Riverbed erosion of the final 565 kilometers of the Yangtze River (Changjiang) following construction of the Three Gorges Dam

The world’s largest hydropower dam, the Three Gorges Dam (TGD), spans the upper Yangtze River in China, creating a 660-km long and 1.1-km wide reservoir upstream. Several recent studies reported a considerable decline in sediment load of the Lowermost Yangtze River (LmYR) and a rapid erosion in the subaqueous delta of the river mouth after the closure of the TGD in 2003. However, it is unknown if the TGD construction has also affected river channel and bed formation of the LmYR. In this study, we compared bathymetric data of the last 565 kilometers of the Yangtze River’s channel between 1998 and 2013. We found severe channel erosion following the TGD closure, with local riverbed erosion up to 10 m deep. The total volume of net erosion from the 565-km channel amounted to 1.85 billion m3, an equivalent of 2.59 billion metric tons of sediment, assuming a bulk density of 1.4 t/m3 for the riverbed material. The largest erosion occurred in a 100-km reach close to the Yangtze River mouth, contributing up to 73% of the total net eroded channel volume.

The Lowermost Yangtze River (LmYR), starting at Datong (river kilometer, or RK 565) and ending at the river mouth Wusongkou (RK 0), is a tide-affected river reach. Datong has the most seaward gauging station of the Yangtze River with long-term hydrological records, and both water and sediment discharges at this station have been considered as the final input into the East China Sea 22,23 . A few recent studies reported different erosion and deposition patterns near the river mouth, attributing the changes to the reduction in riverine sediment following the Three Gorges Dam closure in 2003 24 . It is important to understand how sediment decline may have influenced the entire channel morphology of the tide-affected river reach from the river's mouth to Datong. Such knowledge is crucial for predicting long-term stability of the large alluvial river and its deltaic and nearshore continental shelf development.
With the above in mind, this study utilized two sets of channel bathymetric charts of the Lowermost Yangtze River from Datong to Wusongkou, produced 5 years before and 10 years after the TGD closure, to investigate changes in channel morphology and riverbed deformation. Our primary goal was to quantify spatially-explicit erosion and deposition rates along this 565-km long alluvial river reach, in order to elucidate the possible future trend of the TGD effects on the Lowermost Yangtze River.

Study Area
The Yangtze River originates in Tibet and enters the East China Sea with a total length of 6300 km. Its upstream, midstream, and downstream reaches extend, respectively, from the headwater to Yichang, from Yichang to Hukou, and from Hukou to the river mouth (Fig. 1). This study investigated the final 565-km river reach of the Yangtze River from Datong to Wusongkou near the river's mouth to the East China Sea (Fig. 1). The Ghiangnania and Huaiyang shields controlled the main geological structure and a series of fracture zones exist in the study area. Thus, the river morphology and movement trend were controlled by these geological structures 25 . In addition to carrying the flow from the upper Yangtze River basin, this river reach also drains approximately 6% of the entire 1.8 million km 2 Yangtze River Basin area with a mean precipitation of approximately 1200 mm per year 13 . This region is one of the most densely populated and industrialized areas in China, contributing 24% of China's national GDP.
Datong is the most seaward gauging station, and discharge at the location is generally considered as the final freshwater flow into the East China Sea (Fig. 1). The annual average discharge at Datong for the period of 1950-2002 is 905 × 10 9 m 3 (or 905 km 3 ), with an average annual sediment load of 0.38 × 10 9 tons (or 368 million tons) (www.cjh.com.cn/pages/nsgb). Due to dam construction and soil conservation in the upper Yangtze River Basin, the annual sediment load has declined sharply 11 , particularly after the closure of the Three Gorges Dam (TGD) in 2003. The average annual sediment load during 2003-2015 was 0.14 × 10 9 tons which is only about one-third of the long-term average sediment load before 2003, while the average annual river flow during 2003-2015 remained similar to the previous long-term average river flow.

