Kideok Do

Optical satellite images are a valuable alternative to conventional surveys by enabling consistent, high-resolution observations spanning more than 40 years. This study presents a comprehensive framework based on optical satellite imagery and CoastSat for analyzing the long-term shoreline changes on the East Coast of the Korean Peninsula, a region increasingly affected by sea-level rise, intensified storms, and human interventions. These pressures have accelerated erosion and destabilized the shoreline dynamics, posing risks to infrastructure and ecosystems. A monthly time series dataset (1984–present) was constructed using the Google Earth Engine with standardized preprocessing, including image enhancement and georeferencing. Shorelines were extracted using CoastSat, an open-source software that automates shoreline detection from multi-source satellite imagery. Nevertheless, the process required manual verification and showed reduced accuracy in anthropogenic coastal environments. Time-series analyses at Gangneung Port–Yeomjeon Beach and Gungchon–Chogok Port, where sequential coastal structures have significantly altered the shoreline dynamics, revealed distinct shoreline seasonality and horizontal shifts. These spatial patterns are closely associated with the construction of breakwaters, groins, and other structures as erosion mitigation strategies. These findings highlight the potential of long-term satellite datasets to enhance shoreline monitoring and improve coastal resilience under growing environmental uncertainties. The framework offers a scalable solution for satellite-based coastal management across evolving global environments.

The behavior of waves, influenced by numerous factors, becomes markedly complex as they approach coastlines and interact with seabed geometries. Simultaneously, this process involves a range of phenomena, including wave breaking, which is one of the most iconic and influential. Wave breaking is not only pivotal in shaping coastal landscapes but also has far-reaching consequences for the design and safety of coastal infrastructure. Traditional research methods, including theoretical, experimental, and numerical approaches, have sought to elucidate the intricate relationships involved in wave breaking. However, the inherent nonlinearity and transient nature of this phenomenon pose substantial challenges to understanding its natures (Goda, 2010). Departing from conventional methodologies, our research leverages the advanced capabilities of deep learning to refine data acquisition techniques and decode the intricate patterns of wave breaking, thereby advancing predictive models and management strategies for coastal environments.

Various structures and coastal developments have been installed to mitigate the increases in coastal erosion caused by climate change, but numerical modeling studies to predict the changes in coastal morphology induced by these structures have been insufficient. This study analyzed the morphological changes in the complex coastal environment of Wonpyeong Beach, Gangwon-do, where coastal maintenance projects are underway, using XBeach simulations based on observational data. For accurate modeling, the geomorphic characteristics, such as the subsidence of submerged structures, sediment layer thickness, and spatial varying bed friction coefficients, were incorporated using satellite imagery and field survey data. The results showed that accounting for the subsidence of submerged structures represented the formation of coastal sandbars while adjusting the sediment layer thickness reduced excessive erosion around the structures. In addition, incorporating spatial variations in the bed friction coefficients improved the representation of localized erosion and deposition, even though the overall impact was limited. This study emphasizes the importance of incorporating geomorphic features for accurate predictions of morphological changes and shows the value of using satellite imagery as a supplementary tool in data-limited environments. Nevertheless, additional observational data are necessary to validate and refine the proposed approach.

Coastal erosion has caused significant ecological damage and substantial social and economic losses, emphasizing the necessity for long-term monitoring of shorelines. Video monitoring systems have emerged as effective tools for observing shorelines changes. However, the shoreline extracted from video monitoring systems represents the waterline, defined as the boundary where waves reach the beach. The extraction of waterlines using these systems often requires manual parameter adjustments to account for dynamic coastal environments. Furthermore, the extracted waterline is influenced by variations in waves and currents, posing challenges for consistent and reliable monitoring. To address these issues, this study utilizes the deep learning model ‘U’-shaped Neural Network (U-Net) to analyze images of Chuam Beach. A dataset comprising 204 pairs of Time-exposure (TimeX) images was developed to train the U-Net model, which was subsequently utilized to propose an automated method for waterline extraction. Using the extracted waterlines, the beach slope was calculated, enabling the estimation of a tide-based shoreline unaffected by wave and currents variations. Comparative analysis with UAV (Unmanned Aerial Vehicle)-based shoreline data revealed a maximum vertical error of 0.18 m for the waterline and 0.27 m for the shoreline, demonstrating high accuracy. These findings indicate that the U-Net model effectively and accurately detects waterlines, while the proposed shoreline estimation method ensures consistent monitoring of shoreline changes, including advances and retreats.

In coastal video monitoring, the direct linear transform (DLT) method with ground control points (GCPs) is commonly used for geo-rectification. However, current practices often overlook the impact of GCP quantity, arrangement, and the geographical characteristics of beaches. To address this, we designed scenarios at Chuam Beach to evaluate how factors such as the distance from the camera to GCPs, the number of GCPs, and the height of each point affect the DLT method. Accuracy was assessed by calculating the root mean square error of the distance errors between the actual GCP coordinates and the image coordinates for each setting. This analysis aims to propose an optimal GCP placement method. Our results show that placing GCPs within 200 m of the camera ensures high accuracy with few points, whereas positioning them at strategic heights enhances shoreline extraction. However, since only fixed cameras were used in this study, factors like varying heights, orientations, and resolutions could not be considered. Based on data from a single location, we propose an optimal method for GCP placement that takes into account distance, number, and height using the DLT method.