Exploration of an urban lake management model to simulate chlorine interference based on the ecological relationships among aquatic species

In eutrophic lakes, algae are known to be sensitive to chlorine, but the impact of chlorine on the wider ecosystem has not been investigated. To quantitatively investigate the effects of chlorine on the urban lake ecosystem and analyze the changes in the aquatic ecosystem structure, a dynamic response model of aquatic species to chlorine was constructed based on the biomass density dynamics of aquatic species of submerged macrophytes, phytoplankton, zooplankton, periphyton, and benthos. The parameters were calibrated using data from the literature and two simulative experiments. The model was then validated using field data from an urban lake with a surface area of approximately 8000 m2 located in the downtown area of Guangzhou, South China. The correlation coefficient (R), root mean square error-observations standard deviation ratio (RSR) and index of agreement (IOA) were used to evaluate the accuracy and reliability of the model and the results were consistent with the observations (0.446 R < 0.985, RSR < 0.7, IOA > 0.6). Comparisons between the simulated and observed trends confirmed the feasibility of using this model to investigate the dynamics of aquatic species under chlorine interference. The model can help managers apply a modest amount of chlorine to control eutrophication and provides scientific support for the management of urban lakes.


Results
Calibration and validation. The calibrated parameters based on the results of the two experimental results or obtained from the literature are presented in Table 1. Model validation was performed using the monitored data from Lotus Lake.
The attenuation model of chlorine according to the results of experiment 1 is presented in Table 2. The R values ranged between 0.9130 and 0.9601, reflecting good agreement between the model and the observations. The results of the regression analysis is showed the linear relationship between C0 and K. A higher initial concentration of chlorine corresponded to slower attenuation. The root mean square error-observations standard deviation ratio (RSR) was 0.55, which indicated good performance. Table 3 shows the results of the model evaluation for each component. The simulated values showed good qualitative agreement with the measurements based on the R, RSR and IOA measures. The R values ranged between 0.446 and 0.985. Except for the simulated biomass densities of M. lopsleuckarti (Claus) and G. compressus (Hütton), the simulated values presented highly significant correlations with the measurements (p < 0.01). In addition, the RSR values for all of the components were less than 0.7 and were thus satisfactory. The values of IOA for all the components were greater than 0.6, reflecting consistency between the model and observations. Simulation of TP in water and sediment. Figure 1 shows the total phosphorus concentrations according to the modeled data and the field data for lake water and sediment.
The TP in water presented a single peak ( Fig. 1(a)) with a value of approximately 1.65 mg L −1 . Although differences were observed between the measurement and simulation values, the trends of TP concentration obtained from the model were generally in agreement with the observations. The increase in TP not only led to the outbreak of eutrophication but also resulted in an increase of TP in sediment ( Fig. 1(b)) due to settling during this period. Maxm 0.03 [e2] 0.007 [e2] 0.007 [e2] 0.007 [e2] 0.007 [e2] 0.005 [e2] 0.005 [e2] 0.005 [e2] 0.14 49 0.14 49 0.14 49 10 −4 49 Table 3. R, RSR, and IOA values indicating the agreement between the measured and simulated values. **Significant at p < 0.01, *significant at p < 0.05, 0 < RSR < 0.5 indicates very good performance, 0.5 < RSR < 0.6 indicates good performance, 0.6 < RSR < 0.7 indicates satisfactory performance, and 0.7 > RSR indicates unsatisfactory performance 61 . A single peak was found for the biomass density of V. natans (Lour.) Hara. The simulated values were analogous to the observed values, with corresponding peaks in May. Although the simulated maximum values were larger than the observed values, the relative error was less than 15% (12.3%). The simulated results were reasonable. However, eutrophication began in the lake on May 10 th , resulting in the decline of V. natans (Lour.) Hara biomass density. Both the observations and simulations followed decreasing trends after mid-June, when phytoplankton and periphyton appeared to be blooming (Figs 3 and 4). Subsequently, the biomass density of V. natans (Lour.) Hara continued to decrease, reaching a measured density of approximately 3000 g·m −2 in September.

