Dianshan Lake is a tidal lake located in the lower reaches of Taihu Lake Basin, which is an important water source in Shanghai(Fig. 1). The shoreline length is 62.3 km and its area is 58.9 km². The mean annual water level of Dianshan Lake is lower than 2.63 m based on the Wusong datum level in Shanghai. The average water depth of this lake is 2.1 m, and the maximum water depth is 3.59 m. The maximum depth is in the main channel between Jishuigang and Lanlugang. Most water depths in the south and east are less than 2 m, while the water depths in the north are 2.3-2.5 m. The main water source of Dianshan Lake is Taihu Lake Basin. The outflow of Dianshan Lake goes through the Huangpu River into the Changjiang River estuary. There are over 59 rivers that influence Dianshan Lake, with the main inflow rivers being Jishuigang, Dazhuku, Baishiji, and Qi and engpu and the main outflow rivers Lanlugang, Xiwanggang, Shitanggang, and Dianpuhe. The outflow rivers also conduct the tidal flow into the lake. According to the hydrology data observed in 2008 and 2009 by Shanghai Municipal Hydrology Station(SMHS), the total inflow of Jishuigang and Dazhusha accounted for 82% of the main four inflows, while the outflow of Lanlugang accounted for 84% of the four main outflows.

Fig. 1 Location, water depth and main inflows and outflows of Dianshan Lake Geographic Reference System: UTM-WGS84, N51 zone.

Figure options

Dianshan Lake has deteriorated to a hypereutrophic lake in the last 20 years because of heavy nutrient inflows around the lake and upstream from industrial, agricultural, and household sewage(Zhang et al., 2006; Cheng and Li, 2008). By analyzing observed water quality data from SMHS in 2008, we found that the inflow flux of total nitrogen reached 5790.7 ton/a, while the inflow flux of total phosphorus reached 419.8 ton/a. The main phytoplankton phyla in Dianshan Lake were Cyanophyta Bacillariophyta and Chlorophyta, although Euglenophyta, Cryptophytak, Chrysophyta, Pyrrophyta and Xanthophyta were also present(Xu et al., 2012). The average annual relative abundances of the main phytoplankton phyla were about 50%, 35%, and 10%-15%, respectively. The population of Cyanophyta increased gradually in summer, reaching the highest levels in August, when they accounted for 90% of the total algal population. In the last 10 years, algal blooms occurred several times in summer, during which time Microcystis was the dominant organism(Zhang et al., 2006; Xu et al., 2012).

Previous studies of eutrophication in Dianshan Lake were primarily experimental data analyses, and the main eutrophication simulation studies in Dianshan Lake have focused on the response relationship between the concentrations of chlorophyll - a and nutrients, in which chlorophyll- a was used to represent algae. The present study was conducted to build an eutrophication model to simulate the seasonal growth of diff erent algae in Dianshan Lake, and to improve the accuracy of simulation of algal blooms. High accuracy eutrophication simulation projects will enable identification of an eff ective method to control algal blooms in Dianshan Lake.

In this study, the hydrodynamic conditions and algal spatial distribution characteristics in the algal bloom period were analyzed by numerical simulation. The relationship between flow stagnant area and algae spatial distribution was found. Two types of algae control methods, reducing nutrients inflow and improving flow state, were simulated to compare their eff ects on algal blooms in Dianshan Lake. The modeling processes of the hydrodynamic model and aquatic ecology model are given in Section 2. The validation and simulation results of the two models are shown in Section 3, as well as the results of diff erent algae bloom control projects. Section 4 is the conclusion.

The hydrodynamic simulation module(FLOW)of Delft3D was applied to build a 3D hydrodynamic model in Dianshan Lake. Non-orthogonal coordinates in the horizontal direction and the σ coordinate in the vertical direction were utilized in the grids of this model. According to the shoreline and bathymetry data of Dianshan Lake, the model grids were designed to be equally distributed in most of the simulated area and only refined around the channel. The horizontal grid size was about 120 to 150 m(Fig. 2). An σ-coordinate transformation was adopted in the vertical direction. The depth proportions of each vertical level were 15%, 20%, 30%, 20%, and 15%. Eight flow boundaries were set into the model according to the main inflows and outflows. The actual data of flow boundaries for June and September 2008 were adopted to calibrate and validate the hydrodynamic model. Because there were no detailed real-time data and the characteristics of the residual flow field in Dianshan Lake could be analyzed more easily without tidal influence, monthly averaged flow data were used to simulate the flow field in 2008. The hydrodynamic results for the whole year were used to support the ecological model. Additionally, actual wind data of Dianshan Lake for 2008 were used in this model. The time step of the simulation was 2 min, and simulated data were output per 6 min.

