Natural Hazards

, Volume 83, Supplement 1, pp 201–222

Shoreline variability of an urban beach fronted by a beachrock reef from video imagery

  • A. F. Velegrakis
  • V. Trygonis
  • A. E. Chatzipavlis
  • Th. Karambas
  • M. I. Vousdoukas
  • G. Ghionis
  • I. N. Monioudi
  • Th. Hasiotis
  • O. Andreadis
  • F. Psarros
Original Paper

DOI: 10.1007/s11069-016-2415-9

Cite this article as:
Velegrakis, A.F., Trygonis, V., Chatzipavlis, A.E. et al. Nat Hazards (2016) 83(Suppl 1): 201. doi:10.1007/s11069-016-2415-9

Abstract

This contribution presents the results of a study on the shoreline variability of a natural perched urban beach (Ammoudara, N. Crete, Greece). Shoreline variability was monitored in high spatio-temporal resolution using time series of coastal video images and a novel, fully automated 2-D shoreline detection algorithm. Ten-month video monitoring showed that cross-shore shoreline change was, in some areas, up to 8 m with adjacent sections of the shoreline showing contrasting patterns of beach loss or gain. Variability increased in spring/early summer and stabilized until the end of the summer when partial beach recovery commenced. Correlation of the patterns of beach change with wave forcing (as recorded at an offshore wave buoy) is not straightforward; the only discernible association was that particularly energetic waves from the northern sector can trigger changes in the patterns of shoreline variability and that increased variability might be sustained by increases in offshore wave steepness. It was also found that the fronting beachrock reef exerts significant geological control on beach hydrodynamics. Hydrodynamic modelling and observations during an energetic event showed that the reef can filter wave energy in a highly differential manner, depending on its local architecture. In some areas, the reef allows only low-energy waves to impinge on the shoreline, whereas elsewhere penetration of higher waves is facilitated by the low elevation and limited width of the reef or by the presence of an inlet. Wave/reef interaction can also generate complex circulation patterns, including rip currents that appeared to be also constrained by the reef architecture.

Keywords

Beach morphodynamicsCoastal video imageryPerched beachesBeachrock reef

1 Introduction

In recent years, there has been an increased interest in the morphological evolution (morphodynamics) of perched beaches, a beach type that proliferates in many parts of the global coastline. Perched beaches are those ‘… underlain and/or fronted seaward by shallow buried and/or outcropping natural or engineered hard structures’ (Gallop et al. 2012). Natural perched beaches can form on all types of rock (Larson and Kraus 2000), such as corals (e.g. McLean and Kentch 2015), beachrocks (e.g. Vousdoukas et al. 2007) or limestone (e.g. Gallop et al. 2011a, 2012).

Nearshore natural reefs, such as submerged beachrocks and fringing coral reefs, can modify nearshore hydrodynamics by dissipating the incoming wave energy and generating complex nearshore flows (e.g. Larson and Kraus 2000; Sheremet et al. 2011; Vousdoukas et al. 2012). In the case that the reef is backed by a deeper body of water (e.g. a lagoon and/or a longshore channel), it has been shown that the architecture of the reef/lagoon system can play a major role in the momentum balances across the system and, consequently, to the magnitude of the wave-driven flows and related coastal flushing rates (Lowe et al. 2010; Gallop et al. 2013). Concerning the morphodynamics, there are many open questions. For instance, studies suggest that perched beaches can show significant intra-annual and inter-annual variability (Gallop et al. 2013; 2015a) and be either less or more impacted by energetic events than open beaches (e.g. Rey et al. 2004; Muñoz-Perez and Medina 2010; Alexandrakis et al. 2013). Post-event recovery is also influenced by the presence of nearshore reefs; perched beaches have been observed to recover relatively slowly (Vousdoukas et al. 2009a; Muñoz-Perez and Medina 2010; Gallop et al. 2011b). Moreover, adjacent sections of the shoreline have been observed to show contrasting patterns of erosion and accretion that may be due to, for example, the development of particular nearshore flows and/or seasonal blockages to local sediment transport pathways (Gallop et al. 2013; Alexandrakis et al. 2013).

It appears that in order to resolve the morphodynamics and assess the erosion risk of perched beaches there is need for monitoring their morphological variability in different spatio-temporal scales (e.g. Gallop et al. 2015b). As the shoreline position is a feature that integrates large part of the information pertaining to beach morphodynamics, information on shoreline retreat (beach loss) and shoreline advance (beach gain) is required with a spatio-temporal resolution to match that of the shoreline change (Boak and Turner 2005). Such information is necessary in order to monitor/understand both the short- and long-term morphodynamics which, in some cases, can show contrasting trends (e.g. Hapke et al. 2006).

The aim of the present study was to investigate in high spatio-temporal resolution the shoreline variability of an urban beach fronted by a beachrock reef that has experienced severe long-term erosion (Ammoudara, N. Crete) (Fig. 1). Towards this objective, a 10-month long time series of images from a ground-based coastal video system were analysed using a novel automated procedure, whereas the influence of the reef architecture on wave dissipation and nearshore circulation was examined through hydrodynamic observations and modelling during an energetic event.
Fig. 1

a Ammoudara Beach, Heraklion, Crete; stars denote the location of the mouths of the four small outflowing rivers, i.e. the mouths of the rivers Giofyros, Xiropotamos, Gazanos and Almyros (from east to west), whereas the position of the offshore POSEIDON E1-M3A wave buoy is shown in the inset with the island of Crete. b The study area (eastern part of the beach), showing the field of vision of the three video cameras and the locations of the instrument deployment; the dark grey zone offshore of the shoreline delineates the submerged beachrock, whereas the longshoredashed line shows the length of the shoreline examined by video imagery. The cross-shore dashed line shows the location of the profile shown in Fig. 4. Satellite image source: Bing Maps, Microsoft

2 The study area

Ammoudara is an urban beach, located to the west of the port of Heraklion, Crete (Greece) (Fig. 1). It has a high socio-economic importance for the city of Heraklion due to its accessibility, recreational facilities and large carrying capacity. The perched beach is about 6.1 km long and can be divided into two main sectors. At the western sector (about 1.5 km long), beachrocks are found at or close to the beach face, whereas at the eastern sector (about 4.6 km long) beachrocks form a submerged reef oriented almost parallel to the coast, the width of which and its distance from the shoreline vary between 15–50 m and 40–70 m, respectively.

