Monitoring of Recovery Process at Yeongildae Beach, South Korea, Using a Video System

Featured Application

The video monitoring system used in this study can be applied for monitoring coastal processes of erosion and recovery and thus can be useful in designing beach protection/prevention plans for damage by storm waves.

Abstract

Once a beach is eroded by storm waves, it is generally recovered under milder wave conditions. To prevent or reduce damage, it is therefore important to understand the characteristics of the site-specific recovery process. Here, we present the results, based on a data set from a video monitoring system and wave measurements, of the recovery process in a pocketed beach located inside a bay where the shoreline retreated harshly (~12 m, on average, of beach width) during Typhoon TAPAH (T1917) in September 2019. It took about 1.5 years for the beach to be recovered to the level before the typhoon. During this period, the erosion and accretion were repeated, with the pattern highly related to the wave power ($P_w$); most of the erosion occurred when $P_w$ became greater than 30 kWatt/m, whereas the accretion prevailed when $P_w$ was no greater than 10 kWatt/m. The recovery pattern showed discrepancies between different parts of the beach. The erosion during storm events was most severe in the southern part, whereas the northern shoreline did not significantly change even during TAPAH (T1917). In contrast, the recovery process occurred almost equally at all locations. This discrepancy in the erosion/accretion process was likely due to human intervention, as a shadow zone was formed in the northern end due to the breakwaters, causing disequilibrium in the sediment transport gradient along the shore. The results in this study could be applied in designing the protection plans from severe wave attacks by effectively estimating the size of coastal structures and by correctly arranging the horizontal placement of such interventions or beach nourishment. Although the application of these results should be confined to this specific site, the method using wave energy parameters as criteria can be considered in other areas with similar environments, for future planning of beach protection.

1. Introduction

Beaches are important for nearby communities, as they provide areas for recreational activities. They are sources of tourism, and thus well-groomed beaches with a nice coastal environment attract many visitors, even from abroad. Therefore, maintaining these beaches in good condition is necessary to improve their value. Apart from economic reasons, beaches are also environmentally important because they are buffer zones between the ocean and land, protecting on-land facilities by mitigating wave energy. Because of this buffer role, beaches themselves are exposed to attacks of extreme waves under storms such as tropical cyclones. When the wave energy is released in the surf zone due to the breaking of waves, sediments in the beach face and the nearshore seabed are dynamically transported. Therefore, one of the serious types of damage as a result of storm attacks is beach erosion.

Under energetic wave conditions, nearshore sediments can move offshore rapidly due to wave-induced underwater currents [1], and the shoreline can significantly retreat, even for a short time of hours or days [2,3]. The volume of erosion is directly related to the maximum wave height during the storm, as the near-bed shear stress initiates sediment motion proportional to the square of the wave height, and more sediment moves under higher shear stress once it exceeds a critical value. In addition, the wave period that contributes to the power of storm waves and the wave direction that influences the distribution of wave energy along the shore are important factors in the amount of eroded sediment. In addition to the wave parameters, the duration of a storm [4] or a sequence of consecutive storms [5,6] could play a significant role in the erosional process.

Once a storm event ends, onshore sediment motion prevails, under mild wave conditions, due to wave nonlinearity such as skewness and asymmetry [7], which results in the natural process of shoreline recovery following storm events. Therefore, studies on the recovery process after storms are as important as studies on the erosional process during storms [8]. The time scales of the post-storm nearshore morphological recovery are varying, from days [5,8] to months [9]. However, large storms can cause rapid erosion locally, from which recovery may take many years or even decades if the impacts are sufficiently large [10]. In extreme cases, the damage can be irreversible if the coastal structures that are built to protect the shore are destroyed by the storms [11].

In the recovery process, the interval between multiple storms is important, as it is inversely related to the storm damage. If the interval between two consecutive storms is less than the recovery time of the beach, the next storm could cause greater damage to the beach [12]. For example, storm clusters with small return periods could impose similar erosion impacts to a single storm that had a much longer return period (i.e., much greater storm). If a beach reached equilibrium after the attack of the first storm, additional erosion would occur only when the intensity of the following storms exceeded that of the first one [5]. Therefore, the recovery after storm events is a complex process, not only depending on the post-storm recovery time but also depending on the intensity of the following storms.

The studies on beach recovery after storms have used various approaches. One of the traditional methods is the direct measurement of the elevation and water depth along the lines set perpendicular to the coastline, which is called profiling. The profiling would be useful for long-term monitoring for beach processes with a low initial cost [13]. The profile data could be also used to validate simulation results calculated from numerical models [14]. The model approach is usefully applied for the prediction of beach recovery for either short-term or long-term processes. Short-term prediction models such as XBeach [15] calculate the rapid changes in the seabed and dunes during storms and the post-storm recovery process in the time scale of months [9]. Due to the high computational cost, however, the application of process models is still restricted for year-scale predictions. For such long-term predictions, line models [16] based on the one-line theory [17] or statistical models such as Monte Carlo have been applied [18].

