Impacts of the Qingcaosha Reservoir on saltwater intrusion in the Changjiang Estuary

: The massive Qingcaosha Reservoir (QCSR) is located in the Changjiang Estuary along the northwest coast of Changxing Island. The reservoir significantly narrowed the upper reaches of the North Channel and deepened the channel near the reservoir. These topographical changes inevitably influenced hydrodynamic processes and saltwater intrusion in the estuary. A well-validated model was employed to investigate the influence of the QCSR on saltwater intrusion in the Changjiang Estuary. The model results showed that the narrowed upper reaches of the North Channel decreased the water diversion ratio and thus increased salinity in the North Channel. During the moderate tide after neap tide, the salinity decreased at the water intake of the QCSR because saltwater intrusion was obstructed at flood slack at the surface, while the salinity increase during the moderate tide after spring tide was mainly due to the intensified saltwater intrusion during spring tide. The deepening of the channel near the QCSR resulted in an increased water diversion ratio, and the salinity in the Eastern Chongming Shoal decreased by more than 0.5 psu during spring tide; however, the saltwater intrusion was enhanced due to the strengthened baro- clinic force, which is proportional to the water depth. During neap tide, the salinity in the entire North Channel decreased because of a 1.4% increase in the water diversion ratio of the North Channel and the relatively weak tide.


Introduction
The 120 km long and 90 km wide Changjiang Estuary is one of the largest estuaries in China. It is distinguished from other estuaries around the world by its complex topography, which features three-level bifurcation and four outlets (Fig. 1a). Chongming Island divides the estuary into the North Branch and South Branch. The upper reaches of the North Branch have been severely narrowed due to natural evolution and the artificial reclamation of the intertidal zone since the 1950s (Zhu and Bao 2016); these factors have made the North Branch almost orthogonal to the South Branch in the bifurcation area and seems to prevent runoff from entering the North Branch. Moreover, the funnel-shaped lower reaches of the North Branch magnify the tidal effect. This unique river regime eventually causes a residual landward flow from the North Branch to the South Branch at low river discharge levels during spring tide; this landward flow is also known as saltwater spillover (SSO), a unique form of saltwater intrusion in the estuary (Shen et al. 2003;Wu et al. 2006Wu and Zhu 2007). The South Branch is bifurcated by Changxing Island and Hengsha Island into the North Channel and South Channel. Finally, Jiuduansha Island divides the South Channel into the North Passage and South Passage. Near the river mouth lies a large area of tidal flats, i.e., from north to south, the Eastern Chongming Shoal, the Eastern Hengsha Shoal, and the Nanhui Shoal. An enormous amount of freshwater, approximately 9.24 × 10 11 m 3 , is discharged into the East China Sea through the estuary each year (Shen et al. 2003). However, the river discharge shows considerable seasonal variation, from a maximum monthly mean of 49 850 m 3 /s in July to a minimum monthly mean of 11 180 m 3 /s in January . Tides contribute the most of any factor to water movement in the Changjiang Estuary, which is a mesotidal estuary with a mean tidal range of 2.84 m at the river mouth (Shen et al. 2003). The tide exhibits semidiurnal, diurnal, and fortnightly spring-neap signals. Due to the subtropical monsoon climate, the estuary experiences a prevailing southeasterly wind with a mean velocity of 5.0 m/s in summer, while a northerly wind with a mean value of 5.5 m/s prevails in winter .
Saltwater intrusion naturally affects stratification (Simpson et al. 1990) and helps to develop estuarine circulation (Pritchard 1956) with baroclinic forces. By influencing mass transport, it can further alter sediment transport (Geyer 1993) and degrade water quality . Generally, there are two forms of saltwater intrusion in the Changjiang Estuary. One is direct landward invasion through the North Channel, North Passage, and South Passage, while the other is the SSO mentioned above. The saline water that spills into the South Branch is transported downstream and arrives at the middle reaches of the South Branch during the following neap tide. The direct saltwater intrusion combined with the SSO severely jeopardizes the security of regional freshwater resources since three reservoirs, namely, the Dongfengxisha Reservoir (DFXSR), the Chenhang Reservoir (CHR), and the Qingcaosha Reservoir (QCSR), are located in the South Branch (Figs. 1a and 1b). Saltwater intrusion in the Changjiang Estuary is the result of a complicated nonlinear system that is affected by many factors, among which tides and river discharge are dominant (Shen et al. 2003;Wu et al. 2006;Li et al. 2010;Zhu et al. 2010Zhu et al. , 2018Qiu et al. 2012), but wind (Li et al. 2012), topography (Li et al. 2014;Chen et al. 2019), the river watershed, and sea level rise (Qiu and Zhu 2015) all deserve serious consideration as well. In addition to all these natural determinants, anthropogenic activities also play a significant role in saltwater intrusion. Estuarine projects directly force topographic change and thereby affect saltwater intrusion. Some previous studies have been conducted to determine the impacts of artificial projects on saltwater intrusion in the Changjiang Estuary. The tidal flat reclamation project in the Eastern Hengsha Shoal was found to weaken saltwater intrusion in the North Channel but to have the opposite effect in the South Channel, North Passage, and South Passage, and the SSO was enhanced after the reclamation project in the Eastern Hengsha Shoal, which degraded the freshwater quality at the water intakes of the three reservoirs . Based on simulated results, Li et al. (2020) concluded that under climatic wind conditions, the Deep Waterway Project abates saltwater intrusion in the North Channel by inhibiting the southward transport of diluted water at the mouth of the channel; however, a horizontal circulation of high-salinity water is produced between the North Channel and South Channel under persistent strong northerly winds, which can result in severe saltwater intrusion.
The massive Qingcaosha Reservoir, which supplies more than 70% of the freshwater for the 13 million residents of Shanghai, began construction in 2007 and was completed in 2010 (Chen and Zhu 2018). As Fig. 1b shows, the QCSR is located along the northwest coast of Changxing Island, and the reservoir area is approximately 66.15 km 2 (Chen and Zhu 2018). Although the QCSR has been in operation for nearly a decade, how it affects saltwater intrusion remains unclear. This paper aims to fill this gap in saltwater intrusion research and provide some insight into the impact of estuarine projects on fundamental dynamics. The remainder of this paper is organized as follows. In Section 2, the numerical model and analytical methods are introduced. Section 3 provides the results of the numerical experiments and some comparative analyses. The ways in which the local topographic changes caused by the QCSR have altered saltwater intrusion are discussed with an extra experiment in Section 4. Section 5 provides some conclusions.

