Assessment of a new nutrient management strategy to control harmful cyanobacterial blooms in Lake Taihu using a hydrodynamic-ecological model

The external nutrient loading significantly affects large shallow lakes, particularly those with intricate rive networks. In Lake Taihu, pollutant discharge standards have traditionally been quantified based on water environmental capacity, while neglecting the response characteristics of algal growth in different regions to boundary inputs. For that analysis, this study first estimated the river pollutant loadings of 16 most significant inflow and outflow rivers of Lake Taihu from 2008 to 2020 and explored the correlations between inflow water quality parameters and lacustrine chlorophyll-a levels. Results highlighted the significant influence of high river input of permanganate values in spring on the chlorophyll-a levels in the lake. Based on this, this study proposed the hypothesis that reducing the inflow concentration of permanganate in spring would inhibit algal growth, which was further validated using coupled hydrodynamic and ecological models. The simulation results indicated that the reduction of permanganate inflow concentration during spring would significantly decrease chlorophyll-a concentration in spring and summer, especially leading to a notable impact on peak values. However, due to variations in background concentrations among rivers, the extent of reduction in lake chlorophyll-a levels showed significant spatial differences. Additionally, analysis of extracted algal bloom areas revealed that there still remained a relatively high risk of algal blooms occurring in the main regions, particularly during autumn when inflow pollutant concentrations increased rapidly. These findings emphasized the importance of formulating reasonable exogenous reduction schemes, which should consider the concentration and variation trend of inflow pollutants, as well as the response characteristics of algal bloom growth in different lake zones to the boundary.


Introduction
The Lake Taihu basin is a renowned water network area in China, and the crisscross river water network serves as both an important transportation hub and a channel for water exchange and resource sharing (Xu et al 2010).Lake Taihu, as the core of the basin, provided domestic drinking water for approximately 10 million people before the water pollution crisis (Guo 2007, Yuan et al 2019).However, since the 1980s, the speed of economic development has outpaced the self-recovery speed of the ecosystem, leading to numerous environmental crises (Romero et al 1999), particularly eutrophication and harmful cyanobacterial blooms (CyanoHABs) (Paerl et al 2011, Xu et al 2021b).Since the drinking water Crisis caused by CyanoHABs in 2007, efforts for water environment governance and ecological restoration of Lake Taihu have been accelerated.However, despite extensive restoration work, Lake Taihu experienced a severe and year-round CyanoHABs in 2017, indicating minimal improvement in water quality and limited inhibition of cyanobacteria growth (Qin et al 2021, Kang et al 2023).
It is widely known that eutrophication results from the increased input of nutrients to waters, therefore, rivers, especially inflows, exert a significant influence on the water quality of lakes (He et al 2020).However, the eutrophication resulting from nutrient pollution from rivers remains a major and unresolved challenge in pollution control and water resource management.Despite ongoing efforts, a fundamental change in addressing this issue has not yet been achieved.In response, in recent years, there has been a growing focus on in-lake methods to expedite the restoration process (Yin et al 2022).However, some case studies of remediation action indicated that in the absence of external control, in-lake methods may not necessarily be effective in controlling CyanoHABs (Bormans et al 2016).For instance, both the dredging in Vajgar fishpond, Czech Republic, and the years-long hypolimnetic aeration in Lake Tegel, Germany, have proven unsuccessful in controlling cyanobacteria and the failure of these measures can be attributed to the lack of concurrent management of external loads (Pokorný andHauser 2002, Schauser andChorus 2009).For Lake Taihu, after studying the influence of sediment dredging on the release of phosphorus, Zhong et al (2008) suggested that dredging could be a possible measure to control CyanoHABs after external loading was efficiently reduced.In conclusion, efficiently decreasing external pollution is the first and indispensable step for CyanoHABs control.
The central, provincial, and local governments have implemented various management strategies for excessive nutrients in Lake Taihu basin, including regulations and rules (Yan et al 2019b).Unfortunately, the pollution load carried by the rivers remains the primary exogenous source of pollution.Taking total phosphorus load as an example, the annual riverine total phosphorus load accounts for 55% to 70% of the total load, serving as the primary pathway for exogenous input (Li et al 2022).One of the reasons for this is that the current studies on the control and reduction of major water pollutants are mainly based on the regional water environment capacity, especially regarding total nitrogen and total phosphorus (Yan et al 2019a, Hu et al 2023).However, in reality, legacy internal nitrogen and phosphorus supplies can lead to a lag between reduction in external loading and decreases in bloom magnitude and areal extent (Xu et al 2021a).Therefore, these studies often overlook the responses of lake water quality improvement and CyanoHABs occurrence to the reduction of river pollutants at different time periods and boundaries.On the other hand, the inflow rivers are located in different administrative provinces and regions, and the pollution sources vary significantly (Zha et al 2018).This leads to variations in the impact of inflow rivers on the water quality of different areas of the lake, making it challenging to balance the effectiveness of source control and pollution interception measures across different lake areas.
To approach these questions, this study first analyzed and understood the spatial distribution, variation trend, and concentrations of nutrients in inflow rivers to identify discrepancies and sensitivities among different lake zone ecosystems.Next, the contribution of water quality parameters from river to lacustrine CyanoHABs in different seasons was quantified to determine the main pollutants that affect algal bloom growth at different stages.A modeling approach was employed to assess the changes and distribution of chlorophyll-a under different degrees of pollutant reduction in different lake zones.The results of this study can help evaluate the effectiveness of pollution reduction in the lake basin and provide a scientific basis for implementing effective and flexible management strategies for water resources control.