Materials and Methods
Datasets. We collected 52 navigational charts of 1998 and 2013 surveyed by the Chang Jiang Waterway Bureau (CJWB) from Datong to Wusongkou. Each chart covered approximately 11 km of the river channel. These navigation charts included water depth points with 500 m distance between two adjacent points ( Fig. 2 and Supplementary Fig. S1). The water depth data for 1998 were generated from the charts of 1994, 1995 and 1998, and the scales of these maps were 1:40,000 and 1:60,000 (Supplementary Tables S1 and S2) which collected according to Specifications for Port and Waterway Engineering Survey published in 1994 (JTJ203-94). The water depth data for 2013 were derived from the charts of 2012 and 2013, and the scales of these maps were 1:20,000 to 1:40,000 according to Specifications for Port and Waterway Engineering Survey published in 2012 (JTS131-2012). All bathymetry data were carried out with a vertical error of about ± 0.2 m when water depth (H) was ≤20 m and of ± 0.01 H when water depth was >20 m.
The reference datum used in the 1998 and 2013 charts are same. For example, the reference datum upstream of Jiangyin (RK 160-565) is the navigational datum of the Lower Yangtze River; The reference datum downstream of Jiangyin (RK 0-160) is the theoretical lowest tide level. Sediment load and river discharge data at Datong Station from 1998 to 2015 were collected from the Changjiang Water Resources Commission of the Ministry of Water Resources (http://www.cjw.gov.cn). All the navigation chart maps were processed in ArcGIS 10.4 (ESRI, Redlands, CA).

Estimation of channel width changes.
Channel width along the Lowermost Yangtze River could have extensively changed during the past two decades due to extensive river engineering work (i.e., shoreline protection and coastal reclamation). Previous studies found that human modifications of a large alluvial river channel could lead to considerable adjustments of bed morphology 26 . Therefore, we first assessed the channel width from Datong to Wusongkou in 1998 and in 2013 using the DEMs data. A total of 167 cross sections were selected along the 565-km river reach (Fig. 1). Downstream RK 70, the channel width is the top width of the cross section when the water level is equal to the theoretical lowest tide level, while upstream RK 70, the channel width is the top width of the cross section when the water level is equal to the navigation datum ( Supplementary Fig. S2).

Assessment of bed deformation.
With the same reference datum, changes of water depth are caused by riverbed elevation changes body 27,28 . Therefore, by comparing the changes in water depth between 1998 and 2013, we can estimate the sediment deposition and erosion on the bed. For instance, the decreased water depth indicates sediment deposition, while the increased water depth indicates eroded riverbed.
In this study, the navigational charts were georeferenced using 5-9 fixed landmarks with an error less than 0.01 cm in ArcGIS 10.4. For generating high-quality digital elevation models (DEMs), in addition to original water depth points, depth points were also extracted from the isobaths of 0 m, −2 m, −5 m, and −10 m with a Scientific REPORts | (2018) 8:11917 | DOI:10.1038/s41598-018-30441-6 spacing of 80-150 m between two points. As a result, each digitized navigational chart has an average density of 90-100 depth points per square kilometer. The water depth data were interpolated to DEMs in 1998 and 2013 with 25 m × 25 m spatial resolution using kriging interpolation technique. For a detailed interpretation of the riverbed elevation changes, the 565-km long DEM of the study reach was further divided into 57 sub-reaches, each presenting a 10-km long channel. The change of average bed elevation and average deposition/erosion depth were quantified for each subreach from RK0 to RK560. The last segment was a 5-km long subreach, i.e., from RK 560 to RK 565.
The error of the estimated results is mainly from navigational charts data and the interpolation, which was estimated to be less than 10% 29,30 . Multi-beam data. Riverbed   sensor, positioning system, depth sensor, and sound velocity to control the data accuracy. The compass is used to get a heading. The motion sensor measures the attitude (roll, pitch, and heave) of the boat. An external access DGPS positioning system was used to collect the position information. The depth sensor is used in the sea level computation for two purposes. One is to obtain the sea level and the other is obtain the depth of the boat. The sound velocity sensors are used to obtain a sound velocity in time. The principle of multi-beam is a proper phase or time delay for signals received from different sources in the acoustic array which can guide the main lobe to a specific direction. This sounder had an operational working frequency of 200 kHz/400 kHz. Under the mode of 400 kHz, the central beam angle is 0.5° and the maximum ping rate is (50 ± 1) Hz. It has 512 beams at 400 kHz and 256 beams at 200 kHz. The theoretical depth resolution is 6 mm, and typical measurement depth is 0.5 m to 150 m at 400 kHz. Under the mode of 400 kHz, the along-track transmit beam-width is 1° and the across-track receive beam-width is 0.5°. During the surveying, a Trimble real-time differential global positioning system (Trimble Inc., Sunnyvale, CA) was used to control the position accuracy at the decimeter level. The traveling speed of measuring boat was controlled to be as constant as possible at 1-3.5 m s −1 . The threshold of maximum ping rate was 20 Hz which can automatically adjust with water depth. The real Ping rate was not less 9 times per second. Thus, the precision of the successive measurements along the direction of the boat during 0.1-0.4 m. The swath width can cover more than 5-6 times of the water depth. The maximum water depth is about 50 m in the lower Yangtze along the main navigation channel. The 400 kHz work mode was chosen in the field, thus the swath width was no more than 300 m with 512 beams. As a result, the precision of the successive measurements perpendicular to the direction of the boat is 0.6 m. The final point cloud data were processed by sound speed correction. Meanwhile, the calibrations of the roll, pitch, and yaw were conducted. Abnormal beams were removed in the editing module using the PDS 2000 software (Version 3.9.1.7). Finally, the mapping of riverbed bedforms was directly built by the processed point cloud data, and the resolution was set to 1 m × 1 m.