Materials and methods
The hydrodynamic kernel of the numerical model applied in this study is a reconstructed and improved version of the original Estuarine, Coastal, and Ocean Model (semi-implicit) (ECOM-si) (Blumberg 1994). A nonorthogonal curvilinear coordinate transformation (Chen et al. 2001) and a vertical σ coordinate transformation were introduced into the primitive ocean equations. The independent variables are distributed on an Arakawa C grid. A third-order oscillation-free advection scheme, HSIMT-TVD , was developed to solve the advection terms in the tracer equations. The horizontal and vertical mixing processes are calculated by the Smagorinsky scheme (Smagorinsky 1963) and the modified Mellor-Yamada level 2.5 turbulence closure model (Mellor and Yamada 1982;Galperin et al. 1988), respectively.
The computational domain covers the region 27.5°N-33.7°N and 117.5°E-124.9°E, as shown in Fig. 2a. In the horizontal direction, the grid is sufficiently refined in the area of interest, with the minimal resolution reaching 300 m at the bifurcation of the North Channel and South Channel, and is carefully adjusted to fit the coastline and the engineering structures, especially the Deep Waterway Project. Ten uniform σ levels were used in the vertical direction. A wet/dry scheme with a critical depth of 0.1 m was adopted for the tidal flats in the estuary, which was found to be important for correctly simulating the initial stage of the Changjiang plume .
Along the open sea boundary, the water elevation is determined by tidal level and residual water level. Tidal level is calculated with the tidal constants of 16 astronomical tidal constituents (M 2 , S 2 , N 2 , K 2 , K 1 , O 1 , P 1 , Q 1 , MU 2 , NU 2 , T 2 , L 2 , 2N 2 , J 1 , M 1 , and OO 1 ), derived from the NaoTide dataset (https://www.miz.nao.ac.jp/staffs/nao99/). The residual water level was obtained by filtering the water level of a model with a larger domain covering the entire East China Sea, the Yellow Sea, and the Bohai Sea, as well as part of the Japan Sea and the Pacific Ocean (Xiang et al. 2009). Moreover, the monthly mean salinity was used as the salinity open boundary condition in this model. The daily river discharge measured at the Datong hydrological station was used as the river boundary condition. The initial salinity distribution beyond the estuary was interpolated from the Ocean Atlas in Huanghai Sea and East China Sea (Hydrology) (Editorial Board for Marine Atlas 1992), and that for the estuary was determined from observed data from the river mouth. Bathymetry data recorded in 2007 were used as the source of the depths of the model grid centroids.
The model was developed to investigate saltwater intrusion and its spatial and temporal variability in the Changjiang Estuary, and it was well validated in previous studies. Chen et al. (2019) used the measured data taken in the South Passage from 9 to 19 March 2018 to validate the model results. Three skill assessment indicators (correlation coefficient, root mean square error, and skill score) showed that the model performed successfully in simulating the hydrodynamic processes and saltwater intrusion in the estuary. Lyu and Zhu (2018) validated the model with the data measured in the North Passage from 19 February to 1 March 2017 by two experiments. Through comparing the skill assessment indicators between the two experiments, the model presented a better ability to capture the temporal variations in salinity in the North Branch based on the Chézy-Manning formula than the logarithmic formula of bottom drag coefficient. Qiu and Zhu (2015) and Li et al. (2012Li et al. ( , 2014 also obtained satisfactory results in validating the model. Two numerical experiments were performed to investigate the direct impacts of the QCSR on saltwater intrusion in the Changjiang Estuary. In numerical experiment 1 (Exp. 1), the QCSR was not considered in the model (Fig. 2b). In experiment 2 (Exp. 2), the reservoir area which contains 356 grid points in the model domain was left blank (Fig. 2c).
The experiments all ran for 92 days starting on 1 December 2007. The model conditions included a constant 5 m/s northerly wind field for surface forcing, and the river flux boundary condition was assigned its monthly mean values since 1950, which are 13 600, 11 100, and 12 000 m 3 /s for December, January, and February, respectively. The analysis was carried out on the last spring-neap cycle.
To analyze the subtidal mechanism, we introduced the residual unit width water flux (RUWF) and the residual unit width salt flux (RUSF). They can be calculated as where V ! is the velocity vector, D is the total water depth, s is the salinity, σ is the relative depth, which varies from 0 at the surface to −1 at the bottom, t 0 is the start time for integration, and T is the total period, which was six semidiurnal tidal cycles in this study. The water diversion ratio (WDR) is a crucial statistical concept in bifurcated estuaries. It can provide useful information about mass transport and exchange events . In this paper, the net transect water flux (NTWF) was used to calculate the WDR, and the locations of the transects are labeled in Fig. 1a. The NTWF is given as where V ! n is the velocity component normal to the transect and L is the width of the transect.

