1. Introduction
South Korea has a high proportion of mountainous terrain; therefore, land reclamation has been continuously carried out to secure agricultural land and to construct ports, industrial complexes, and new towns. Reclaimed land, which is an artificial ground formed through dredging, filling, and compaction, is prone to long-term consolidation settlement. The construction of high-load facilities on such ground directly affects the serviceability and safety of superstructures (Cho, 2008). Consequently, identifying the deformation characteristics of reclaimed ground and quantitatively evaluating the magnitude of long-term settlement are essential not only at the design stage of reclamation projects but also for post-construction maintenance.
Recently, as reclaimed land has been increasingly used for various high-value-added infrastructures such as port facilities, logistics complexes, and residential areas, interest in ground stability and long-term settlement behaviour has grown (An et al., 2025). However, in Korea, most previous studies on large-scale reclamation projects such as Saemangeum and Sihwa Lake have focused on changes in soil environments and agricultural use, and relatively few studies have quantitatively analysed differences in ground behaviour at each construction stage using long-term time series data. In particular, there is still a lack of research that compares and evaluates stage-dependent deformation characteristics in areas where multiple stages of reclamation development coexist (An et al., 2025).
Conventional ground-based monitoring methods such as survey benchmarks and leveling can provide highly accurate displacement measurements, but they require substantial time and cost for the installation and maintenance of observation points and are limited in their ability to monitor large areas over long periods (Sun et al., 2018). Accordingly, satellite-based remote sensing, especially interferometric synthetic aperture radar (InSAR), has been widely used for monitoring surface deformation (Lee et al., 2025). Among these techniques, persistent scatterer InSAR (PSInSAR) extracts reliable persistent scatterers from long-term, repeatedly acquired SAR data and estimates millimetre-level displacement time series, making it particularly suitable for urban areas with dense artificial structures (Ferretti et al., 2001).
However, conventional PSInSAR assumes a single continuous time series over the entire observation period, which limits its ability to fully capture time-varying deformation behaviour in reclamation areas where ground-surface properties and dominant scatterers frequently change due to construction activities (An et al., 2025). To overcome this limitation, the time-segmented PSInSAR (TS-PSInSAR) technique has recently been proposed, in which the analysis period is divided into several segments, and PSInSAR is applied separately to each segment (Moon et al., 2022). By segmenting the time series at points corresponding to changes in loading conditions or construction phases, TS-PSInSAR maintains temporal coherence within each segment, allows independent estimation of displacement rates and patterns, and facilitates comparison of deformation behaviour between segments (Moon et al., 2022).
Meanwhile, for consolidation settlement in reclaimed land, the hyperbolic method and its variants have been widely used to predict settlement as a function of time (Kwak et al., 2025). The hyperbolic method enables estimation of ultimate settlement and residual settlement from displacement data measured over a limited observation period and is thus useful for assessing post-construction ground stabilization (Jang & Han, 2022). However, because the classical hyperbolic method is sensitive to changes in ground and loading conditions and its predictive performance strongly depends on data quality (Jang & Han, 2022), it is necessary to examine its applicability in conjunction with PSInSAR time series.
Mokpo New Port is one of the major logistics hubs on the southwestern coast of Korea and plays an important role in regional maritime transportation and industrial activities. Continuous monitoring of ground deformation is therefore necessary to ensure the long-term stability and safe operation of port facilities. In particular, understanding the spatial and temporal characteristics of settlement can provide valuable information for maintenance planning and infrastructure management.
Several studies have focused on ground deformation associated with reclaimed land in the Mokpo area. Kim et al.(2008) utilized JERS-1 and ENVISAT SAR data to identify long-term subsidence patterns in reclaimed regions of Mokpo and demonstrated the effectiveness of PSInSAR for monitoring settlement in coastal environments. Subsequently, Kim et al.(2010) combined PSInSAR observations with a hyperbolic settlement model to estimate consolidation behavior and predict future settlement trends. These studies provided important insights into the applicability of satellite-based deformation monitoring and empirical settlement analysis for reclaimed land management.
Despite previous achievements, several limitations remain. Most earlier studies in the Mokpo area relied on historical SAR datasets with relatively sparse temporal sampling compared with Sentinel-1 observations. In addition, many studies focused primarily on line-of-sight(LOS) displacement or vertical settlement, while horizontal deformation components received less attention. Furthermore, the long-term deformation characteristics of Mokpo New Port, which consists of reclaimed land developed during different construction stages and subjected to varying loading conditions, have not been comprehensively evaluated using recent Sentinel-1 time-series observations.