Fig. 2 Model grids and observation sites F1-F3: hydrology sites; W1-W6: water quality sites.

Figure options

Using the water quality simulation module(WAQ)of Delft3D(Los et al., 2008), the ecological model in Dianshan Lake was constructed with the same grid of the hydrodynamic model. The time step of WAQ simulation was 6 minutes, and simulated data were output per 1 hour. The main ecological processes in this model are shown in Fig. 3. Dissolved oxygen, ammonia nitrogen, nitrate, phosphate, dissolved silicon, organic detritus(both in water and sediment), biological oxygen dem and (BOD5) and three algae(Bacillariophyta, Chlorophyta, and Cyanophyta)were simulated in the ecological model.

Fig. 3 Primary ecological processes in the model

Figure options

Microcystis was selected to represent Cyanophyta because it is the dominant genera of Cyanophyta in Dianshan Lake(Zhang et al., 2006; Xu et al., 2012). In the model, each species was divided into types that represented diff erent states of a given species; namely, being energy-limited(E-type), nitrogen-limited(N-type), or phosphorus-limited(P-type). E-type is the main type because of the high concentrations of nitrogen and phosphorus in Dianshan Lake(Kung and Ying, 1991; Cheng and Li, 2008). Wind, solar radiation, water temperature, zooplankton grazing, and coupled hydrodynamic condition were set as external driving variables in this model.

Only the nutrient loading from upstream was considered in the model because it is the most important source entering Dianshan Lake. Based on the historical nutrient loading data(1984-1985), surface waters were estimated to contribute up to 82% in nutrient loadings, whereas the other main sources(precipitation(6%) and direct farml and runoff(4%)made much lower contributions(Kung and Ying, 1991). The results of the present study(2008-2009)from Shanghai Environmental Monitoring Center(SEMC), Shanghai Municipal Hydrology Station(SMHS), and Shanghai Academy of Environmental Sciences(SAES)were in agreement with these findings(Table 1). The monthly observed water quality data for 2008 off ered by SEMC were used to set the nutrient flux boundaries of four inflows and four outflows. Because the available data do not comprise some important parameters such as phytoplankton concentrations and zooplankton concentrations, some assumptions must be made to estimate the average values and seasonal profiles for these concentrations. Phytoplankton biomass was derived from chlorophyll- a concentrations, relative abundance of species and species-specific Chl:C ratios. The relative abundance of phytoplankton species was assumed to be constant in time at the inlets(Zhang et al., 2006). Zooplankton biomass was assumed based on other studies(Zuo et al., 2008; Feng et al., 2011). Driving forces such as wind speed, water temperature, and solar radiation were set according to the local real-time data in 2008. The meteorological observed data were collected hourly at the lakeside by an automatic weather station maintained by the Meteorological Bureau of Qingpu District.

Table 1 Annual mean loading of nutrients into Dianshan Lake(2008-2009)

Table options

There were 62 main process coefficients in the ecological model(Table 2). The reasonable ranges of these coefficients were determined by experiments and articles(Escaravage and Soetaert, 1995; Boers et al., 1998; Søndergaard et al., 1999; Pei and Wang, 2003; Robson and Hamilton, 2004; Arhonditsis and Brett, 2005; Schindler, 2006; Xiao et al., 2011; Wu et al., 2011; Bone and Altaweel, 2014; Chung et al., 2014), and their values were adjusted and confirmed in the calibration.