Another tool that is useful for the long-term observation of the beach process is remote sensing. The satellite imagery can be regularly collected over a long period, and it has been used to derive shorelines and identify their spatial patterns [19,20]. However, it is sensitive to cloud cover and has a limited temporal resolution. Similarly, aerial photos are useful to detect geographical changes over decades, although they cannot provide detailed information on detected variation [21]. On the other hand, airborne and in situ LiDAR also provides detailed information on the beach geography, for comparison and validation of other measurements [22,23]. Despite their advantages, the application of aerial photos and LiDAR data is restricted in measuring the beach recovery process continuously due to the high cost. For this reason, Video Monitoring Systems (VMSs) installed at beaches have been successfully applied to observe the recovery process in specific sites [8,9]. The VMS is not only useful for long-term monitoring [24] but also for short-term observation of the recovery process after storms [5] and for estimation of the wave parameters [25] as well as wave-induced currents [26].

In this study, we also investigated a recovery process after an attack of storm waves, when a severe erosion occurred, by analyzing the shoreline variations in a pocketed beach located inside a bay in the southeast coast of Korea. Our study focused on the conditions that determined the erosional or depositional process at the beach using data sets from a VMS and wave measurements. As previously described, it is a natural process that beaches undergo erosion by attacks of storm waves, and then they are slowly recovered under milder conditions. However, the criteria that decide the shoreline erosion and accretion are still unclear. Although such information obtained at a beach is site-specific and thus cannot be generally applied for other beaches, the analytical methods may still be applied to beaches with different conditions. In addition, if a criterion is provided for a specific beach, then plans can be designed to protect the beach from severe damage that can lead to a loss of its value.

Another focus was examination of the locality in the recovery pattern between different locations along the beach. Previous studies using the VMS generally focused on the time scale of the recovery process [8]. However, it is also important to analyze the discrepancy in the recovery process between different locations, if it exists, because such disequilibrium may lead to the concentration of damage at specific locations if subsequent storms attack the site before fully recovered. The results of the present study are, therefore, useful not only to understand the criteria for erosion/accretion processes at the study site, but also to analyze the causes of disequilibrium during the recovery process. These outcomes are then applied to suggest measures for conservation of the beach, considering the characteristic erosion and recovery pattern, generally or locally.

The paper is organized as follows. The information of the study site and the damage as a result of Typhoon TAPAH (T1917) are described in Section 2.1. The VMS used in this study and its data are introduced in Section 2.2. The general trend of the beach process is described in Section 3.1, and the locality (discrepancies in the process between areas of the beach) is analyzed in Section 3.2. The discussion on the results is provided in Section 4, and the conclusion of the study in Section 5.

2. Materials and Methods

2.1. Study Sites

Yeongildae Beach is in the southeast of the Korean Peninsula (Figure 1a), inside Yeongil Bay, which faces northeast with a mouth width of ~10 km (Figure 1b). Due to its location in the west corner of the bay, the beach has been protected from severe erosion because the wave energy is generally attenuated when reaching the site. Yeongildae Beach is a ~1.7 km long sandy beach where the sediment size ranges from about 0.15 to 0.35 mm. It is a pocketed beach facing ESE (East-Southeast), bounded by two ports—Pohang Port at the southern end and Duho Port at the northern end. Due to the breakwaters that were built for these ports, the sediments were expected to remain within the coastal cell without loss in the longshore direction. However, the beach has been classified as an erosive beach because input sources of sediment to the beach have been lost due to the construction of the ports. In addition, after the breakwaters of Duho Port were constructed, a shadow zone could have been formed in the lee area of the breakwaters, reducing the wave energy by wave diffraction [27]. The shadow zone has caused redistribution of sediments along the beach shore so that the shoreline width has become narrower near the Yeongil Bridge area but thicker at both ends of the beach near the two ports (Figure 1b). The aerial photograph and satellite image in Figure 2 compare the shorelines measured in 1977 (before the construction of the Duho Port breakwaters) and in 2019 (after construction). It shows that the shoreline severely retreated in the middle of the beach, whereas it accredited at both ends in 2019, compared to 1977, indicating the redistribution of sediment to reach an equilibrium by the changed wave energy field after the construction of Duho Port.

The wave data were measured at two locations, M1 and M2, as shown in Figure 3a. At M1, a pressure transducer was moored at the bottom, at an 8.5 m depth, ~1 km away from the shore of the Yeongildae Beach, measuring the wave height ($H_s$) and period ($T_p$). Because the pressure transducer could not provide information on the wave propagation ($D_p$), it was measured at M2, where an Acoustic Wave and Current Profiler (AWAC) was moored at the bottom, at a 21.6 m depth, located in nearly the center of Yeongil Bay, ~7 km away from Yeongildae Beach. The tidal range in the study area was no greater than 0.3 m, and the tide-induced current was no greater than 0.2 m/s. Therefore, the effect by the tide on the beach processes may not be significant, and thus was not considered in this study.