Saltwater intrusion before QCSR construction
3.1.1. During spring tide Figure 3 shows the distribution of RUWF, RUSF, and tidally averaged salinity during the spring tide in Exp. 1. The North Branch was nearly perpendicular to the main axis of the river in the bifurcation area, which to some extent obstructed the river runoff that discharged into the North Branch. The weak river discharge and the funnel shape of the North Branch (Qiu and Zhu 2015) eventually caused landward RUWF and RUSF in the North Branch. The RUSF was especially high at the mouth of the North Branch and brought a large amount of high-salinity water into the North Branch (Figs. 3c and 3d). Through the North Branch, the highly saline water finally spilled over into the South Branch. The NTWF in the upper reaches of the North Branch (section 1 in Fig. 1a) was −390 m 3 /s, and the WDR was −3.9% (Table 1); the negative sign of the WDR indicates that the water was transported from the North Branch to the South Branch.
The RUWF in the North Channel, North Passage, and South Passage was obviously seaward at the surface, while its magnitude sharply decreased at the bottom because of baroclinic force and bottom friction; the RUWF turned landward at the river mouth (Figs. 3a and 3b). The RUSF exhibited in a similar direction as the RUWF (Figs. 3c and 3d). Most of the river water (75.6%, Table 2) flowed into the sea through the North Channel rather than through the South Channel. However, the WDR of the North Passage was slightly larger than that of the South Passage (Table 3). The diluted water extended northeastward to the east of Chongming Island due to the large amount of NTWF flushing through the North Channel all the way to the sea (Figs. 3e and 3f). In contrast, the saltwater intrusion into the North Passage and South Passage was much stronger than that into the North Channel and South Channel.

During neap tide
During neap tide, the NTWF through section 1 became positive, with a value of 138 m 3 /s (Table 1), indicating that the SSO had disappeared. However, the saline water that spilled over from the North Branch during spring tide was advected downstream and influenced the salinity in the South Branch (Figs. 4e and 4f). Compared with those during the spring tide, the WDR of North Channel decreased by 14%, but the magnitude of the NTWF increased by almost 2000 m 3 /s, which meant that more net flux was diverted into the South Channel. In the river mouth area, a distinct salt wedge formed owing to the weak tide-induced mixing (Fig. 5). The steep salinity gradient strengthened baroclinic force and thus intensified the landward bottom RUWF and RUSF in the North Channel, North Passage, and South Passage (Figs. 4b and 4d). A stronger salinity front, which blocked the downstream river flow, also accounted for the large decrease in WDR in the South Passage, which changed from 48.9% to 23.8% (Table 3). The RUWF and RUSF flowed southward outside the river mouth because the northerly wind effect became dominant when the tide weakened (Figs. 4a-4d).