Phytoplankton.
The main species of phytoplankton in the lake were M. aeruginosa, A. flos-aquae, E. gracilis, and M. granulata (Ehr.) Ralfs. The simulated and observed biomass densities of the four main species of phytoplankton are presented in Fig. 3.
Three peaks were found in the biomass density dynamics of M. aeruginosa, whereas only one peak was found for each of the remaining three species. The simulated and observed patterns of the densities of the four phytoplankton species presented similar trends. M. aeruginosa was clearly very sensitive to chlorine. Decreases in the biomass density of M. aeruginosa and E. gracilis were observed on May 8 th , 19 th , and 29 th , June 4 th , July 17 th and August 4 th and August 28 th based on both the observations and simulations due to the intense inhibitory effects of chlorine. In addition, both the simulated and observed values for A. flos-aquae and M. granulata (Ehr.) Ralfs. presented increasing trends after the application of chlorine on May 8 th , 19 th , and 29 th and June 4 th , reflecting a weak effect of chlorine on these species. The peaks in density for four species of phytoplankton presented distinct delays in the simulations relative to the observed values. Although the modeled values agreed well with the observations, the observed maximum values were smaller than the corresponding simulated values.
Periphyton. O. chlorine, U. tenerrima (Kütz.) Kütz, and S. acus were the dominant species of periphyton in the lake. Figure 4 shows the simulated and observed biomass density values for these three main species.
Unlike the biomass densities of phytoplankton, the biomass densities of periphyton displayed a single peak according to both observation and simulation. Although differences were observed between the observations and simulations, the trends and density values obtained from the model were generally consistent with the  observations. In addition, the simulated maximum biomass densities of each of the three species of periphyton agreed well with the observations; however, the observed duration of the O. chlorine bloom was shorter than the simulated duration, whereas the opposite pattern was observed for U. tenerrima (Kütz.) Kütz. Among the three species, U. tenerrima (Kütz.) Kütz. had the strongest restoration ability and exhibited an increasing trend after August 21 st .
Zooplankton and Benthos. The main species of zooplankton in the lake were B. plicatilis, M. leuckarti (Claus) and D. brachyurum (Liévin). Chironomid larvae, P. canaliculata (Caenogastropoda and Ampullariidae) and G. compressus (Hütton) were the dominant species of benthos in the lake. The simulated and observed biomass density values of the three main species of each zooplankton and benthos are presented in Fig. 5.
Three peaks were observed in the biomass density of all three main species of zooplankton ( Fig. 5(a)), with similar values observed between the simulations and observations. A decrease in the biomass density of each of the three species occurred when chlorine was applied on May 8 th , 19 th , and 29 th , June 4 th , July 17 th , and August 4 th and August 28 th . The modeled peaks of B. plicatilis were obviously higher than the corresponding observed values, and this pattern was also observed in the other two zooplankton species, but the discrepancies were not obvious. Despite that, the relative error between the simulations and observations was relatively small. A single peak in the biomass density of each of the benthos species was observed ( Fig. 5(b)). Although differences were observed between the measurements and simulations, the trends and values obtained from the model were generally in agreement with the observations. The modeled peaks for the Chironomid larvae and G. compressus (Hütton) were much higher than the observed peaks.
Overall, this study explored the responses of the aquatic species to chlorine interference with two initial concentrations, i.e. 0.045 and 0.188 mg/L. The results showed that, within this concentration range, chlorine has a remarkable inhibitory effect on the dominant species (M. aeruginosa), thus control eutrophication in the urban lake.