Beach widths range between 22 and 75 m, with the inner beach associated with low sand dunes (average height of 2 m) as well as extensive human development. Beach face gradients vary, with the steeper slopes (5°–8°) found at the east. At the eastern sector of the beach, the subaerial section consists of sands/gravely sands, whereas the seabed between the shoreline and the reef consists of poor/moderately sorted, gravely sands and sandy gravels (d50s between 1.8 and 2.7 mm). The coarsest sediments (d50 of about 2.7 mm) are found offshore of the Xiropotamos River mouth, while the area offshore of the reef consists of mostly sands (Alexandrakis et al. 2006). Seabed sediment thickness varies, with the sediment wedge thinning offshore towards the outcropping beachrock. Two distinct beachrock layers have been identified (Alexandrakis et al. 2013): an upper layer, about 0.3 thick, consisting of cemented coarse-grained beach material and showing erosional features consistent with beach face beachrocks (potholes, grooves and runnels); and a thicker, lower layer consisting of finer-grained sands exhibiting cross-bedding.

Ammoudara is a microtidal beach with a tidal range not exceeding 0.1 m (HNHS 2005). It is exposed to winds (and waves) from the northern sector, i.e. from the NW (28.7 % frequency of occurrence), the N (16.9 % frequency of occurrence) and NE (4.0 % frequency of occurrence) (Alexandrakis et al. 2013). Offshore significant wave heights (Hs) and periods (Tp) under energetic conditions from the NW have been estimated as 1.8 m and 7.2 s, respectively; both the northerly and (particularly) the north-easterly waves are less energetic (Alexandrakis et al. 2013).

Historical aerial photographs and repeated topographic levelling have shown that Ammoudara Beach has experienced significant shoreline retreat/beach loss since 1960; shoreline position has retreated by 10–60 m, with the longer retreats (>1 m/year the last 55 years) found at the east (Alexandrakis et al. 2013). It appears that the now submerged reef was originally part of the beach face and was left offshore following the erosion of fronting/overlying unconsolidated sediments and shoreline retreat. The extensive beach erosion could be partly due to the presence of the beachrock itself, which can cause sediment loss at the beach face by constraining beach profile adjustment, influencing swash infiltration and impeding onshore sediment movement at the recovery phase of the beach through the development of a ‘scour step’ at its offshore margin (Vousdoukas et al. 2007, 2009a, 2012). This erosion process had been certainly aggravated by beach sediment abstraction that took place in the 1950s and 1960s, the river management scheme that trained Giofyros River outflow (see Fig. 1) starving the eastern part of the beach and extensive coastal development that consumed most of the dune fields shown on a 1960 aerial photograph (Rentzepopoulou 2014). It has been suggested that, in recent years, erosion at the eastern sector of the beach has slowed down to about 0.7 m/year due to decreasing coastal development and to the protective influence of the now offshore beachrock reef (Alexandrakis et al. 2013).

3 Data acquisition and analysis

3.1 Coastal video imagery and shoreline detection

Coastal video imagery has been increasingly used for beach morphodynamic monitoring, as it allows low-cost acquisition of long-term, nonintrusive observations at high spatio-temporal resolution (e.g. Plant and Holman 1998; Pearre and Puleo 2009). It can provide high-resolution information on intertidal bathymetry (e.g. Uunk et al. 2010), swash processes (e.g. Vousdoukas et al. 2009b; Vousdoukas 2014), shoreline position (e.g. Senechal et al. 2015), nearshore bar morphology (Balouin et al. 2013), surf zone bathymetry (Catálan and Haller 2008) and beach use (e.g. Balouin et al. 2014). Main coastal video products include the time-averaged coastal imagery TIMEX images and the greyscale variance imagery SIGMA images (Holman and Stanley 2007). TIMEX images are defined on the red–green–blue (RGB) colour model and are time averages of time series of snapshots of the coast from a stationary field of view. SIGMA images represent the distribution of the standard deviation of pixel intensity along image frames for the acquisition period. Video images are usually acquired in a burst mode and must be translated from pixels to geographical coordinates to yield useful quantitative information (Hartley and Zisserman 2004).

In TIMEX and SIGMA image time series, the dynamics of significant beach features can be recognized/recorded, such as those of the ‘wet’/‘dry’ beach interface, the most inshore wave breaking line, or the upper swash zone limit; such features can be used as proxies for the shoreline horizontal position. Plant and Holman (1997) based shoreline extraction on the distinct shore-parallel bands of high light intensity generated by swash zone foam (shoreline intensity maximums—SLIMs). Turner and Leyden (2000) used a colour channel divergence (CCD) model that utilizes colour discrimination procedures focusing on the intrinsic differences produced by the differential light reflectance of the ‘wet’ and ‘dry’ beach. Aarninkhof et al. (2003) introduced a pixel intensity clustering approach, using the hue–saturation–value (HSV) colour model on TIMEX images. Rihouey et al. (2009) used a k-mean function to the parameters obtained after combining different RGB and HSV channels to define the shoreline, whereas Osorio et al. (2012) proposed a shoreline definition approach that uses image segmentation algorithms. However, there are still challenges associated with the different automated approaches. Plant et al. (2007) reviewed four different algorithms on image data sets from four experimental sites and found that the development of a universal, robust shoreline extraction procedure from coastal images faces significant challenges due mainly to the high temporal variability of the beach environmental, hydrodynamic and morphological conditions; this has prompted attempts to handle the shoreline extraction problem using artificial neural networks—ANNs (Kingston 2003; Vousdoukas et al. 2011; Rigos et al. 2014, 2016).