Site description
The Lake Taihu Basin is characterized by a complex and high-density system of river networks, with 172 surrounding rivers and tributaries draining into the lake (Xu et al 2010).Lake Taihu, a large and shallow lake, experiences severe disruptions in water supply and economy due to the occurrence of CyanoHABs, which have become increasingly frequent and spatially expanding as a result of nutrient inputs that have been increasing since the 1980s (Wu et al 2022).These algal blooms are primarily concentrated in the northwest and centre of the lake (Zhang et al 2016, Janssen et al 2017).
According to the Report of Water Regime released by Taihu Basin Authority of ministry of water resources (http://www.tba.gov.cn/), the surrounding areas of Lake Taihu are generally divided into five hydrological zones (figure 1).Despite being located in the Wuchengxiyu Region, the Wangting hydro-junction project, which diverts water from the Yangtze River to the Taihu Basin through the Wangyu River (Wang et al 2019), was listed separately in this study due to its unique function.Lake Taihu is generally divided into nine lake zones, which are further classified into algae-dominated lake regions and macrophyte-dominated lake regions (Wu et al 2021).The algae-dominated lake regions mainly include Meiliang Bay, Zhushan Bay, Gonghu Bay, the western coastal area, Gonghu, and the central area (figure 1).

Data sources
Currently, Jiangsu Province has established a relatively comprehensive network of water quantity and quality monitoring stations, which includes national basic hydrological stations and patrol stations for the rivers surrounding Lake Taihu.The water quantity data were provided by Jiangsu Province Hydrology and Water Resources Investigation Bureau.The water quality data were provided by Suzhou Branch, Wuxi Branch, and Changzhou Branch of Jiangsu Provincial Survey Bureau of Hydrology and Water Resources.Water level data were collected from the Hydrologic Yearbook.The collected data have been compiled, verified, and deemed reliable, ensuring the accuracy of the analysis results.
Given the complexity and diversity of the river channels around Lake Taihu, smaller rivers were merged into adjacent main rivers.Additionally, certain gates that were strictly controlled due to the enhanced water environmental governance in the basin were not considered in this study.Consequently, this analysis and simulation focused on the connection between Lake Taihu and the 16 most significant rivers for inflow and outflow (figure 1).
The water inflow, outflow, and water quality data of Lake Taihu from 2008 to 2020 were based on routine monitoring data obtained from the lake and its surrounding rivers.The water quality data were analyzed using monthly average concentrations from monitoring sections (If the monitoring frequency was twice a month, the monthly average concentration was calculated as the average of the two measurements; If the frequency was once a month, the monitoring data served as the monthly average concentration).The monthly inflow and outflow discharge were derived by converting the daily flow from each survey section.Eight parameters, water temperature (WT), dissolved oxygen (DO), permanganate values (COD Mn ), pH, total phosphorus (TP), total nitrogen (TN), ammonia nitrogen (NH 3 -N), and chlorophyll-a (Chl.a), were adopted as indicators of lake water quality, with Chl. a specifically used as an indicator of CyanoHABs severity.
In the simulation, the topographic data of Lake Taihu (measurement scale: 1:250000) was obtained from the topographic survey conducted in 2009 (http://www.geodata.cn/index.html).The inflow and outflow of the lake, as well as atmospheric conditions and surface wind in 2018, were selected as dynamic boundary conditions.Hydrodynamic calibration data were derived from daily water level data and monthly temperature data recorded in 2018, while water quality calibration data were obtained from monthly water quality data in 2018.

