Monitoring Coastal Land Reclamation Progress in Busan-Jinhae Area, Korea, Using Sentinel-1 SAR Images

Abstract

Reclamation of public water causes permanent loss of public water that destroys coastal environments, including tidal flats. Korea has established legal and administrative systemsto preserve the coastal environment and manage public water sustainably. Several previous studies suggested remote sensing methods to monitor coastal land reclamations using optical satellite sensors, but this makes it difficult to apply in Korea due to the concentration of rain in the Korean summertime. In this study, we proposed a method for efficient monitoring of coastal land reclamation in complex public water areas regardless of weather conditions using Sentinel-1 synthetic aperture radar (SAR) images. We estimated the reclamation rate and classified planned reclamation projectsinto three classes(Not initiated, Ongoing, Completed) using 183 Sentinel-1 SAR images acquired in the Busan-Jinhae area from 2017 to 2023. The accuracy of classification results was 97.14%, and we monitored the reclamation progress of the planned public water reclamations with the estimated reclamation rate. From February 2017 to December 2023, the ongoing reclamation continuously decreased, while the reclamation area was steadily increased. Most of the reclamation projects were completed during the study period. On December 21, 2023, it was estimated that 81.93% of the total planned area had been reclaimed. The proposed method can efficiently survey coastal land reclamation in the wide and complex Korean coastal environment.

1. Introduction

Public water refers to a government-owned water surface or water stream that is used by the public. Public water is used for public interest purposes such as agriculture, fishing, and industry. Reclamation of public water is a way selected to create new land and to develop the local economy of coastal cities (Sengupta et al., 2019; Shin, 2023). These public water reclamations destroy coastal environments, including tidal flats, and make public water difficult to use sustainably (Ai et al., 2022; Liu et al., 2023; Murray et al., 2019). The Korean Ministry of Ocean and Fisheries is trying to prevent damage to the coastal environment and loss of public water. They survey the actual situation of public water reclamation and its legality every year to monitor whether the public water reclamation is going as planned (Ministry of Government Legislation, 2024). In addition, newly created lands can be monitored to help establish and implement marine space policies or national land policies in coastal cities (Eom et al., 2012). However, Korea has a complex and lengthy shoreline and numerous islands, making it difficult to accurately survey the public water reclamation situation. The satellite image-based public water reclamation monitoring method can be used as a tool to efficiently monitor public water reclamation because of the characteristics of the satellite images, which allow periodic observations of large areas to be made.

Several studies were conducted previously to monitor the coastal land reclamations using satellite images (Huang et al., 2025; Senputa et al., 2018; 2023; Shi et al., 2022. Huang et al. (2025) proposed a method to monitor the coastal reclamation with denser temporal resolution and without any detection of natural landcover changes using optical satellite images. Sengupta et al. (2023) mapped worldwide reclamation projects of the 21st century using optical satellite images. Sengupta et al. (2018) analyzed the coastal reclamations of the world’s 16 cities using optical satellite images. Shi et al. (2022) tracked temporal changes of a coastal city due to the reclamation projects using optical satellite images. However, Korea has a concentration of rainfall in summertime, and optical satellites are limited in observing through the clouds. The exact date of the coastal reclamation is important when following administrative procedures for public water reclamation or determining the legality of the reclamation project, which is a major obstacle to monitoring public water reclamation.

In this study, we proposed a method for monitoring public water reclamation using synthetic aperture radar (SAR), which can observe the Earth’s surface regardless of weather conditions. We monitored the basic public water reclamation plans held in the Busan-Jinhae area using time-series Sentinel-1 SAR Images. First, we estimated the reclamation rate, which indicates how much progress the reclamation project has made compared to the basic public water reclamation plan, then we classified the reclamation status into three classes (Not initiated, Ongoing, Completed) for every single reclamation project of each scene. In Section 2, the study area and data used in this study will be introduced. The method for monitoring the basic public water reclamation plan is detailed in Section 3. Sections 4 and 5 present the results and their discussion, while Section 6 presents the conclusion of the study.