In this study, we analyse long-term ground deformation in the reclaimed land of Mokpo New Port, where multiple construction stages coexist, by applying the TS-PSInSAR technique to Sentinel-1 SAR data acquired between 2017 and 2021. LOS displacements derived from ascending and descending tracks are decomposed into vertical and east–west horizontal components. The time series of representative persistent scatterer (PS) groups with different construction stages and land-use types are compared to characterise long-term consolidation behaviour within the reclaimed land. Furthermore, hyperbolic regression analyses are applied to representative PS time series to estimate ultimate settlement and settlement progression, thereby providing insights into the long-term stabilization level of the Mokpo New Port reclamation area and supporting the prioritization of future maintenance actions.
2. Materials and Methods
2.1. Study Area and Data
The study area is the reclaimed land of Mokpo New Port located in Yudal-dong, Mokpo City, Jeollanam-do, on the southwestern coast of South Korea (Fig. 1). Mokpo New Port is a national trade port constructed by reclaiming the area around Gohado Island at the estuary of the Yeongsan River, during which the adjacent Heosado and Janggu-do islands were merged into a single island through reclamation (Mokpo Port Authority, 2026). This area is a typical mixed-use reclamation site where port facilities, logistics, residential areas, and green spaces coexist on land formed by dredging and filling at the Yeongsan River estuary, and its ground properties vary markedly in space depending on the reclamation stage and land-use type.

Fig. 1. The study area and Google Earth optical images from 2002 to 2020, showing the sequence of the Mokpo New Port reclamation area.
The reclamation processes and subsequent changes in surface utilization were identified using high-resolution optical satellite images acquired between 2002 and 2020 from Google Earth (Fig. 1). Each image covers the same spatial extent and geographic coordinates, allowing time-series comparison of reclamation expansion, construction of roads and buildings, and changes in land cover. These optical image interpretations were then used to relate SAR-derived ground deformation patterns to the different reclamation stages.
The remote sensing data used to derive ground deformation in this study are SAR images acquired by the Sentinel-1 satellites operating at C-band (5.4GHz). In this study, we used VV-polarized single look complex (SLC) data obtained in the Interferometric Wide (IW) mode with a VV+VH polarization combination. All SAR datasets were freely downloaded from the Copernicus Data Space Ecosystem of the European Space Agency (ESA). The observation interval is 12 days for the orbit and frame covering the study area. The basic acquisition parameters of the Sentinel-1 data used in this study are summarized in Table 1.
Table 1. Sentinel-1 SAR datasets used in this study

These acquisition geometry parameters are used to construct the coefficient matrix for decomposing the ascending and descending LOS displacements into east–west horizontal and vertical displacement components. Although Sentinel-1 continued to acquire images only from the ascending track after 2022, these data were excluded from the present time-series analysis because relying on a single track would increase uncertainty in the separation of vertical and horizontal components and would hinder consistent interpretation over the entire study period.
2.2. TS-PSInSAR Data Processing
We used the 2017–2021 Sentinel-1 SAR datasets where both ascending and descending tracks were available. The TS-PSInSAR technique was applied by dividing the PSInSAR analysis period into one-year segments so as to secure statistically sufficient image numbers for PSInSAR while maximising temporal coherence (Moon et al., 2022). The overall processing workflow is illustrated in Fig. 2.

Fig. 2. Workflow of the TS-PSInSAR processing. The annual PSInSAR results for the ascending (asc17–asc21) and descending (des17–des21) tracks over 2017–2021 were first temporally linked using an in-house software called the tscon program and then converted into east–west (x) and vertical (z) displacement components using the psad and psxz programs.
First, for each year and for each track, namely ascending (asc17, asc 18, asc 19, asc 20, asc21) and descending (des17, des18, des19, des20, des21), PSInSAR pre-processing begins in ESA’s Sentinel Application Platform (SNAP) software (European Space Agency, 2026) by selecting a master date using the InSAR Stack Overview module, and then the data were processed through S-1 Back Geocoding, S-1TOPS Deburst, Product Subset, Interferogram Formation, and StaMPS Export. PSInSAR processing was then carried out for each year using the Stanford Method for Persistent Scatterers(StaMPS). StaMPS is free software to extract ground deformation from time series of synthetic aperture radar datasets (StaMPS, 2026).