In the present study, coastal imagery has been provided by a video system consisting of three PointGrey FLEA-2 cameras, installed on the seaward roof of the Heraklion Olympic Stadium (Figs. 1, 2); the elevation of the centre of view of the cameras was about 26 m above the mean sea level. The cameras were connected to a field station PC, and hourly 10 min bursts were obtained during daylight with an image acquisition rate of 5 Hz. Following automated pre-processing at the field PC, TIMEX and SIGMA imagery is transferred via an internet link to the laboratory for final processing and georectification.
Fig. 2

Installation of the optical system and view of the Ammoudara Beach

The system was set to monitor a beach stretch approximately 2400 m long at the eastern, most dynamic part of Ammoudara Beach. However, as the accuracy decreases with the distance from the camera due to the increase in the pixel footprint, only images from a beach stretch 1400 m long are presently analysed (x-coordinates between about UTM 326500 and 327900, Fig. 1). In this area, pixel footprint (and thus accuracy of detection) is mostly 0.25 m and always <0.5 m.

All images were georectified through standard photogrammetric methods, including the calibration of the cameras’ intrinsic parameters (distortion) and estimation of extrinsic parameters (Bouguet 2007) on the basis of a set of ground control points—GCPs—collected during dedicated RTK-DGPS topographic surveys. All images were projected to real-world coordinates, and since more than one cameras was used, geo-referenced mosaics were generated with 0.5 m resolution, both alongshore and cross-shore (Vousdoukas et al. 2011).

As image intensity at the shoreline is associated with the wave foam on the swash zone, the weighted mean position of the innermost zone of high intensity (as expressed in the TIMEX imagery) is used to record the average shoreline position during the 10-min burst. In order to facilitate the shoreline detection monitoring procedure, an automated coastal feature detector was developed. The detector is based on a localized kernel that progressively ‘walks’ along the feature of interest on the raw or georectified TIMEX imagery, automatically following the high-intensity zone along the shoreline. The site-specific configuration parameters of the detector are: (1) the preferable, in terms of shoreline following, general direction of kernel movement along the imagery (right to left for the Ammoudara mosaics analysed herein, Figs. 1 and 3) and (2) a corresponding user-defined ‘root’ cross-shore transect at the rightmost part of the image which spans across the feature of interest (in this case, the shoreline).
Fig. 3

a Georectified TIMEX mosaic of Ammoudara Beach (2 June 2014, 08:00 h); the rectangle (centred at x = 327680) shows the region magnified in panel (b), and the dotted line indicates the shoreline location (x0) examined in panels (e, f). b Magnified region of the TIMEX image, showing manual and automated detections of the shoreline (daily means). c Standard deviation of the cross-shore shoreline position (22 daily means) for the entire longshore analysis span. d Histogram of the difference of the cross-shore shoreline position detected through the automated and manual procedures; differences were calculated for each day separately (22 daily means) across the entire longshore analysis space (2725 points, spaced 0.5 m apart). e Difference between the daily and running means of manual detections (location x0), for seven randomly selected days; for each day, the running mean is calculated by incorporating a progressively increasing number of hourly detections (2 ~ 08:00–09:00, 3 ~ 08:00–10:00, 4 ~ 08:00–11:00…). The daily mean is the average of all hourly detections (08:00–16:00, 10-min long bursts). f As is in e, but for automated detections

Fig. 4

Location of the instrument deployment (black triangle) in relation to a cross-shore elevation profile showing the reef crest and flat (data collected in this study); the extent of the beachrock outcrop is shown by small vertical lines. Note that the three sensors were deployed close to each other and at the same water depth (see also Fig. 1)

Both configuration parameters are set once and apply to all images under analysis. For each image, the detection line initializes as a single ‘anchor’ pixel Apx(1), which is the pixel with the maximum backscatter intensity along the root cross-shore transect. Following the automatic detection of Apx(1), a local kernel centred to Apx(1) is spawned (typical kernel size: 5 × 5 pixels), and within this kernel, the detector scans—towards the direction of movement—for the pixel with the maximum backscatter intensity Apx(2), excluding Apx(1). When detected, Apx(2) becomes the new anchor point of the detector for the next pass, and the process is iteratively repeated, progressively moving the detector along the shoreline; the latter is represented by merging the successive Apx(i) points by order of detection. A linear interpolant is then applied to the raw Apx(i) series, yielding a shoreline that has a uniform longshore resolution which matches that of the underlined mosaic (0.5 m per mosaic pixel). It must be noted that by focusing the detection on a local window, the process is robust to intensity variations and does not require image normalization (e.g. Rigos et al. 2014). The duration of the overall detection process is <0.5 s per TIMEX image. The detector can be used on TIMEX and SIGMA imagery, given that the local maximum value of the underlying image within the sliding kernel can either express pixel intensity (TIMEX) or variance (SIGMA).

Optical data from Ammoudara Beach were analysed with the detector, using all available hourly TIMEX mosaics between 1 January 2014 and 5 November 2014 (N = 1430 mosaics); dates of system downtime were ignored. For each image, the raw detector shoreline was smoothed using a 30-point moving average window. In order to assess the performance of the detector, automatically extracted shorelines were compared against manual shoreline detections (Fig. 3) that used the same shoreline definition criterion: shoreline was defined as the mean position of the innermost zone of high intensity (as expressed in TIMEX imagery).

Manual shoreline detections were performed by a single human operator on 143 hourly TIMEX mosaics (for the period 28 May–19 June 2014). The raw manual detections were interpolated to the same longshore analysis span and resolution of the automatic detections, and smoothed using a 30-point moving average window.