2. Materials

2.1. Study Area

The Busan–Jinhae region has been designated as a free economic zone (Busan-Jinhae Free Economic Zone), and various development projects have been established to promote the regional economic growth. Multiple port construction projects involving reclamation of public water have been undertaken, and reclamation projects covering approximately 25.96 km² have been set for the period from 2020 to 2030 in the Hwangpo and Dongseon area (Busan-Jinhae Free Economic Zone Authority, 2023). Projects include the construction of the Jinhae Myeong-dong Marina Port and Cheonseong Port, the ongoing development of Busan New Port, and the planned development of the Waseong District for neighborhood facilities and industrial use.

2.2. Data

To monitor reclamation projects planned for the Busan–Jinhae area, accurate spatial information was analyzed using the basic public water reclamation plan and the basic national port plan provided by the Ministry of Ocean and Fisheries. The basic national port plan, which governs port construction projects involving reclamation, holds the same legal authority as the basic public water reclamation plan (Ministry of Government Legislation, 2025b). In this study, both the basic public water reclamation plan and the basic national port plan are referred to as the basic public water reclamation plan. The study area was limited to the Hwangpo and Dongseon area, which includes the 3rd and 4th generation of basic public water reclamation plans, planned from 2010 to 2030, of the Busan-Jinhae free economic zone. Fig. 1 shows the Cadastral map of South Korea, the basic public water reclamation plan, and the coastal grid map of Dongseon and Hwangpo, South Korea.

Fig. 1. Cadastral map of South Korea. The basic public water reclamation plans implemented in Dongseon and Hwangpo, South Korea.

In this study, to monitor the temporal changes of reclamation progress planned in the Busan–Jinhae area, 183 Sentinel-1A images with a 12-day repeat cycle, acquired from 2017 to 2023, were collected. The images were acquired in ascending orbit geometry on track 54, frame 111. Only Sentinel-1A data are used, due to the end of the mission of Sentinel-1B since December 2021. We pre-processed single look complex images acquired with interferometric wide swath mode. In the Busan-Jinhae area, the water and land are clearly distinguished, so only VV polarization images were used.

3. Methods

3.1. Pre-Processing

The Sentinel-1 SAR data were preprocessed using the SNAP software provided by the European Space Agency. Thermal noise removal, radiometric calibration, co-registration, multi-looking, and terrain correction were applied sequentially to generate Gamma Nought backscattering coefficient images. Co-registration was carried out based on the cross-correlation method. The image acquired on February 19, 2017, was set as the reference image. To reduce computational load, the preprocessed Sentinel-1 images were clipped to the Hwangpo and Dongseon area from the coastal grid map provided by the Ministry of Ocean and Fisheries. According to the legal definition of public water, land was masked out using the 2017 cadastral map provided by the Ministry of Land, Infrastructure, and Transport (Ministry of Government Legislation, 2025a).

3.2. Framework

Reclamation of public water was defined as the conversion of water into non-water (Ministry of Government Legislation, 2025a; Shi et al., 2022). And the reclaimed land was defined as new land created by the reclamation activity on public water. We assumed that only cadastral registered land, reclaimed land, and public water exist in the SAR images. First, reclaimed land detection was carried out. We detected reclaimed land by classifying water and non-water via thresholding the modeled probability density function (PDF) of sampled pixels from the public water region that is not registered on the cadastral maps and not included in the public water reclamation planned area. Reclaimed land was detected in each public water reclamation project and each SAR image, and then the reclamation rate of each public water reclamation project was estimated using all 183 reclaimed lands detected. We estimated the initiation date and completion date of each reclamation project using the change validation factor (CVF). Reclamation projects were classified into three reclamation status categories, as follows, based on estimated initiation time and completion time.

• Not initiated: A reclamation project for which construction has not yet begun.

• Ongoing: A reclamation project for which construction has started but is not yet completed.

• Completed: A reclamation project in which no public waters remain within the designated area.

Here, “Completed” refers to the physical loss of public water in the planned area and is not directly related to administrative procedures. Finally, we adjusted the reclamation rate using the reclamation status, and a temporal analysis of reclamation progress was conducted. The detailed methods for each step are described in Sections 3.3 (Reclamation Detection) and 3.4(Estimation of Reclamation Rate and Reclamation Status). Fig. 2 is the flow chart of the method suggested in this study.