Data analysis and evaluation
The TFPW-MK (Trend-free Pre-whitening MK) statistic value (Z) was used to analyze the variation trend of pollutant flux.To explore the relationship between river and lake water quality, as well as the relationship between inflow water quality and CyanoHABs at each corresponding lake site, Spearman correlation analysis was performed using SPSS Version 24.In this analysis, Chl. a was used as an indicator of CyanoHABs severity.Multiple stepwise linear regression was conducted with SPSS Version 24 to quantify the influence of seasons and joint riverine water quality factors on CyanoHABs at each specific nearby lake site.
The results of the hydrodynamic calibration were evaluated using the Nash-Sutcliffe efficiency coefficient (NSE).A higher NSE value closer to 1 indicated a greater consistency between the simulation results and reality, thus indicating a more reliable model.Generally speaking, the simulation result is adequate when 0.5 NSE 0.7, superior when 0.7 NSE 0.9 and outstanding when NSE 0.9.The calibration results of the water quality simulation were evaluated using the percent bias (PBIAS).This coefficient represents the average tendency of simulated values to be underestimated or overestimated relative to the actual observed values.Generally speaking, the simulation result is adequate when 40% PBIAS 70%, superior when 25% PBIAS 40% and outstanding when PBIAS 25%.NSE: where OBS i is the measured values, SIM i is the simulated values, and ̅ OBS is the average measured value.

Model introduction
Herein, the hydrodynamic and water quality modeling were conducted using the MIKE 3 model.This is a finite element model widely used in fluvial and lake modeling studies, which contains several modules, including the hydrodynamic model (HD), advection diffusion model (AD), ecological model (ECO Lab), and sand transport model (Weng et al 2020, Weng et al 2021).The HD module simulates changes in water level and flow resulting from different forces, providing a computational foundation for environmental hydrology.The ECO Lab module describes the components of the water environment ecology through a series of mathematical or differential equations, incorporating physical pollutant settling and suspension, chemical reactions, substance transformations, as well as biological and ecological processes (Butts et al 2012).Consequently, it can be integrated with the HD module to depict the interaction of multiple substances and the cyclic processes occurring within aquatic ecosystems (Khwairakpam et al 2021).
This study defined the calculation formulas for biochemical terms of DO, COD Mn , TP, TN, NH 3 -N, and Chl. a in the material transport and diffusion model based on the standard template provided by ECO Lab manual.For more details, please refer to the supplementary material.