Figure 3b shows a wave rose diagram that represents the wave conditions at M2. Inside Yeongil Bay, the waves mainly propagate from the NE direction, which is similar to the wave propagation outside the bay. Wave conditions in M2 are moderate because the significant wave heights ($H_s$) are generally less than 2.5 m, except for several extreme storm cases. The dominant wave directions in M2 are NE and NNE, which indicates that the wave propagates inside the Yeongil Bay generally in the normal direction to the mouth of the bay.

Due to the geographical characteristic of Yeongildae Beach being located at the northwest end, inside Yeongil Bay (Figure 1b), the impact of storm waves has not been significant compared to other beaches located outside the bay. In September 2019, however, Typhoon TAPAH (T1917) attacked the study site. Although TAPAH (T1917) was a Category 1 typhoon (max. wind speed 35 m/s), it caused serious damage to many beaches located on the southeast coast of Korea due to the proximity of its path (Figure 4). The damage was also significant at Yeongildae Beach because the beach face was rapidly eroded during the attack of TAPAH (T1917), which provided the motivation of the present study to investigate the characteristic pattern of its recovery process.

2.2. Video Monitoring System

At Yeongildae Beach, two VMSs were constructed at S1 and S2, as shown in Figure 1c. The monitoring by the VMS was initiated in November 2018 and in February 2019 at S1 and S2, respectively. The VMSs were built at the top of buildings in both locations to save the cost of constructing new towers to mount the VMSs. At S1, four video cameras were mounted to cover the central and southern part of the beach, and two additional cameras were mounted at S2 to cover the resting northern end of the beach.

The VMS consists of the cameras, camera controller, data processing, and data transfer systems that send processed data (averaged images) to the main server. Every 30 min from 7:00 a.m. to sunset, each camera took snapshot pictures for 3 min every half-second. In Figure 5a,b, examples of the snapshot images taken at S1 and S2 are shown. The 180 snapshot images collected for 3 min were then averaged to produce an ‘averaged image’; in this way, the averaged images were provided twice per hour. The averaged images stored in the controller were transferred to the main server one time per day in the nighttime, when the internet network was not busy. The snapshot images were not saved in the system unless specified to do so, to save the cost of running the VMS. Once the averaged-image data were transferred to the main server, the locations of the images were corrected in the x-y coordinate system. Because the images were obliquely measured with angles that were different between the cameras, the data needed to be converted into orthogonal images to the ground. After this, the images from each camera were combined into one-image data that covered the whole beach. Figure 5c shows the data in which the orthogonal images are combined into one.

The shoreline positions and beach width were estimated from the image data of the whole beach. For this process, a set of baselines needed to be established, as shown in Figure 5d. The baselines started from the inner line set, along the landward end of the backshore of the beach. Each baseline was set perpendicular to the inner line to be ended in the water so that the position of the shore could be determined manually by comparing the color of the pixels along the baseline. Once the shoreline positions were determined, the beach width could be calculated by connecting the shoreline positions of a baseline to that of the adjacent baselines. Similarly, the position where a baseline crossed the inner line could be connected to the adjacent positions to form a polygon, as marked with the blue solid lines in Figure 5. The beach width was then obtained by calculating the area inside the polygon. To measure the time variation of the beach width, the inner line and baselines were fixed once determined. In the case of Yeongildae Beach, a total of 34 baselines were established.

3. Results

3.1. General Pattern of Shoreline Recovery

In Figure 6, time variations of the four parameters measured in the site are compared for about 33 months, since the initiation of the VMS measurement in November 2018. The significant wave height, $H_{m0}$, and the significant wave period, $T_s$, were measured at M1 (Figure 3). The time variations of $H_{m0}$ and $T_s$ in Figure 6a show complex patterns, as they highly fluctuate, which may cause difficulties in understanding the impact of these wave parameters on the shoreline change. Therefore, another parameter, wave power ($P_w=\frac{\rho g^2}{64\pi }{H}_{m0}^2{T}_s$) was calculated (Figure 6b). $P_w$ is an index that directly measures the wave energy flux and can be used as a parameter that combines the effects of the wave height and period. The time variation of $P_w$ is more clearly distinct from those of $H_{m0}$ and $T_s$. Its magnitude is usually less than 10 kWatt/m, with an average value over the 33 months of ~0.9 kWatt/m. However, there were times when $P_w$ sharply increased, exceeding 20 kWatt/m, with its maximum value reaching ~75 kWatt/m.