During spring tide
The distributions of the differences in RUWF, RUSF, and tidally averaged salinity between Exp. 2 and Exp. 1 during spring tide are depicted in Fig. 6. The difference in RUSF in the North Branch was landward (Figs. 6c and 6d), which explains the small increase in salinity in this area (Figs. 6e and 6f). QCSR had a considerable impact on the residual water and salinity transport in the upper reaches of the North Channel. The differences in RUWF and RUSF to the north of the QCSR were negative, which indicates that the NTWF is restricted by the QCSR to some degree; however, they turned positive immediately behind the reservoir (Figs. 6a-6d). In fact, the WDR in the North Channel decreased by 0.9% after the construction of the QCSR ( Table 2). The smaller amount of river water that bifurcated into the North Channel resulted in an increase in salinity of >0.5 psu to the north of the Eastern Chongming Shoal at the surface (Fig. 6e) and an even larger increase at the bottom (Fig. 6f).

During neap tide
The distribution patterns of the differences in RUWF, RUSF, and tidally averaged salinity at the surface and bottom during neap tide were similar to those during spring tide, except that the differences in RUWF and RUSF in the waters near the QCSR decreased to <0.1 m 3 /s (Figs. 7a-7d). The WDR in the North Channel in Exp. 2 was 1.7 lower than that in Exp. 1, and this value was greater than that during spring tide; these conditions augmented the    increase in salinity over the Eastern Chongming Shoal found during spring tide, especially at the bottom, and extended the 0.5 isohaline southerly to the northern dyke of the reclamation project in the Eastern Hengsha Shoal (Figs. 7e and 7f).

Impact on freshwater resources
Saltwater intrusion directly influences the security of freshwater resources during the dry season in the Changjiang Estuary, which is crucial for maintaining city operations. Previous studies concluded that the salinity at the water intakes of the DFXSR and CHR is completely determined by the SSO, while the salinity at the water intake of the QCSR is primarily controlled by the SSO but is also impacted by saltwater intrusion from the North Channel (Chen and Zhu 2014;Zhu 2018, 2019). Figure 8 illustrates the temporal variations in the surface salinity at the water intakes of the three reservoirs. As the closest reservoir to the bifurcation of the North Branch and the South Branch, the DFXSR is the earliest to experience the SSO, and the duration of the period during which the salinity is lower than 0.45 psu, the salinity standard for drinking water, is approximately half of the spring-neap period (Fig. 8d). The occurrence times of the salinity maxima at the water intakes of the DFXSR and CHR suggest that it takes approximately 2 days for the SSO to move from the DFXSR to the CHR, and the duration suitable for water intake lasts for nearly 10 days in one spring-neap cycle at the water intake of the CHR (Fig. 8c). The most significant difference of the three experiments in terms of salinity was at the water intake of the QCSR (Fig. 8b). During the moderate tide after neap tide (MTAN), the salinity peaks, which are mainly controlled by the intensity of the saltwater intrusion at flood slack, decreased sharply to lower than 0.45 after the engineering project, which meant the water was available for the reservoir during this whole period. As the salinity distribution along the profile during the MTAN (Figs. 9a and 9c) shows, the surface saline water extends farther landward in Exp. 1 than in Exp. 2 at flood slack; this reveals that the engineering structure slightly hinders the saltwater intrusion at flood tide. The situation reverses during the moderate tide after spring tide (MTAS) (Fig. 8b), when the salinity at the water intake of the QCSR in Exp. 2 is higher than that in Exp. 1. The duration during which the salinity was higher than the drinking water standard was extended after the QCSR, reducing the time suitable for water intake in other words. On the one hand, the impact of QCSR on the SSO is negligible, as the WDR decreased by only 0.1 (Table 1). On the other hand, as mentioned above, the saltwater intrusion became stronger after the project. These two aspects combined can explain the salinity increase during the MTAS.
In summary, the engineering project benefits the QCSR water intake during the MTAN but deteriorates the salinity conditions during the MTAS.