Table 2 The values of main process coefficients in the ecological model

Table options

Bottom roughness, wind drag coefficients, and vertical and horizontal viscosity were calibrated according to the real data observed in June 2008 off ered by SMHS. The model was validated with the real data in September 2008(Fig. 4). Analysis of the real and simulated results revealed that the velocity in Dianshan Lake is no more than 0.08 m/s, and that there was an obvious tidal phenomenon resulting in change of flow. However, some irregular periods exist in this tidal change because of the influence of wind force. The calibration and validation results showed that the hydrodynamics characteristics in Dianshan Lake were simulated well with the hydrodynamic model, and the simulated results could off er reliable hydrodynamic conditions for the ecological model.

Fig. 4 Calibration and validation of elevation, surface velocity, and direction at F2 Left panel: calibration results from June 15, 2008 to June 22, 2008; right panel: validation results from September 24, 2008 to September 28, 2008; dots: observed data; lines: simulated data.

Figure options

The simulated hydrodynamic results for Dianshan Lake in 2008 revealed that, when the influence of tide was not considered, the flow field was dominated by the wind. The predominant wind is from the northwest in November to February, the averaged wind velocity was 2.2 m/s and the main direction of surface flow is east-southeast. Additionally, there are clockwise circulations in the mid and bottom layers(Fig. 5), and the predominant wind is from the east in March to June, with an average velocity of 2.6 m/s.

Fig. 5 Flow fi elds in the mid layer under diff erent predominant winds a. northwest wind; b. southeast wind.

Figure options

The surface flow direction is east-southeast in the channel, while it is west and southwest in other areas. There are counterclockwise circulations in the mid and bottom layers. The predominant wind is southeast from July to October, with an average velocity of 2.3 m/s. The surface flow is small in the channel, and the main flow directions are west-northwest and westsouthwest in Dianshan Lake, except in the channel.

There are counterclockwise circulations in the mid and bottom layers(Fig. 5), although the main circulations are all in the north of Dianshan Lake. According to the continuous change in wind direction, the flow fields for the entire year are separated into three time intervals: November to February, March to June, and July to October. The areas are considered stagnant areas if the maximum velocities are less than 0.01 m/s(Fig. 6). There were no stagnant areas in the surface layer for any intervals because of the obvious wind influence. The main direction of bottom currents is almost opposite to the surface owing to the water level gradient and water compensation. The largest stagnant areas are in the mid layer for the restriction between wind force and water level pressure gradient. The main direction of flow without wind is east, which is almost opposite to the main direction of flow from March to October because the predominant winds are east or southeast during these months. Therefore, the stagnant areas from March to October are larger than those from November to February, and they primarily gather in the south, middle and northeast portions of Dianshan Lake.

Fig. 6 Distribution of stagnant areas(covered by grey triangles)in the mid layer under diff erent predominant wind a. east wind; b. southeast wind.

Figure options

The ecological model was validated using monthly observed data from 2008(observation sites are shown in Fig. 2). Chlorophyll- a(Chl- a), ammonia nitrogen(NH₃ -N), total nitrogen(TN), total phosphorus(TP), dissolved oxygen(DO), BOD5 and Secchi depth(SD)were validated in this model(Fig. 7). With the exception of Chl- a, the nutrients in the shallow lakes showed no significant diff erences vertically(Zhu et al., 2004). The observed and experimental data obtained at a depth of 0.5 m below the surface according to the Criterion of Investigation on Lake Eutrophication(Jin and Tu, 1990)were adopted as the real data in this validation, while the top layer simulated results were used as the simulated data. The values of real data and simulated data do not fit with each other well, while their monthly variations were almost the same. The simulated errors of each parameter were calculated to evaluate the accuracy of this ecological model. The equation adopted from other similar studies was as follows(Pei and Ma, 2002):

Fig. 7 Comparison of simulated results and real data Left panel: site W1; right panel: site W5; lines: simulated results; dots: real data.

Figure options

The maximum errors are from Chl- a and BOD5(averaged errors=67.6% and 31.7%, respectively), while the minimum errors are from SD and DO(averaged errors=8.11% and 8.75%, respectively). The averaged errors of other parameters are less than 30%. The errors of NH₃ -N and TP(52.56% and 44.83%)calculated from site W6 were larger than from other sites. This phenomenon is considered to be caused by deviations in the nutrient inflow of Qi and engpu(north boundary in Dianshan Lake). As shown in Table 3, when compared with the calibration errors in other studies(Pei and Ma, 2002), the errors of simulated results are acceptable to describe the variations in water quality and algae for the whole year. The ecological model sensitivity and uncertainty in shallow lakes have been introduced in other similar studies(Pei and Wang, 2003; Li et al., 2014).

