It is instructive to investigate the potential contribution of the proposed Huangpu Gate to the flood control in the Lake Taihu basin, which is currently still in the preliminary demonstration-of-benefit stage. The main research strategy used in this study is the scenario analysis based on numerical model simulations.

In total there are five different scenarios considered in this study as summarized in Table 1. Among them, the first scenario considers the case without gate construction, while all the other four scenarios consider the case with gate construction but with different operation modes. Scenario Base A is used as the baseline case for comparison with other scenarios, which represents the case where the estuary gate is not constructed at the Huangpu River mouth. Scenarios A1 and A2 are two cases where the proposed gate begins to operate when Lake Taihu was under a severe flooding situation. Scenario A3 is designed to analyse its contribution to prevent tidal water intrusion from exceeding a pre-defined threshold. The last scenario, A4, is the case for analysing its contribution to blocking all tidal water intrusion, which is the potential maximum contribution of gate construction to flood control of the Lake Taihu basin.

For scenario Base A, the estuary gate is not constructed at the outlet of the Huangpu River. Thus, the water in the Huangpu River and the Yangtze River estuary can exchange naturally. For scenario A1, the proposed gate would be operated in the rising stage of the lake levels according to weather forecasts. In the model simulation of the 1999 flood, the gate began to operate 7 days in advance before the lake level reached its peak value.

For scenario A2, the proposed gate would be operated when a large basin-wide flood occurs with the lake level higher than 4.50 m, indicating a severe and urgent flooding situation in the Lake Taihu basin. All drainage rivers linking the lake and the coast (bay, estuary) require the acceleration of floodwater drainage, including the Huangpu River.

For scenario A3, a portion of tidal water intrusion would be blocked by the gate, and the gate will remain open until the tide rises to a threshold (defined here as 4.0 m). That is, the gate would not be closed for blocking tide intrusion for each day; instead, it would be closed only under the situation when the high water exceeds the tide threshold (4.0 m) and is also forecasted to continue rising.

For the last scenario, A4, the gate would prevent all tidal water intrusion during the entire flooding period. It represents a hypothetical extreme case since it is not practical in implementation owing to the difficulty in overly frequent operation of such a huge gate with a width of about 400–500 m. This scenario is just a hypothetical case for analysing the potential maximum benefits to flood control of the basin. Indeed, it is not necessary to block all tidal water intrusion since under such operation it is also likely to produce negative impacts on both waterway transportation and the water environment system.

Considering the time needed for policy-making and gate construction, it is highly likely that the Huangpu Gate will not be completed and start operation until after 2025. For this reason, the proposed flood control projects in the plan designed by MWR (2008) will also be incorporated as the scenarios of flood defense in this study. Hu (2006) proposed the anchorage ground located at the mouth of the Huangpu River as the best site for gate construction since the negative impacts to shipping and navigation are least due to its location. Cui et al. (2012) and Lu (2008) proposed the same location for gate construction. Accordingly, in the following numerical model simulations the estuary gate will be located at the anchorage ground (shown in Fig. 1), which is about 5–6 km from the Huangpu River mouth.

The Huangpu River is the main shipping and drainage route in the Lake Taihu basin. In general, the embankment of the Huangpu River can protect against the occurrence of normal floods, while the proposed gate will be operated only under severe flooding situations. The following numerical model simulations for these five scenarios are all based on the condition during the 1999 flood event, which is the largest flood in history for the study area.

The HOHY model developed by Hohai University in China will be used in this study. This model has been tested in numerous regional applications since the 1970s, and was also applied at the Lake Taihu basin since 1997. It is one of the main products of a 3-year water quality study at the Lake Taihu basin, supported by the World Bank loan and jointly undertaken by Hohai University and Delft Hydraulics in the Netherlands. The HOHY model can simulate the cycle of floodwaters well. Meanwhile, the model can provide a broad-scale simulation of the flood control system in the Lake Taihu basin. It can simulate not only the complex hydro-systems with numerous interlaced rivers and lakes, complicated relationships between river nets, hilly topography, and tidal boundaries, but also the complex operational rules of control structures such as sluices, pumps, and siphons. This model has been utilized in a variety of past studies, such as the preliminary demonstration-of-benefit stage of water works in the Lake Taihu basin. In particular, the model has been successfully applied in the flood control planning of the Lake Taihu basin as approved by MWR (MWR, 2008).