Results and discussions
6.1.Multi-year variation analysis of pollutant flux in and out of Lake Taihu The monthly inflow and outflow water volumes were multiplied by the corresponding water quality monthly average concentrations to obtain the pollutant load within each surveyed section.By summing these loads, the annual inflow and outflow loads were determined and spatially aggregated to calculate the overall river inflow and outflow loads within the water resource zone.As a result, the variation trends are presented in table 1, while the annual average pollutant loads are displayed in figure 2. Overall, apart from a slight increase in total phosphorus load, the annual pollution load entering Lake Taihu exhibited a significant decreasing trend to varying degrees.Meanwhile, the outflow of annual pollution loads displayed an increasing trend to varying degrees.Consequently, it is reasonable to observe that the net inflow of various pollution loads experienced different degrees of decrease, supported by the negative TFPW-MK statistical values presented in table 1.These findings indicated the effectiveness of pollution control in Lake Taihu, particularly in relation to ammonia nitrogen and total nitrogen, which exhibited a positive response.
However, it is important to note that despite the improvement, a significant amount of pollutants still entered the lake each year, posing a serious threat to the water quality of Lake Taihu.When considering the division of water resources, it was observed that the pollutants carried by rivers flowing into the Western Region were generally higher compared to other regions, accounting for more than 70% of the total pollutants (figure 2).This observation was consistent with the higher inflow quantity in the Western Region (Li et al 2022).Furthermore, the interannual variability of pollutant fluxes in the Western Region was the lowest among all the lake regions, indicating the need for strengthened pollution control measures in this area.Additionally, three main bays in Lake Taihu, namely Zhushan Bay, Meiliang Bay, and Gonghu Bay, are wellknown for their severe eutrophication and occurrence of CyanoHABs, particularly during the summer-autumn seasons (Wu et al 2013, Wang et al 2021).Among the three bays, Zhushan Bay exhibited the highest pollutant fluxes, primarily due to inputs from the Western Region, followed by Gonghu Bay and Meiliang Bay.Hence, Lake Taihu is still experiencing a severe eutrophication phenomenon due to the persistently high nutrient input load, resulting in a high incidence of CyanoHABs.

Contribution of joint water quality of inflow rivers and lake on lacustrine CyanoHABs in Lake Taihu
Previous studies mainly focused on the effect of individual environmental factors on the growth of CyanoHABs, while CyanoHABs are typically not caused by a single environmental driver but rather by multiple factors that may co-occur (Heisler et al 2008, Paerl and Otten 2013, Baig et al 2016).The drivers also differ from seasons and regions.Besides, there is a concentration dilution effect by the water quantity.Therefore, based on multiple years of data, Spearman correlation analysis was conducted to reveal the contribution of monthly water quantity (WQ), WT, and riverine water quality factors, including DO, pH, COD Mn , TP, TN, and NH 3 -N in river to Chl. a in different regions (figure 3).Furthermore, multiple stepwise linear regression was conducted to identify the water quality factors significantly related to Chl. a at different seasons (table 2).
Among the inflow rivers, COD Mn was identified as the most frequent contributor to lacustrine CyanoHABs among inflow rivers (details refer to figure 3).Moreover, the results of multiple stepwise linear regression showed that Chl. a was significantly positively correlated with COD Mn (r = 0.658 ** ) in spring, followed by pH (r = 0.434 ** ) (table 2).It was observed that the concentration of COD Mn in rivers was generally higher than that in the lake during spring, primarily due to the increased non-point source pollution resulting from crop fertilization during spring cultivation.Therefore, the river's terrestrial input of COD Mn in spring had a significant impact on the concentration of COD Mn in the lake and the growth of algae, and this effect may become more pronounced in the context of global climate change.In fact, according to the health status report of Lake Taihu, the occurrence of CyanoHABs has gradually transitioned from seasonal outbreaks in summer and autumn to year-round blooms starting from March.This indicated that the high biomass of cyanobacteria in spring would lay the foundation for the summer blooms.
Furthermore, the correlation between COD Mn and pH with Chl. a in spring not only reflected the triggers of cyanobacterial outbreaks but also the outcomes.Over time, cyanobacteria would rapidly absorb carbon sources in the water for reproduction and growth, leading to a sustained high pH in the water, which enters a vicious cycle of "cyanobacterial outbreaks → reduction of carbon source → high pH → promoting the growth of cyanobacteria" (Qin et al 2013).This cycle provides a strong foundation for summer CyanoHABs.Moreover, the shallow water depth makes the increased pH value in the water body affect the bottom and exacerbate the release of phosphorus from sediments (Tao et al 2020).Table 2 shows that TP was a significant factor both in summer and autumn.Hence, in early summer, the vicious cycle may become "cyanobacterial outbreaks → high pH → the release of phosphorus in sediments → algal blooms aggravate".Consequently, controlling COD Mn in spring makes a lot of sense for algae control.

