The in situ hydrodynamic observations (including water level, velocity, flow direction, turbidity, and salinity) in the YRE during August 2009 at YLJ-02 (124.3028°E, 39.9203°N) and YLJ-03 (124.278°E, 39.8396°N) were used in this study. Continuous observations were performed at these two sites for 25 h during the spring and neap tide, respectively ( Figure 1C ). Flow velocity and direction data were obtained using an Acoustic Doppler Current Profiler fixed at 0.1 m from the bottom of the vessel (the acoustic frequency was set at 1,200 kHz and the flow velocity resolution was 0.001 m/s). Suspended sediment concentration (SSC) was obtained from on-site sampling. We used the YSI6600 multi-parameter water quality detector for turbidity measurements and transformed turbidity into SSC according to the relationship between the two (Cheng et al., 2016). Bed sediment column samples were collected from 19 sites around the estuary in 2010 and 2014, and the accumulation rates of the bottom sediments were obtained from the chronological results (Cheng et al., 2019).

This study also analyzed the runoff and sediment discharge data of the Yalu River from 1958 to 2017 (daily data for 2009–2017; annual data for 1958–2008) recorded at the Huanggou and Lishugou hydrological stations ( Figure 1C ). According to anomalies in the river discharge data, years were classified into three categories: dry years, normal years, and wet years ( Figure 2A ). The classification method is as follows:

where P is the average anomaly level given in percentage, Flux_n is the total runoff for one year, and Flux_average is the multi-year averaged total runoff. When P< −20%, the year is classified as a dry year; when −20%< P< 20%, the year is classified as a normal year; and when 20%< P, the year is classified as a wet year.

We took the average flow rate (1,000 m³/s) during the wet season in 2015 (dry year) as a representative runoff value for dry years, 2,000 m³/s in 2011 (normal year) for normal years, and 4,000 m³/s in 2013 for wet years. Furthermore, a value of 6,000 m³/s was taken as the average runoff rate for a typical flooding event in this area ( Table 1 ). Meanwhile, the hourly hydrological data from July–September 2010 was used to study the sedimentation change during the 50-year flood event (extreme flooding event) in the YRE ( Figure 2B ).

To improve the accuracy of coastal model results, this study adopted satellite-derived bathymetry (SDB) and coastlines with high resolution. Images and band reflectance from the satellites Landsat 7 ETM+ (Enhanced Thematic Mapper), Landsat 8 OLI (Operational Land Imager), and TIRS (Thermal Infrared Sensor) were used to calculate the estuarine geometry with a resolution of 30 m. In this study, the shoreline of the YRE was extracted using ArcGIS from satellite images taken by Landsat 7 in July 2009, and the bathymetry was obtained from the reflectance of Landsat 8 in August 2013.

The band ratio model, a commonly used SDB method, is adopted in this study. As the depth of shallow water increases (H < 15 m), the ratio of the blue reflectance band to the green reflectance band increases (Philpot, 1989) (Traganos et al., 2018). In this study, the band ratio model (Eq. 2) was used to obtain the water depth of the study area based on the corresponding band reflectance obtained from Landsat 8 (Stumpf et al., 2003). By relating the band ratios to the water depth and deriving the logarithmic relationship between them, an SDB equation for coastal estuaries can be established by means of linear regression. It can be expressed as follows:

where H is the actual water depth, R _rsi is the green band reflectance, R _rsj is the blue band reflectance, n is a positive integer (usually taken as n = 1,000), and m ₁ and m ₂ are constants derived from the linear regression fit.

For model validation, we compared the model results of case 0 with the measured data at YLJ-02 and YLJ-03 stations for water elevation, current speed, current direction, salinity, and SSC. The correlation coefficient (R) and the root mean square error (RMSE) between the two datasets were also calculated ( Table 5 ). The validation results showed that the model successfully reproduced the hydrodynamic conditions and sediment transport in the YRE ( Figure 4 ); it is thus reliable to be used to investigate changes in hydrodynamics and bottom sediments during wet seasons.