Fig. 2. Flow chart of the method suggested in the study.

3.3. Reclamation Detection

Detection of reclaimed land was carried out by applying a threshold to the statistical distribution of pixel values corresponding to public water in the SAR images. Because water has very low values in SAR images, pixels with values greater than the set threshold were treated as reclaimed land.

In SAR images, the region where they are not included in both the cadastral map and basic public water reclamation plans, it was assumed that only public water existed. Pixels representing public waters in the SAR images were sampled to model a PDF. Since the backscattering coefficient of SAR images follows a normal distribution, the PDF was modeled as a normal distribution using the mean and standard deviation of the sampled backscattering coefficient, and thresholds were set at a significance level of 0.1%. Moreover, because the statistical distribution of water in SAR images varies with diverse environmental conditions (Alsdorf et al., 2007), a threshold was established for each image individually.

3.4. Estimation of Reclamation Rate and Reclamation Status

To classify the status of public water reclamation into the three predefined categories (Not initiated, Ongoing, Completed), the reclamation rate was estimated for each reclamation project for each scene. The reclamation rate of public water reclamation was calculated as the ratio of reclaimed land area to the planned reclamation area, as shown in Eq. (1).

\(\begin{align}R=\frac{N_{L}}{N} \end{align}\)       (1)

R represents the reclamation rate, N_L is the number of pixels that are detected as reclaimed land within the reclamation project, and N is the total number of pixels in the reclamation project. R was calculated for all 183 images and takes values from 0 to 1, where 0 indicates that reclamation has not been initiated and 1 indicates that reclamation has been completed.

The initiation and completion of reclamation were determined based on changes in the reclamation rate over time. To identify the initiation date (t_i) of reclamation, the cumulative mean of R was used. The cumulative mean was calculated from the beginning until the time reclamation was completed, and the time at which R first exceeded 0.98 was regarded as the completion date (t_c) of reclamation. Since R increases sharply once reclamation begins, the difference between the cumulative mean and the value of R becomes large. Therefore, the last time at which R was smaller than the cumulative mean was considered as a candidate of the t_i (t_ic). CVF was then computed with Eq. (2) to validate if the t_ic truly indicates the reclamation (Colin Koeniguer and Nicolas, 2020).

\(\begin{align}C V F=1-\frac{\operatorname{mean}\left(R \in\left\{1, \ldots, t_{i c}\right\}\right)}{\max \left(R \in\left\{t_{i c}+1, \ldots, N_{s c n}\right\}\right)}\end{align}\)       (2)

N_scn represents the total number of scenes. R∈{1, …, t_ic} denotes the values of R from the first image to the ic-th image, and R∈{t_ic + 1, …, N_scn} denotes the values of R from the t_ic+1-th image to the last image. If the CVF is greater than the threshold, the t_ic is determined as the t_i, if it is smaller than the threshold, it is judged that reclamation began earlier than the first image (February 19, 2017). In cases where reclamation was not completed, t_c was set to the last image, and the CVF was calculated. If the CVF was greater than the threshold, it was determined that reclamation had not started. The threshold was empirically set to 0.3. The reclamation rate was adjusted such that all values before the start of reclamation were set to 0 and all values after completion were set to 1.

4. Results

Using the method proposed in Section 3, the reclamation status and the reclamation rates of the basic public water reclamation plans were estimated. Fig. 3(a) shows the average of 183 Sentinel-1 VV polarization Gamma Nought backscattering coefficient images acquired in the Hwangpo and Dongseon area, where the red box indicates an area not included in both the cadastral map and the basic public water reclamation plan, thus representing permanent public water. It was therefore assumed that the public water in the SAR images follows the statistical distribution of the pixels sampled within the red box in Fig. 3(a).

Fig. 3. Sentinel-1 SAR images acquired from the study area. (a) The average of 183 Sentinel-1 backscattering coefficient images from February 19, 2017, to December 21, 2023. The red box is the region where sampling of the public water pixels was performed. (b) RGB image of Sentinel-1 backscattering coefficient images that shows temporal changes (red: February 19, 2017, green: October 13, 2020, blue: December 21, 2023).