The mean beach width, 〈$y$〉, was estimated by integrating the beach width, $y$, along each baseline, from baseline #1 to #34. Here, the $\mathrm{y}$ magnitude of each baseline was obtained by subtracting the average value during the first 2 years from the original VMS beach width (i.e., y is the average removed beach width). As shown in Figure 6d, the beach was significantly eroded when TAPAH (T1917) attacked the site in September 2019. During the period of ~2 days, 〈$y$〉 was reduced by ~12 m. In this study, the time of 23 September 2019 was set as T1, as it was the initial time for the erosion to occur due to the typhoon. Since the severe erosion, the beach width hardly recovered, as 〈$y$〉 could not reach the original value in T1 (marked with the dashed magenta line in Figure 6d) until 1 March 2021, and this time was set as T3 as an indication of the time of shoreline recovery, although 〈$y$〉 touched the original value only for a short time at T3 and decreased again sharply after that. The time variation of 〈$y$〉 from T1 and T3 shows an interesting pattern. Once the severe erosion occurred in T1, 〈$y$〉 tended to increase slowly (i.e., with much lower speed than the erosion in T1). However, there were times when 〈$y$〉 decreased back to the level when the most severe erosion occurred in T1. This accretion and erosion pattern was repeated until 5 October 2020 (this time was set as T2). After T2, 〈$y$〉 continued to increase gradually (without severe erosion) until T3, when 〈$y$〉 recovered to the level before the erosion in T1. Once it reached the maximum level in T3, 〈$y$〉 decreased again, and the gradually increasing process stopped. After this, 〈$y$〉 remained level, without showing a clear pattern of erosion/accretion until July 2021, the time of the most recent data.

The pattern of 〈$y$〉 shows a high correlation with that of $P_w$, especially when the erosions occurred from T1 to T3. To increase visibility, four red rectangles are marked to show the time periods when 〈$y$〉 and $P_w$ are significantly correlated. In these four periods, the beach width showed a significant decrease (or the increasing trend in 〈$y$〉 was halted, as shown in T3), and the maximum magnitude of $P_w$ was observed to exceed 20 kWatt/m, which was significantly higher than observed at other times. In addition to the wave power, the amplitude of the near-bed velocities, $U_m$, was estimated using the formula ($U_m=0.0116H_{m0}/T_s$) by Soulsby (1986) [28]. $U_m$ could be an important factor for sediment motion because cross-shore sediment fluxes are calculated from bed shear stresses that are based on near-bed velocities [29]. As shown in Figure 6c, the time variation pattern of $U_m$ was similar to that of $P_w$; hence, $U_m$ increasing with increasing $P_w$ might trigger the sediment motions. However, the correlation between 〈$y$〉 and $U_m$ is less relevant compared to that between 〈$y$〉 and $P_w$; that is, $U_m$ peaked on 3 September 2020, not on 23 September 2019, when 〈$y$〉 declined considerably. Furthermore, $U_m$ was an indirect parameter estimated from the empirical formula using the measured wave heights and periods from the wave sensor. Therefore, it may not be appropriate to apply $U_m$ to directly measure the erosion/accretion processes in the beach face. In contrast, $P_w$ was directly estimated from the wave measurements and thus can be suggested as a controlling indicator in this study.

The correlation between the beach width and wave power can be more clearly investigated in Figure 7, in which the time variation of $P_w$ is also compared with that of 〈$y$〉 for the same observation period. To increase the visibility, however, the range of $P_w$ in the y-axis is adjusted so that its upper limit is set to 30 kWatt/m (Figure 7a). In addition, four green rectangles are added in the figure to mark the times when the magnitude of 〈$y$〉 increased (i.e., when the shoreline was advanced). Red/green rectangles are marked over the whole observational period of 33 months so that the erosion/accretion trend can be examined even during the time before Typhoon TAPAH (T1917). Compared to the pattern in the red rectangles, the times of shoreline accretion are also closely related to the wave power in the green rectangles. The maximum magnitude of $P_w$ was generally higher than 20 kWatt/m in the periods of five red rectangles, whereas it was less than 10 kWatt/m in the periods of four green rectangles. In particular, at four out of five periods of the red rectangles, the maximum magnitude of $P_w$ exceeded 30 kWatt/m and could reach up to ~75 kWatt/m. During the longest time of the gradual shoreline accretion for ~4 months, which started from T2, the maximum magnitude of $P_w$ was generally lower than 10 kWatt/m. This indicates that a continuation of low-powered wave conditions is necessary for a recovery process to occur consistently. In contrast, the erosion process occurred in relatively shorter periods, when high-powered waves attacked the site. Specifically, it is noted that the gradual increase of 〈$y$〉 stopped in T3, when other high-powered waves with $P_w$ of ~50 kWatt/m (Figure 6b) attacked the site, which indicates that the high-powered waves not only caused the rapid erosion but also could change the accretion trend and bring it to a halt.

