Effects of stepwise tidal flat reclamation on tidal evolution in the East China and Yellow Sea

Because many coastal developments have been continuously occurred in the Yellow and the East China Sea, it is necessary to analyze the effect of persistent topographic change. This study simulated the tidal change in response to stepwise tidal flat reclamation in East China and the Yellow Sea using the MOdelo HIDrodinâmico (MOHID) ocean model. Based on previous studies and historical coastal information maps, we conducted several numerical experiments with reliable coastal topography changes around two areas (Jiangsu Shoalwater and Gyeonggi Bay) from 1990 to 1994 when the most active development took place. The results show that, unlike other components (S2, O1, and K1), the simulated amplitude of the M2 constituent significantly increased with the disappearance of the tidal flat in the Yellow Sea. At the same time, it decreased in the East China Sea. These results are consistent with the quantile regression analysis using observational data. We also found an accumulating effect of tidal energy flux when the reclamation continued, which does not appear in the previous studies. These results indicate persistent man-made tidal flat reclamation in a specific area can cause more remarkable regional tidal changes through tidal energy redistribution and modification.


Introduction
The coastal area of the East China Sea including the Yellow Sea, has experienced many regional topographic changes due to tidal flat reclamation, sea-walls, or dyke construction.The blue dots in figure 1(a) represent areas where major reclamation projects have been undertaken.The Caofeidian Reclamation Project, Tianjin New Coastal Zone Project, and Jiangsu coast reclamation project are representative historical examples in China (Li et al 2004, Wang et al 2011).The western coast of South Korea also has similar projects such as the construction of the Saemangeum dike and the reclamation of Yeongjongdo and Songdo (Choi 2014).From these coastal construction and reclamation projects, 40% of tidal flats have disappeared and the coastal environment including the sea level has significantly changed (Yang et al 2010, Song et al 2013, Choi 2014, Suh et al 2014, Tian et al 2016).
Anthropogenic coastal topography changes lead to changes in water depth and coastline, which is a main factor that can cause changes in local tides.Since these changes have occurred continuously and steadily, changes in regional tidal component are occurred (Lim and Chang 2018, Haigh et al 2020, Pan et al 2022).It is also well known that related sea level changes threaten major coastal cities (Hinkel et al 2014, Neumann et al 2015).Therefore, many attempts have been made to simulate the impact of local topographic change on the tidal patterns, which is the most important factor affecting tidal evolution in the East China Sea, the Yellow Sea and the Bohai Sea (Pelling et al 2013, Song et al 2013, Gao et al 2014).
Song et al (2013) showed that the removal of major tidal flats increased mean M2 amplitudes about 0.11 m in the East China Seas.Gao et al (2014) also investigated the effects of reclamation in the Jiaozhou Bay, China.The simulated M2 decreased by approximately 4%, and M4 increased by 38%.Suh et al (2014) showed that several tidal reclamations increased the amplitude of M2 on the west coast of Korea, and the construction of the Saemangeum sea-wall decreased the amplitude of M2 around the Gunsan of South Korea (Kang et al 2013).Li et al (2018) also investigated the effect of accumulated reclamation on tides in the Xiangshan Bay, China, resulting in decrease in M2 amplitude by 6% and increase in M4 amplitude by 27%.Based on in-situ tide gauge data in the western coast of South Korea, Lim and Chang (2018) showed that the regional difference in amplitude change of the total sea surface height is strongly related to tidal change.
Although several numerical studies have shown the tidal response due to coastal topographic changes around the East China Sea, most studies used an artificially modified (shallower or deeper) bathymetry around tidal flat areas, assuming that tidal flat reclamation or the dyke construction was carried out only once during the model simulation period.However, it is necessary to look at the tidal response due to topography change with time because reclamation is persistently made over a long period of time.
Therefore, this study sequentially changed the coastal topography during the simulation periods.We used MOdelo HIDrodinâmico (MOHID) ocean model, which has been widely used to simulate the coastal marine environments (Lee et al 2018).The model configuration and experimental description are presented in section 2. Section 3 describes the results of model validation and changes in the tidal characteristics in the Yellow Sea and the East China Sea.A summary and discussion are presented in section 4.