Discussion
The model developed in this study coupled dynamic modules of biomass density of the main aquatic species with environmental factors. Table 4 lists the similar models from other studies [41][42][43] . Most of these models focused on the dynamic simulation of nutrients and phytoplankton or chlorophyll-a. Marchi developed a model to simulate the seasonal dynamics of major components of Lake Chozas, when an invasive fish was introduced into the lake. Compared with the previous studies, the model developed in this study focused on species instead of genera. Addressing only the dynamics of a category of aquatic organisms, obviously, ignores the ecological functions of certain important species. Although a simpler genera-specific model may be easier to use, calibrate and validate, a species-specific model based on the dynamics of primary aquatic species provides a tool for the researchers and managers who aim to explore the dynamic of species and to understand the mechanisms among aquatic species. Therefore, the constructed model in this study is essential to help the manager to complete and strengthen the urban lake management.
In addition, using chlorine is an effective method to manage the lake eutrophication, hence it is necessary to comprehensively understand the ecological dynamics of aquatic ecosystems under chlorine interference in urban lake management. This model, using the exponential function to couple the chlorine interference and accurately simulating the responses of the species to chlorine interference, can help managers apply a modest amount of chlorine to control eutrophication and provides scientific support for the management of urban lakes.
The dynamic model presented in this study explained the radical changes observed in the shallow urban lake. Both the simulated and observed results showed that the algae with cystic structure and zooplankton were sensitive to chlorine, whereas the algae with filamentous structure and submerged macrophytes showed strong resistance to chlorine. The increase in TP in the water between June 12 th and 22 nd clearly promoted the growth of phytoplankton and periphyton, which resulted in the breakout of eutrophication. The results confirmed that abundant TP nutrients provided the foundation for the phytoplankton outbreak. However, the TP additions were necessary but not sufficient 44 , as evidenced by the low TP concentrations between May 2 nd and June 12 th . However, the biomass of A. flos-aquae and M. granulata (Ehr.) Ralfs presented increasing trends in the measured experiment. The biomass densities of phytoplankton and periphyton species presented decreasing trends from July 12 th to August 10 th , resulting in a regime shift in the lake ecosystem from a turbid phase to a clean phase. During this period, the TP in water showed a decreasing trend (Fig. 1), thereby limiting the growth of phytoplankton and periphyton. Moreover, the intense interference of chlorine inhibited the growth of M. aeruginosa, which was an active species on July 17 th and August 4 th .
By comparing the trends and biomass density values of the main aquatic species under the chlorine interferences between the simulations and observations, we found that the modeled results were acceptable with certain discrepancies. Chlorine can decay in a few hours. The attenuation model of chlorine ran with a time interval of one hour. However, the daily amount of chlorine in water used as input in the model was calculated using the exponential equation to produce a daily average. Thus, the calculation method and the time interval of chlorine concentrations might have introduced errors into the simulations. In addition, the density peaks of the four species of phytoplankton presented distinct delays in the simulations relative to the observed patterns, which was likely because the simulations were run without considering the hydrodynamic effects. Studies have found that small-scale motion or low flow velocity were generally favorable for algal growth and aggregation in monoculture or mixed culture 45 . Despite these discrepancies between the simulated and observed values, good agreement was observed for both the values and trends, and the model satisfactorily described the dynamics of aquatic species under chlorine interference. The structure and parameter values of this model can provide a reference for the study of urban lake models in the near area.
The model accurately simulated the dynamics of aquatic species with the chlorine interference in an urban lake. Differences between the observed and simulated values were inevitable because some important components of the ecosystem were ignored in the model, including fishes and bacteria species and hydrodynamic influences 46,47 . In addition, because of the limitations of study time and location, the further validation of the model on different lakes and other seasons are recommended. Nonetheless, the model incorporated the main species in each class of aquatic organism that performed critical functions in the lake ecosystem and exhibited good performance. Therefore, this study provides strong support for the use of the model for urban lake eutrophication management. The application of constructed model should be based on the construction of new ecosystem. To extend this model to other similar urban lake ecosystems, the following steps are needed: (1) conducting ecological investigation of the lake, (2) identifying the dominant species, (3) calibrating the relevant parameters in the model.

Conclusions
By now, urban lakes management remains rough, especially for the problem of lake eutrophication. The associated management measures often lack scientific support and are thus not effective and sufficiently. Here, an urban lake management model was developed, and the model provides scientific support for the management of urban lakes.
For the management and control of eutrophication problem, a comprehensive understanding of the ecological dynamics of aquatic ecosystems must be developed. A simulative model of aquatic ecosystem dynamics could assist managers in understanding the processes of eutrophication. In this study, a dynamic model of a shallow urban lake was constructed based on the biomass density of the aquatic species subjected to frequent and strong chlorine interference. The test results showed that the simulation was highly accurate and consistent with field variations. Thus, this model is appropriate for addressing urban lake eutrophication.
The effect of chlorine on different aquatic species was studied and found to help regulate the balance of aquatic species, and the model simulation could reveal the direction of the balance of the aquatic ecosystem. The model developed in this study can provide guidance on the interactions between aquatic species growth and chlorine applying. The consistency between the field investigation and model outputs provides strong support for the use of the model for urban lake eutrophication management, including the application time and chlorine dosages based on aquatic organism dynamics.
The model was established via laboratory tests, experiments, field investigation and previous literature, and the statistical tests showed appropriate results. The reliability of the model also provides support for the management of urban lakes and shows that simple and effective chemical methods, such as the application of chlorine, should be considered. The structure and parameter values of this model can provide a reference for the study of urban lake models in the near area.
Further refinements of urban lake eutrophication management should be coupled with the use of bio-regulators.