Results
Changes in channel width and bed elevation. Based on 167 cross-sectional measurements in 1998 and 2013, we found that on average, the 565-km Lowermost Yangtze River has narrowed by 370 m in the past 15 years. The change in channel width mainly occurred in the lower reach of the Lowermost Yangtze River. Specifically, the channel width of the river reach from RK 130 to RK 30 decreased by 1609 m (i.e., from 8070 m to 6461 m) (Fig. 3B), while the channel width between RK 565 and RK 130 only slightly reduced, i.e., from 2410 m in 1998 to 2317 m in 2013. The river reach below RK 30 showed little change in its channel width, i.e., from 11,300 m in 1998 to 11,220 m in 2013.
There was a significant reduction in bed elevation from 1998 to 2013 along the entire 565-km Lowermost Yangtze River (Fig. 3A). Overall, the average bed elevation decreased by 1.  Multi-beam riverbed measurements. The 681-km multi-beam measurements along the main navigation channel of the Lowermost Yangtze River were used to verify the current riverbed topography. Approximately 63.3% of the surveyed reach developed dunes, 21.8% showed eroded surface, and 14.9% was a flat riverbed 31 . Dunes distributed along the entire studied reach and the distribution density of very large dunes (wavelength greater than 100 m) showed a decreasing trend from Datong to the river mouth. Statistics data show that the flat riverbed mainly developed downstream of Xuliujing (RK 70-0). In this segment, the length of flat riverbed accounted for 72.2% of the total flat riverbed length in the 565-km Lowermost Yangtze River. More than 10-m deep riverbed erosion was observed at several segments (Fig. 4). For instance, an average erosion depth of 11.5 m was found at RK 470 (Fig. 4B), whereby the riverbed depth surrounding the erosional holes was 8.9 m, while the maximum depth of the erosional holes was 20.4 m with a discharge of 25,000 m 3 s −1 recorded at the Datong station. Similar large erosional phenomena were also found in other river segments, e.g., RK 490, RK 160, and RK 65 (Fig. 4).

Riverbed erosion and deposition estimates.
The findings from this study show that the final 565 km of the Lowermost Yangtze River has undergone excess riverbed erosion in the past 15 years (Fig. 5), resulting in an average erosion of 0.05 m per year in the Lowermost Yangtze River (Fig. 5A). However, the erosion rates varied largely from 0.01 to 0.19 m per year along the river reach. The average bed degradation is 0.04 m per year in the upstream 465 km, while the value increases to 0.08 m per year in the final 100 km (Fig. 5B). Based on the changes in river width and bed elevation, we quantified a total erosion volume of 2.157 × 10 9 m 3 . The deposition was found only in two reaches, i.e., RK 100-150 and RK 380-480, with a range of 0-0.17 m. The total deposition volume was much smaller (0.310 × 10 9 m 3 ) than that of erosion (Fig. 5B). Hence, a net change in volume of the riverbed was 1.847 × 10 9 m 3 from 1998 to 2013, i.e., 3.3 × 10 6 m 3 /km. This enormous erosion would represent an equivalent of 2.59 billion metric tons of sediment, assuming a bulk density of 1.4 t/m 3 for the riverbed material. It is worth noting that the erosion volume downstream of RK 100 accounted for 72.7% (or 1.342 × 10 9 m 3 ) of the entire net erosion volume (1.847 × 10 9 m 3 ) of the Lowermost Yangtze River.