Following the time variation of the total mean beach width 〈$y$〉 in Figure 7b, c displays the time variations of the beach widths in three different parts of the shoreline. In the figure, the solid blue line shows the time variation of averaged beach width for the baselines #2–#5, which is symbolled as ${\langle y\rangle }_s$ as it represents the pattern in the southern end area. Similarly, the orange line, ${\langle y\rangle }_m$, represents the variation in the middle of the beach as averaged for the baselines #15–#17. The yellow line, ${\langle y\rangle }_n$, represents the variation in the northern end as averaged for the baselines #30–#33. The variations of each group of beach widths show different changes from time to time. However, all the time variations of each group of beach widths follow the trend of the total mean beach width 〈$y$〉 in T1, T2, and T3. That is, ${\langle y\rangle }_s$, ${\langle y\rangle }_m$, and ${\langle y\rangle }_n$ started to decline rapidly at T1 and formed troughs around T2, and then they all gradually increased until around T3.

3.2. Locality in the Erosion/Accretion Process

The time variation in Figure 6 and Figure 7 shows the general pattern of the erosion and recovery processes of the averaged beach width of the 34 baselines. As described in Section 2.1 with Figure 1c and Figure 2, however, the beach widths show high locality—i.e., a discrepancy in $\mathrm{y}$ between different baselines. Due to the construction of Duho Port, a shadow zone was formed in the area near the baselines #29–#34 (Figure 5), where the shoreline was advanced. In contrast, the shoreline at the area near baselines #21–#27 severely retreated even before Typhoon TAPAH (T1917) attacked the site. Similarly, the beach widths in the southern end (near baselines #1–#10) were thicker compared to those in the middle of the beach (near baselines #11–#20). This high locality in the beach width between different areas of the beach might affect the erosion and recovery processes, which is analyzed in this section.

Figure 8 shows the time variation of parameters that are related to the locality in terms of wave power. In Figure 8a, the colored contours represent the beach width ($y$) variations of the 34 baselines, as the numbers in the y-axis denote the baseline numbers marked in Figure 5. Here, it is noted that the colors in the contours show the variation in $y$ along each baseline but cannot be used to compare the magnitude between the baselines because each y is the average removed beach width using the first 2-year average. It is clear that the $y$ values rapidly dropped after TAPAH (T1917). Compared to the northern area (# of baseline >25), however, the $y$ values decreased more severely in the southern end (# of baseline <6). The locality was also observed after the typhoon, as $y$ values were even reduced in the southern end near baselines < #5 until the recovery process prevailed, starting in November 2020. In contrast, $y$ values slightly increased in the northern end (# of baseline >28) by that time. This pattern continued during the recovery process after November 2020, because the $y$ values in the northern end became even greater than those before the typhoon since January 2021, whereas they were smaller than those before the typhoon in the southern end. This pattern of locality can be more clearly quantified through ${\langle y\rangle }_s$, ${\langle y\rangle }_m$, ${\langle y\rangle }_n$, and 〈$y$〉 in Figure 8c, in which the time variations of the beach widths calculated for three different groups are compared to the mean beach width. In the southern end, the ${\langle y\rangle }_s$ rapidly decreased ~30 m after TAPAH (T1917) and reached a minimum on 3 September 2020. The ${\langle y\rangle }_s$ magnitude could not be recovered even during the recovery process, since November 2020, as the maximum ${\langle y\rangle }_s$ observed in 2021 was still ~10 m smaller than that before the typhoon. In the northern end, ${\langle y\rangle }_n$ was not significantly reduced by the typhoon and gradually increased after that time as it became greater than the value before the typhoon during the recovery process, since November 2020. In the middle of the beach, ${\langle y\rangle }_m$ shows the middle course between the two previous cases. It is also noted that the pattern of ${\langle y\rangle }_m$ is similar to that of 〈$y$〉, indicating the time variation in the middle of the beach can represent the pattern for the whole beach.

In Figure 8c, the beach width change pattern is compared at two additional time periods, marked with circles. The grey circle covers the period from October 2019 to January 2020, when the beach settled down after TAPAH (T1917). The purple circle covers the recovery period from October 2020 to February 2021. These two periods were chosen because the beach width change pattern shows a discrepancy between the southern and northern ends. During the period of the grey circle, the magnitude of ${\langle y\rangle }_s$ decreased, whereas that of ${\langle y\rangle }_n$ increased, showing that the shoreline retreated in the southern end but advanced in the northern end. These results may be interpreted as an indication of a long sediment movement from the southern area toward the northern area. In contrast, both ${\langle y\rangle }_s$ and ${\langle y\rangle }_n$ increased during the recovery process, as marked with the purple circle, which indicates that the shoreline was advanced in both areas during the period.

