Listen

Affected by urban construction and industrial discharge, the diffuse source pollution of river water environment in urban sections is becoming more and more serious. The use of numerical models can accurately and efficiently simulate the process of diffuse source pollution and analyze its evolution. Based on MIKE21, the urban river water environment model of the Bai river is constructed. After parameter calibration, it is proved that the model has high accuracy. Based on the discharge of this river section in 2018, the river diffuse source pollution is simulated under three hydrological conditions, and the BP neural network is used to simulate the river diffuse source pollution. Network assessment of water quality in dry years. The results showed that the four water quality indicators showed the spatial distribution characteristics of upstream>urban section>downstream, and the river water quality in the year showed the temporal distribution characteristics of flood season better than a non-flood season. The mean value of the coefficient of certainty R2 reaches 0.94 when comparing the simulated and measured values, which indicates that the model has high applicability and satisfactory fit. If the water quality of the water year does not meet the standard, the diffuse source pollution discharge should be reduced from May to October. This study can provide reference and guidance for urban river diffuse source pollution control.

Listen

  • A two-dimensional hydrodynamic model was developed using MIKE software and validated by field test data to determine the two-dimensional model parameters, including flow rate, flow velocity, eddy viscosity coefficient, boundary conditions, roughness, etc.

Graphical Abstract

Graphical Abstract
Graphical Abstract
View largeDownload slide
Graphical Abstract
Graphical Abstract
View largeDownload slide
BP neural network, diffuse source pollution, numerical simulation, water environment
Listen

In recent years, urban river water pollution has been aggravated due to human activities, and diffuse source pollution is the main source of urban pollution discharge. Therefore, accurately analyzing the process of diffuse source pollution and putting forward improvement suggestions based on this is a major scientific problem that needs to be solved in the development and management of the water environment. The randomness of water flow and the complexity of pollutant diffusion make diffuse source pollution analysis difficult. Common water pollution research methods include the empirical formula method, physical model method, and numerical calculation method. The numerical calculation model of water environment has been widely used in the fields of water pollution control and water environment impact assessment due to its unique characteristics of high efficiency and high precision (Li et al. 2021a, 2021b; Shahady & Cleary 2021).

Domestic and foreign researchers have done a lot of research on urban river water environment simulation. Xue et al. (2021) used a coupled model to model the water quantity and water quality of a typical agricultural plain watershed in North China. Yang et al. (2021) applied the MIKE 21 model to set different working conditions and studied the impact of water diversion in typical urban artificial lakes on water quality. Taking Wuhan East Lake as the research area, Li et al. (2020) proposed a model framework that coupled the physics-based model and MIKE 21 and evaluated the change process of its water quality. Zhu et al. (2021) used MIKE11 to analyze the effect of water diversion from the Yangtze River on the improvement of water quality in the receiving area of the eastern route of the South-to-North Water Diversion. Lin & Li (2015) used the spatial data management function of GIS and the water environment simulation function of the distributed hydrological model to construct a water environment management system for the watershed and took the Jiulong River Basin as the experimental watershed to realize the water environment of diffuse source pollution discharge. Simulation, assessment of environmental effects of non-diffuse source pollution management measures, and release of water environment information. Based on the MIKE SHE, Li et al. (2021a, 2021b) established the response relationship of the Yanghe water environment to the drainage in the basin, simulated and predicted the temporal and spatial variation of pollutants in the mainstream of the Yanghe River, and compared and analyzed the water quality status of key sections under different water pollution control schemes, to provide technical support for the improvement of water environment in Zhangjiakou City. Zhang et al. (2020) artificially solved the problems of poor river water quality and water environment degradation, constructed a coupled model of watershed hydrology, hydrodynamics, and water quality, and further evaluated the improvement efficiency of river water quality under different control measures.

