3.1. Wave and Hydrodynamic Measurements
In panels a–c of
Figure 4,
Figure 5 and
Figure 6, three wave parameters—significant wave height (
), Peak wave period (
), and peak wave direction (
)—are plotted for selected periods (T1, T2 and T3) of Exp#1 and Exp#2. The wave parameters were estimated based on the wave spectra calculated from the burst data, which gave averaged value per burst. The time is set zero at December 20, 2016 for Exp#1 and
Figure 4 shows the time variation of the wave data for T1. In this period, the data in L1 are characterized with the distinctive periods of storm waves (
Figure 4). Three extreme events (t = 3–6, 8–10) with
> 3 m in L1 were clearly identified. The wave period also varied similarly with the wave height as it increased with increasing height. Specifically,
became maximum reaching to 15 s at t = 5–6, corresponding to the high
in this time. The wave propagation direction shows different pattern as the waves generally approached the shore in the normal direction (
= 0) during the storm periods except for the sharp changes in t = 2–3. In
Figure 5, the same wave data are plotted for T2 during which another storm wave attacked the beach with
> 3 for longer time (t = 26–30) with wave periods of ~10 s and with nearly normal wave direction. For Exp#2, the initial time was set at 21 December 2018, and
Figure 6 shows the three wave parameters for T3. In L3, the wave condition was less distinctive compared to that in L1 of Exp#1 as
was significantly lower than those in L1, indicating that the wave condition was milder during Exp#2, and storm wave events in which
exceeded 2 m over a period longer than 1 day did not occur (
Figure 6a).
In
Figure 4, the temporal variations of the five parameters during T1 of Exp#1 are compared between L1 and L2. As shown in
Figure 4d,
sharply increased during the two storm periods in t = 3–6 and 8–9 in both locations, but
in L2 was lower than that in L1 likely because the wave orbital velocity magnitude decreased with decreasing water depth. It might be also because the wave energy in L3 decreased due to the wave breaking in L2, though there was no clear indication of the breaking during these storms. It should be noted that
in L2 sharply increased at t = 3–4 in the early stage of the first storm when
remained close to zero in L2 (
Figure 4e), which indicates that strong onshore current developed in L2 when the storm developed. In L1, the onshore current was not observed. Instead, longshore current developed to the right of the beach at t = 3 (
Figure 4f). Once the storm was fully developed at t = 4–5, the direction of the current in L1 rapidly changed as it
became negative and
changed to positive, showing that the current flowed offshore to the left of the beach. It should be also noted that
of cross-shore velocity shows opposite pattern between L1 and L2 as it became positive in L1 but negative in L2 (
Figure 4g). As analyzed in the previous study [
19], the high positive
in L1 likely increased the bed stress to cause the seabed erosion in this time. The negative
in L2 could be understood that the wave steepness increased at L1 due to shoaling of storm waves decreased by wave breaking when they reached to L2. Contrary to the significant changes in
,
remained close to zero in both L1 and L2. This low magnitude in
during extreme wave condition was not clearly understood as the wave asymmetry usually increases under breaking wave condition (
Figure 4h).
Once the storm period ended at t = ~6, the statistics of data in L2 was not available due to contamination until t = ~8 when the second storm waves occurred for shorter time period until t = ~9. The flow pattern was different from that in the first storm as
in L2 was lower during this second storm period while
in L1 was similar between them (
Figure 4d). The direction of the cross-shore current was also different as
in L2 directed onshore. The most significant difference can be found in
as it remained positive not only in L2 but also in L1, which indicates that the wave shoaling might be dominant instead of breaking in both locations (
Figure 4g). In addition,
shows clear difference between L1 and L2 at t = 8–9 as the magnitude of
increased in L2 as it was dispersed in wider range, compared to that in L2 where
remained close to zero similarly to that observed in t = 3–5 (
Figure 4h). Since the positive/negative sign of
means that the waves were tilted backward/forward [
27], the higher
during the second storm indicates that the waves were deformed when they approached from L1 to L2. Another period examined in
Figure 4d is t = 12–18 when the wave energy became low after the storms.
in L2 shows more severe fluctuations compared to that in L1. It should be also noted that both
and
in L2 show higher dispersion around zero that led to their higher magnitude compared to those in L1 (
Figure 4g,h). Specifically,
in L2 tended to be positive, indicating that the waves became steeper as they approached the shore from L1. For
, this tendency was not found as negative values were also observed, showing that the waves were tilted but not necessarily forward.