For a better understanding of the locality in the shoreline variation pattern, a parameter was developed using the formula $L_t=\int \left(\partial y/\partial t\right)dx$, as plotted in Figure 8d. The dots in the figure show the magnitude of $L_t$ at each time step, and their colors also indicate the range of $L_t$ magnitude, with reds for positive and blues for negative values. In the formula of $L_t$, $\mathrm{x}$ is the alongshore direction toward the northern end. Therefore, the negative value of $L_t$ indicates that $\partial y/\partial t$ magnitude decreased in the positive x-direction. In other words, the positive/negative values of $L_t$ indicate the beach width change rate ($\partial y/\partial t$) increased/decreased toward the northern end of the beach. For example, the minimum $L_t$ value (~−4) during the time of TAPAH (T1917) indicates that the erosion rate was greater in the southern area and decreased in the increasing $\mathrm{x}$-direction so that beach might be most severely eroded in the southern end. Figure 9a shows the distribution of $L_t$ in terms of wave power. Although its probability of occurrence is low (Figure 9b), $L_t$ usually became negative with increasing magnitude when $P_w$ was greater than 10 kWatt/m. In particular, the extreme $L_t$ values (<−3) occurred when TAPAH (T1917) afflicted the beach. This pattern of $L_t$ distribution indicates that the shoreline was more severely eroded in the southern part of the beach, compared to that in the northern part, when storm waves attacked the site. In Figure 9c, the distribution of wave direction, $D_p$, is also plotted in terms of $L_t$. The x-axis denotes the wave propagation angles, which increase clockwise from the origin (0$°$) at the north. Therefore, 45$°$ and 90$°$ denote the NE and E, respectively. The high-powered waves generally approached the beach from ~NE. In the case of Typhoon TAPAH (T1917), the wave propagation direction ranged between 50$°$ and 60$°$. Considering the shoreline in Figure 1c and Figure 3a, the storm waves during the typhoon directly attacked the southern part, with a slight slope from the normal direction, whereas a shadow zone was likely formed in the northern part of the beach due to the breakwaters of Duho Port.

The discrepancy in the erosion/accretion pattern between the southern and northern parts of the beach can be also observed in the plane view of shoreline changes. In Figure 10, the beach widths of all 34 baselines are compared at three different times, as they correspond to the times of T1, T2, and T3 in Figure 6 and Figure 7. The figure clearly shows that the reduction in the beach width from T1 to T2 was greatest in the southern end (# of baseline <6), where the maximum reduction distance along baseline #2 was ~40 m. In contrast, the beach width in the northern end (# of baseline >28) was not significantly changed from T1 to T2, as the maximum change in the beach width was no greater than 5 m along baseline #29. At the baselines higher than #30, the change was minimal and not observed. During the recovery process from T2 to T3, the beach width increased at most of the baselines. However, the locality is also clearly observed, as its magnitude at T3 was smaller than that at T1 for the baselines #2–#20, indicating the shoreline was not fully recovered to the level before the typhoon attack in the southern part of the beach. In contrast, the beach width at T3 was greater than that at T1 for the baselines higher than #28. This result indicates that the shoreline in this northern area was not eroded significantly during the attacks of storm waves but advanced under milder wave conditions, leading to an accretion of the shoreline in general.

In the present study, the wave data were measured in location M1, which led to the assumption that the wave conditions were uniform along the coast of the beach. Therefore, alongshore variation of sediment transport caused by the spatial difference in the wave conditions could not be detected using the present data sets. As described, the locality in the results was induced by the shadowing by breakwaters and the curved shoreline, which might produce alongshore discrepancies in the shoreline evolution pattern, as implied from the locality parameter, $L_t$, in Figure 8d. The impacts of alongshore variation of wave conditions can be established, in future studies, by employing an array of wave gauges in the longshore direction.

4. Discussion

From the shoreline erosion and accretion patterns shown in Figure 6 and Figure 7, a consistent description can be provided for the recovery process, after the severe erosion by TAPAH (T1917) occurred in T1. The beach width decreased under high wave conditions and increased under low conditions, which corresponds to the general understanding of the shoreline evolution as previously described in Section 1. However, the criteria that determine the range of accretion and erosion are still unclear. The results in the present study can provide these criteria, although their application should be only confined to this study site.

Once the severe erosion by TAPAH (T1917) occurred in T1, the recovery process mainly started in T2, about 1 year after the erosion, and continued until T3, about 5 months after T2. The factors that affected short-term (days) processes can be further examined by posing a question about the recovery process—why didn’t the recovery process start earlier rather than starting ~1 year, until T2, after the typhoon? The answer may be simple: there were earlier times in which the recovery actually occurred under mild wave conditions. However, it could not continue because another set of high waves attacked the site, restraining the shoreline accretion. Some of the high waves were powerful enough to reverse the accretion trend into erosion. Therefore, the recovery process could not continue further, while the accretion and erosion processes were repeated until T2. During the period of ~5 months from T2 to T3, the wave power remained low so that the recovery process could continue without severe disturbances.