Subsequently, the annual PSInSAR results were temporally integrated over the entire 2017–2021 period using in-house software, tscon. To maintain spatial continuity of the linked PSs for each reference-year PS (2021 in this study), the closest PS within a horizontal distance of 100m was identified and connected to form the annual time series for the ascending, asc (17–21), and descending, des (17–21) datasets.
Because the locations of PSs selected from the two tracks may differ even during periods when both ascending and descending datasets are available, each PS extracted from the descending track was used as a reference to connect the nearest ascending-track PS within a horizontal distance of 100 m using in-house software, psad. This procedure yielded two LOS displacement time series estimated from different viewing geometries, which were then converted into east–west horizontal displacement components, x (17–21), and vertical displacement components, z (17–21), using another in-house software, psxz. The algorithm implemented in psxz is described in detail in the following section.
2.3. Conversion of Displacement Components
To convert the ascending and descending LOS displacements obtained from PSInSAR into vertical and horizontal displacement components, a linear transformation based on the acquisition geometry and orbit azimuth was applied. Fig. 3 conceptually illustrates the geometric relationship between the ascending/ descending LOS displacements and the east–west and vertical components, showing that the LOS displacements have different directions depending on the viewing geometry of the ascending and descending tracks and that the actual displacement vector is decomposed into horizontal and vertical components accordingly. Using the incidence angles and orbit azimuth angles of Sentinel-1, the relationship between the ascending and descending LOS displacements dasc and ddes, the east–west horizontal displacement component dx and the vertical displacement component dz can be expressed as follows(Ferretti et al., 2001; Fuhrmann & Garthwaite, 2019).
\(\begin{align}\binom{d_{a s c}}{d_{d e s}}=\left(\begin{array}{ll}-\sin \theta_{a s c} \cos \alpha_{a s c} & \cos \theta_{a s c} \\ -\sin \theta_{d e s} \cos \alpha_{d e s} & \cos \theta_{d e s}\end{array}\right)\binom{d_{x}}{d_{z}}\end{align}\) (1)

Fig. 3. Schematic illustration of the conversion of two-dimensional displacement (dx, dz) using ascending and descending LOS displacements (dasc, ddes) . The vector \(\begin{align}\vec d\end{align}\) is the displacement projected onto the x-z plane, while θasc and θdes are the incidence angles.
Here, θasc and θdes denote the incidence angles of the ascending and descending tracks, respectively, and αasc and αdes denote the corresponding orbit azimuth angles. The above relationship is valid only at low to mid latitudes where the azimuth angles are close to the north–south direction and the values of sinαasc and sinαdes are sufficiently small(Lee et al., 2022). As the orbit azimuth angles are 349.86097° and 190.13646°, respectively (Table 1), this condition is satisfied for the study area.
2.4. Settlement Model and Prediction
To predict the long-term settlement behaviour of the reclaimed land, a hyperbolic empirical model was applied to the cumulative vertical displacement time series derived from PSInSAR. Tan et al. (1991) and Jang and Han (2022) transformed the original hyperbolic regression equation into a linear form and estimated the parameters using linear regression. In this study, however, a hyperbolic-type nonlinear regression model was adopted so that the onset time and magnitude of settlement could be specified more intuitively, as shown in Eq. (2).
\(\begin{align}d_{z}(t)=\frac{b t}{t-a}+c\end{align}\) (2)
where dz (t) denotes the cumulative vertical settlement (in mm) at elapsed time t measured from the reference epoch. The coefficients a, b, and c are determined through nonlinear regression, where a can be interpreted as a virtual origin on the time axis, b characterizes the evolution pattern of the settlement rate, and c reflects the initial vertical displacement.
Because of the asymptotic nature of the hyperbolic function, dz (t) converges to a constant value as time becomes sufficiently large, and this limiting value is defined as the predicted ultimate settlement dz,∞. In other words, dz,∞ represents the long-term consolidation settlement estimated from the fitted hyperbolic curve.
\(\begin{align}d_{z, \infty}=\lim _{t \rightarrow \infty}\left(\frac{b t}{t-a}+c\right)=b+c\end{align}\) (3)
The degree of settlement progression was defined as an index indicating how far the cumulative vertical displacement at the end of the observation period has progressed towards the predicted ultimate settlement. Let dz,obs denote the cumulative vertical settlement at the final observation epoch; then the settlement progression P was computed as a function of these two quantities.