BP neural network has good durability, real-time performance, high self-learning, self-adaptation, and fault tolerance. The model also has the advantage of a simple structure, which is very convenient and fast in actual operation. A series of studies have been carried out at home and abroad in applying artificial neural networks (ANNs) to water quality evaluation. Sun et al. (2004) conducted a further discussion based on the traditional ANN method, based on the BP algorithm, and adjusted the ANN output structure to make it more adaptable. The improved ANN was applied to the evaluation and classification of groundwater quality and compared with the evaluation results of fuzzy comprehensive evaluation. Wang et al. (2020) used the fuzzy BP neural network method to evaluate the water quality of the Liaohe Estuary wetland in different periods and regions and combined the fuzzy mathematics comprehensive evaluation method to analyze the membership degree of the output results. An optimization algorithm combining particle swarm optimization (PSO), genetic algorithm (GA), and BP neural network is proposed by Yan et al. (2019) and applied to predict water quality in Beijing Haihu Lake and achieved the desired results. Chen et al. (2020) established a new type of BP neural network, the IABC-BP network, and found through practice that the model has more predictive capability and is approximately 25% more accurate than the traditional BP neural network.

Although scholars have achieved fruitful results in water environment simulation and water quality assessment, there is a lack of research in some key aspects. Researchers usually only consider water environment simulation unilaterally and ignore the evaluation and analysis of calculation results, or only evaluate the water quality of rivers without studying the changes in river water environment caused by changes in hydrological conditions. The innovation of this study lies in the targeted simulation and analysis of the water quality results of diffuse source pollution in urban rivers under different hydrological conditions, and at the same time, the water quality grade of the results is evaluated by combining machine learning methods, which can well provide new insights for future diffuse source pollution research. ideas. The structure of the research content of this paper is as follows: Taking the urban section of Bai river as the research object (1) Establish the hydrodynamic and water quality model of the Yangcheng section of the Bai river, and calibrate the model parameters. (2) Based on the relevant hydrology and water quality data, the ECO module in MIKE21 is selected for simulation calculation, and the biochemical process of each state variable is carried out through differential equations, and the numerical solution of the study area is carried out. (3) Use BP neural network to evaluate the water quality grade, analyze the water environment characteristics of Bai river Basin based on water quality evaluation, and give reasonable suggestions. The research results can provide a theoretical basis and reference for urban river water environment improvement and diffuse source pollution control.

Listen

The Mike21 model is a plane two-dimensional numerical simulation model based on hydrodynamic and water quality research. Compared with empirical formulas and physical models, this model greatly improves the calculation speed and the accuracy of the results. Compared with other numerical models, the Mike21 model is developed and evolved based on a lot of engineering experience, and it has better performance in the research of large-scale water flow fields and water quality changes and is widely used. The Mike21 water environment model mainly includes the hydrodynamic HD and water quality Ecolab simulation process, which can well simulate the diffuse source pollution process.

Hydrodynamic governing equations

Listen
The governing equations of the three-dimensional flow can be integrated along with the water depth and averaged along with the water depth, and the two-dimensional shallow water flow averaged along the water depth can be obtained. The equations are expressed as (Manache & Melching 2008; Yazdi et al. 2019):
where, ⁠, ξ is the free water level, m. t is time, d. x, y are spatial coordinates, m. p, q is the flow densities in the x and y directions, respectively, m2/s. d is the time-varying water depth, m. h is the water depth, m. g is the acceleration of gravity, m2/s, g=9.8 m2/s. C is the Chézy coefficient, m1/2/s. ρ is the density of water, kg/m3. τxx, τxy, τyy are the horizontal shear stress in the x-direction, the vertical shear stress in the x-direction, and the vertical shear stress in the y-direction, Pa. Ωq is the Coriol coefficient. f is the wind resistance coefficient. V, Vx, Vy is the wind speed and the wind speed components in the x and y directions, m/s. P is atmospheric pressure, Pa.

Water quality control equation

Listen
MIKE Eco Lab module is brand-new water quality and water ecological tool developed by the Danish Hydrodynamic Research Institute DHI on the basis of the traditional water quality module concept. The establishment of the hydrodynamic module is the basis of water quality simulation (Freni et al. 2011). The transport equation in a body of water is as follows:
where, Cs—scalar variable concentration. Dh—horizontal Diffusion Coefficient. t—time. kp—scalar solution coefficients.
The water quality model calculation formula is as follows:
where, c—concentration of Ecolab state variable. u,v,w—The flow velocity component of the convection term. Dx,Dy,Dz—The dispersion coefficient of the diffusion term. Sc—Source and sink items. Pc—Biochemical reactions of Ecolab.