Model configuration
This study uses the MOHID ocean model developed at Marine and Environmental Technology Research Center (MARETEC) of the Instituto Superior Ténico, University of Lisbon.The governing equations of MOHID are resolved based on the finite volume method with the Arakawa-C grid.The Boussinesq approximation is assumed, as well as the Hydrostatic equilibrium.The MOHID model uses the ADI (Alternate Direction Implicit: semi-implicit algorithm) method.This method calculates alternatively one component of horizontal velocity implicitly while the other is calculated explicitly (Abbott andBasco 1994, Vaz et al 2009).The MOHID model has been mainly used for environmental change studies in coastal and estuary areas related to tide, tsunami, etc (Martins et al 2001, Leitão et al 2005, Sobrinho et al 2021).
The model domain of this study covers the area from 117 • E to 144 • E and 28 • N to 48 • N, as shown in figure 1 (a).This area includes the East/Japan Sea, East China Sea, Yellow Sea, Bohai Sea and part of the Northwestern Pacific Ocean, so it can simulate the propagation patterns of tidal waves over a wide range.The horizontal resolution is 1/12 • latitude by 1/12 • longitude, and the topographic data are based on a 15 s spatial resolution from The General Bathymetric Chart of the Oceans.The water level of the model is started with a uniform value 2.08 m (cold start) with a one day spin-up period (omitted in the analysis).The model time step is 30 s, and the bi-harmonic coefficient is 10 11 m 2 s −1 with the constant horizontal viscosity (5 m 2 s −1 ), as well as the bottom drag coefficient that is calculated using the von Karman constant and the bottom roughness length (0.0025 m).As an open boundary tidal forcing, we used the NAO.99Jb regional tidal model result provided by the National Astronomical Observatory (NAO) with 16 major constituents (M2, S2, K1, O1, N2, P1, K2, Q1, M1, J1, OO1, 2N2, Mu2, Nu2, L2, and T2) with a spatial resolution of 1/12 • (Matsumoto et al 2000).The analysis area is limited from 117 • E to 130 • E and from 28 • N to 42 • N in order to focus on the tidal dynamics in the Yellow Sea and the East China Sea as shown in figure 1(a).

Numerical experiment
To investigate the tidal evolution related to coastal topographic changes, we must first obtain a reasonable basis for how much and where topographic changes should be made.Therefore, we collected a coastal information map from the Ministry of Oceans and Fisheries.They provide detailed information on the tidal flat position and corresponding water depth at a scale of 1/25 000 via a public data portal (www.data.go.kr/en/).The average depth of the tidal flat around the Korean Peninsula was estimated to be approximately 5.5 m and it is considered as the standard depth for our numerical experiments.
Song et al (2013) simulated sea level change by reclaiming Gyeonggi Bay and Jiangsu Shoalwater, and showed the redistribution of tidal energy around this area.Based on this previous result, we simulated tidal change by similarly reducing water depth in the same area, but the difference was that we varied the reclamation depth and applied topographic change over time.The simulation period was set in the 1990s when coastal construction and reclamation were most active on the west coast of Korea (Choi 2014).To determine whether a rapidly and stepwise changing coastal water depth impacts regional tidal evolution, the amount of changing depth is approximately twice the average tidal flat depth around the Korean peninsula.
The first simulation was conducted in which the water depths of two regions as shown in figures 1(b) and (c) were sequentially reduced by 3 m on January each year for 5 years from 1990 to 1994 (hereafter 'stepwise run').3 m corresponds to about half of the actual average depth (5.5 m) of the tidal flat we surveyed based on coastal information map as we previously mentioned.During the model simulation period, the water depths of the two regions decreased at a rate of 3 m per year.After three reclamations, the water depths of the two regions change as shown in figures 1(d) and (e).Through this simulation, we analyzed the effect of stepwise tidal flat reclamation that gradually occurred in the two regions.
From the model result, harmonic analysis was applied to get tidal components.We used T_Tide version 1.5 provided by Rich Pawlowicz (www.eoas.ubc.ca/∼rich/) with yearly data to see the tidal distribution change from year to year.