The model is composed of two parts: a hydrologic part for simulating runoff generation and routing, and a hydraulic part for simulating channel flows. Each of them can run independently. The schematization of the model is shown in Fig. 2; more details of the model can be found in Cheng et al. (2006) and Jin (2009).

Runoff is generated when precipitation exceeds the total of infiltration, interception and depression storage. The basin land use is classified into four types: water surface, paddy field, non-irrigated farmland, and constructed land. Each of them employs different parameterizations to calculate runoff generation, and then the total runoff is routed according to basin topography. In hilly areas, the instantaneous unit hydrograph method is used, considering the storage and drainage processes of reservoirs and large ponds. In plain areas, the method of runoff curve number is used for each computed area.

After the runoff from the hilly and plain areas flows into river networks, the hydraulic method is applied for simulation of river flow. Only the lakes with a larger surface area are considered to possess the function of storing floodwaters, while other smaller lakes are considered intersections like the links among rivers. The operation of water-engineering works such as gates, pumping stations, and siphons is also simulated in the model. The Saint-Venant equations are used as the governing equations for the one-dimensional unsteady open channel flow, including the continuity equation (1) and the momentum equation (2) as follows: ∂Q∂x+∂A∂t=qL,∂Q∂t+∂∂xαQ2A+gA∂Z∂x+gAn2QQR1.333=qLνx, where x (m) is the distance along the channel; t (s) is the time; A (m2) is the cross-section area; Q (m3s-1) is the flow rate; Z (m) is the water level; α (-) is the momentum correction coefficient; R (m) is the hydraulic radius; qL (m2s-1) is the lateral inflow per unit length of channel; νx (ms-1) is the velocity of the lateral inflow in the x direction; and g (ms-2) is the gravity acceleration.

The model parameters of the numerical simulations in this study are specified to be the same as those used in the design plan by MWR (2008). The model calibration data are from 2 consecutive years, 1984 to 1985, and the validation data are from 1995 and 1996. The model has also been validated by Ou and Wu (2001) using the observed water level and river flow data of the 1999 flood. Figure 3 compares the difference in water level simulations with observations during the 1999 flood at eight representative stations of the basin (see Fig. 1 for their locations). Figure 4 shows the differences in river discharges between observations and simulations at the Taipu Gate and the Wangting Siphon (see the location shown in Fig. 1). These comparisons of simulated water levels and discharges with observations demonstrate that the simulations of the HOHY model are of sufficient accuracy to be used in the following scenario analyses.

Among the five scenarios considered, scenario A3 is the most complex to simulate since different operational rules of the gate are applied for the flood tide and ebb tide, respectively. If the high water in the flood tide is higher than the tide threshold, the gate would be closed. Once the gate is closed, it will not be re-opened until it has the natural water-expelling ability to drain floodwaters in the ebb tide (until the tide level falls to be lower than the water level in the upstream of the gate). Hence, the HOHY model needs to be modified in order to enhance its capability for this purpose.

The model modification is based on the flowchart given in Fig. 2, focusing on the flood routing part related to the algorithms of unsteady open channel flow, and the control rules of gates related to the tidal condition. The main program of the HOHY model is improved by adding a function to judge the stage of the tide before running the gates (i.e. in the flood or the ebb tide), which makes the specification of the control rules of gates more flexible.

The modified HOHY model is tested by using a simple case in which the tide threshold is assumed to be 4.0 m. The simulation results are presented in Figs. 5 and 6. Figure 5 describes the case when the gate remains open since the high water in the tidal period is always lower than 4.0 m. Figure 6, as a comparison, is the case where the gate will be closed when the rising tide is higher than 4.0 m. The gate will re-open to drain floodwaters when the gate has the natural water-expelling ability to evacuate floodwaters in the ebb tide.