Comparison between the automated and manual detections gave generally good results (Fig. 3). Detections from the automated procedure were found to be systematically located further offshore than the manual detections, having also lower cross-shore variability (Fig. 3b, c). The automated shoreline detection appears to give more conservative estimations of beach loss than those of the manual detection. In addition, the automated detection proved to be more robust with the difference between the daily and running means being more consistent as well as smaller than that of the manual detection (Fig. 3e, f).

3.2 Hydrodynamic observations and modelling

High-resolution nearshore bathymetric/topographic information was collected during very calm wave conditions on the eve (end of October 2014) of the hydrodynamic observations. The nearshore bathymetry of the nearshore area (down to water depths of about 6 m) was resolved through the surveying of a dense grid (10–20 m) of nearshore cross- and longshore bathymetric transects; further offshore (to a depth of about 12 m) a much coarser survey grid was used (of about 50 m). Data were obtained through a single-beam digital Hi-Target HD 370 echo-sounder and a Differential GPS (Topcon Hipper RTK-DGPS) deployed from a very shallow draft inflatable boat. The morphology of the ‘dry’ beach (down to water depths of about 0.8 m to overlay the topographic and bathymetric transects) was recorded through a dense grid of RTK-DGPS cross- and longshore topographic transects. Additional information on seabed elevation was collected over areas of very shallow/outcropping reef crests by divers.

Wind data to assess the wind regime affecting the beach were obtained from the Heraklion meteorological station of the Hellenic National Meteorological Service (HNMS) located at the coast to the east of the beach (UTM northing: 3912269, easting, 334603, elevation: 39 m); translation of the recorded wind speeds to speeds at 10 m elevation (U10) followed Thomas et al. (2005). Offshore wave information was obtained from an offshore buoy (POSEIDON E1-M3A buoy) located about 35 km to the north of the study area (35.66°N and 24.99°E) at 1440 m water depth, installed/operated by the National Centre for Marine Research (NCMR). In this area, this is the only available long-term wave information.

High-frequency hydrodynamic information was collected during a relatively high-energy event in early November 2014. RBR virtuoso D/wave recorders, operating at 4 Hz, were deployed offshore and inshore of the reef, at 10 and 1.4 m water depths, respectively (1–5 November), with an ADV—Nortek Vector deployed very close (and at the same water depth) to the inshore wave recorder to collect concurrent wave/current information (1–3 November). In addition, a sideward looking ADCP (Nortek Continental) was deployed between 10:30/1 November and 12:50/3 November 2014 on a raised platform on the seabed about 20 m to the west of the other sensors (UTM northing: 3912292, easting: 327908 m) and at the same water depth to monitor cross-shore current distribution inshore of the reef (cell size 2 m, sampling frequency 1 Hz, profile average of 60 measurements). Due to limited water depth in this area that may influence the effective horizontal range of the records, only currents recorded to a distance of about 14 m from the ADCP deployment site are used in the present analysis.

In order to gain further insights into the nearshore hydrodynamics of the beach, an advanced hydrodynamic model was employed. The model solves high-order Boussinesq equations to describe nearshore hydrodynamics; the classical Boussinesq equations have been extended so as to include higher-order nonlinear terms that can describe better the propagation of highly nonlinear waves in the shoaling zone. Detailed description of the model has been provided elsewhere (e.g. Karambas 2006, 2012; Vousdoukas et al. 2009a). The model was set up using the very detailed beach elevation/water depth information collected in the present study and forced by offshore wave conditions recorded in the field experiment. The model was run in a stationary mode, and validation was provided using the wave and current information recorded by the sensors inshore of the reef (Fig. 4).

4 Results

4.1 Shoreline position changes in the 10-month monitoring period

The nearshore bathymetry of the monitored section of the eastern Ammoudara Beach examined here is dominated by a beachrock reef with complex architecture (Fig. 5a). The reef’s distance from the shoreline varies along the beach, decreasing towards the west. Water depths inshore of the reef are shallow (in all cases <2 m), whereas the elevation of the reef crest varies being, in few cases, close to the mean sea level. The bathymetry offshore of the reef is also complex (due to the formation of ‘scour steps’ along its offshore toe, see, e.g. Vousdoukas et al. (2007)), having the potential of influencing the alongshore wave distribution before wave breaking. There are also reef inlets, the most significant of which is located about 500 m to the west of the camera station (at about x = UTM 327260–327430). Formation and maintenance of this inlet may have been influenced by the Xiropotamos River outflow (Fig. 1) that may have constrained beachrock formation in this area; a trace of a submerged river channel can still be discerned on the seabed offshore of the beachrock reef opening (Fig. 5a).
Fig. 5

a Nearshore bathymetry (in m) of Ammoudara Beach. b Georectified TIMEX image mosaic (2 November 2014, 10:00 h) from the eastern part of Ammoudara Beach; note the turbid plume at the area of the reef inlet (location x4). c Spatial distributions of the standard deviation of cross-shore shoreline position and differences between the most offshore (max) and most inshore (min) shoreline position detected during the 10-month period (1 January–5 November 2014) and the period of hydrodynamic observations (1–5 November 2014). d Temporal changes in cross-shore beach gain/loss at eight representative locations (see panelb); changes are relative to the shoreline position in the beginning of the monitoring period (1 January). Light grey stripes indicate the timing, duration and speeds of energetic wind events (winds from the northern sector with speeds ≥13.9 ms−1 at 10 m and duration >6 h)

Shoreline position showed significant variability during the 10-month monitoring period. At any shoreline section 0.5 m long, the minimum cross-shore change detected (the difference between the most inshore and most offshore shoreline position) was 3.1 m for the monitoring period and the maximum about 8 m; the change appeared more prominent to the west of the major reef inlet (mainly at x about UTM 327000–327300). Three main areas of increased variability were identified: one along the western margin of the monitored beach section and two immediately behind and to the west of the reef inlet, with the latter being more extensive and with the highest variability (Fig. 5c).