We obtained the sediment accumulation rate from 19 collected column samples around the estuary ( Figure 1C ) and plotted the distribution of sedimentation rate in the YRE using Kriging interpolation ( Figure 5A ). A comparison was made between these results and the modeled bed thickness change over 3 months ( Figure 5B ). For the distribution pattern, the overall similarity between the model results and the sampled results is 80%, indicating that the model can generally capture the patterns of erosion and siltation of bottom sediments in the YRE. However, large differences are observed in the magnitude between the two two methods, especially near the top of Chouduan Island. This is because fewer sampling sites are present there ( Figure 5 ).

Figure 6 shows the response of bed thickness to different runoff conditions in the YRE during the wet season (July to September). The results show that the bottom sediments are strongly eroded in the upper part of the main channel (39.9°N–40.05°N) in dry and normal years. A sand bar is formed between 39.83°N and 39.93°N, which is highly consistent with field observation and previous studies (Gao et al., 2012).

During dry and normal years (runoff< 2,000 m³/s), sediments from the upper estuary are transported seaward with runoff, and those from the lower estuary are transported landward with the tide. The net transport of those two groups of sediments determines the location of the sand bar. The sand bar center moves seaward, with a higher peak when the runoff increases. During wet years (4,000 m³/s), the increased runoff leads to stronger scouring, which hinders the formation of the sand bar and sediments accumulate further south. However, the sand bar does not move further south when the runoff continues to increase. During the flood event (6,000 m³/s), the even larger velocity generates extremely large bottom shear stress, resulting in a stronger resuspension of bottom sediments in the upper estuary, which leaves the location of the sand bar unchanged. The tidal limit in the YRE is generally at Douliupu; the range of runoff influence (see Figure 6B , where the variation is shown to be close to 0) is 12 km south of the tidal limit during dry years, 13 km during normal years, and 14.5 and 17 km during wet years and flood events, respectively.

Figure 7 shows the distribution of the morphological change in the YRE after 3 months under different runoff conditions. During dry and normal years ( Figures 7A, B ), large amounts of sediment silt in the southwest area, close to Xia Island. During wet years and flood events ( Figures 7C, D ), high flow rates allow the sediment to be transported further south in the estuary. The southwest area near Xia Island shifts from siltation to erosion after the runoff exceeds 4,000 m³/s. As the tidal direction of the YRE is southwest–northeast, some sediments accumulate on the west side of Chaolong and Westlake Islands for all experiments. However, this patch of accumulated sediments becomes weaker during flood events ( Figure 7D ), possibly due to the stronger flow velocity that washes large amounts of sediments from this area to the south.

In general, the bed thickness in the YRE exhibits an erosion pattern during wet years and siltation in dry years. The runoff threshold for this change from siltation to erosion is approximately 4,000 m³/s. After reaching the threshold, the runoff dominates the bottom sediment distribution in the YRE, indicating a delicate balance in the estuarine morphology, which can be easily damaged by flooding events.

The maximum flow velocities in the YRE occur between 39.95°N and 40°N under different runoff conditions. This is mainly due to the rapid convergence of the estuary width at this location, which leads to tidal wave breaking and stronger instability, increasing the local tidal energy and current velocity. The increasing runoff enhances ebb currents, while there is a significant decrease in flood currents during wet years. The ebb tide velocity is on average 1 m/s greater than the flood tide ( Figure 8A ); however, the degree of this discrepancy is greatly influenced by the runoff condition ( Figure 8B ).

The bottom shear stress along the main stream section shows the same pattern as the current velocity, with the maximum shear stress located between 39.95°N and 40°N ( Figure 9A ). However, a stronger ebb current velocity induces a larger bottom shear stress, and the maximum ratio of the ebb/flood shear stress occurs from 38.85°N to 39.90°N ( Figure 9B ). The river-induced uneven pattern in bottom stress ranges south to 39.8°N, generally corresponding to the location of the entrance sand bar. As the runoff increases, the ratio gradually increases, indicating that the bottom shear stress at ebb is greater than that at flood. This could result in larger amounts of sediments moving southerly from land to sea, while such a transport pattern would be hindered at 39.8°N. This seaward asymmetry becomes significantly stronger during larger runoff (blue line in Figure 9B ). This indicates that the shape of the entrance sand bar cannot be maintained during wet years or flood events. The decreased ebb/flood bottom stress ratio after 39.8°N is caused by the sudden decrease in the ebb bottom stress, as the area is out of the range of the river runoff influence. This range can expand south to the sea when the river runoff increases.

