Beach Area Changes and Resilience of the Eastern Coasts Before and After Typhoon Goni

2.1 Study Site

The Ministry of Oceans and Fisheries has gradually expanded the number of video monitoring sites in Korea since 2003, starting with Haeundae Beach in Busan and Daecheon Beach in Chungnam. As of the present study, 42 monitoring sites are operational. For this study, 14 target beaches located along Korea’s east coast were selected: Chodo, Gonghyinjin, Gyoam, Bongpo, Sokcho, Sodol, Youngjin, Gyengpodae, Gangmoon, Yeomjeon, Hamaengbang, Wonpyeong, Wolsongri, and Goraebul. Real-time image data were collected through video monitoring and were utilized in the analysis of beach area changes (Fig. 1).

Table 1 provides detailed information for each beach, including shoreline length, number of video monitoring installations, number of cameras, observation range, and installation year. Shoreline lengths ranged from 590 m to 4,520 m, and the number of video monitoring installations varied from one to three per site. The number of cameras ranged from two to ten, enabling the monitoring of over 85.1% of each beach.

2.2 Video Monitoring System

The video monitoring system was used to observe changes in sandy beach areas before and after a typhoon. This technique, which allows continuous real-time monitoring of sandy beach changes using video cameras, can operate even under adverse weather conditions, including typhoons (Thuan et al., 2016). Typically, video monitoring systems are installed on tall buildings or towers located behind the beach being monitored. The installation site is selected based on the shape, length, and width of the beach. Additionally, the number of cameras is determined by the camera installation height, position, and viewing angle, with overlapping images from different angles to minimize errors (Kang et al., 2017).

The video monitoring system comprises an imaging unit, a camera shooting scheduler, and image processing and data transmission modules (Fig. 2). The effective observation range per camera was set to minimize the coordinate error in the captured images to less than 1 m.

2.3 Image Processing

Video monitoring was used to analyze the beach areas of each target site. Given that the monitoring installation environment differs by region, the data must be standardized and accurately analyzed using a unified coordinate system. This study analyzed beach erosion and recovery speed by extracting shoreline data and calculating beach area through a stepwise analysis technique (Fig. 3).

Since the shoreline on wave-affected coasts, like the east coast, changes constantly, a mean image must be created to accurately extract the shoreline boundary from the video data. This was done by averaging red-green-blue pixel color values from 180 images captured over three minutes (one image per second). In Fig. 4, the swash zone appears as dark white in the mean image, making the shoreline easily distinguishable (Lippmann and Holman, 1989; Plant and Holman, 1997).

The correlation between the images obtained from video monitoring and the on-site measurement coordinates was established, and ground control point (GCP) measurement was performed to enable processes such as coordinate transformation and distance conversion. This allowed for the extraction of image information necessary to convert the captured images of the target area into an actual plane coordinate system. The accuracy of the coordinate transformation depends on the image information extraction coefficient, which is directly influenced by the location, number, and configuration of the GCPs (MOF, 2015).

For image rectification, the direct linear transform (DLT) technique was employed. The DLT technique performs the transformation using GCPs measured within the spatial coordinates of the imaging range, even when the camera’s position and installation angle are unknown (Holland et al., 1997). This method describes the relationship between the image coordinates and the GCP coordinates while accounting for errors arising from the orthogonality of the image’s x and y axes, as well as differential linear distortion. The DLT technique minimizes complex computational steps, such as initial value setting and iterative calculations; however, it requires at least six GCPs. It also necessitates careful consideration because the analysis results can vary based on the number and position of the GCPs in the image (Holland et al., 1997; Cho et al., 2001; Lee et al., 2024).

Using beach information (shoreline and beach width) obtained from the video images, the variability of the sandy beach was identified by performing tidal correction, time series analysis, and trend analysis. Since the beach width extracted from the image varies with the tidal height, tidal correction was necessary. However, on the east coast, the tidal variation is minimal (approximately 30 cm), and the beach slope is relatively steep. As a result, shoreline analysis was conducted using the mean sea level as a reference.

During the synthesis of the coordinate-converted orthoimages, at least two GCPs from each image were overlapped to clearly observe changes across the entire shoreline. The synthesized orthoimages, captured at the same time, were mapped to a digital map (1:5,000 scale). The accuracy of the coordinate transformation was verified by examining the continuity of the surf zone at the junction of each image and by comparing the GCP coordinates from the images with those on the digital map (Fig. 5).