Adjacent sections of the shoreline also showed very different temporal patterns of beach retreat loss and beach advance gain. The spatial variability of the cross-shore shoreline position started to increase in April–May (Fig. 5d), probably in response to the energetic events of this period. The increase continued till about the beginning of July; variability was then stabilized through the rest of the summer (until the end of September), before it started decreasing towards the end of the 10-month monitoring period. The most consistent trends were recorded along the western part of the monitored shoreline (locations x1 and x3) to the west of the reef inlet (Fig. 5). In these locations, continuous beach loss was observed from the beginning of April, with a recovery trend discerned towards the end of the monitoring period (from the end of September 2014); between these two locations (location x2), beach loss appeared with a time lag of about 3 months and more or less continued until the end of the monitoring period. In comparison, the neighbouring shoreline to the east that was directly fronted by the reef inlet (locations x4 and x5) showed relative stability and even beach gain. Further to the east (locations x6, x7 and x8) the shoreline showed less cross-shore variability (Fig. 5c, d).

In order to improve our understanding of the temporal evolution of the spatial variability of the shoreline position, this was evaluated against the available wind and wave information. Analysis of long-term meteorological data from the Heraklion coastal meteorological station shows that gale force winds (wind speeds at 10 m elevation—U10 > 13.9 ms−1) that can generate substantial waves affecting the beach are mostly from NW and N directions. In the monitoring year (2014), a significant decrease in the energetic winds from these directions was observed; in 2014 only 7 % of the winds from these directions recorded speeds U10 > 13.9 ms−1, whereas in the previous years 2012 and 2013 the percentages were 10.4 and 9.1 %, respectively. During the 10-month monitoring period, the beach was affected by 19 such wind events from the northern sector (315°–45°N) with a duration of more than 6 h; ten of these events were 30–81 h long (Fig. 5d).

The offshore characteristics of the waves from the northern sector as recorded at the POSEIDON E1-M3A buoy are shown in Fig. 6. A weak trend may be discerned; offshore wave heights and periods appear to be generally decreasing and the wave steepness (S0) increasing during spring–early summer (April–July 2014), i.e. the period with a generally increasing trend in shoreline variability. Following this period, these offshore wave trends steadied or, even, reversed. These trends partially correspond to the tendency identified in shoreline variability which exhibited an increase between April and July, followed by a period of relative stability until the end of September (Fig. 5d).
Fig. 6

Monthly means and standard deviations of the characteristics of the offshore waves from the northern sector (315°–45°N) as recorded at the POSEIDON Wave buoy (for location see Fig. 1): a Hs (in m); b Tp (in s) and c wave steepness S0 (HO/LO). Original wave data averaged over 3 h

4.2 Hydrodynamic observations and modelling during an energetic event

The field experiment took place in early November (1–5 November) 2014, when there were energetic winds blowing from the N-NNW (mean direction of about 3500) for about 60 h. Although this event was quite energetic (see, e.g. Fig. 6), it caused limited changes in shoreline position (Fig. 5c); cross-shore changes during the event ranged between 0.2 and 0.5 m, with variability being generally higher to the west of the inlet. Although this pattern is generally similar to those observed during the whole monitoring period, it shows also significant differences. First, shoreline variability during this event was higher in the area immediately behind the reef inlet (at locations x4 and x5, see Fig. 5). Secondly, shoreline variability appears to be mostly associated with small beach gains as is also suggested by the early November trends in the time series shown in Fig. 5d.

Energy inputs fluctuated during the event. The wind picked up on 1 November and recorded increased speeds until late on 3 November when it started to subside; this event although did not qualify as a gale was, nevertheless, energetic. Wave energy also varied during the observation period showing a correlation to the wind intensity at both the POSEIDON buoy and the wave recorder offshore of the reef (Fig. 7). In the beginning (1 November), the significant height (Hs) and peak period (Tp) of the waves recorded offshore and inshore of the reef were 0.9 m and 6 s and 0.5 m and 6.1 s, respectively. Wave energy increased the following day, reaching its peak at between 10:00 and 11:00 h on 2 November: Hs of up to 1.5 and 0.55 m and Tp of 6.6 s and 7.1 s were recorded offshore and inshore of the reef, respectively. Following this peak, wave energy progressively decreased, showing converging offshore and inshore wave heights and periods towards the end of the observation period (24:00 h, 4 November 2014). During the event, the waves were impinging almost perpendicular to the coastline (mean wave direction of about 1700 N inshore of the reef, identified through the ADV records). It should be noted that the POSEIDON buoy and the offshore RBR recorded waves show a good correlation regarding their variability, with the significant wave heights recorded at the RBR sensor (10 m water depth) being substantially reduced relative to those recorded at the deep water buoy.
Fig. 7

Wind speed (a) and direction (b) (3 h averages) and corresponding significant wave height (Hs) (c) and peak period (Tp) (d) from the POSEIDON buoy (3 h averages) and as estimated from hourly RBR wave records offshore and inshore of the beachrock reef at 10 and 1.4 m water depth, respectively

Generally, energetic winds from the northern sector can be translated to energetic waves which are then dissipated over the beachrock reef. The results suggest that the reef damps wave energy very effectively. Not only it reduces significantly the incoming wave energy, but also ‘filters out’ the variability of offshore wave heights; waves inshore of the reef (Figs. 1, 4) had very similar significant wave heights (about 0.5–0.55 m) during most of the event, irrespective of the offshore wave conditions (Fig. 7). It appears that the reef (at least at the observation site) dissipated disproportionately more wave energy as incident wave energy increased (e.g. Ferrario et al. 2014). Although there was also wave dissipation offshore of the reef which was modelled (see below) to reduce wave heights to about 1.1 and 1.3 m at the offshore toe of the reef (at about 2.8 m water depth), interaction of the incoming waves with the reef crest/flat reduced the wave height to 0.5–0.55 (transmission coefficient (Kt) 0.4–0.5 under the observed conditions); these results compare well with the dissipation recorded in similar studies relating to coral reefs (Van Dongeren et al. 2013; Ferrario et al. 2014).