In summary, during wet years and flood events, larger runoff causes stronger erosion in the inner estuary, while landward flow in the outer estuary is weakened, resulting in intensive erosion in the estuary. 4.4 Baroclinic effects on estuarine sedimentation during flooding events

The model results show that the maximum extent of saltwater intrusion into the YRE is located at approximately 39.9°N during dry and normal years. As runoff increases, the salinity front continues to move seaward ( Figures 10A–D ). It is noteworthy that vertical stratification becomes significantly stronger when the runoff reaches 6,000 m³/s. This change in vertical structure does not change the location of ETM in the main stream. It is caused by the unique geometry of the YRE; large amounts of sediments are accumulated at the river–estuary junction around Douliupu, providing an almost infinite supply for resuspension here. As a result, vertical stratification has a limited effect on the location of this high SSC patch, but it greatly reduces the magnitude of its peak value ( Figures 10E, F ). With larger runoff, the sediments accumulate slightly further south–center, while resuspension is hindered ( Figures 10G, H ). The time series of the Richardson number in the vertical water column at the center of the estuary at 39.7°N shows that the estuary forms a strong mixing at flood tide and stratification at ebb tide. Vertical mixing is more intense during spring tide than during neap tide ( Figure 11 ). With the increase of runoff, this stronger mixing during neap tide significantly weakens. Overall, baroclinic effects in the YRE are not dominant factors during dry and normal years; however, they have a significantly larger impact on vertical mixing during flooding events. Previous studies have shown that the YRE is mostly well-mixed due to its shallow water depth and strong currents. These results indicate that the YRE is not always well-mixed, and that vertical circulation needs to be considered in estuarine morphology research.

The tidal skewness at the center of the ETM under different runoff conditions is shown in Figure 12 . There is a slight decrease in the amplitude of the M₂ and M₄ components with increasing runoff. The tidal skewness remains positive, indicating that the tidal asymmetry is still flood dominant.

However, the degree of this flood dominance decreased. Here, flood dominance does not necessarily mean that the net sediment transport is landward, as other factors also contribute to the net sediment transport, such as Stokes drift, Lagrange asymmetry, and Eulerian asymmetry (Brown and Davies, 2010). Stokes drift exists under traveling tidal wave conditions when there is a phase advance/lag between amplitude and current flow. Stokes drift may produce a net seaward sediment flux with a limited magnitude (Stokes, 1847). The Lagrange and Eulerian asymmetries in estuaries can generate a larger net transport flux than Stokes drift. Lagrange asymmetry controls the transport of finer sediments by influencing the settling time and the resuspension process (Friedrichs, 2012), while Eulerian asymmetry controls the variation in bottom shear stress by influencing the current speed at flood and ebb times (Dronkers, 1986; Friedrichs and Aubrey, 1988; Van Maren and Winterwerp, 2013).

To further analyze the impact of runoff on these factors, the Eulerian residual flow from model results is decomposed using Eq. 9. It is decomposed into ENT, EIA, and TIA, as described in “Section 3.4.3” ( Figure 13 ). As total sediment transport is proportional to residual flow, it can be better understood by decomposing the residual flow (Van de Kreeke and Robaczewska, 1993; Dronkers, 2005; Guo et al., 2014). The results show that the ENT has a small effect on sediment transport; it only changes the transport during wet years (4,000 m³/s) and flood events (6,000 m³/s). However, the EIA, which is caused by the interaction between the Eulerian residual flow and tides, appears to have a more significant impact on sediment transport. The EIA term is markedly larger than the TIA term during all experiments. The increase in river flow mainly leads to a gradual increase in the ENT and EIA. When runoff reaches a certain level, the total value of ENT and EIA will exceed TIA. This means that the direction of net sediment transport may shift from landward to seaward during flooding events. This explains why the tidal skewness of the Yalu River estuary is positive (flood dominant), but the net sediment transport is mostly seaward during wet seasons ( Figure 13 ).