Model verification
The calculation area was divided into 7341 grids using the unstructured Flexible Mesh (FM) triangular grid, where the area of a single triangular grid was controlled within 500000 m 2 (figure 4).The simulation period was from 8:00 a.m. on January 1 to 8:00 a.m. on December 31, 2018, and the simulation was conducted in a twelvehour step, resulting in 729 simulations.Herein, the dry-wet boundary method was adopted to improve the computational efficiency of the model, and the horizontal eddy viscosity coefficient (AHO) was solved by the Smagorinsky sub-grid method.The wind field parameters, lake bottom roughness, and horizontal eddy viscosity coefficient were calibrated by the water level data throughout 2018 of two monitoring points (DPK and WTLJ Station) and temperature data of two monitoring points (THL9 and THL13), in which the parameters were 0.003, 0.024 and 0.8, respectively.Through model debugging and running, the measured and simulated values of water-level and temperature time series are presented in figure 5.It was verified that the NSE of simulated and measured water level was 0.888 (DPK Station) and 0.865 (WTLJ Station), while the PBIAS of simulated and measured temperature was 10.87% (THL9) and 7.78% (THL13), indicating that the simulation results were acceptable and could meet the requirements (figure 5).After completing the hydrodynamic model calibration, the Eco Lab model was calibrated.These parameters remained constant throughout the simulation period, with their variations in different seasons primarily controlled by hydrometeorological conditions such as temperature function and wind speed (see the formula in supplementary material for details).It is important to note that only the verification results of COD Mn and Chl. a were presented as they were the focus of the subsequent simulation.The calibration results are shown in table 3, and the model parameters are shown in table S1.
From the simulation results, it is evident that the calculated values can accurately reflect the spatial and temporal distribution characteristics of the elements in the lake area, both in terms of numerical magnitude and distribution trends.Furthermore, the PBIAS values of the four monitoring points were all less than 40%, indicating that the selected model and parameters were reasonable.Therefore, the established models can be used to predict the response of cyanobacteria to the reduction of pollutant loads entering the lake.

Response of CyanoHABs to pollutant reduction of inflow rivers
According to the statistics of the monthly CyanoHABs areas in Lake Taihu in 2018 using satellite remote sensing images (sourced from National Earth System Science Data Center, http://www.geodata.cn/index.html),the results revealed an extremely rapid expansion process of CyanoHABs in spring.Specifically, the average CyanoHABs area in early spring (March) was only 11 km 2 , which then increased to 107 km 2 in April and further reached 369 km 2 in May.
Therefore, to reduce the intensity and area of algal blooms more feasible and faster, combined with the results of 3.1 and 3.2, we believed that adequate control of the main nutritional factors affecting algae in spring may help inhibit the outbreak of CyanoHABs in late spring and early summer.Consequently, based on various existing research on pollution load reduction in Lake Taihu (Yan et al 2019b), the reduction gradients of COD Mn entering the lake were selected as 10%, 30%, and 50%, and the response of Chl. a concentration and distribution to COD Mn reduction in spring was simulated.Four monitoring sites located in three bays and the western coastal zone were selected, where the CyanoHABs were most likely to occur, and the simulation results were shown in figure 6.
The simulation results demonstrated that the higher the reduction of COD Mn , the more significant the improvement in water quality, with notable variations in the water quality improvement effects across different regions of the lake.When COD Mn was reduced, the improvement in Chl. a concentration was more pronounced in Zhushan Bay and the west coastal zone.Taking a 50% reduction as an example, the annual average decrease in Chl. a concentration was 11.51% and 14.46%, respectively.In contrast, the adjacent Meiliang Bay, due to poorer hydrodynamic conditions, showed a decrease of 10.19%.Meanwhlie, the improvement capability in Gonghu Bay was also relatively moderate, primarily due to the background concentration of the rivers entering the lake.Additionally, the spatiotemporal heterogeneity of lake water quality also influenced the response of Chl.a, confirming the gradual decrease in pollutant concentrations from northwest to southeast in Lake Taihu.
Compared to the concentration of COD Mn in rivers during April and May, it was relatively lower in March.Therefore, despite the implementation of pollutant reduction measures throughout the entire spring, the reduction of pollutants in March had a relatively minor impact on Chl.a.In contrast, the significant decline in pollutants during April to May resulted in a pronounced inhibitory effect on algal growth.At this time, the decrease of Chl. a concentration in the lake area, particularly in the coastal area, increased with the reduction percentage of COD Mn .This decrease was evident not only in the reduction of peak values but also in a faster decline as the pollutant concentration decreased.These simulation results strongly indicate that the reduction of COD Mn in spring played a significant role in inhibiting algae growth.
The areas with Chl. a concentration exceeding 0.03 mg l −1 were identified as having a significant extent of algal blooms, and the extraction of bloom area was completed, as shown in figure 7. It is observed that in Gonghu Bay, although the reduction in Chl. a concentration may not be as noticeable, there was a significant decrease in the area of algal blooms.In contrast, despite the significant decline in Chl. a concentration in the other three regions, they remained the primary areas of algal bloom occurrence due to their high background concentrations.Despite the sensitivity of algae growth in the coastal area to the decrease in pollutant concentration, it also exhibited a strong response to the continuous increase in pollutant concentration entering the lake at the end of July and the beginning of August.For instance, in Zhushan Bay, the average COD Mn concentration was 4.15 mg l −1 in June, 5.5 mg l −1 in July, and 8.4 mg l −1 in August.This increase in COD Mn concentration played a role in promoting algae growth, leading to a relatively high increase in algae within a short period of time.Furthermore, if the temperature was moderate and the wind speed was low during this time, there was a risk of algal blooms.Even so, the concentration of Chl. a in the lake during this time would still be lower than before the reduction measures were implemented.