As the wave energy was shown to be changing during the event, spectral analysis of the wave records was carried in 1-h windows (Brodtkorb et al. 2000). Significant reduction (about an order of magnitude) in spectral energy density was found inshore of the reef in all 1-h windows (Fig. 8); such substantial reduction in the energy fluxes across the reef is probably due to the dominant effect of wave breaking across the beachrock reef crest/flat which in this area has a considerable width (Fig. 4). Inshore of the reef, a lower frequency energy component (at 0.04 Hz) appears significant, suggesting preferential short-wave attenuation over the reef as well as nonlinear wave interactions (e.g. Pomeroy et al. 2012; Van Dongeren et al. 2013). These changes were more pronounced during the most energetic periods of the event.
Fig. 8

Wave spectral density: a offshore and b inshore of the beachrock reef (at 10 and 1.4 m water depths, respectively). Note the two significant peaks at (b) (inshore records). Analysis shown is for the most energetic 1-h window of the event (10:00–11:00 h, 2 November 2014)

Concerning wave-generated currents, both the ADV and the ADCP recorded weak onshore flows. At the ADV site, the recorded flow was towards the shore (with a direction of about 170°N) having mean speeds of about 0.07 ms−1 at 0.38 m above the seabed at the beginning of the event. ADCP records have also shown mean onshore flows of about 0.08–0.10 ms−1 during the more energetic part of the event, which appear to increase towards the channel between the reef and the shore (at about 12 m from the ADCP deployment site). It seems that cross-shore current jets are generated by wave set-up over the reef crest (Johnson et al. 2005; Gallop et al. 2013), at least during the monitored conditions.

The hydrodynamic model was set at 2.5-m grid using the detailed bathymetric/topographic information collected in the present investigation, forced by the offshore wave conditions observed during the event (in stationary mode) and run until stabilized. Two conditions were tested. In the first experiment, wave forcing was according to the conditions observed during the beginning of the event (i.e. Hs of 1.1 m, Tp of 6.2 s at 10 m water depth and mean wave direction of 350°N). In the second experiment, forcing was according to the wave conditions observed during the most energetic part of the event (i.e. Hs of 1.4 m, Tp of 6.6 s at 10 m water depth and mean wave direction of 350°N). In order to investigate further the effect of wave direction on the nearshore hydrodynamics, two additional experiments were undertaken, forced by waves with wave heights/periods as those recorded at the peak of the event (Hs of 1.4 m and Tp of 6.6 s), but with different directions (315°N and 45°N).

Model results show significant wave attenuation over the reef (Fig. 9) that is in good agreement with the hydrodynamic observations, particularly when the domain complexity is considered. Under the observed wave conditions, significant wave heights inshore of the reef at the instrument deployment site (Figs. 1, 4) were modelled as 0.45 m (Fig. 9), a value close to the field observations (wave heights of 0.5–0.55 m) (Fig. 7). Modelled flows also compared well with the field observations. In the area of field observations, mean currents were simulated to be towards the shore, having speeds of 0.10–0.15 m, whereas recorded currents had very similar directions and mean speeds between 0.09 and 0.13 m (Fig. 10) under the two wave conditions of the event tested. The other two tests (wave forcing with similar conditions to those observed during the event, i.e. Hs = 1.4 m and Tp = 6.6 s, but with wave directions from the NW and NE) also showed generation of complex inshore water circulation which, although very similar in magnitude, also differed to that induced by the northerly waves in terms of the alongshore wave energy distribution and the form/interaction of the inshore current jets (see, e.g. Fig. 9b, d).
Fig. 9

Hydrodynamic model results. a Wave height distribution (in m) and b mean wave-induced flow (vector scale in the inset) under Hs of 1.1 m and Tp of 6.2 s and mean direction of wave approach from 350°N. c Wave height distribution (in m) and d mean wave-induced flow under Hs of 1.4 m and Tp of 6.4 s and mean direction of wave approach from 45°N. e Bathymetry of the study area showing also the location of the ADCP deployment. f Distribution of the mean flow velocity inshore of the reef (distance in m from the ADCP deployment site); results shown are mean flows for the period 1 November–3 November 2014)

Fig. 10

Comparison of the mean flow velocity and direction from the Boussinesq model (circles) and the ADCP observations (triangles) along a cross-shore transect towards the shore (Fig. 1). a ADCP flow averaged over the period 11:00/1 November–05:00/2 November 2014 and model results for significant wave height/period and direction of 1.1 m, 6.2 s and 350°N, respectively (at 10 m water depth), and b ADCP records averaged in the period 05:00/2 November–05:00/3 November 2014 and model results for significant wave height/period and direction of 1.4 m, 6.6 s and 350°N, respectively. The reef is to right of the figures

The model suggests complex distribution of wave energy inshore of the reef. Under all tested conditions, nearshore wave heights show high spatial variability (Fig. 9); there are many areas where wave heights change in short alongshore distances. This variability is due to the complicated architecture of the beachrock reef that results in highly differential wave dissipation as well as wave train interactions. There are two main areas where wave dissipation is substantially decreased: the area behind the main reef inlet (at x about 327200–327500) and an area to west of the instrument deployment (at x about 327750, Fig. 9). In these areas, detailed bathymetric information showed that the beachrock reef is either absent or intermittent and lower in elevation and fronted by increased water depths.

The complexity of the hydrodynamic conditions inshore of the reef is also reflected in the mean flow circulation patterns. Converging and diverging nearshore jets were simulated in all experiments; in some cases, these attained significant velocities (up to 0.3–0.35 m). Water appears to flow over the reef into the nearshore areas, the mean sea level of which was found to increase by up to 0.15 m during the event (at the RBR wave recorder—1.4 m water depth); this piled water seems to be drained through the reef inlets by offshore flows (rip currents) fed by converging longshore currents (Fig. 9).