As runoff continues to increase, the seaward transport peak becomes increasingly stronger. Furthermore, the total transport direction completely shifts at 39.83°N, and this turning point corresponds with the sand bar location as discussed above and shown in Figure 6 . This turning point is located in the upper portion of the estuary during dry and normal years and is then pushed to approximately 39.8°N during flooding events, with the sand bar moving seaward and a significantly reduced thickness.

In conclusion, the impact of the EIA term becomes increasingly dominant with the increase of runoff. This can explain the stronger erosion pattern during wet years as described in “Section 4.2” since a larger EIA contributes to a stronger net seaward sediment transport.

The fixed runoff used in the model can provide a theoretical basis for investigating the mechanism behind the estuarine morphological change, but this is not representative enough for understanding the conditions during extreme flooding events. Therefore, we simulated the impact of an extreme flooding event on estuarine morphology using real runoff data for the Yalu River recorded at the two hydrological stations in 2010 ( Figure 2B ). Flooding events often exhibit distinct flood peaks; therefore, we studied the results at four moments: the initial morphological pattern (07/01), before the flood event (08/01), after the flood event (09/15), and 15 days after the flood event (09/30) ( Figure 14 ) and systematically analyzed the impact of the flood process on the estuarine morphological change.

During the period before the flood event (07/01 to 08/01), the average runoff of the Yalu River basin is 1,070 m³/s. A sand bar is formed between 39.85°N and 39.9°N ( Figure 14A ). During the flood event period (08/01–09/15), the average runoff in the Yalu River basin is 6,028 m³/s. This sand bar is strongly scoured during flood events; large amounts of sediments are washed further south and settle between 39.8°N and 39.9°N. At 15 days after the flood event (09/30), the average runoff in the Yalu River basin during this period decreases to 2,124 m³/s. The sediments previously deposited in the south gradually move back to an upper location, forming a new sand bar with time. In addition, the strong flush caused by the extreme flooding increase the depth of the main stream channel from 39.9°N to 40°N, which is another dramatic change in estuarine morphology after the flooding. When the runoff returns back to normal levels during normal years, the bottom shear stress decreases, leading to a change in the sediment transport pattern. The newly formed sand bar has a different shape than before the flood event. This indicates that the estuarine morphology can be considerably changed after only one extreme flooding event ( Figures 14A, B ).

The results of bed thickness change under real runoff and sediment load, i.e., case 6, are shown in Figures 14C, D and Figure 15 . A comparison is made between cases 6 and 5 (identical to case 6 but with a fixed sediment load). For the distribution of bed thickness change under real conditions ( Figures 15D–F ), before the flooding, the runoff rate is approximately the average rate during dry years with limited sediments. Therefore, this does not significantly influence bed thickness. During the flooding event, the substantially higher runoff enhances erosion in the main stream. Despite the fact that the sediment load is also large, it does not change the erosion pattern in the main stream; however, more sediments are deposited in the southeast area of the estuary. This pattern of morphological change remains almost the same 15 days after the flooding.

Comparing the pattern with a fixed sediment load ( Figures 15A–C ), the bed thickness does not drastically change under real sediment load conditions. Sediment supply from the river has always been a key factor controlling estuarine morphology; however, it does not appear to have an obvious impact in the YRE. The reason for this is that the Yalu River is generally a clear river, containing limited amounts of sediments. Thus, the fluvial sediment supply does not considerably alter the morphology pattern here, but the impact of runoff volume is significant ( Figures 15G–I ).

In summary, the results show that, prior to the flood event, most of the sediments in the YRE were silted at the south area close to Xia Island ( Figures 15D–F ). This could block the water channel between the West and Middle Rivers, decreasing water exchange between them. Such a process will, in turn, enhance siltation in the West River and further weaken its dynamics. This pattern did not change after extreme flooding, indicating that the West River may always experience siltation even upon flushing by a significantly large runoff. Thus, natural conditions cannot change its siltation; human activities, including dredging, should be carried out to maintain the water depth in the West River. Otherwise, it may eventually become a tidal creek.