Fig. 3(b) is the RGB image of Sentinel-1 backscattering coefficient images that show temporal changes in the scene (red: February 19, 2017, green: October 13, 2020, blue: December 21, 2023). Fig. 4(a) shows the histogram of pixels sampled from February 19, 2017, image (blue dots), the modeled PDF (red line), and the set threshold (red dashed line). Fig. 4 shows the thresholds calculated for the 183 images. The thresholds are high when the waves are strong and low when the sea is calm. This seems to be due to the weather conditions, such as wind (Alsdorf et al., 2007).

Fig. 4. Threshold estimation based on PDF modeling. (a) The histogram of public water pixels (blue dots), modeled probability density function (red line), and estimated threshold of February 29, 2017, SAR image. (b) The thresholds that are estimated in each scene are estimated in order.

Fig. 5 shows, for one example of an arbitrary reclamation project, the calculated R (blue line), the cumulative mean (orange line), and the reclamation period (pink region) from the initiation time (black dashed line) to the completion time (red dashed line). The estimated t_i and t_c were 2 December 2020 and 13 April 2023, respectively. The cumulative mean was only calculated before t_c, because the purpose of the calculation of the cumulative mean is to find out tic. In 2023, the reclamation rate seems to be decreasing. It seems that, after the reclamation is completed, drying and consolidation work are performed, and as the reclaimed land dries and even, the backscattering coefficient gets low (Ezzahar et al., 2020). And the missed detection of the reclaimed land due to the highly set thresholds to avoid false alarms could lead to such a decrease in the reclamation rate.

Fig. 5. Estimated R (blue line) of each scene,cumulative mean (orange line), and the period (pink region) from ti (black dashed line) to tc (red dashed line). The decrease in the reclamation rate after mid–2023 seemed to be due to soil drying and consolidation processes following the completion of reclamation construction.

Fig. 6(a) shows the reclamation rates of December 21, 2023, where the status of each plan is represented by a value between 0 and 1. Fig. 6(a) R₂ and R₃ represent reclamation projects in progress, with reclamation rates of 0.1 and 0.83, respectively. Fig. 6(b) illustrates the temporal variation of reclamation rate for Fig. 6(a) R₁ (blue line) along with the reclamation period (pink area). Fig. 7(a) shows the temporal variation of the number of completed (blue), not initiated (orange), and ongoing (green) reclamation projects classified for each image. Fig. 7(b) represents the reclaimed area of each SAR image (blue).

Fig. 6. The reclamation status estimation result. (a) Result of estimation of reclamation status on December 21, 2023. R₂ and R₃ are ongoing reclamation projects, and the reclamation rate is 0.1 and 0.83, respectively. (b) Time-series profile of reclamation rate of R₁ in (a). The start date of reclamation is April 4, 2020, and the end date of reclamation is October 20, 2021.

Fig. 7. The time-series profile of (a) the number of completed (blue), not initiated (orange), ongoing (green) reclamation projects, and (b) reclaimed area (blue).

Between 2017 and 2023, 49 planned reclamation projects were completed out of 70 sites designated in the basic public water reclamation plan for the Busan–Jinhae area. Most reclamation projects were completed during the study period, and Fig. 7(a) clearly shows that the number of ongoing reclamation projects gradually decreased over time. However, Fig. 7(b) indicates that reclaimed land was created at a relatively constant rate on average from early 2017 to the end of 2023. This suggests that while the number of small-scale reclamation projects in the Busan–Jinhae area has decreased, large-scale reclamation projects have continued steadily. The decrease of small-scale reclamation project is due to the early completion of the Ungdong reclamation site, located in upper Fig. 6(a) R₁ and lower Fig. 6(a) R₂, where small-scale reclamation projects are distributed in the intervals of large-scale reclamation projects. The phenomenon, steadily continuing large-scale reclamation, appears to reflect the local characteristics of the Busan–Jinhae area, where there are many harbor constructions, including the reclamation projects of Busan New Port and Jinhae New Port, planned by 2030. Little reclamation activity occurred in 2019, resulting in no significant increase in reclaimed land, while in 2020, although the number of ongoing reclamation projects was small, the area of reclaimed land increased sharply.