\(\begin{align}P=\frac{d_{z, o b s}}{d_{z, \infty}} \times 100(\%)\end{align}\) (4)
The degree of settlement progression was expressed as a percentage, and values closer to 100 were interpreted as indicating that the settlement of the corresponding PS group is approaching a long-term equilibrium state.
3. Results
3.1. Surface Displacement of Mokpo New Port during 2017–2021
Fig. 4 shows the spatial distribution of PS showing displacement vectors in vertical and east–west horizontal displacement rates for the Mokpo New Port reclamation area, estimated using the TS-PSInSAR technique for the period 2017–2021. Over most of the study area, the vertical displacement is dominated by settlement, and even within the same reclamation area, the magnitude and direction of displacement vary spatially between different zones. These spatial differences can be interpreted as reflecting variations in reclamation timing, surface land-use types, superstructure loading conditions, and the degree of consolidation of the underlying ground.

Fig. 4. TS-PSInSAR-derived displacement and displacement vectors for the Mokpo New Port reclamation area during 2017–2021. The vectors have two components: the east–west displacement rate (mm/yr) and the vertical displacement rate (mm/yr) at each PS. The colour of each PS indicates the cumulative vertical displacement (mm) over the five years.
When the study area shown in Fig. 4 is divided into four quadrants, the northern sector exhibits relatively large vertical displacement rates indicative of ongoing settlement, whereas the southern sector maintains comparatively stable ground conditions. In the north-western zone around area A and in the northeastern zone around areas B1–B3, zones of locally concentrated vertical displacement rates are observed, and the horizontal displacement vectors tend to exhibit a coherent directional trend. In particular, the displacement patterns change along road networks, infrastructure corridors, and reclamation boundaries, indicating that ground behaviour is heterogeneous even within individual subareas of the reclamation site.
By contrast, the south-western zone, including area C, and the south-eastern zone, including area D, show relatively stable displacement behaviour. The south-western sector, which is characterised by extensive paved surfaces and open spaces, generally exhibits displacement vectors of small magnitude with no pronounced directional pattern. Although some localised deformation is observed in the south-eastern sector, the overall displacement appears scattered, reflecting differences in surrounding land use and reclamation history.
Fig. 5 presents the cumulative vertical displacement (mm) time series of representative PSs for groups A, B1, B2, B3, C, and D in a single plot, allowing comparison of relative settlement behaviour among major zones within the reclaimed area. The horizontal axis indicates the elapsed days from the reference epoch, February 11, 2017, the beginning of the satellite data used in this study.

Fig. 5. Comparison of cumulative vertical displacement time series for PS groups at locations A, B1, B2, B3, C, and D. The time axis represents the elapsed days from the reference epoch.
In groups A, B1, B2, and B3, settlement within the reclaimed land is evident, but the onset time and degree of settlement progression differ among the groups. For example, group A exhibits a typical hyperbolic-type vertical settlement pattern from the early observation period, whereas settlement in groups B1, B2, and B3 starts after approximately 730 days (February 2019). The B1 curves continue to decrease in a hyperbolic fashion towards the end of the observation period, while the B2 and B3 curve enters a mild convergence phase around day 1095 (February 2020) and then show renewed settlement, indicating a more complex behaviour. In contrast, the C and D curves remain close to zero displacement or show only small-magnitude movements over the entire period, suggesting that the ground in these areas has largely stabilised.
3.2 Prediction of Settlement Progress
To quantitatively evaluate the long-term settlement behaviour of each group, the coefficients of the hyperbolic regression model described above were estimated by nonlinear regression using the cumulative vertical displacement time series at the representative PSs. Figs. 6–9 present the spatial distribution and settlement prediction results for groups A, B1, B2, and B3, respectively, and Table 2 summarises these results to facilitate comparison among the groups. On this basis, both the predicted ultimate settlement and the degree of settlement progression were assessed for each PS group.
Table 2. Settlement prediction results for groups A, B1, B2, and B3 within the Mokpo New Port reclamation area

RMSE: root mean squared error, Regr.: regression.
Fig. 6 illustrates the spatial distribution of cumulative vertical displacement and the settlement prediction results for group A. In the left panel, the colour of each PSInSAR point represents the vertical displacement rate estimated for 2017–2021, and the circular outline indicates the extent of group A. This area exhibits a pattern of concentrated settlement relative to the surroundings, and the representative PS lies within this concentration zone, indicating that group A corresponds to a sector of pronounced long-term settlement within the reclaimed land.