Discussion
Impact factors of riverbed deformation. Riverbed deformation of an alluvial river is subject to governing conditions that include human activities and changes of sediment load, river discharge, and climate [32][33][34] . As a large alluvial river, the final 565 kilometers of the Yangtze River can be affected by many factors and the riverbed deformation mechanism is complicated [35][36][37] . The results from this study demonstrate that the Lowermost 565km Yangtze River experienced excess riverbed erosion from 1998 to 2013. Before the construction of the Three Gorges Dam (TGD), channel bed erosion downstream of it was predicted extending to the river mouth (i.e., 1600km long channel). However, a recent study found that the bed erosion downstream of the TGD would very likely cease at the point about 400 km downstream of the dam 21 . Much coarser bed sediment and significantly reduced bed slope are main contributors to the cessation of the further downstream erosion. A previous 1-D modeling analysis of bed profiles downstream of dams also revealed that the bed erosion owning to sudden sediment reduction propagates extremely slow (i.e., several centuries) to the downstream in a large alluvial river 38 . Located about 1000 km downstream of the TGD, the observed excess bed erosion in this study seems unlikely to be caused solely by the closure of the TGD in 2003. The largest deepening of riverbed occurred between RK 130 and RK 30, where the channel also narrowed greatly. Narrowing channels leading to bed erosion has been observed in many alluvial river systems throughout the world [39][40][41] . In fact, the river channel width of the entire Lowermost Yangtze River reduced by 9.1% from 1998 to 2013 (Fig. 3B) because of channel bar growth, reclamation of shoreline, and construction of dikes and revetments (Supplementary Tables S3 and S4). Between RK 130 and RK 30, the channel width reduced by nearly 20% (or 1609 m). This should have significantly increased the flow velocity, resulting in the bed degradation in this river reach.
In terms of erosion volume, our result shows that the channel bed downstream of RK 100 was mostly eroded, accounting for 72.7% (or 1.342 × 10 9 m 3 ) of the entire net erosion (1.847 × 10 9 m 3 ) of the Lowermost Yangtze River. Except for the channel width reduction, the bed degradation could also be closely related to the backwater effects. In large alluvial rivers, the Lowermost reaches tend to be scoured during high flow due to the drowdown effects. This can be seen from the modeling studies 42,43 and the recently observed bed erosion in the Lowermost Mississippi River 44 . In fact, the channel bed in the last 50 km of the Yangtze has experienced erosion since 1978, according to a long-term bathymetry study by Luan et al. 45 . The largest erosion was found between 1986 and 1997 when frequent floods occurred. Therefore, the flood effects on bed morphology in this reach could not be neglected. Particularly, historical floods occurred in 1998 and 1999 with the peak discharge over 80,000 m 3 /s. These floods may have also contributed to the bed erosion downstream of RK 100. In addition, this reach is influenced by both channel flow and tide current in the same direction during ebb tide 46 . This can aggravate the  Supplementary Fig. S3). The large eroded amount of sediment in the Lowermost Yangtze River should have been an important source for the accumulation. It is also plausible that sediments eroded from the riverbed are coarser and heavier than suspended sediment, making them more likely to deposit near the river mouth. This was recently observed by Yang et al. 49 , whereby they found that the average median size of the sediment in the river mouth increased from 8.0 µm in 1982 to 15.4 µm in 2012.
While average channel width of the reach RK 130-100 reduced by 20% (or 1,609 m) from 1998 to 2013 (Fig. 3B), a deposition of 0.121 × 10 9 m 3 occurred in the 30-km reach (Fig. 5B). Several engineering projects in the past 15 years (Supplementary S4) may have contributed to the deposition. For instance, a sandbar consolidation project at Rugaoqunsha (Supplementary S3) caused kilometers of sedimentation along the left bank of the 30-km reach. As a result, the right bank occurred excess riverbed erosion up to 3 m (Fig. 3A). A recent study on foundation scours of the large bridges in the studied river reach pointed out that the riverbed erosion phenomenon after the TGD closure has been stronger 31 . Although we are unable to do any comparison on grain size before and after the TGD construction, a study by Luo et al. 11 has reported such a change and has suggested that the impact of the Three Gorges Dam on grain size in the middle and lower Yangtze River would continue after 2011. It is therefore likely that excess bed erosion in this river reach would continue in the future.
Channel evolution mechanism response to natural and human activities. The sedimentation mechanism of upstream erosion and downstream-ward aggradations is weakened in the Lowermost Yangtze River in the past 15 years. This reach is influenced by both channel flow and tide current in the same direction during 46 . However, the erosion and deposition pattern found in our study shows that the volume of erosion in upstream reach is 1.5-9.3 times than the volume of deposition in its downstream (Table 1). This phenomenon suggests that scoured sediment from the riverbed became the source of suspended sediment in the LmYR. Hence, a large amount of sediment was eroded from upstream and just part of them deposited in the downstream river channel. As a result, the sedimentation mechanism of upstream erosion and downstream-ward aggradations in the LmYR is weakened owing to the sediment decline in recent years. A detailed analysis of the erosion and deposition process in different water levels shows that the river reaches of deeper than 10 m (Fig. 6) in the last 100 km were the main erosion areas in the Lowermost Yangtze River in the past 15 years, accounting for 49% of the total erosion volume ( Table 2). Detailed erosion and deposition volume data of 0 -−5 m, −5 -−10 m, and below −10 m also showed similar results (Fig. 7). This suggests that reaches deeper than 10 m were the main erosion areas of the riverbed.
Contributors to the riverbed erosion in the last 565-km Low most Yangtze River can include channel width reduction, backwater effects, tide and wave, and in-channel sand mining activities. In addition, due to the construction of TGD, a large amount of sediment was trapped in the upstream of the YR after 2003, which could contribute to increasing sediment transport capacity in the downstream river reach 45,50,51 . It is not clear, however, how long it will take until a new hydromorphological equilibrium is reached in the Lowermost Yangtze River under such a complex of human and natural impacts. Modeling analysis in the future could probably help answer the questions of how channel width change can affect bed morphology and at what point bed grain size distribution will become stable enough to withstand further scouring tendencies in this large alluvial river.