This is likely due to the progress of large-scale reclamation near Fig. 6(a) R₁. In this region, the reclamation had been stopped and delayed due to environmental problems. Most of the reclamation projects in the study area are composed of the large-scale reclamation project for the construction of the Busan New Port, and the reclamation projects of the Busan New Port are initiated sequentially. So, delay of a single reclamation project could affect the whole trend of the reclamation area, as shown in Fig. 7(b). As of December 2023, 21.27 km² out of a total planned area of approximately 25.96 km² in the study area had been reclaimed, corresponding to 81.93% of the planned area.

5. Discussion

The reclamation status classification results for December 21, 2023, were evaluated using the 2024 cadastral map, the construction safety management integrated information, serviced by the Korean Ministry of Land, Infrastructure and Transport, and Google Earth time-series images. Fig. 8 shows the variation in overall accuracy with the CVF threshold. In all cases, overall accuracy was higher than when CVF was not used, which is represented as ‘none’ in Fig. 8. Overall accuracy is expected to be low if the CVF threshold is too low due to the over-acceptance of the changes, and vice versa. We used a threshold of 0.3 to get the highest performance of the methodology and avoid the previously described effect. The results are summarized in Table 1. Among 70 reclamation projects in the basic public water reclamation plan, 68 were correctly classified, while two “Not initiated” plans were misclassified as “Completed.” Consequently, the overall accuracy was 97.14% with the CVF threshold of 0.3. The overall classification accuracy was 97.14%.

Fig. 8. Overall accuracy variation due to the set of CVF thresholds. In all cases, the overall accuracy seemed to be much higher than that of CVF, which was not used, represented as‘None’. Overall accuracy gets lower with the increase of the CVF threshold.

Table 1. Result of the reclamation status of basic public water reclamation plans

The proposed method did not show specifically high performance compared to other methods. The method suggested by Ai et al. (2022) showed overall accuracy from 83.05% to 88.57%, Shi et al. (2022) showed 99%, and Huang et al. (2025) showed 97.04% of sensitivity and 96.76% of specificity. However, while the method proposed in other studies does not provide the result with higher than 1 month of temporal resolution, the method proposed in this study provides 12 days of temporal resolution, that of a repeat cycle of Sentinel-1A. Using only SAR images, which can acquire images regardless of the weather conditions, unlike optical sensors, it was able to get such high temporal resolution.

As shown in Fig. 5, the reclamation rate decreases after mid-2023. This phenomenon appears to be associated with construction processes such as soil drying and consolidation. These processes are required to form stable ground conditions that can be used after reclamation, therefore are not expected to affect the determination of the completion date of reclamation. In future research, such behavior of the reclamation rate could be used to define a new stage between the physical completion of the reclamation and the subsequent administrative procedures, allowing the monitoring and evaluation of the stability of the newly created land.

We classified the reclamation status of the basic public water reclamation plan with the proposed method; two “Not initiated” reclamation projects were misclassified as “Completed”. In these misclassified cases, 152 out of 203 pixels in one plan and 26 out of 48 pixels in the other appeared to be pixels representing beaches. Both areas were relatively small in scale, and they were classified as completed in the consecutive images acquired on February 21, 2019, and March 5, 2019. The thresholds applied for reclamation detection in the 44th image (February 21, 2019) and the 45th image (March 5, 2019) out of 183 SAR images were –17.99 and –19.05, respectively, values close to –18.04, which corresponds to 3σ within the 183 thresholds. This indicates that the thresholds set during this period were outliers.

Therefore, in the two reclamation projects that were small in scale and contained a number of beach pixels, numerous false alarms occurred, leading to misclassification as completed reclamation. Such misclassifications are attributed to the spatial analysis of satellite images. Future work will aim to enhance performance by detecting reclaimed land through temporal analysis of SAR backscatter signals. Temporal analysis may enable accurate detection of reclaimed land regardless of the spatial scale of the reclamation site.

The reclamation detection method proposed in this study is a simple approach that distinguishes water from non-water using a threshold calculated from the modeled PDF of water. Therefore, it is likely to be less applicable in regions such as the west coast of Korea, where tidal flats are well developed. In addition, in the estimation of reclamation progress, noise caused by ghost phenomena and coastal facilities can occur, which may significantly affect the classification of reclamation status in small-scale reclamation sites. Finally, because this method monitors only the reclamation projects included in the basic public water reclamation plan, it has limitations in detecting unplanned reclamation activities in public water. These issues could potentially be mitigated by employing a method that directly analyzes the time-series signals of SAR backscatter.