Fig. 6. Displacement vectors and settlement prediction for group A. The left panel shows the location of the group with displacement vectors in vertical and east-west directions, and the right panel shows the vertical displacement time series derived from TS-PSInSAR, with the solid line indicating the fitted hyperbolic regression curve.
In the right panel, the points show the cumulative vertical displacement time series dz (t) at the representative PS, and the solid line denotes the fitted hyperbolic regression curve. The time series displays a consolidation-type curve in which relatively rapid settlement occurs in the early observation period, followed by a gradual decrease in slope and asymptotic convergence over time. The regression curve reproduces this change in curvature consistently over the entire period, and the model predictions remain close to the observations even in the gently varying segment towards the end of the record. The dashed horizontal line indicates the predicted ultimate settlement dz,∞, which is estimated to be –176 mm for group A. The cumulative settlement at the end of the observation period is –133 mm, corresponding to 76% of the predicted ultimate settlement, implying that a substantial portion of the settlement has already occurred in group A but that some residual settlement potential remains.
Fig. 7 shows the results for group B1, where the representative PS is located near an eastern road intersection. In the left panel, zones of relatively high vertical displacement rates are distributed around the road, and group B1 lies along the boundary between this settlement concentration and the road infrastructure, indicating that significant settlement is occurring where road loads and reclamation boundary conditions interact.

Fig. 7. Displacement vectors and settlement prediction for group B1. The left panel shows the location of the group with displacement vectors in vertical and east-west direction, and the right panel shows the vertical displacement time series derived from TS-PSInSAR, with the solid line indicating the fitted hyperbolic regression curve.
In the right panel, the cumulative vertical displacement time series is initially stable but begins to show settlement after about 700 days, followed by a gradual flattening towards the end of the observation period. The fitted hyperbolic regression curve approximates this nonlinear settlement pattern with a single continuous function, simultaneously capturing changes in settlement rate during the early, middle, and late stages. The predicted ultimate settlement dz,∞ for B1 is –272 mm, while the cumulative settlement at the end of the observation period is –165 mm, corresponding to 61% of the predicted ultimate settlement and representing the lowest degree of settlement progression among the four groups (Table 2). Thus, despite having already experienced substantial settlement during the observation period, group B1 is evaluated as a zone with a high potential for further long-term settlement.
Fig. 8 shows the results for group B2; in the left panel, group B2 covers an area where settlement is dominant, but westward horizontal displacement is also observed. This indicates that even within the same reclamation area, differences in loading conditions and subsurface structure can lead to spatially varying magnitudes of settlement.

Fig. 8. Displacement vectors and settlement prediction for group B2. The left panel shows the location of the group with displacement vectors in vertical and east-west direction, and the right panel shows the vertical displacement time series derived from TS-PSInSAR, with the solid line indicating the fitted hyperbolic regression curve.
In the right panel, the time series exhibits little displacement change during the initial observation period, followed by an increase in settlement rate after about 700 days, a gentle convergence up to around day 1070, and then a renewed acceleration, suggesting the occurrence of two distinct settlement-inducing events. The fitted hyperbolic curve represents these two stages of settlement rate change within a single functional form, encompassing both the early quasi-stable segment and the subsequent phase of accelerated settlement. The predicted ultimate settlement dz,∞ for B2 is –103 mm, while the cumulative settlement at the end of the observation period is –89 mm, corresponding to 86% of the predicted ultimate settlement and indicating that group B2 is the closest to long-term equilibrium among the groups analysed (Table 2).
Fig. 9 presents the settlement prediction results for group B3. Group B3 is located in the reclaimed area adjacent to B2 on the eastern side, and the left panel shows that the vertical displacement rates in this zone are slightly larger than those in B2. While B2 exhibits westward horizontal displacement, B3 shows displacement towards the east, highlighting the spatial heterogeneity in settlement magnitude and direction even within the same reclaimed fill.

Fig. 9. Displacement vectors and settlement prediction for group B3. The left panel shows the location of the group with displacement vectors in vertical and east-west directions, and the right panel shows the vertical displacement time series derived from TS-PSInSAR, with the solid line indicating the fitted hyperbolic regression curve. The predicted ultimate settlement is approximately –167 mm, and the settlement progression is 81.4%.
As with B2, B3 appears to be influenced by two distinct settlement mechanisms. The fitted hyperbolic curve approximates this staged settlement behaviour through two changes in slope, and the predicted ultimate settlement dz,∞ is estimated to be –144 mm. The cumulative settlement at the end of the observation period is –136 mm, corresponding to a settlement progression of 95% (Table 2), indicating that although group B3 has reached a slightly more advanced level of convergence than group A, a non-negligible amount of residual settlement remains.