Materials and Methods
Dynamic model of biomass density. A simulation model for the lake is developed by coupling the biomass density dynamic modules of the main aquatic organisms and environmental factors. The model is developed using Stella software (version9.1.4). The system dynamics of the model presents a strong modeling environment and a simple operation mode 48 . The model structure is presented in Fig. 6. Vallisneria natans (Lour.) Hara, phytoplankton, zooplankton, periphyton, benthos, detritus and the total phosphorus in water and sediment are considered the main components of the model. The phosphorus in water, as a linkage in the system, is directly related to the growth of submerged macrophytes, phytoplankton and periphyton and to the sediment of the lake. The food chain, competition, predation, respiration, and mortality are the main processes in the model. Chlorine is the only artificial intervention included in the system. Tables 5 and 6 show the different parameters and equations in the model.  Table 1.   Parameter calibration. Calibration data are obtained from the literature and two experiments in the laboratory. Data on the ingestion rate, the assimilation efficiencies of zooplankton and benthos, and the respiratory cost for grazing by zooplankton are obtained from the literature 51 . The inhibitory rate of chlorine on phytoplankton and zooplankton is obtained from the literature 52,53 . The half saturation constants for phosphorus uptake from water (mg m −3 day −1 ) and the half saturation constant for solar radiance are obtained from the literature 54 . The attenuation rate of chlorine and the inhibitory rate of chlorine to submerged macrophytes are obtained from the data from experiment 1. The daily maximum growth, mortality and respiration rates of aquatic species, and the attenuation coefficient caused by the self-population density, as well as the attenuation coefficients of phytoplankton and periphyton due to V. natans (Lour.) Hara, are obtained from data in experiment 2.

Experiment 1.
To explore the inhibitory rates of chlorine to filamentous algae and submerged macrophytes, a co-culture experiment is performed with Spirogyra.spp. and V. natans (Lour.) Hara at various initial concentrations of chlorine solution in the laboratory. The fresh weight of Spirogyra.spp. and V. natans (Lour.) Hara is 10 g and 15 g, respectively, and the organisms are cultured in a container (18.0 × 12.0 × 6.5 cm) containing 750 mL To understand the dependence between the initial concentration of chlorine solution (C 0 ) and attenuation rate (k), a regression analysis is used. In addition, the root mean square error-observations standard deviation ratio (RSR, equation 23) is used to evaluate the accuracy of the curve estimation in the regression analysis. Experiment 2. To explore the model parameters of primary producers and consumers, the effects of submerged macrophytes and daphnia magna on lake eutrophication are investigated in the laboratory. White polyethylene      average rainfall of approximately 1736 mm. The submerged macrophyte, V. natans (Lour.) Hara is planted in January 2015 and covers the whole lake bottom by March 2015.
The study period of the field investigation is from March 15 th to September 30 th , 2015. The surface shape of the lake is rectangular and divided into four parts by two intersecting diagonals. Samples of water, phytoplankton, zooplankton and periphyton are collected weekly, and sediment, submerged macrophyte and benthos samples are collected monthly at five points: one point at the center of the lakes and four other pointy, each in the center of one of the four parts. All samples are held at 4 °C in an insulating container and were taken back to the lab immediately.
Eutrophication in the lake appears repeatedly after May 10 th , 2015. Exogenous pollutants are input into the lake on June 10 th , resulting in an outbreak of eutrophication. To control eutrophication, 5000 g of calcium hypochlorite containing 30% chlorine (i.e. 0.188 mg/L chlorine), is applied to the lake on May 8 th , 19 th and 29 th ; June 4 th ; July 17 th ; and August 4 th ; and 1200 g of calcium hypochlorite containing 30% chlorine (i.e. 0.045 mg/L chlorine) is applied on August 28 th . The daily concentration of chlorine in the water is calculated using the exponential equation and then averaged over one hour. where Ci is the concentration of chlorine at a specific time; Co is the initial concentration of chlorine, which was calculated from the applied amount of calcium hypochlorite that was applied; k is the attenuation rate; t is time (h); and Ci is the daily average concentration of chlorine. The R, RSR and IOA values are calculated to assess the reliability and accuracy of the model. The RSR varies from 0 to a large positive value. A lower RSR indicates better simulation performance. The IOA is developed by Willmott 60 and measures the degree of model simulation error in a range from 0 to 1. When this measure is close to 1, the measurement present high consistency. The detailed calculation formulas for R, RSR and IOA are shown in Equations 23, 24 and 25. Data availability statement. The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.