6. Conclusions

This study proposed a method to monitor basic public water reclamation plans using SAR images to address the limitations of optical satellite imagery, which is constrained during the summer season due to concentrated rainfall. A total of 183 Sentinel-1 SAR images acquired between 2017 and 2023, together with the basic public water reclamation plan, were used to detect reclamation of public water in each Sentinel-1 image and to estimate the reclamation rate for each reclamation project. Reclaimed lands were detected using a thresholding approach, where thresholds were calculated from the PDF modeled as a normal distribution of SAR backscatter coefficients sampled from areas identified as permanent public water.

Based on the estimated reclamation rate, reclamation status for each plan was classified approximately every 12 days into three categories: Not initiated, Ongoing, and Completed. The classification results for December 21, 2023, were evaluated using the construction safety management integrated information system, the 2024 cadastral map, and Google Earth time-series images. The overall accuracy was 97.14%, indicating that the proposed method is suitable for use in monitoring reclamation. The overall accuracy was always higher than when CVF was not used, regardless of the CVF threshold. And, as the CVF threshold gets higher than 0.3, the overall accuracy gets lower.

Using the proposed method, reclamation progress was analyzed for each of the 70 basic public water reclamation plans from 2017 to 2023. Over time, the number of ongoing reclamation projects decreased, while the total area of reclaimed land steadily increased. This indicates that small-scale reclamation projects have diminished, whereas large-scale reclamation projects were steadily ongoing from 2017 to 2023. The decrease in small-scale reclamation projects is because of the completion of the Ungdong area, where small-scale reclamation projects are distributed in intervals between large-scale reclamation projects. The ongoing large-scale reclamation projects are due to the sequential initiation of the Busan New Port's large-scale reclamation projects.

As shown in Fig. 5, the decrease in reclamation rate after mid-2023 is likely related to soil drying and consolidation. This behavior does not appear to affect the determination of the completion date of the reclamation and could be utilized in future studies to define an intermediate stage between physical completion and administrative procedures, enabling the assessment of land stability.

As a result of the reclamation status classification, two “Not initiated” reclamation projects were misclassified as “Completed”. This was determined to have occurred because the threshold used in the reclamation detection process produced outlier values, which introduced significant errors in estimating the reclamation rate for small-scale reclamation projects with a high proportion of pixels representing beach.

The method proposed in this study can facilitate efficient and timely surveys of reclamation in public water along Korea’s coasts, which are characterized by long and complex shorelines. Given Korea’s seasonal characteristics of concentrated rainfall during the summer, it appears feasible to monitor reclamation of public water using satellite systems. Furthermore, the approach is expected to contribute to the sustainable development of public waters and support decision-making in marine spatial policy.

Author Contributions

Conceptualization: Jung HC, Hwang D-H, Lee SJ, Kim SJ; Data curation, Formal analysis: Jung HC, Lee SJ; Funding acquisition: Jung HC, Hwang D-H, Investigation, Methodology: Jung HC, Hwang D-H, Lee SJ, Kim SJ; Project administration, Supervision: Jung HC; Validation: All authors; Writing–original draft: Lee SJ; Writing–review & editing: All authors.

Conflicts of Interest

No potential conflict of interest relevant to this article was reported.

Funding

This research was supported by the National Research Foundation of Korea (NRFK) grant funded by the South Korea Government (RS-2021-NR058189), the Korea Institute of Marine Science & Technology Promotion by funded the Ministry of Oceans and Fisheries through the “Development of simulation technology for maritime spatial policy” (RS-2022-KS221620), Yonsei University Future-Leading Research Initiative (2025-22-0042), and Global-Learning & Academic research institution for Master’s, PhD Students, and Postdocs(LAMP) program of the National Research Foundation of Korea (NRF) grant funded by the Ministry of Education (RS-2024-00442483).

Data Availability Statement

Data sharing is not applicable to this article as no new data were created or analyzed in this study.