Model validation
Figure 2 shows the co-tidal and co-amplitude distributions of the M2, S2, O1, K1 constituent generated by the 0 m run.Locations of amphidromic point in 0 m run are matched well inside the Yellow Sea.However, the amplitude in Korea's western and southern coast is relatively larger than previous cotidal charts (Lee and Beardsley 1999, Fang et al 2004, Xia et al 2006).It may be from different bathymetry, boundary conditions, and model configuration whether data assimilation was applied or not.In this study, the data assimilation was not applied because the artificial depth change had to be tested, so some degree of model bias is inevitable.
The harmonic constants were compared with those of 17 tide gauge stations, as shown by the red dots in figure 1(a).The observed tidal data were collected from the Korea Hydrographic and Oceanographic Agency (KHOA) and the University of Hawaii Sea Level Center (UHSLC).Figure 3 depicts the M2 phase and amplitude at each tide gauge position calculated from the observation and model.Calculated mean absolute errors (MAEs) of the 4major constituents between 17 tide gauges and simulation results are also listed in table 1.Compared to recent assimilation models (EOT20, FES2014, NAO.99Jb), the MAEs of 0 m run is about twice, but our simulate results are not bad compared to other models (FES2012, TPX08, HAMTIDE 11a, DTU10, EOT11a) (Lee et al 2022, Liu et al 2023).Therefore, it can be considered that simulation error of MOHID model does not have a significant effect on various reclamation experiments.It is also expected that follow-up studies will be able to improve the model's performance by applying spatially varying bottom friction coefficients and using other assimilation products (EOT20, FES2014) as a boundary data (Qian et al 2021, Liu et al 2023).
We analyzed the extreme long-term sea level variation of the observed tidal gauge data around the Korean marginal seas by addressing the quantile Sea shows an increasing trend in both the lower and upper quantiles, but the increasing trend of the upper quantile is higher than that of the lower quantile in this area, which is defined as a divergence trend in a previous study (Lim and Chang 2018).Meanwhile, the South Sea of Korea shows a convergence trend, because the increasing trend of the upper quantile is lower than that of the lower quantile.This revealed the regional difference in quantile regression values between the Yellow Sea and the South Sea of Korea, which is the same result as that of Lim and Chang (2018).
The process diagram from the stepwise run of the model simulation shows a tendency very similar to that of the observations (figures 4(c) and (d)).Because the amount of changing depth is approximately twice the average tidal flat depth in    the stepwise run, the difference between the high and low quantiles is larger than that of the observations.This similarity in quantile regression means that the model simulates the dynamics similarly in the Yellow Sea and the East China Sea, although the MAEs is relatively large.However, in the model, the linear regression values of the two regions shown by the red line are approximately zero because the tidal forcing only affects the simulation.This means that local topographic changes cannot cause absolute sea level change but can affect relative sea level change in the two regions.In the Yellow Sea, the lower quantile has a negative value, while the upper quantile has a positive value, and vice versa in the South Sea.These tendencies also imply an amplitude change in regional sea surface height.The divergence tendency of the Yellow Sea indicates an increase in amplitude, and the convergence tendency of the East China Sea indicates a decrease in amplitude.As expected, the simulation results show a similar regional difference tendency, as shown in figure 5.In the early 1990s, the amplitude is 1.18 m in the Yellow Sea and 1.22 m in the East China Sea; however, in late 1994, after the four times topographic changes in two regions, it is 1.30 m in the Yellows Sea and 1.08 m in the East China Sea.We perform a harmonic analysis around the two regions to determine why these regional differences in amplitude change appeared.