The model results, including the gate discharge, the tide water level at the estuary, and the difference in water levels between the upstream and downstream of the gate, show the reasonable relationships of the operational rules of the gate. Figure 5 demonstrates that the discharge at the gate resembles a sinusoidal curve as affected by the tidal boundary. It is likely that the gate does not need to be closed since the high water during the tidal period is less than the tide threshold of 4.0 m. Figure 6 is another case of the gate operational rules of which the high water is about 4.70 m. At 02:30, 15 August 1999, the tide level at the river outlet in the flood tide slightly exceeded the tide threshold of 4.0 m, and the gate has to be closed. It was not re-opened in the ebb tide until 08:15, 15 August 1999, when the water level in the upstream is higher than that in the downstream near the gate location, meaning that the gate has the natural water-expelling ability to drain floodwaters at this moment (see the red bars in Fig. 6). Overall, the modified HOHY model has demonstrated its ability to simulate the complicated operational rules of the proposed Huangpu Gate well.

Table 2 summarizes the peak values of lake water level and the duration (the number of days) when various control levels were exceeded during the June–August period in 1999 for the five scenarios considered in this study. Figure 8 plots the simulated water levels at Lake Taihu corresponding to five scenarios during the 1999 flood event from June to August. As seen, the lake levels in scenarios A1, A2, A3, and A4 were all lower than that in scenario Base A. Similarly, the duration when the water level is higher than a certain control level was also reduced. Compared with the maximum daily water level of 5.03 m (that occurred in early July) in scenario Base A, the maximum water levels in other scenarios were decreased by 0.04, 0.01, 0.03, and 0.12 m, respectively, for scenarios A1, A2, A3, and A4. Thus, these four scenarios contribute to the flood control capacity of Lake Taihu and its adjoining low-lying areas to the west.

It should be noted that the differences in the design water levels corresponding to different return periods are not significant for such typical shallow lakes located in the low-lying plains. For instance, the design water level of the 100-year return period is 4.80, 0.15 m higher than that corresponding to the 50-year return period. For this reason, the decrease in the peak lake level by 0.04 m in scenario A1 as well as by 0.12 m in scenario A4 is significant for flood control of the lake. Additionally, the western adjoining floodplains would also benefit from the gate construction. Due to the relatively lower flood control capacity of the western adjoining areas, those regions are likely to be inundated when the sluices cannot yet control the water intrusion from the lake to the adjoining areas once the lake level is too high. The flooding condition in the western adjoining areas will be even worse once the lake breaches the dike.

From the viewpoint of flood control of the lake, it can be concluded that the Huangpu River with an estuary gate is more effective than without a gate. The gate operation would prevent or reduce the amount of tidal water from entering the Huangpu River that already has high water levels caused by increased river flows from the lake and the surrounding areas. The extent of the gate contribution to flood control depends largely on its operation mode and duration. The longer the gate is operated, the less tidal water will intrude into the Huangpu River estuary, and the more floodwaters in the lake will be drained to the Yangtze River via the Taipu Canal and the Huangpu River. Overall, scenario A1 is a nice example to examine the potential contribution of the proposed gate. In the simulation of the 1999 flood, the Huangpu Gate is more effective at reducing flood risk in the lake by operating the estuary gate in advance. Even with the case of operating the gate by a relatively short duration such as 1 week as assumed in scenario A1, the contribution to reduce the peaks and the rising rate of lake levels is rather significant.

The Huangpu River also receives the floodwaters drained from the following surrounding areas, including the (1) Yangchengdianmao, (2) Hangjiahu, (3) Puxi, and (4) Pudong catchments, as shown in Fig. 7. Therefore, the safety of these four catchments against flooding is also closely linked to the capacity of the Huangpu River. Table 3 lists the peak water levels at the four representative stations (S1 to S4, shown as the orange circles in Fig. 1), each for one of the above four surrounding catchments. Figure 9 plots the simulated daily water levels during the 1999 flood at these four stations, from which a similar trend in the water level to that in Lake Taihu (Fig. 8) can be observed. Scenario A4 represents the potential maximum contribution of the gate, i.e. the maximum decrease in the daily peak level at the four stations in surrounding areas is 0.32, 0.19, 0.39, and 0.05 m, respectively. In contrast, the improvement in flood control capacity at station 4 located in the Pudong catchment is the smallest among the four stations due to its unique terrain. The local floodwaters in the Pudong catchment have the priority of draining to the East China Sea over that draining to the Huangpu River due to its natural water-expelling ability. Generally, the flood capacity of station 4 does not depend as much on the drainage capacity of the Huangpu River as the other three stations.