Table 3 Errors of simulated water quality parameters at each observation site

Table options

Chl- a values were used to present algae in the validation because there are no observed data for algal species biomass in 2008. However, the seasonal growth regularity of the three algae could still be analyzed using the simulated results of algal species. The simulated algal variation processes at observation sites are shown in Fig. 8. The results revealed that the algal growth regularities at the observation sites are similar to each other. The predominant algae is Bacillariophyta from February to March, while it changes to Chlorophyta from April to May, and Cyanophyta from July to August. This regularity is consistent with the results of previous studies(Kung and Ying, 1991; Zhang et al., 2006; Wu et al., 2011; Hu et al., 2012). Considering the hyper-eutrophic environment in Dianshan Lake, nutrients are not the limiting factors for algal growth, while seasonal variations in water temperature and most suitable temperatures for diff erent algal species growth are essential to determination of the predominant algae.

Fig. 8 Variations in algal processes at observation sites Red solid line: Bacillariophyt a ; green dashed line: Chlorophyta ; blue dashed-dotted-line: Cyanophyta.

Figure options

Comparison of the species peaks in diff erent sites revealed that the peaks of Chlorophyta were lower than those of Bacillariophyta in the north(W1, W2, W3), while the peaks of Chlorophyta were higher than Bacillariophyta in the south(W4, W5, W6). The peaks of Cyanophyta were much higher than the others for the entire lake.

As shown in Fig. 8, two large-scale algal blooms occurred in 2008. The first bloom occurred in early July(close to July 3), during which time the main bloom area was in the south portion of Dianshan Lake(W1 and W3). The second bloom was observed in mid-August(close to August 15), when the main bloom area was in the northeast portion of Dianshan Lake(W5 and W6). In the algal bloom period, the biomass of Cyanophyta grew quickly, reaching levels much higher than those of Bacillariophyta and Chlorophyta. The simulated Cyanophyta distributions of the surface layer during the peak periods of the algal bloom were extracted(Fig. 9)to compare with the EOS/MODIS satellite pictures from SEMC(Fig. 10). The simulated results matched the bloom distribution shown in the satellite pictures well, and the ecological model was shown to be reliable to simulate the algal bloom in Dianshan Lake. The main bloom areas are in the south and northeast of Dianshan Lake, which corresponded to the stagnant areas(Fig. 6). Therefore, the flow residence time is speculated to be a key factor influencing algal accumulation and growth, with a longer residence time corresponding to more rapid algal growth.

Fig. 9 The distributions of simulated Cyanophyta during peak periods of algal blooms in 2008 a. July; b. August; unit: g/m3.

Figure options

Fig. 10 The distributions of algae based on EOS/MODIS satellite images

Figure options

Five scenarios were simulated to validate the speculation. Scenario 1 was the benchmark scenario, in which nothing changed. In Scenario 2, all nutrient inflows were reduced by 30%, while in Scenario 3 they were reduced by 50%. In Scenario 4, 10 m³ /s discharge was added to the Baishiji inflow boundary(almost 100% of the annual flow in Baishiji), while Scenario 5 added 10 m³ /s discharge to the Dazhuku inflow boundary(almost 50% of the annual flow in Dazhuku). The discharges of outflows in Scenario 4 and 5 were added in proportion to ensure mass conservation according to the simulated results of particle tracing. The reduction in the algae bloom in Dianshan Lake under each scenario was evaluated by the statistical areas of the Cyanophyta peak concentration(Table 4). According to the results of Scenario 3 shown in Table 4, the high concentration area of Cyanophyta(higher than 2.3 g/m³)could only be reduced by 3.7%(from¹⁰.9 km² to 10.5 km²), and the lower concentration areas were almost unchanged. These results show that the algal growth is not obviously limited by reducing the nutrient inflows by 50%, which was likely because of the hyper-eutrophic environment in Dianshan Lake.