Tidal characteristic change of stepwise run
Figure 6 depicts the time series of the amplitudes of the four major constituents (M2, S2, O1 and K1) obtained from harmonic analysis.As the topography changes every January 1, there are corresponding changes in the tidal amplitude in the Yellow Sea and the East China Sea.The diurnal constituents, K1  A detailed inter-comparison among the six different experiments was conducted to determine the effect of depth change on the Yellow Sea and the East China Sea. Figure 7 elucidates the amplitude of the M2 constituent of two regions (The Yellow Sea and the East China Sea) when depth changes in both regions (Jiangsu Shoalwater and Gyeonggi Bay) is applied.The black line is the result of the stepwise run, the red line is the result of the 0 m run, and the blue (purple, green, navy blue) line is the result of the 3 m (6 m, 9 m, 12 m) run, respectively.
The greater the depth change in Jiangsu Shoalwater and Gyeonggi Bay, the higher the M2 amplitude in the Yellow Sea and the lower the M2 amplitude in the East China Sea, which is consistent with figures 5 and 6(c)).The amplitude difference between the 6 m and 12 m runs is smaller than the difference between 0 m and 6 m run.As reclamation progresses to deeper waters, more areas become land.
Therefore, the overall tidal response is smaller because the area of the ocean that is relatively affected by reclamation is reduced.
There is no amplitude change in the M2 constituent over time in any of the experiments, except for the stepwise run.There is a significant amplitude change in the stepwise run as the depth of the two regions changes.In 1990, the M2 amplitude of the stepwise run is the same as that of 0 m run.As the topography changes from 0 m run to 3 m run in January 1991, the amplitude of the M2 constituent of the stepwise run shift equal to that of 3 m run.As the simulation progressed and the depth changed, the M2 amplitude of the stepwise run increased in the Yellow Sea and decreased in the East China Sea.This aspect of the change in M2 amplitude appears to follow the original M2 amplitude value at the changed depth.
Interestingly, this change in the M2 amplitude does not exactly follow the simulation with the change in depth.In other words, the M2 amplitude of the 12 m run in 1994 and the M2 amplitude of the stepwise run in 1994 are not the same.In the Yellow Sea, the M2 amplitude in 1994 was greater than that in the 12 m run.In the East China Sea, the M2 amplitude in 1994 was lower than that of the 12 m run.This result implies that sequential topographic changes can cause larger changes in sea level owing to tidal modification.

Tidal energy flux analysis
In this section, we investigate the tidal energy flux to quantitatively confirm the reason for the tidal amplitude difference between constant and stepwise runs.The tidal energy flux within one tidal cycle, which is adopted from Song et al (2013), can be calculated as follows: where ν is the velocity vector, ρ is the water density, D = H + ζ, H is the water depth, ζ is the tidal elevation, g is the gravitational acceleration and T is the tidal cycle.
Figure 8 shows the differences in the tidal energy flux distribution between each simulation.The upper panel shows the difference between each constant simulation, and the lower panel shows the difference between the stepwise and constant runs.The divergence and convergence of the tidal energy flux differences are shown in red and blue, respectively.Vector represents the directional difference in the tidal energy flux between each simulation.The tidal energy flux difference between constant simulations (0 m, 3 m, 6 m, 9 m, and 12 m run) shows that there are divergences in the energy flux in the Gyeonggi Bay and the Jiangsu Shoalwater.This released energy is either redistributed inside the Yellow Sea from Gyeonggi Bay or outside of the Yellow Sea, mainly from Jiangsu Shoalwater.
To identify the flow of the released energy, we calculated the amount of difference in the tidal energy flux in three areas indicated in figure 8(a) and tables 2, 3.In January 1991, the difference between the 3 m run and the 0 m run indicates that the energy released from the Gyeonggi Bay is 2.61 GW and the energy released from the Jiangsu Shoalwater is 5.72 GW (figure 8(b)).Most of the released energy (6.58 GW) is dispersed to the eastward and southward direction, open sea.The remaining 1.75 GW of energy is redistributed inside the Yellow Sea, which leads to sea level rise.In January 1992, from the difference between the 6 m run and the 3 m run, the released energy from the two regions is 5.82 GW (Gyeonggi Bay) and 3.64 GW (Jiangsu coast), respectively.The dispersed energy is 5.98 GW, and the difference (redistribution energy flux) is 3.48 GW (figure 8(c)).In January 1993, the difference between the 9 m run and the 6 m run shows that Gyeonggi Bay releases 0.79 GW and the Jiangsu Shoalwater releases 1.3 GW.The dispersed energy is 2.44 GW, and the remaining energy is −0.35GW (figure 8(d)).In January 1994, similar to 1993, the Gyeonggi Bay releases 2.44 GW, the Jiangsu Shoalwater releases 1.98 GW, and the dispersed energy is 2.92 GW, 1.5 GW remained (figure 8(e)).
When the depth was changed sequentially, a different pattern was observed (table 3, figures 8(f)-(i)).In 1991 and 1992, the stepwise run was simulated with a shallower topography of 3 m and 6 m in two regions (Gyeonggi Bay and Jiangsu Shoalwater).Therefore, the release of tidal energy in the two regions was similar to the existing 6 m run-3 m run and 3 m run-0 m run cases.However, in 1993 and 1994, as the degree of topographical change in the stepwise run increased, strong outflows from Jiangsu   5.14 GW in 1992, 1.59 GW in 1993, and 2.79 GW in 1994.This is related to the stronger outflow in Gyeonggi Bay.The net flow of released energy in the stepwise run contributes to a decrease in energy flux in the East China Sea including the South Sea of Korea and an increase in amplitude in the Yellow Sea.Interestingly, compared with the constant run, the stepwise run efficiently simulated the accumulating effect of energy flux over multiple reclamations.