Table 2 summarises the time-series and regression results described above. In the table, a, b, and c denote the hyperbolic regression coefficients for each PS group; the observed settlement dz,obs represents the cumulative settlement from the reference year 2017 to the end of 2021; and dz,∞ (= b + c) corresponds to the predicted ultimate settlement. The settlement progression P is expressed as a percentage, where values closer to 100% indicate that settlement is approaching completion. The settlement progression values for groups A, B1, B2, and B3 are 76%, 61%, 86%, and 95%, respectively, showing that B3 has the highest progression and B1 the lowest. Accordingly, group B3 is interpreted as being closest to long-term equilibrium, whereas group B1 requires the greatest attention for future settlement management, with groups A and B3 exhibiting intermediate levels of progression.
Taken together, these observation- and prediction-based results clearly demonstrate that the major PS groups within the Mokpo New Port reclamation area exhibit different ultimate settlement magnitudes and degrees of settlement progression. This differentiation provides a critical basis for assessing zone-dependent settlement risk in the design and maintenance of structures within the reclamation area and for prioritising intensive monitoring and reinforcement measures in sectors with low settlement progression.
4. Discussion
4.1. Spatial Patterns of Displacement Vectors
The spatial distribution of vertical displacement rates and east– west horizontal displacement vectors observed in the Mokpo New Port reclamation area during 2017–2021 demonstrates that ground deformation across the reclamation site does not evolve in a uniform manner. Even within the same reclamation area, zones of concentrated settlement coexist with relatively stable zones, which can be interpreted as reflecting spatially non-uniform reclamation timing, filling history, ground improvement conditions, superstructure loading, and surface land-use patterns. In artificially constructed grounds such as reclamation fills, differences in embankment thickness, distribution of soft layers, and drainage conditions typically lead to spatial variations in consolidation rates and residual settlement, and the present results are consistent with these general characteristics.
The east–west horizontal displacement vectors are important because they reveal lateral deformation that cannot be identified from vertical settlement information alone. In general, the east–west component derived from two-dimensional InSAR decomposition provides key clues for interpreting lateral movement along embankments, roads, and reclamation boundaries, and the directionality and clustering of vectors may reflect horizontal deformation driven by load redistribution or contrasts in ground stiffness. In the present study area, vectors in some zones are aligned in a consistent direction, suggesting that the ground is undergoing systematic lateral deformation rather than exhibiting purely random local noise; such patterns, especially where they occur near roadways, infrastructure, or reclamation boundaries, can be interpreted as the result of heterogeneous deformation induced by superstructure loads and property contrasts between the reclamation fill and the original ground.
However, the interpretation of horizontal displacement vectors must also account for the geometric limitations of InSAR (Fuhrmann & Garthwaite, 2019). The two-dimensional displacement fields reconstructed by combining ascending and descending LOS measurements are intrinsically most sensitive to east–west and vertical components, whereas north–south motions are poorly resolved. Accordingly, the horizontal displacement vectors presented in this study should be viewed as representative of east–west horizontal deformation rather than the full horizontal displacement field. Nevertheless, in environments such as coastal reclamation sites, where deformation tends to be aligned with shorelines or structural layouts and thus is often pronounced in the east–west direction, these vectors alone provide meaningful evidence for spatial heterogeneity and help identify potentially vulnerable zones.
4.2. Applicability and Limitations of the Hyperbolic Regression Model
The hyperbolic regression model applied in this study is considered a useful empirical tool for interpreting long-term behaviour from settlement time series and for estimating ultimate settlement or convergence trends. Because the hyperbolic method assumes that settlement rates decrease progressively with time, it has been widely used to predict future settlement in soft ground where consolidation is the dominant mechanism. In particular, when a sufficiently long observation period is available, and the settlement curve exhibits a gradual flattening, the method allows the long-term trend to be estimated relatively robustly even from early-stage data.
However, the applicability of the hyperbolic method is not uniformly guaranteed for all time series. The method implicitly assumes that settlement is governed by a single dominant mechanism, namely a relatively continuous consolidation process, and its suitability decreases when abrupt changes in loading conditions, local re-filling, groundwater fluctuations, or disturbances induced by nearby construction occur. Time series in which settlement rates first decrease and then increase again, or remain stagnant for long periods before re-accelerating, are difficult to describe adequately with a single hyperbolic function; in such cases, a high coefficient of determination does not necessarily imply a physically sound prediction.