Table 4 Statistical areas of microcystis peak concentration in scenarios

Table options

According to the results of Scenario 4 and 5 shown in Table 4, the high concentration area of Cyanophyta(higher than 2.3 g/m³)could be reduced by more than 88%, while lower concentration areas were obviously reduced. The stagnant areas during the algal bloom periods in Scenario 4 and 5 were clearly narrower than those in the benchmark scenario(Figs.6 and 11), which matched the statistical distributions of maximum Cyanophyta concentration well(Fig. 12). The results of scenario simulations demonstrate that the accumulation and growth of algae can be inhibited eff ectively by adding inflow discharges and activating the flow field.

Fig. 11 Distributions of stagnant areas(covered by grey triangles)in the mid layer in Scenario 4(a) and 5(b)during the algal bloom period

Figure options

Fig. 12 Statistical distributions of maximum Cyanophyta concentration in Scenario 4(a) and 5(b)during the algal bloom periods(unit: g/m3)

Figure options

The flow module of Delft3D was applied to build a 3D hydrodynamic model in Dianshan Lake. There is an obvious tidal phenomenon that leads to changing flow based on both the observed and simulated results. However, some irregular periods exist during this tidal change because of the influence of wind force. Without considering the tidal influence, the flow field is dominated by the wind. The surface currents are obviously influenced by the wind, and the main direction is consistent with the wind. The main direction of the bottom currents is almost opposite to the surface because of the water level gradient and water compensation. There are clockwise circulations in the mid and bottom layers when the predominant wind is from the west, while they are counterclockwise circulations when the predominant wind is from the east. The largest stagnant areas are in the mid layer for the restriction between wind force and pressure gradient, which is primarily in the south, middle and northeast portions of Dianshan Lake.

The ecological model was built with the WAQ module of Delft3D. Upon validation, the maximum errors were observed for Chl- a and BOD5(average errors=67.6% and 31.7%, respectively), while the minimum errors were for SD and DO(average errors=8.11% and 8.75%, respectively). The average errors of the other parameters were less than 30%. The concentration of Chl- a is obviously aff ected by the growth of algae(Hu et al., 2012), and the growth of algae fluctuates significantly due to variations in water temperature and solar radiation(Li et al., 2014). As a result, it is difficult for the monthly observed data to fully represent the status of Chl- a in a month, which might lead to more uncertainty regarding the simulation of Chl- a. SD, DO and other nutrients change seasonally with the upstream input and water temperature, but their fluctuations are small within a month. According to the simulated results, the predominant algae is Bacillariophyta from February to March, while it changes to Chlorophyta from April to May, and Cyanophyta from July to August. This regularity is consistent with the results of previous studies(Zhang et al., 2006; Hu et al., 2012; Xu et al., 2012). The simulated Cyanophyta distributions of the surface layer during the peak periods of the algal bloom matched the bloom distribution shown in the EOS/MODIS satellite pictures well, which can be attributed to the stagnant areas. Five scenarios were simulated to evaluate diff erent methods to reduce the algae bloom.

According to the scenario in which nutrient inflows were reduced by 50%, the high concentration area of Cyanophyta(higher than 2.3 g/m³)could only be reduced by 3.7%, while the lower concentration areas were almost unchanged. According to the scenarios of increasing inflow discharges(Baishiji or Dazhuku), the high concentration area of Cyanophyta could be reduced by more than 88%. These findings indicate that the accumulation and growth of algae can be eff ectively inhibited by increasing inflow discharges.

The biomass of Cyanophyta will increase greatly with suitable growth temperature during summer; however, the accumulation of Cyanophyta can be restrained by adding inflow discharges, narrowing stagnant areas and activating the flow field. By adding inflow and outflow discharges to Dianshan Lake, blue algae can be transported out of the lake before they have the opportunity to bloom. Conversely, Microcystis, the dominant genera of Cyanophyta in Dianshan Lake, can control its floating and sinking by intracellular gas vesicles(Komárek, 2003). Although this provides a great competitive advantage in stagnant flow environments, this advantage can be obviously weakened by increasing flow velocity and turbulence.