In
Figure 5, the same five parameters are plotted but for t = 24–32 (T2) during which the third storm had lower
compared to the previous two storms developed for longer period (t = 26–29). In this storm period,
increased to ~0.5 m/s in both L1 and L2. Interestingly, a strong onshore current was observed in L2 while no cross-shore currents were detected in L1 as
~ 0 during the storm period. Difference between L1 and L2 was also observed in the longshore current as
= −0.2 m/s in L1 at t = 26 but
~ 0 m/s in L2 at the same time. On the contrary,
~ 0 m/s in L1 at t = 29 while it increased to ~0.2 m/s in L2. Except for the storm period,
and
were close to zero in both L1 and L2. As for the skewness,
slightly increased in L1 during the storm as its maximum value reached to ~0.25 at t = 26 while
~0 before and after the storm period. In L2,
sharply increased to ~0.5 just before the storm at t = 25–26. During the storm period, however, it became negative as it ranged between −0.1 and −0.2 at t = 26–29. Therefore, the pattern of positive skewness in L1 and negative skewness in L2 observed in the first storm (t = 3–5) repeated in the third storm even though its magnitude was lower. Except for the storm period,
was remained close to zero in L1 but it was dispersed around zero in L2, increasing its magnitude, as also shown in
Figure 4. The dispersion in L1 outside the storm period was more clearly observed in the wave asymmetry. During the storm period,
was close to zero in both L1 and L2, which was similarly observed during the first storm (t = 3–6 in
Figure 4). Except for the storm, however,
was strongly dispersed and its magnitude increased in L2 while it remained close to zero in L1.
The characteristic pattern of
is also clearly observed in Exp#2 as shown in
Figure 6 in which the five hydrodynamic parameters are compared for T3 (December 21–28, 2018). In this time, a period of high waves when
reached ~2 m was observed at t = 2–3 (
Figure 6a). Correspondingly,
increased to reach ~0.5 m/s in both L3 and L4. In case of currents, however, difference was found between L3 and L4 as strong cross-shore and longshore currents were observed at t = 2–3 in L4 while their magnitude was significantly smaller in L3. During the high wave period, the positive/negative
in the outer/inner measurement location that was observed in Exp#1 was not detected, but the stronger scattering of
in L4, the shallower location closer to the shore, was repeated. As for wave asymmetry,
was significantly reduced during the high wave period in both L3 and L4 as
~ 0. However, it increased with strong dispersion around zero at other times, as also observed in Exp#1.
3.2. Shoreline Positions Measured by VMS
The shoreline changes measured by VMS are analyzed based on the hydrodynamic measurements.
Figure 7 shows the shorelines measured on four different times around Exp#1, namely 19 December 2016; 29 December 2016; 10 January 2017; 19 January 2017. The shore was retreated between 19 December 2016 and 29 December 2016 in the majority of the sectors along the beach. This data can be compared with the hydrodynamic data in
Figure 4 (t = – 1–9) in which two severe storms attacked the beach with strong nearshore currents and high wave nonlinearity, indicating that destructive forces might be dominant. However, there were locations where the shoreline was advanced. Specifically, the shoreline in the middle of the beach was significantly advanced during the storm. Comparing the satellite image in
Figure 3, this location corresponded to the protruded area connected to the horn of the crescentic sandbar. Therefore, the shoreline accretion can be interpreted as the wave energy was reduced over the shallow horn area due to enhanced wave breaking while the wave energy at other areas was relatively higher and the sediments were transported to be cumulated in the horn area.
Figure 7 also shows the shorelines measured by VMS at 29 December 2016 and 10 January 2017. This period corresponds to the time at t = 9–18 in
Figure 4 when the wave energy became much lower without storm event. The shoreline change also shows no clear pattern of retreat or accretion. However, the northern part of the protruded area (left side in the figure) was eroded while the other side of it was advanced, which was likely related to the oblique wave incidence during t = 12–17 as shown in
Figure 4b. The shorelines measured at January 10 and 19, 2017 shows the shoreline change pattern corresponding to the period of T2 of Exp#1. Interestingly, the shoreline was generally advanced in most of the sectors along the shore. The accretion of the shoreline in this period could not be understood considering the high wave conditions at t = 26–29 was considered (
Figure 5). It may be related to the abnormally strong onshore currents observed in L2 at t = 26–29 (
Figure 5b). However, the hydrodynamic condition in this period was similar to that observed at t = 3–6 when the storm waves attacked the shore, but the response of shoreline was opposite as it was advanced at t = 26–29 while it was retreated at t = 3–6, which may require further investigation.