Summary and discussion
In this study, we analyzed tide modification phenomena by changing the coastal topography around the East China Seas and the Yellow Sea.By changing the topography sequentially using the MOHID model, we reproduced the quantile regression trend that appeared in a previous study (Lim and Chang 2018).We also determined the tidal constituents that responded to the changes in these tidal characteristics.
In addition, we analyzed the energy flux distribution using model inter-comparison.
From the tidal characteristic analysis, we found that the regional sea level different tendency is affected by the topographic change associated with two major reclamations.Also, these tendencies are influenced by the change in semidiurnal constituents, especially M2 constituent.From the simulation inter-comparison using the M2 time series, we found that the amplitude of the M2 constituent of the stepwise run changed abruptly.This result implies that if major reclamation continues several times, unpredictable changes will occur, which does not appear in the previous study assuming that the change was carried out only once during the simulation period.
As a result of tidal energy flux analysis, we found that as topography changes, there are divergences in the energy flux in the Gyeonggi Bay and the Jiangsu coast.Their energy is redistributed or moves away from the East China Sea, which appears in a somewhat consistent form in the constant runs.However, when the depth is changed sequentially, a different pattern is simulated, which is not observed in comparisons between constant runs.In the first 2 years, there was no significant difference from the constant runs, but in the latter 2 years, a strong outflow from the Jiangsu Shoalwater into the open sea appeared.This outflow, caused by the accumulating effect of energy dissipation over multiple reclamations, reduces the total energy flux inside the East China Sea.Meanwhile, because outflow from Gyeonggi Bay is still redistributed in the interior of the Yellow Sea, the net energy flux in the Yellow Sea is positive.As a result, the difference in the regional response to multiple reclamations caused the regional difference tendency shown in a previous study (Lim and Chang 2018).
Even though this study focuses on tidal change related to reclamation, the motivation for this study is partly based on a literature in the area of sea level rise.Therefore, it is need to discuss projected sea level rise due to global warming as a factor that has an impact on coastal areas.Many modeling studies have presented projections of future changes in tide caused by mean sea level rise (Greenberg et al 2012, Ward et al 2012, Su et al 2013, Luz et al 2015, Arns et al 2017, Idier et al 2017).Figure 9 show the amplitude change between 1994 and 1990 from stepwise simulation.The amplitude in most areas of the Yellow Sea region increased after the tidal reclamation, and the amplitude in the East China Sea including the Yangtze River and the South Sea of Korea is decreased by 0.2 m at most.Interestingly, this change pattern and amplitude is similar as the mean sea level rise (0.8 m) experiment conducted by Su et al (2013).This result could be explained that relatively shallow bathymetry due to reclamation can raise mean sea level in the East Sea and Yellow Sea.Considering the combined effects of tidal flat reclamation and local sea level change due to global warming, if coastal development proceeds indiscriminately, the risk experienced by coastal areas will be more than twice that of other regions.
Although the degree of topographic change we applied to the model is based on the tidal flat information map and previous studies (Song et al 2013, Lim and Chang 2018), it may differ slightly from the actual values.Murray et al (2019) calculated and analyzed the change in the global tidal flat from 1984 to 2016 as a 3 year average value.They provided a tidal flat information map from 1984 to 2016 and argued that there was a 16.02% loss in the global tidal flats between 1984 and 2016.Therefore, further analysis with a more realistic dataset is required to estimate the effects of actual bathymetric changes.