Under the conditions tested, nearshore flows have been found to be of limited intensity along the beach. Such flows on their own can have only limited effects on nearshore seabed sediment mobility, due to the rather coarse texture of the nearshore seabed sediments (d50s of between 1.8 and 2.7 mm); this explains the limited impact observed along the shoreline during the examined event (Fig. 5c). Nevertheless, mobility/resuspension of the finer fractions of the seabed sediments should take place under different conditions, particularly in the areas exposed to higher nearshore waves as, for example, the area behind the major reef inlet fronting the Xiropotamos River mouth (Fig. 1). This area has been observed on TIMEX images to be occasionally associated with turbid water plumes, even in periods with very limited or cut-off river outflow, as was the case during the field observations when the river mouth was observed to be completely blocked by a beach sand barrier. Despite the absence of riverine supply, a turbid water plume was observed in this area during the most energetic period of the event (Fig. 5b), the shape of which appeared to have been modified by the currents flowing out of the inlet (Fig. 9). Resuspension/winnowing of the finer fractions of the seabed sediments under energetic events might explain the increased grain sizes of the seabed sediments found in this location (Alexandrakis et al. 2013).

Model results show that the area to the west of the beachrock inlet [which is characterized by high long-term shoreline position variability (Fig. 5)] is associated with relatively strong nearshore flows feeding the inlet rip current from the west, as well as increased nearshore wave energy under the conditions tested (Fig. 9). Under more energetic conditions, wave/current interaction in this area might result in nearshore sediment erosion, beach face sediment ‘draw-downs’ and, ultimately, shoreline retreats as those identified through the TIMEX imagery (e.g. at location x3, Fig. 5).

In summary, both hydrodynamic observations and modelling suggest: (a) differential wave dissipation along the reef which is controlled by the elevation and width of the reef crest/flat; (b) generation of a complex circulation regime inshore of the reef consisting of slow converging and/or diverging jets that feed rip currents transporting water offshore, mainly through the reef inlet; and (c) that the wave conditions observed/tested were associated with relatively low cross-shore spatial shoreline variability and beach recovery.

It appears that the beach response to energetic events is complicated. Beach rotation depends not only on the magnitude of the wave energy inputs but also on their sequencing. It may well be that the cumulative effect of sequences of lower, higher steepness waves that can impinge on the coastline without significant dissipation over the reef (Fig. 7) interlaced with more energetic events and the effect of the seasonal, intermittent sediment supply from Xiropotamos River is behind the observed intriguing patterns of beach variability.

5 Discussion

Our results suggest a highly variable shoreline at the eastern part of the reef-fronted Ammoudara Beach. Cross-shore shoreline change during a 10-month period in 2014 was more than 3.1 m along the shoreline; in comparison, maximum tidal shoreline displacement in this area has been estimated as <1.2 m, due to the negligible tides (0.1 m) and high beach face slopes. The difference between the minimum and maximum cross-shore shoreline displacement was found as between 3.1 and 8 m along the monitored shoreline; in some areas, temporary beach losses of up to 6 m relative to the beginning of the monitoring period were observed at a time. These mostly transient changes are of similar magnitude to that suggested by Alexandrakis et al. (2013) to be the recent long-term beach loss in this area (5 m between 2005 and 2012). It appears that long-term, high-resolution monitoring of the beach is required to resolve its long-term morphodynamics.

Adjacent sections of the shoreline showed contrasting patterns of beach loss or gain. Generally, spatial variability appeared increasing in the spring following a sequence of energetic events and the trend was sustained through early summer, in the beginning of the Etesian wind season (a local wind system in the Aegean Sea generating strong winds in the area from the northern sector (Poulos et al. 2013). At the end of the summer, spatial variability started decreasing, with many areas of the shoreline experiencing accretion; our hydrodynamic observations and modelling refer to this accreting period.

The reason behind this temporal (seasonal) change in the beach loss/gain patterns is not straightforward to explain. Unfortunately, long-term directional wave data are not available closer to the shoreline. Comparison of the available information on the characteristics of the waves recorded at the offshore POSEIDON buoy does not provide any clear evidence. Nevertheless, some patterns can be discerned. First, changes in the general patterns of the shoreline spatial variability might be triggered in periods associated with particularly energetic waves from the northern sector (Figs. 5, 6). Secondly, although there have been energetic wind/wave events recorded during summer (the period of the Etesian winds), a reduction in the overall offshore wave energy from the northern sector may be recognized in this period that is accompanied by a small overall increase in the offshore wave steepness. How these changes in the offshore wave regime are related to the observed increases in shoreline variability during spring–early summer (which in some areas was due to increasing beach losses) cannot, however, be ascertained on the basis of the available information. It should be also noted that during winter (January–March) riverine sediments are supplied to the system from the Xiropotamos River (Kontogeorgos 2014); this sediment supply might affect the shoreline variability, which in the winter period appears to be small.

Short- and medium-term beach loss and recovery at the eastern section of Ammoudara Beach appear to be controlled by the architecture of the fronting beachrock reef which influences the alongshore distribution of the wave energy and generates complex nearshore circulation. The reef does not only damp energetic waves very efficiently due to wave breaking on its steep offshore slope (e.g. Pomeroy et al. 2012), but also filters wave energy due to its large, but varying, width allowing energetic waves to reach only certain parts of the shoreline. At the same time, lower frequency waves (0.04 Hz) appeared to become significant inshore of the reef. Mean wave-driven circulation cells were also indicated for the study area, with cross-shore flows becoming more longshore-dominated before exiting the system through morphologically controlled conduits in the reef and, particularly, through the reef inlet (e.g. Van Dongeren et al. 2013).

