoring year, a general migration pattern could be established based on most of them. A clear change in the summer and winter migration patterns was observed, according to which the majority of the bar sections tended to migrate offshore during the winter months (November 2022–March 2023, and September–November 2023) and nearshore during the summer months (April–August 2023) (Figure 8 and Figure 9A). The only time that all bar sections had the same migration direction was in April-May 2023, where the entire bar was migrating towards the coast (Figure 8). A comparable migration pattern has also been described by [12,20,57] for non-tidal and micro-tidal beaches, suggesting that during periods of increased wave action (e.g., winter), the bar tends to migrate offshore due to undertow intensification, while during low-energy wave action, the bar migrates nearshore [6,46,58]. The difference in the migration direction between parts of the sandbar might be explained by local sandbar morphology that produces differential sediment transport (thus favouring sandbar migration) under the same wave forcing [20,56]. This behaviour indicates that different parts of the sandbar are at different stages of the net offshore migration (NOM) cycle [1,2,56,59]. The NOM cycle is typical in barred beaches, and it is used to describe the development of a sandbar in the surf zone and its migration and decay offshore [1,56]. The diagram in Figure 8 captures a general pattern that could be attributed to a distinct NOM cycle. However, there is no clear evidence that the intermediate sandbar decays when it reaches its offshore position (Figure 7). In the case of Episkopi Beach, it seems that some parts of the intermediate sandbar do not decline offshore, but they bounce back shoreward before they weld onto the nearshore platform. This is probably explained by the alongshore variability of the multiple sandbars at Episkopi beach that interferes with the typical NOM migration pattern occurring in single-barred beaches [2]. According to the diagram in Figure 8, it is inferred that the duration of the NOM cycle at Episkopi Beach is approximately one year.

The maximum migration rates for the Episkopi sandbar reach more than 60 m.month⁻¹ (near- or offshore direction) during wave activity peaking (throughout January and February 2023, and from September to October 2023). However, the average monthly migration rate does not exceed 10 m.month⁻¹. This is due to the diverging migration directions between different sections of the sandbar. The lowest migration rates (<2 m.month⁻¹) fall within the spatial uncertainty of the datasets (4 m), and they were observed during the summer months (May 2023–August 2023), perhaps as a result of prolonged, low-wave conditions (Figure 4) resulting in morphology arrestment. The migration rates observed in this study are comparable with those reported by [2] for the Gulf of Lions (West Mediterranean).

4.2. Workflow Applicability

The SRBI is a novel concept based on optical imagery, enabling multi-temporal analysis of PBMs in the absence of ground-truth depth measurements. The SRBI in this study was produced from very high-resolution orthomosaics with 3–4 m overall spatial accuracy. The resulting SRBI data contain relatively low positional errors comparable to those reported in timex studies [21]. Nevertheless, it is suggested that when drone imagery is coupled with ground control points and/or real-time kinematics (RTK) positioning, a superior level of spatial accuracy (centimetre scale) can be achieved. Maximizing horizontal accuracy would result in improved sandbar localization and support better discrimination of fine-scale migration rates. Resolving meter or sub-meter differences in sandbar movement (over a certain time) would enable the monitoring of nearshore seafloor in weekly or even daily intervals. Currently, only drone-based bathymetric LiDAR sensors can yield data with centimetre accuracy (both horizontal and vertical) and high spatial resolution [38,39,40] that allow for a detailed and precise monitoring of nearshore seafloor change. Despite their considerable cost, drone-based LiDAR sensors hold a significant potential for shallow seafloor monitoring, as they provide an effective tool for obtaining direct bathymetric measurements.

In contrast to other optical methods suggested by [4,17], the SRBI offers the possibility to obtain a surface grid that can be used as a surrogate of nearshore bathymetry. In this way, the 3D geometry of the nearshore seafloor is highlighted, assisting in more detailed and accurate interpretations. In addition, the SRBI allows the extraction of cross-shore profiles, thus supporting morpho-bathymetric analysis of nearshore features. The SRBI grids can be utilised in the same way as conventional DEMs for performing map algebra calculations such as grid subtraction, thus obtaining erosion-deposition maps. An important advantage of the presented method over the timex approach is that it allows for identifying not only the crest of the sandbar, but also the trough and surrounding bathymetry, highlighting important 3D information about sandbar morphology [6]. Sandbar morphometric parameters such as width, slope, height from seafloor base, and crest depth influence sandbar migration rates [2]. Applying the SRBI approach can assist in delineating these parameters more effectively and accurately, manually or with automated profile processing [57]. The SRBI cannot be used for direct bathymetric or volumetric estimations without being calibrated first with actual bathymetric data. Calibrating with in situ bathymetry could result in vertical errors of less than 0.5 m, as reported in earlier, similar studies [28,43]. The SRBI method depends on input data quality and can be applied at any location with acceptable water clarity. It requires RGB or multispectral imagery as input, captured by any platform (e.g., satellite, airborne, drone), and it is easily implemented within any GIS software. Multispectral imagery is expected to provide more accurate results, as the narrow-band wavelengths show better correlation with depth than broad-band RGB data [60]. Compared to satellite imagery, using drones in coastal monitoring provides better spatio-temporal resolution than satellite imagery, as it allows frequent revisit times according to project requirements. In addition, drones operate at close range without being influenced by clouds or other atmospheric effects [28,61,62]. The exploitation of drone imagery for monitoring shallow-water, coastal processes is benefited by the SRBI concept, as it allows for a rapid assessment of nearshore morpho-bathymetry without the need for ground truth measurements.

5. Conclusions

The SRBI is a novel concept developed in this study offering a useful transformation of optical sensor pixel values to relative bathymetric information. Such an approach is necessary for producing and comparing temporally-normalised datasets of surrogate bathymetry when in situ depth measurements are unavailable. In this study, the SRBI assisted in evaluating the morphodynamic behaviour of nearshore sandbars on a monthly basis for one year. Monthly PBMs are utilised for extracting cross-shore profiles and the position of the bar crest. The results from this study suggest that sections of the crescentic sandbar at Episkopi Beach are at different stages of the NOM cycle. The characteristic NOM cycle at Episkopi Beach is driven by the annual variability of wave activity during the winter and summer periods. This is further supported by temporal stacking of SRBI profiles where the sandbar’s development, migration, and decay/welding are observed. In particular, extreme wave conditions during the winter time altered the crescentic shape of the sandbar, while decreasing wave action during spring and summer resulted in a re-appearing of the crescentic geometry. During periods of minimal wave activity, the sandbar morphology remained stable, as reported similarly in other Mediterranean beaches. It is noteworthy that the SRBI profile stacks reveal the morpho-bathymetric evolution of sandbar components such as the trough and the slope. Future studies should exploit detailed measurements of these components for examining their relation with different coastal and/or hydrodynamic settings. This study highlights the applicability of the SRBI concept in coastal seafloor monitoring when in situ bathymetry data is difficult to obtain. The SRBI concept can be extended to similar study areas, providing useful insights for the morphodynamic processes occurring on nearshore seafloor.