Conclusion
For Lake Taihu, prioritizing the control of external loading remains crucial in the overall strategy to address the disparity between the current water quality in the inflow rivers and the lake.Analysis of multi-year data revealed that high CODMn concentrations in the rivers during spring promote algae growth and localized CyanoHABs in Lake Taihu every May.Therefore, a valuable strategy focusing on reducing summer cyanobacterial outbreaks by controlling the elevated CODMn concentration in spring has been proposed.This strategy's effectiveness was subsequently verified through the established mathematical model of water environment.The modeling results demonstrated significant spatial and temporal variations in the response of Chl. a to the reduction of boundary loads.The reduction of pollutants showed significant effects on the water quality in the less-polluted Gonghu Bay, resulting in a remarkable decrease in algal bloom areas.However, for Zhushan Bay, the west coastal zone, and Meiliang Bay, although there was a noticeable decreasing trend in Chl. a concentration, they still remained the main areas of algal blooms due to their higher background concentrations.Additionally, the model captured that the rapid rise in inflow pollutant concentrations during autumn could increase the risk of CyanoHABs occurrence.In summary, the results of this study provided a new idea for exogenous load reduction, that is, the response characteristics of Chl. a to the boundary in different lake zones should be determined first, so as to formulate reasonable reduction schemes according to the concentration and variation trend of inflow pollutants.In this way, the cost of load control could be saved and the algal bloom outbreak can be more effectively dealt with.

Figure 5 .
Figure 5. Measured and simulated values of the time series of daily water level and temperature of monitoring station in 2018: (a) Water level of DPK Station; (b) Water level of WTLJ Station; (c) Temperature of THL9; (d) Temperature of THL13.

Figure 7 .
Figure 7. Algal bloom area and distribution map of Lake Taihu: (a) No pollutant reduction on 22 May; (b) Pollutant reduction on 22 May; (c) No pollutant reduction on 11 June; (d) Pollutant reduction on 11 June.

Table 1 .
The TFPW-MK statistical values of annual pollutant fluxes in and out of Lake Taihu from 2008 to 2020.

Table 2 .
Contribution of lacustrine joint water quality factors to CyanoHABs in different seasons.(In the table, B represents the regression coefficient, Std.Error represents the standard error, Beta represents the standardized regression coefficient, T represents the t-statistic, Sig.represents the significance level, and VIF represents the variance inflation factor).

Table 3 .
Simulation results and accuracy evaluation of water quality model (MMV-measured mean value, SMV-simulated mean value).