The hyperbolic regression model applied to the InSAR time series proved to be a useful tool for empirically estimating long-term settlement trends at consolidation-dominated sites. At such locations, the difference between the ultimate settlement predicted by the hyperbolic fit and the InSAR-derived observed settlement at the end of the monitoring period is on average only about 5 mm, providing sufficient agreement for a quantitative assessment of settlement progression. However, at some sites the explanatory power of a single hyperbolic model is limited owing to changes in loading conditions, structural influences, and constraints on the length and quality of the observation records; in these cases, the hyperbolic method is more appropriately interpreted as a supplementary tool for evaluating settlement stage and relative stability rather than as a definitive predictor of ultimate settlement.
These limitations must also be considered when interpreting the present dataset. At some locations, good agreement between the hyperbolic regression and the observations suggests that settlement is largely governed by relatively simple consolidation behaviour. In contrast, where the slope of the time series changes markedly or variability increases towards the end of the observation period, a single hyperbolic curve is insufficient to fully capture the settlement process, implying that additional factors such as secondary consolidation, load redistribution, changes in drainage conditions, or local structural influences may be contributing to the observed behaviour.
4.3. Implications for Reclamation Management
The spatial distribution of vertical displacement rates and horizontal displacement vectors, together with the hyperbolic regression–based settlement predictions derived in this study, provide practical criteria for developing management strategies for reclaimed land. Zones exhibiting concentrated settlement or large horizontal deformation can be interpreted as areas with relatively high risk of additional settlement and structural damage over the service life of overlying facilities, and these risks are not easily captured by conventional global safety factors considered only at the design and construction stages. Such zones should therefore be prioritised in management plans for measures such as pavement maintenance, drainage improvement, load redistribution (e.g. reconfiguration of embankment fills or superstructure layouts), and restrictions on land use.
Finally, if an analysis framework similar to the present InSAR–hyperbolic approach is implemented from the early stages of reclamation development, the accumulation of time-series data over several years after construction can be used to continuously compare and calibrate design-stage assumptions against actual ground behaviour. This feedback can support more conservative yet economical decisions on design loads, the extent of ground improvement, and drainage conditions in future reclamation projects, and provides a rational basis for determining which zones should be managed with higher priority and under which specific strategies.
5. Conclusions
This study used satellite-based InSAR to monitor long-term ground deformation in the Mokpo New Port reclamation area and reconstructed the spatial distribution of vertical displacement rates and east–west horizontal displacement vectors, thereby revealing the heterogeneous nature of ground behaviour within the reclamation site. Within the reclaimed area, zones with cumulative settlement up to approximately –165 mm coexist with relatively stable zones where settlement remains below about –50 mm, indicating that the magnitude of settlement can differ by more than a factor of three even within a single reclamation project. In some areas, the horizontal displacement vectors exhibit consistent directionality, suggesting that deformation is governed by complex mechanisms involving both vertical and lateral components rather than by simple one-dimensional consolidation alone, although the cause of the lateral displacement cannot be identified from the available data.
The results of this study offer several practical implications for reclamation management. Zones where settlement and horizontal deformation are concentrated can be identified as potential weak areas with elevated risk of structural damage and greater long-term maintenance demands for overlying facilities, and the combination of InSAR monitoring and hyperbolic regression can support rational decisions on maintenance timing, inspection intervals, and prioritisation of reinforcement measures. Furthermore, integrating wide-area InSAR observations with local instrumentation enables the establishment of a feedback system in which design-stage assumptions are continuously verified and refined during the operational phase.
In conclusion, the combined use of satellite InSAR and hyperbolic regression presented in this study demonstrates the potential for efficiently characterising long-term ground deformation over wide spatial extents in reclamation areas and for contributing to the development of maintenance strategies based on these deformation insights. Future work should integrate multiple satellite datasets, three-dimensional displacement reconstruction techniques, and in situ measurements to provide a more quantitative understanding of the three-dimensional ground behaviour of reclaimed land.
Author Contributions
Conceptualization: Lim JM, Lee HY; Data curation: Lim JM; Formal analysis: Lim JM, Oh KS; Funding acquisition: Lee HY; Investigation: Lim JM, Oh KS; Methodology: Oh KS; Project administration, Supervision: Lee HY; Validation, Writing– original draft: Lim JM; Writing–review & editing: Lee HY.