The next question on the process is which wave condition determines the trend of erosion or accretion? Based on the results of this study, a rough estimation can be provided: the shoreline was advanced when the wave power was no greater than 10 kWatt/m. In contrast, the erosion likely occurred when the wave power became greater than 20 kWatt/m. In fact, most rapid erosions occurred when the wave power exceeded 30 kWatt/m. During the total observational period of 33 months, the maximum wave power reached up to 75 kWatt/m, whereas the average wave power during the period was ~0.9 kWatt/m. When the wave power was too low in this site, the sediments did not move, and neither erosion nor accretion could occur. However, the criteria for this ‘no sediment motion’ condition could not be determined based on the results of this study.

It should be noted that there are critical cases in which the criteria may not be applicable. For example, once the beach width reached the minimum value at T1, its magnitude did not fall below this value during the rest of the observation period. This indicates that the erosional status reached a critical limit with this minimum width, and thus further erosion was prevented unless more extreme events occurred to beak the equilibrium again. On the other hand, there was a period of ~3 months from mid-February to mid-May 2019 (between the first green rectangle and the first red rectangle in Figure 7) when the wave power was generally no greater than 10 kWatt/m, but the beach width did not increase. This is likely because the beach width was thicker than the other period, and thus additional accretion of the shoreline might be disturbed, i.e., the beach width already reached saturation.

The results from locality analysis on the erosion/accretion process showed that the southern part of the beach was more severely eroded, especially when high-powered waves afflicted the site, whereas the shoreline in the northern part of the beach was relatively well protected. During the recovery process after October 2020, however, the imbalance between the two ends was reduced, and the shoreline was advanced almost equally at all parts—southern, middle, and northern—of the beach. The reason for this is still unclear, but the northern end could be protected from erosion due to the shadow zone formed by the breakwaters of Duho Port, in which the wave energy was attenuated, especially when the high-powered storm waves attacked the site.

In the southern end, the breakwater of Pohang Port does not form a shadow zone, considering that most of the waves approached from the NE or NNE, as shown in Figure 3b. Figure 9c also shows that the propagating direction of waves during Typhoon TAPAH (T1917) or other storm waves ranged from 50$°$ to 60$°$, which confirms that the damage by storm waves could be focused on the southern part of the beach. In contrast, the recovery process occurred under milder wave conditions, by transporting the sediments equally at all parts of the beach. However, there was a possibility of alongshore sediment movement from the southern part to the middle and northern areas. As shown in Figure 8c, ${\langle y\rangle }_s$ decreased but ${\langle y\rangle }_m$ and ${\langle y\rangle }_n$ increased during the period just after TAPAH (T1917), as marked with the grey circle. This locality of different shoreline evolution processes implies that the sediments could be transported in the longshore direction from the south to the north.

The results of this study can be usefully applied in designing the protection/prevention plans against damage to the beach and nearby coastal areas, not only from the disastrous storm events but also from the processes that may break the equilibrium in the beach status. For example, the suggested criteria that could determine the erosion/accretion conditions could be useful in planning coastal structures. The size of these structures can be effectively estimated if information on the wave conditions that affect the erosion or accretion in this specific site is known in advance. If the size of such a structure is too large, it may be cost-ineffective. In addition, the beach recovery process could be disturbed if the structure reduces too much wave energy; in this case, the minimum power required for accretion cannot be met. Similarly, the structure would not correctly function if its size is too small to reduce the wave power below the criteria for erosion. The information on the locality in the shoreline evolution process could be also useful in planning hard and soft coastal structures. If the longshore transport from the southern to the northern part is quantitatively understood (although it has not been done in this study), structures such as groins can be designed to reduce the loss of sediment in specific areas. In addition, the results of the locality analysis in this study would be useful in planning beach nourishment—in determining where to put the sand to maximize the effect of the beach fill by considering the imbalance in the beach process. For example, beach nourishment and placing some coastal structures have been planned in the study area, although they have not been finalized. The results of this study are, therefore, expected to provide useful information in determining such plans, to maximize the effect of the structures and nourishment.

It should be noted that the results from this study are confined only to this specific site. This small beach is characterized as a pocketed beach, where the sediments cannot be added from or lost to other coastal cells. In addition, the beach is located inside a bay, so the waves are usually attenuated when reaching the study site such that the shoreline can reach equilibrium under lower wave energy compared to beaches on the open coast. Figure 11 compares the wave power between Yeongildae Beach (8.5 m depth) and Hwajin Beach (32.0 m depth), located ~20 km north of Yeongildae Beach. Because Hwajin Beach is located on the open coast outside Yeongil Bay, its wave force is much greater than that observed at Yeongildae Beach despite, the difference in water depth.