Figure 1 .
Figure 1.(a) Study areas of the Yellow and East China Sea with model topography (in m, the contour interval is 20 m from 0 to 100 m, 200 m from 200 to 4000 m).The analyzed study area is a portion of the total model domain (117 • E-144 • E, 28 • N-48 • N).Red dots indicate tidal gauge stations, and blue dots represent areas where reclamation projects have been done.(b), (c) Model coastal line and bathymetry map of (b) Gyeonggi Bay and (c) Jiangsu Shoalwater.(d), (e) the same as (b), (c) except when 9 m shallower bathymetry is applied.The contour interval is 3 m from 0 to 12 m, representing the area reclaimed sequentially in this study.

Figure 2 .
Figure 2. Co-tidal chart of (a) M2 (b) S2 (c) O1 and (d) K1 constituents from 0 m run.Height (color) is shown in m (note that the scales differ), and phase (labeled contours) in degrees.

Figure 3 .
Figure 3.The M2 constituent (a) phase and (b) amplitude at each tide gauge position calculated from KHOA and UHSLA data and the model simulation (0 m run).The dashed line marks the zero difference line, and the dotted lines mark one standard deviation of the difference.

Figure 4 .
Figure 4. Regional mean process diagram of in-situ observation sea surface height of (a) the Yellow Sea and (b) the South Sea of Korea.The regression value in each quartile is displayed as a black dot, and the 90% confidence limit is colored in grey.Ordinary linear regression result calculated using the least squares method is represented by a red line, and a red dotted line represents the 90% confidence limit.(c) and (d) Provide the same information except of stepwise experiment of model simulation.

Figure 5 .
Figure 5.Time series of 30 days boxcar filtered amplitude of simulated sea surface height of the Yellow sea (black line) and the East China Sea (red line).
To confirm and analyze the divergence and convergence trend shown in the process diagram in figure 4, this study calculated tidal amplitude from the stepwise run averaged around two regions; the Yellow Sea region (121 • E-126 • E, 34 • N-40 • N) and the East China Sea region including the South Sea of Korea (123 • E-130 • E from 32 • N to 34 • N).This spatiallyaveraged area was determined based on the previous results using process diagram (Lim and Chang 2018), but for accurate confirmation, tidal changes in M2 amplitude at observed points shown in figure 1(a) were checked.In the stepwise run, all 4 stations in the Yellow Sea showed increasing trend with similar variability in the M2 amplitude, while M2 amplitude decreased with similar variability at 9 stations in the East China Sea including the South Sea of Korea (see supplementary figures S1 and S2).Based on this result, the tidal amplitude change in the two areas were analyzed in this study.

Figure 6 .
Figure 6.Time series of amplitude of four main tidal constituents in the stepwise run ((a) K1, (b) O1, (c) M2 and (d) S2).The Yellow Sea is the black line and the East China Sea is the red line.

Figure 7 .
Figure 7. Tidal amplitude time series of M2 constituent in (a) the Yellow sea and (b) the East China Sea at each experiment.

Figure 8 .
Figure 8.(a) The area where the difference in tidal energy flux is calculated (A: Gyeonggi Bay, B: Jiangsu Shoalwater, C: Open ocean).(b)-(i) The difference in monthly mean tidal energy flux between each simulation.The month to be compared is shown at the top of each figure and the name of the simulation used to calculate the difference is shown in the lower left part of each figure.Colors and vector indicate the magnitude and direction difference between simulations.

Figure 9 .
Figure 9. Spatial distribution of M2 amplitude difference (unit: m) between 1994 and 1990 in stepwise run.

Table 1 .
Mean absolute errors of amplitude (unit: cm) and phase (unit: • ) of 4 major harmonic constants of tidal elevation between 17 tide gauge and simulation results.

Table 3 .
The same as table 2 except for between stepwise run and constant run.