It has been previously observed that variations in reef architecture can drive alongshore variations in the mode/magnitude of seasonal erosion and accretion at perched beaches (Gallop et al. 2013); our results not only support these observations, but also show that reef architecture can affect beach loss/gain patterns in much finer spatio-temporal scales. These findings may have a bearing for the design of certain erosion adaptation options (e.g. Ranasinghe and Turner 2006; Harley and Ciavola 2013). More detailed assessments are, however, required to evaluate the role and effectiveness of these reefs for natural hazard mitigation that can inform appropriate policies for coastal defence (Temmerman et al. 2013; Ferrario et al. 2014).

The high spatio-temporal variability of the shoreline position of perched beaches may present a significant challenge to the implementation of regulatory regimes designed to moderate the already significant exposure of coastal populations, activities and assets which is likely to increase even more in the future due to the projected mean sea level rise (e.g. Allenbach et al. 2015; Vousdoukas et al., 2016). For example, set back zones are increasingly used as an adaptation tool (see, for example, the Art. 8(2) of the 2009 ICZM Protocol to the Barcelona Convention, http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:L:2009:034:0019:0028:EN:PDF). Their planning and implementation require determination of baselines as well as of historic trends in shoreline evolution; this appears to be a rather difficult exercise for perched beaches (e.g. Gallop et al. 2015a). In addition, climate projections suggest that global sea level may rise on the order of 0.26–0.98 m by 2100 (IPCC 2014). Such changes are likely to impact significantly the flow and sediment dynamics of coastal areas with fringing reefs. Recent studies suggest that an increase in water depth of this order on an 1–2-m-deep, fringing reef crest/flat would result in higher inshore waves and set-up, increased water depths and near-bed shear stresses which, in turn, would result in increases in the size/quantity of resuspended sediments and of those eroded from the adjacent shoreline (e.g. Storlazzi et al. 2011).

The automated approach developed to extract shoreline positions from TIMEX images showed to be an efficient tool in resolving beach variability in fine spatio-temporal scales. The procedure is fast, processing large numbers of video imagery mosaics in negligible time, and its results were found to compare well with those obtained from a manual shoreline detection procedure. It should be, however, noted that the criterion used to define the shoreline, i.e. that the detector kernel grows along the line representing the weighted mean of the most inshore high-intensity zone formed by the wave-generated foam, may generate some uncertainty. The method identifies a feature that under low-energy conditions is related to the swash zone, whereas under high-energy conditions might also include the most inshore wave breaking zone; thus, the method may overshoot seaward the shoreline position under high-energy conditions. This tendency is supported by the results of the comparison between the automated and manual shoreline detections which showed that although the automated detection was more robust, it was also more conservative in recording beach loss (Fig. 3).

6 Conclusions

A high spatio-temporal resolution investigation of the shoreline variability of the perched Ammoudara Beach using video imagery showed that the fronting beachrock reef exerts significant geological control on beach morphodynamics. Cross-shore shoreline change during the 10-month monitoring was, in some areas, up to 8 m. Adjacent sections of the shoreline showed contrasting patterns of beach loss or gain and spatial variability increased in spring/early summer, stabilized until the end of the summer when partial beach recovery commenced. It appears that in order to resolve the morphodynamics of perched (and not only) beaches, information on shoreline retreat and advance is required with a spatio-temporal resolution to match that of the shoreline change.

Association of these intriguing patterns of beach change with wave forcing (as recorded at an offshore wave buoy) is not straightforward, with the only discernible trends being as follows. Changes in the general patterns of the shoreline variability might be triggered in periods of particularly energetic waves from the northern sector and that the period of increasing spatial variability might be correlated with a period of a small overall increase in the offshore wave steepness.

Hydrodynamic modelling and observations showed that the reef can filter wave energy in a highly differential manner depending on its local architecture. In some areas, the reef allows only low-energy waves to impinge on the shoreline, whereas elsewhere penetration of higher waves is facilitated by the low elevation and limited width of the reef or by the presence of an inlet. Wave/reef interaction can also generate complex circulation patterns, including rip currents that appeared to be also constrained by the reef architecture.

Acknowledgments

This work was supported by the projects BEACHTOUR (11SYN-8-1466) of the Operational Programme ‘Cooperation 2011, Competitiveness and Entrepreneurship’ and AKTAIA (09SYN-31-711) of the Operational Programme ‘Operation 2009—Partnerships of Production and Research’ co-funded by the European Regional Development Fund and the Greek State. NCMR and HNMS are thanked for the provision of wave and wind data. M.I. Vousdoukas acknowledges funding from the European Union Seventh Framework Programme FP7/2007–2013 under Grant Agreement No. 603864 (HELIX: ‘High-End cLimate Impacts and eXtremes’; www.helixclimate.eu), as well as by the JRC institutional project Coastalrisk. The authors would like to thank G. Alexandrakis, S. Poulos and S. Petrakis for their help during fieldwork as well K. Chatzikonstantinou for his assistance in the generation of Fig. 1. The two anonymous reviewers, whose comments improved substantially the quality of the manuscript, are gratefully acknowledged.

Copyright information

© Springer Science+Business Media Dordrecht 2016

Authors and Affiliations

  • A. F. Velegrakis
    • 1
  • V. Trygonis
    • 1
  • A. E. Chatzipavlis
    • 1
  • Th. Karambas
    • 4
  • M. I. Vousdoukas
    • 1
    • 2
  • G. Ghionis
    • 3
  • I. N. Monioudi
    • 1
  • Th. Hasiotis
    • 1
  • O. Andreadis
    • 1
  • F. Psarros
    • 1
  1. 1.Department of Marine SciencesUniversity of the AegeanMytileneGreece
  2. 2.European Commission, Joint Research Centre (JRC), Institute for Environment and Sustainability (IES)Climate Risk Management UnitIspraItaly
  3. 3.Faculty of Geology and GeoenvironmentNational and Kapodistrian University of AthensAthensGreece
  4. 4.Department of Civil EngineeringAristotle University of ThessalonikiThessalonikiGreece