As a result, the sediments in this site could respond more easily at lower wave energy levels, which might lead to a different pattern of erosion and recovery process. For example, the sand size in the study site ranged from 0.15 to 0.35 mm, as it is categorized as fine to medium sand, which was smaller than the sand size measured in the two beaches located in the open coast outside the bay, where it was ~0.5 mm at both beaches (categorized as medium to coarse sand) [30,31]. The criteria for erosion/accretion suggested in this study can hardly be determined in other beaches on the open coast, where nearshore crescentic sandbars are actively developed. Because the horns of the sandbars are coupled with the shoreline pattern [32], the erosion/accretion of the shoreline has a strong locality rather than showing the general trend as observed in this study site, where no crescentic sandbars developed. For this reason, the information on the locality from these study results should be also confined to this specific site.

5. Conclusions

In this study, we employed a data set of VMS and wave parameters measured over 33 months to investigate the erosional and depositional processes in Yeongildae Beach, located inside Yeongil Bay on the southeastern coast of the Korean Peninsula. The beach was severely eroded when Typhoon TAPAH (T1917) hit the site in late September 2019, during which the beach width retreated ~12 m on average; it took about 1.5 years for the beach width to be recovered to the level before the attack of the typhoon. The study then analyzed the recovery process, during which shoreline erosion and accretion repeated, corresponding to the wave conditions.

The study site is a pocketed beach, as the breakwaters of two ports were built at both ends of the beach, blocking the input or loss of sediments in the longshore direction crossing the lateral boundary of the beach. The beach is also located at the west corner, inside the bay, and the distance from the beach to the mouths of the bay is 8–15 km. Due to its location on the beach, the wave energy was lower and the sediment size was finer, compared to those observed on the open coast outside the bay, because the waves were attenuated when reaching the site.

The results of the analysis corresponded to the general understanding of beach processes—that beach widths decreased/increased under high/low wave energy conditions. In this study site, however, the pattern of shoreline evolution was found to be highly correlated with $P_w$, the wave power that combined the impacts of wave height and period, rather than the wave propagating direction, which was observed to be similar for high-powered waves. In particular, the beach width generally increased when $P_w$ was no greater than 10 kWatt/m and tended to decrease under the condition of $P_w$ greater than 20 kWatt/m. However, most erosional events occurred when $P_w$ exceeded 30 kWatt/m. Especially after TAPAH (T1917), it took a long time (~1.5 years) for the beach width to be fully recovered to the previous level because high-powered waves occasionally disturbed the recovery trend. Therefore, the full recovery of the beach width could be reached over the last five months in the recovery period, during which $P_w$ was kept continuously lower than 10 kWatt/m.

The locality of beach recovery process was also analyzed by examining the discrepancy in the erosion/accretion pattern between different locations within the beach. It was observed that the erosion by TAPAH (T1917) was most severe in the southern part of the beach, whereas the change in the beach width was minimal in the northern part during the same period. This pattern of locality was similarly observed under other storm conditions. Considering the wave propagation direction was similar (~NE) for the storm waves, it is likely that a shadow zone was formed in the northern end due to the breakwaters of Duho Port, such that the sediments in the lee area of the breakwaters were protected because the wave energy was attenuated. On the contrary, the breakwater of Pohang Port located in the southern end did not form a shallow zone, and thus the storm waves directly attacked the southern part, causing erosion. In the phase of recovery under milder wave conditions, the shadow zone did not play an important role, and the shoreline was advanced almost equally at all parts of the beach. However, it was also observed that there was a period just after TAPAH (T1917) when erosion occurred in the southern part while accretion occurred in the northern part, indicating a possible alongshore sediment movement from the south to the north.

The results from this study can be usefully applied to design the protection and prevention plans of the beach from storm damage, considering the general and local erosion and recovery patterns. For example, the criteria that determined the trends of erosion and accretion based on wave power can be used to effectively estimate the size of coastal structures, such as detached breakwaters for shore protection. In addition, the results of locality analysis can be useful to plan the horizontal arrangement of such hard structures or beach nourishment, because they can provide information on where to place the structure or sand to maximize their ability to stabilize the beach. It is noted that the application of these results should be confined only to this specific site and may not be directly applicable to other beaches on the open coasts. For example, the suggested criteria are valid where the wave energy is relatively lower than that on the open coasts. The conditions in this specific beach might result in characteristic shoreline responses. In the case of other beaches located on the open coasts, where various nearshore sandbars develop, the shoreline could be coupled with sandbar positions, and the erosion/accretion pattern would be more complex than that observed in this study site. Regardless of these limitations, the method in this study can be usefully applied in areas with a similar environment for designing prevention plans for disasters due to storms.