Conflicts of Interest
No potential conflict of interest relevant to this article was reported.
Funding
This work was supported by the Technology Innovation Program (RS-2025-02633135) funded by the Korea Planning & Evaluation Institute of Industrial Technology and the Ministry of the Interior and Safety. This research was also supported by the National Research Foundation of Korea (RS-2019-NR040076) and by the Ministry of the Interior and Safety as a Human Resource Development Project in Disaster Management.
Data Availability Statement
None.
Acknowledgments
None.
Supplementary Materials
None.
참고문헌
- An, J., Kim, D., Lee, C., & Kim, D. (2025). Applicability of InSAR for monitoring ground movements in reclaimed land. Journal of the Korean Geo-Environmental Society, 26(11), 27-41. https://doi.org/10.14481/jkges.2025.26.11.27
- Cho, S.-W. (2008). Influence cf settlement of soft ground on adjacent areas and study 01n settlement prediction [Master's thesis, Yeungnam University], Research Information Sharing Service (RISS). https://www.riss.kr/inkfid=T11648483
- European Space Agency. (2026). SNAP Download. https://step.esa.int/main/download/snap-download/
- Ferretti, C., Prati, C., & Rocca, F. (2001). Permanent scatterers in SAR interferometry. IEEE Transactions 01 Geoscience and Remote Sensing, 39(1), 8-20. https//doi.org10.1109/36.898661
- Fuhrmann, T., & Garthwaite, M. C. (2019). Resolving three-dimensional surface motion with InSAR: Constraints from multi-geometry data fusion. Remote Sensing, 11(3), 241. https://doi.org/10.3390/rs11030241
- Jang, S.-M., & Han, H.-S. (2022). Modification of the hyperbolic method for staged fill. The Journal of Engineering Geology, 32(4), 513-523. https://doi.org/10.9720/kseg.2022.4.513
- Kim, S.-W., Wdowinski, S., Dixon, T. H., Amelung, F., Kim, J.-W., & Won, J.-S. (2010). Measurements and predictions of subsidence induced by soil consolidation using persistent scatterer InSAR and a hyperbolic model. Geophysical Research Letters, 37(5), L05304. https://doi.org/10.1029/2009GL041644
- Kim, S.-W., Wdowinski, S., Dixon, T. H., Amelung, F., Won, J.-S., & Kim, J.-W. (2008). InSAR-based mapping of surface subsidence in Mokpo City, Korea, using JERS-1 and ENVISAT SAR data. Earth, Planets and Space, 60(5), 453-461. https://doi.org/10.1186/BF03352812
- Kwak, T. Y., Hong, S., Lee, J. H., & Woo, S. I. (2025). Revised hyperbolic method for predicting settlement of soft clayground based on weighted nonlinear regression. Geomechanics and Engineering 43(1), 041-57. https://doi.org/10.12989/gae.2025.43.1.041
- Lee, H., Moon, J., & Lee, H. (2022). Activity of Okgye Limestone Mine in South Korea observed by InSAR coherence and PSInSAR techniques. Remote Sensing 14(24), 6261. https://doi.org/10.3390/rs14246261
- Lee, K. R., Ahn, T. Z., Ryu, H. H., & Choi S. K. (2025). Development of satellite-based ground displacement monitoring system for power transmission tunnel construction. Journal of Korean Tunnelling and Underground Space Association, 27(4), 271-285. https://doi.org/10.9711/KTAJ.2025.27.4.271
- Mokpo Port Authority. (2026). Mokpo New Port overview and hinterland development. https://www.mokpoport.com
- Moon, J., Lee, H., & Lee, H. (2022). Elevation change of CookE2 subglacial lake in East Antarctica observed by DInSAR and time-segmented PSInSAR. Remote Sensing, 14(18), 4616. https://doi.org/10.3390/rs14184616
- StaMPS. (2026). Stanford method for persistent scatterers. https://github.com/dbekaert/StaMPS
- Sun, Q., Jiang, L., Jiang, M., Lin, H., Ma, P., & Wang, H. (2018). Monitoring coastal reclamation subsidence in Hong Kong with distributed scatterer interferometry. Remote Sensing, 10(11), 1738. https://doi.org/10.3390/rs10111738
- Tan, T. S., Inoue, T., & Lee, S. L. (1991). Hyperbolic method for consolidation analysis. Journal of Geotechnical Engineering, 117(11), 1723-1737. https://doi.org/10.1061/(ASCE)0733-9410(1991)117:11(1723)