Conclusions
This study assessed river width and bathymetric changes in the last 565 kilometers of the Lowermost Yangtze River between 1998 and 2013. The assessment is the first comprehensive analysis of channel deformation of this large alluvial river following the Three Gorges Dam closure in 2003, after which riverine sediment loading declined significantly. We found that on average, the river reach reduced by approximately 370 m (or 9.1%) in width and scoured about 1.2 m in bed elevation from 1998 to 2013. The total net erosion volume from the 565-km riverbed amounted to 1.85 billion m 3 , an equivalent of 2.59 billion metric tons of sediment assuming a bulk density of 1.4 t/m 3 for the riverbed material in the past 15 years. Of the entire studied river reach, 425 kilometers have undergone substantial erosion (2.16 billion m 3 , annual average erosion rate ranging between 0.01 and 0.19 m) while the other 140 kilometers have experienced marginal deposition (0.31 billion m 3 , annual average deposition rate ranging between 0 and 0.17 m). We attribute this excess erosion to the rapid reduction in sediment load following the closure of the Three Gorges Dam in 2003. The average bed elevation and deepest trough curves changed following the erosion and deposition patterns. Severe channel erosion up to 10 m deep was observed at many locations, especially in the last downstream 100 kilometers, showing the seriousness of channel erosion of the Lowermost Yangtze River in the recent 15 years. The findings strongly indicate that the channel erosion will continue progressing, potentially threatening the long-term stability of this large alluvial river and its deltaic and continental shelf development.