The accuracy of the coordinate transformation was further validated through orthoimage synthesis and mapping to the digital map. Additionally, image data extraction coefficients for each camera were calculated, including factors such as the scale factor, focal length, lens distortion, axis movement distance, and axis rotation angle (Table 2). The validation process included calculating the root mean square error (RMSE) between the fixed points in the images and the corresponding coordinates on the digital map, specifically comparing the actual coordinates of GCPs with the transformed coordinates. The XY RMSE value was set to achieve an accuracy of approximately 1 m for each image.

The shorelines of the target coasts were extracted by analyzing the color information contained in each pixel of the images. Since there is a clear distinction in pixel color characteristics between the sea and land sections along the shoreline in the images (Fig. 6), the first point where pixel color characteristics changed sharply was determined as the shoreline. This was done by identifying the pixel color characteristics required for shoreline extraction at each baseline and analyzing the pixel information pattern. The pixel information pattern analysis technique, specifically the pixel intensity moving average coastal shoreline (PIMAES) method proposed by Kim (2014), which utilizes the moving variance of pixel characteristic values, was used for shoreline location determination. PIMAES calculates the moving average for each pixel’s characteristic values and determines the difference between the moving average and the pixel characteristic value. It then extracts the shoreline by selecting the first pixel where the characteristic value is higher or lower than the overall average among the pixels for which the difference between the moving average and the pixel characteristic value exceeds the threshold. This method allows for accurate shoreline extraction regardless of imaging time (e.g., sunrise or sunset) or weather conditions (e.g., cloudy or clear days) (Kim, 2014).

The beach area (A) was calculated by multiplying the daily average beach width of the target area’s baselines (B_i) by the total shoreline length (L), as shown in Eq. (1) (Kang et al., 2017). For the calculation of the beach area (A ), observation baselines (B_i) were set at 50 m intervals, considering the shoreline length and shape of the target area. The vertical distance between the shoreline of the high water ordinary neap tide (H.W.O.M.T) at the set baselines and the safety line provided on the digital map by the National Geographic Information Institute was measured as the beach width.

2.4 Typhoon Goni

The typhoons that impacted the east coast in 2015 were analyzed using data from the Korea Meteorological Administration (KMA) (Fig. 7). In 2015, a total of 26 typhoons occurred near the Korean Peninsula, including in the Yellow Sea, the East China Sea, and Japan (KMA). Of these, 13 occurred during the summer months (July to September) (KMA). Among them, Typhoon Goni was classified as a “small” tropical storm with a central pressure of 1,002 hPa, a maximum wind speed of 18 m/s, and a strong wind radius of 110 km. It originated in waters approximately 370 km east of Guam, USA, and was the 15th typhoon of 2015. It developed in a west-to-west-northwest direction and made landfall on the Korean Peninsula after rapidly passing through Kyushu, Japan, on August 25.

Typhoon Goni directly affected the east coast as it passed through the area, as shown in Fig. 7, before dissipating in the northeast through the central part of the East Sea (KST, August 26, 06:00). According to the KMA data, the typhoon recorded a maximum wind speed of 35 m/s, a strong wind radius of 280 km, and a travel speed of 38 km/h in the southern part of the Korean Peninsula. In the eastern and central parts, it recorded a maximum wind speed of 29 m/s, a strong wind radius of 200 km, and a travel speed of 19 km/h.

Fig. 7 shows the wave buoys near the study sites on the east coast operated by the KMA. Waves and wave energy were analyzed at the peaks of the Tosung, Yeongok, Jukbyeon, and Hupo wave buoys throughout the year, particularly during the Typhoon Goni invasion period, with the results presented in Fig. 7. The wave energy, which is a function of the wave height, is calculated using Eq. (2)

where E_b is the wave energy (J/m²), ρ is the density of seawater (kg/m³), g is the gravitational acceleration (m/s²), and H is the significant wave height (m). According to the wave data during the typhoon period for each peak shown in Fig. 8, the typhoon exhibited the largest wave and wave energy peak on August 25, when it moved to the central part of the east coast with significant size. The timing of peak occurrences varied depending on the typhoon’s path. The maximum wave heights recorded by the Tosung, Yeongok, and Jukbyeon wave buoys were 5.1, 4.8, and 5.2 m, respectively, with corresponding wave energy values of 32.6, 28.9, and 33.9 kJ/m². The maximum wave height and wave energy recorded by Hupo were 5.9 m and 43.7 kJ/m², respectively. During the period affected by the typhoon, from August 22 to 29, the durations of high waves (2 m or higher) recorded by the Tosung, Yeongok, Jukbyeon, and Hupo wave buoys were 64, 63, 55, and 49 h, respectively. The average heights of these high waves were 3.3, 3.3, 3.4, and 3.7 m, while the average wave cycles were 8.4, 8.7, 9.1, and 9.1 s, respectively.