Discussion
Some large reclamation projects have been implemented in the Changjiang Estuary since 2007, such as the QCSR, the reclamation projects in the Eastern Hengsha Shoal and Nanhui Shoal, and others. These projects changed the local topography and indirectly drove topographic evolution throughout the entire estuary by altering its hydrodynamic processes. In addition to narrowing the river channel, the QCSR also deepened the channel due to seabed erosion. An extra experiment (Exp. 3) was conducted for discussing the indirect impact of the QCSR on the saltwater intrusion. In Exp. 3, the depth of Exp. 2 in the selected area (Fig. 1b) from the lower reaches of South Branch to the middle reaches of The distributions of the differences in RUWF, RUSF, and tidally averaged salinity between Exp. 3 and Exp. 2 during spring tide are depicted in Fig. 10. In the lower reaches of the South Branch and the upper reaches of the North Channel, the differences in RUWF and RUSF were landward, whereas they turned seaward in the middle reaches of the North Channel (Figs. 10a-10d). The NTWF in the North Channel was 7991 m 3 /s, and the WDR was 77.7%, which is 3.0% higher than that in Exp. 2. The larger amount of river water flowing through the North Channel into the sea induced a decrease in salinity of more than 0.5 psu over the Eastern Chongming Shoal. However, the salinity still increased in most areas of the North Channel (Figs. 10e and 10f). As shown in Figs. 11a-11d, at flood and ebb slack during spring tide, the saline water of the Exp. 3 intruded farther landward than that of the Exp. 2, with the surface layer 0.45 psu isohaline reaching the position of approximately 57 km on the x axis at flood slack (Fig. 11c). The deepened channel favored saltwater  intrusion during spring tide. The increases in salinity in the South Channel, North Passage, and South Passage were caused by the decreased NTWF and WDR.
During neap tide, the distributions of the differences in RUWF and RUSF were mostly the same as those during spring tide (Figs. 12a-12d), but the salinity in the whole North Channel showed a significantly decreasing trend (Figs. 12e and 12f). The WDR in the North Channel increased by 1.4% from its value of 59.9% in Exp. 2.
Considering the temporal variation in the surface salinity at the water intake of the QCSR, the salinity in Exp. 3 was higher than that in Exp. 2 and almost rose to the level in Exp. 1 during MTAN (Fig. 8b). This change was due to the stronger saltwater intrusion at flood slack and ebb slack during MTAN in Exp. 3 than in Exp. 2 (Figs. 9c-9f). The figure shows that the 0.45 psu isohaline at the flood slack moves farther landward over the position of 50 km on the x axis, which is the approximate location of the water intake. There were negligible differences in the temporal variation between Exp. 2 and Exp. 3 during the other tidal phases. On the whole, the new topography degrades freshwater security at the water intake of the QCSR to some extent but not substantially.

Conclusions
In this paper, a well-validated numerical model was applied to investigate the influences of the QCSR reclamation project and the topographic erosion induced by the QCSR on saltwater intrusion and freshwater resources in the Changjiang Estuary by comparing the results of three numerical experiments. In Exp. 1, namely, before the construction of the QCSR, the SSO occurred during spring tide due to weak river discharge and the topography of the North Branch. Most of the NTWF was diverted into the North Channel rather than the South Channel, and the North Passage and South Passage consequently experienced stronger saltwater intrusion than the North Channel. During neap tide, the SSO vanished, whereas the South Branch was affected by the saline water spillover from the previous spring tide. The WDR in the North Channel decreased to 61.6%, and the weaker tide facilitated the formation of a salt wedge at the river bar area, which blocked the river flow and steered the RUWF and RUSF landward in the bottom layer.  After the construction of the QCSR, the WDR in the North Channel decreased by 0.9% during spring tide, which led to the noticeable increase in salinity in the Eastern Chongming Shoal; this difference expanded further to 1.7% during neap tide, which amplified the increase in salinity. Regarding freshwater security at the water intake of the QCSR, the engineering project considerably decreased the salinity during MTAN because the saltwater intrusion was hindered at flood slack, especially at the surface, whereas the salinity increased by nearly 0.1 during the MTAS due to the combined effects of the SSO and the intensified saltwater intrusion during the spring tide as mentioned above.
In addition to the channel becoming narrower, the channel topography also experienced erosion due to the project. The influence of riverbed erosion in the area from the lower reaches of the South Branch to the middle reaches of the North Channel on saltwater intrusion was simulated. The comparisons between Exp. 2 and Exp. 3 showed that this topographic change decreased the salinity by more than 0.5 psu over the Eastern Chongming Shoal during spring tide by increasing the WDR in the North Channel, while the saltwater intrusion was strengthened, which resulted in increases in salinity in most areas of the North Channel despite the increasing NTWF. The 1.4% increase in WDR in the North Channel and the relatively weak tide resulted in a decrease in salinity through the entire North Channel during neap tide. Moreover, freshwater resources experienced slight degradation due to the erosion of the topography near the QCSR.