The shoreline change observed by VMS in Exp#2 is shown in
Figure 8 where the shoreline positions are compared between 20 December 2018 and 02 January 2019 as it corresponds to the period in
Figure 6. It should be noted that the shoreline feature was clearly different from those observed in Exp#1, which indicates that the shoreline had been actively changed during the ~1-year period between the two field experiments. It is also interesting to find out that the curved shoreline at 20 December 2018 became flattened at 2 January 2019 as the protruded areas were eroded while the caved areas were filled with the eroded sediments. Specifically, the eroded sediments moved south to fill the caved areas, which was likely due to the obliquity of the incident waves as it can be presumed from the wave direction in the corresponding period (
Figure 6b).
3.3. Seabed Movement Detected by Satellite Images
The Sentinel-II data in
Figure 3 were plotted in RGB colors. In order to enhance the visibility of the underwater crescentic sandbars, the image was processed by applying the algorithm developed in [
21] by which the position of nearshore sandbar crests was extracted from Sentinel-II images. In this study, the same algorithm was applied for the detection of sandbar positions. In this approach, calibrations based on in-situ measurements were not necessary to detect the positions of underwater sandbars. In addition, corrections of absolute values between different satellite images that might have different reflectance were not necessary either because the purpose was to extract the shape of response over underwater profiles perpendicular to the shoreline.
The readers are referred to [
21] for the details of the sandbar extraction algorithm from Sentinel-II images, and it is only briefly summarized here. The first step was to extract the shoreline using a simple threshold applied to the short-wave infrared (SWIR) band because the spectral response of the water in the SWIR domain is almost equal to zero in low to moderate turbid waters from the resampling domain of 10 m spatial resolution. Then, a new raster,
, was defined by multiplying all visible bands in order to augment the increases in the spectral response over sandbars, such that
where
,
, and
represent the bands of red, green and blue respectively. Once
distribution was calculated, a network of profiles perpendicular to the shoreline is created. The profiles started from the estimated shoreline, and the distance between two adjacent profiles was set to 10 m, which used total 210 profiles in this study. Along each profile, an exponential model,
, was fit to the profile where coefficients
,
and
were computed for each profile. After that, normalization of the profile was performed by subtracting the exponential model from the original profile. The shape of sandbar was then extracted along the profile using moving window.
Figure 9 shows the converted Sentinel-II image of
on 11 December 2016. In the water, the white dots denote the locations of the crest of sandbars as they were selected along each profile, and they generally correspond to the positions of sandbars expressed by the
distribution. It shows that five crescentic sandbars of similar size were developed along the shore in the shallow nearshore area of the beach at depth of 3–5 m (the depths of bars were observed from the field measurements, not from satellite images). In the southeast part of the beach, sandbars could be hardly formed due to the existence underwater rocks. The horns of the bars (area where the bar is peaked toward the land) extended to the land as they were connected to the shore, on which the waves were breaking.
The image in
Figure 9 was taken on 11 December 2016, nine days before the field experiment started on 20 December 2016. Although the Sentinel-II images are available for interval of 5–10 days, the sandbar data in Hujeong Beach were available on clear days only when the visible light could go through the water column to detect the seabed topography. For this reason, the next Sentinel-II image available in this site was measured on 31 December 2016, after the attacks of the two previous storms. The third Sentinel-II image available in the site was on 31 January 2017. Applying the same algorithm, the sandbar crest positions were extracted from the latter two images.
Figure 10 compares the locations of sandbar crest extracted following the algorithm by [
21] between the three Sentinel-II images taken on 11 December (red dots), 31 December 2016 (blue dots), and 31 January 2017 (green dots). The bar crest locations of December 11 and 31 show that some parts of sandbars had moved offshore for maximum 70 m, which indicates the offshore sediment transport as it might be caused by the high storm waves that attacked the area on 23–25 December and 28–29 December 2016. The offshore movement of the sandbars corresponds to the results in the previous study [
19] that reported a severe seabed erosion in L1 during the storm period on 23–25 December. It is also closely related to the hydrodynamic conditions in
Figure 4 as it shows that the magnitude of wave skewness and asymmetry increased during the two storm periods. This indicates that the waves were likely breaking in L2 during the storms and the destructive forces such as undertow currents could become stronger to migrate the sandbars offshore. Once the wave condition became milder after the severe storms, the movement of the seabed contours was reduced. The sandbar crests measured on 30 January 2017 shows no clear difference with that on 31 December 2016, indicating that the seabed became stable without severe deformation though high wave conditions were observed during the period.
In
Figure 11, the sandbar crest positions are compared between the Sentinel-II images taken on two days, 26 November 2018 (blue lines), 26 December 2018 (red lines) by applying the same algorithm. The two times were chosen at the dates when the Sentinel-II images were available around Exp#2. The figure shows that, though there was a one-month time gap between the two satellite images, the locations of seabed contours were similar indicating that the seabed morphology was stable during the period, regardless of the high wave conditions observed in
Figure 6.