3.1 Analysis of Beach Erosion

The video monitoring method was employed to identify beach changes before and after the typhoon’s invasion. The target sites included 14 beaches on the east coast, as shown in Fig. 1. Quantitative erosion changes were identified by comparing the beach area before and after the invasion of Typhoon Goni in 2015 (Table 3). Fig. 9 presents image data from some of the target sites before the typhoon, after the typhoon, and after beach restoration, which were used in the analysis. In Table 3, the target sites are listed in order from Gangwon-do (north) to Gyeongbuk (south).

According to the analysis results, Bongpo Beach exhibited the largest erosion change due to Typhoon Goni, with 3,407 m² of its sandy beach eroded, corresponding to 38% of the entire beach. Significant shoreline retreat mainly occurred in the central 250 m section (Fig. 9). Chodo Beach showed the second-largest erosion change, with 4,317 m² of its sandy beach eroded, accounting for 34% of the entire beach, and significant shoreline retreat primarily occurred in the central 350 m section (Fig. 9). Goraebul Beach exhibited the largest erosion area, with 7,441 m² of its sandy beach eroded, which is 22% of the entire beach. The sites that showed the smallest damage were Gyoam (871 m², 7%), Gangmoon (743 m², 17%), Hamaengbang (951 m², 6%), and Wolsongri (827 m², 8%) (Table 3).

3.2 Analysis of Restoration

The restoration of the 14 sites on the east coast due to the invasion of Typhoon Goni was analyzed. Restoration was evaluated based on the restoration period and daily recovery speed (m²/d) by comparing the sandy beach area before and after the typhoon invasion. The date when Typhoon Goni began to affect the east coast was August 25, and it dissipated near Ulleungdo Island on August 26. However, after the typhoon’s dissipation, high waves continued to affect the east coast until September 11 (Fig. 7).

In general, beaches experience short-term erosion due to prolonged exposure to typhoons or high waves, but they typically return to their original sandy conditions through natural resilience. Therefore, in this study, the sandy beach area was compared for each beach before and after the typhoon invasion to evaluate the restoration period and daily recovery speed (m²/d) (Table 4). The restoration rate represents the rate of recovery following the typhoon, and both the restoration period and restoration rate were calculated based on achieving 90% recovery of the beach area prior to the typhoon. Sites with a restoration rate of less than 100% indicate that the beach had not fully recovered to its pre-typhoon condition. Most sites showed natural resilience in recovering their beach area over approximately a month, with five sites, including Gyengpodae Beach, achieving 100% recovery of their beach area before the typhoon invasion. The sites with the longest restoration periods were Gonghyinjin and Goraebul Beaches, while Sodol and Yeomjeon Beaches had the shortest restoration periods.

3.3 Analysis of Beach Erosion

The average wave energy and beach erosion area of each beach during the erosion period caused by the typhoon invasion were analyzed (Fig. 10). As shown in Fig. 7, Chodo, Gonghyinjin, Gyoam, Bongpo, and Sokcho Beaches were analyzed using the Tosung wave buoy, while Sodol, Youngjin, Gyengpodae, Gangmoon, and Yeomjeon Beaches were analyzed using the Yeongok wave buoy. Hamaengbang and Wonpyong Beaches were analyzed using the Jukbyeon wave buoy, and Wolsongri and Goraebul Beaches were analyzed using the Hupo wave buoy. An average wave energy of 5.0 kJ/m² or higher (wave height: 2 m or higher) was applied for the beach erosion period, while the overall average was applied for the restoration period. The correlation was assessed using the coefficient of determination (r²). An r² value of one indicates a strong correlation, while a value of zero indicates no clear correlation.

The correlation (r²) between wave energy and erosion area during the erosion period was found to be 0.1 or less, indicating no correlation. Conversely, r² was relatively high (0.3) during the restoration period. For the restoration period, the average energy distribution was somewhat large for beaches with a small erosion area of 1,000 m² or less (Wolsongri, Hamaengbang, Gangmoon, and Gyoam), but a relatively high correlation was observed at sites with larger erosion areas. The wave height, period, and direction change when deep-sea waves reach the shallow water zone in front of the beach due to morphological characteristics. To better identify these effects, it is necessary to analyze long-term correlations with wave data near the target area and further examine the typhoon invasion period, focusing on direct impacts.

The erosion area and restoration period of the sandy beach were compared before and after Typhoon Goni, and the results are presented in Fig. 11. The beach erosion area and restoration period showed no clear correlation. This lack of correlation appears to be due to the limited number of target beaches and typhoons analyzed. In particular, the restoration period and speed are predominantly affected by the waves that invade after the typhoon. Therefore, it is necessary to analyze the wave information near the target area along with beach erosion data during additional typhoon events in the future.