In contrast, for scenario A1 in which the gate is operated in advance, the gate can play a notable role in reducing the peak water levels by the amount between 0.15 and 0.35 m, except for station 4. For scenario A2, the gate can decrease the peak water levels only by 0.07–0.15 m. However, scenario A2 has more advantages in speeding up the drainage rate of floodwaters at the recession stage and shortening the time of waterlogging.

Figure 10 plots the simulated daily water levels in the Taipu Canal and the Huangpu River at the seven cross sections (as marked by the purple rectangle in Fig. 1). The daily water levels at the Taipu Canal and the Huangpu River decrease to various extents when the gate is in operation. Scenario A4 represents the potential maximum contribution of the proposed gate due to complete prevention of tidal water intrusion. In this scenario, the maximum reduction in the peak water level is 0.26–0.37 m for the Taipu Canal and 0.46–0.60 m for the Huangpu River.

The Huangpu River benefits more from gate construction than the Taipu Canal because the latter is located relatively farther away from the gate. The potential contribution of the gate can be attributed to the reduction of tidal water intrusion during the flood period. Generally, the tidal intrusion is mainly concentrated on the lower reach of the Huangpu River, although the intrusion can propagate upward as far as more than 100 km from the estuary. The water level will rise to different extents in the Taipu Canal and the Huangpu River when the gate is closed, and then the discharge rate will increase when the gate re-opens again due to the relatively large difference in water levels between the upstream and downstream sides near the gate. Therefore, the gate can decrease the water levels of the Huangpu River more markedly than that of the Taipu Canal.

In scenario A1, the gate is operated in advance during the rising stage of the lake level, and the peak flood level in the Taipu Canal and the Huangpu River can be decreased considerably due to the enlargement of the drainage capacity of the Huangpu River. In scenario A2, the gate is operated when the lake level is higher than 4.5 m, and its contribution to the peak water levels is less than scenario A1, while the draining rate in the recession stage is faster. If the gate is operated by blocking the high tide during the flood period (scenario A3), the peak water levels at the Taipu Canal and the Huangpu River are decreased during the spring tides. This conclusion is completely consistent with those discussed in the previous sections on the contribution of the proposed gate to the flood control of the lake and the surrounding areas.

Table 4 describes the inflow volumes from the upstream tributaries to the Huangpu River during the flood period. In addition to the Taipu Canal, there are many upstream tributaries originating from the north-western and south-western upstream areas of the Huangpu River (Fig. 7). In scenario A4, the inflow volume from the south-western tributaries into the Huangpu River is up to 3.25 billionm3, more than twice that in scenario Base A (1.50 billionm3). The inflow from the north-western upstream areas in scenario A4 is about 1.05 billionm3, increasing by 78 % in comparison to that in scenario Base A (0.59 billionm3). The inflow volume from the Taipu Canal is about 5.0 billionm3, only increasing by 27.2 % compared to that in scenario Base A (3.93 billionm3). In terms of the major inflows into the Huangpu River, the inflow volumes from the south-western and north-western upstream areas increase significantly in comparison to that from the Taipu Canal, suggesting that the Huangpu River plays a dominant role for these two upstream subareas.

Table 5 describes the tide intrusion and outflow volume at the site of the proposed gate during the flood period. The proposed gate helps to improve the drainage efficiency of the Huangpu River by protecting the river from tidal water intrusion. Compared to scenario Base A, the net outflow volume at the gate site during the entire flood period under the other four scenarios is increased by 4 % (A1), 8 % (A2), 22 % (A3), and 52 % (A4), respectively. Figure 11 shows the comparison of simulated river discharge at the site of the proposed gate between scenarios Base A and A1 from 27 June to 3 July in 1999. The difference in river discharge between these two scenarios clearly reflects the difference in the drainage efficiency of the Huangpu River. Although the river discharge in scenario A1 is only increased by 3.5 % for the entire flood period (increased from 7.20 to 7.45 billionm3), it should be noted that the influence on the flood control during the gate operation period (from 27 June to 3 July) is more significant. The net outflow volume is nearly doubled by changing the bi-directional flow to the unidirectional flow (as shown in Fig. 11).