Model step

Listen
The calculation steps of Mike21 water environment model are divided into two parts: hydrodynamic (HD) model and water quality (Ecolab) model. The water environment (Qiang et al. 2020; Yang et al. 2021) simulation process is shown in Figure 1.
Figure 1
MIKE21 water environment simulation process.
View largeDownload slide

MIKE21 water environment simulation process.

Figure 1
MIKE21 water environment simulation process.
View largeDownload slide

MIKE21 water environment simulation process.

Close modal

Model evaluation

Listen
In this study, a test method to verify the fitting degree of the hydrological model was introduced to evaluate the certainty coefficient R2 of the accuracy of the model. The calculation formula is as follows (Ahn et al. 2019):
where Pm represents an actual measured value. Ps represents the model calculated value. n represents the number of data, and m represents the number of monitoring points.

Result evaluation

Listen

The evaluation standard adopted in this study is China Surface Water Environmental Quality Standard (GB3838-2002), see Table 1 for details.

Table 1

Surface water environmental quality standard (GB3838-2002) unit: mg/L, pH dimensionless

serial numberstate variablesIIIIIIIVV
1 DO≥ 7.5 6 5 3 2 
2 NH3-N≤ 0.15 0.5 1 1.5 2 
3 TP≤ 0.02 0.1 0.2 0.3 0.4 
4 COD 15 15 20 30 40 
serial numberstate variablesIIIIIIIVV
1 DO≥ 7.5 6 5 3 2 
2 NH3-N≤ 0.15 0.5 1 1.5 2 
3 TP≤ 0.02 0.1 0.2 0.3 0.4 
4 COD 15 15 20 30 40 
View Large

In the practical application of artificial neural networks, the BP network is widely used in function approximation, pattern recognition clustering, data compression, regression and fitting, optimization calculation, etc. It can approach any nonlinear characteristics and has strong adaptive and self-learning processing capabilities. The topology of the BP network is shown in Figure 2.
Figure 2
BP network topology diagram.
View largeDownload slide

BP network topology diagram.

Figure 2
BP network topology diagram.
View largeDownload slide

BP network topology diagram.

Close modal
BP network learning rules: The correction of network weights and thresholds should be carried out in the direction of the negative gradient with the fastest decline of the reaction function.
where xk represents the current weight value and threshold matrix; gk represents the gradient of the current function and represents the learning rate. The calculation formula of the 3-layer BP neural network model is as follows:
Node output of hidden layer:
Node output of hidden layer:
Error at output node:

Error reverse transfer: Use the gradient descent method to correct the weight values of all layers. The learning algorithm of the weight values will be introduced in detail below.

Listen

Study area and data source

Listen

Study area

Listen

Bai River is an important mainstream of the Yangtze River in China (Aviation et al. 2021; Kumar & Hong 2021). Bai River has a total length of 264 kilometers and a drainage area of 12,270 square kilometers. It is located in Nanyang City, Henan Province, central China. It has a high strategic position in the optimal allocation of water resources, flood control, and waterlogging prevention in the Yangtze River Basin. This research area selects the Nanyang urban section of Bai River as the research object, with a length of 46.3 km. The proportion of water functional areas with water quality above Class II in the Bai River Basin is 27.8%. The proportion of III water quality is 38.9%, accounting for the largest proportion. The proportion of IV water quality is 22.2%, and the proportion of V water quality is 11.1%. As the main urban landscape water body in Nanyang City, with the urbanization process and rapid social and economic development of the Bai River Basin, the deterioration of water quality will greatly affect the urban landscape function and the living environment of urban residents (Zuo et al. 2020). Therefore, it is of great significance to study the water pollution characteristics of urban rivers, establish a mathematical model of the water environment in the urban section of Bai River, and conduct an overall analysis of the characteristics of the water environment in the study area.

Due to the long river in the urban section of Bai River, the water level data of a single station cannot reflect the simulation situation. After analysis and data screening, three hydrological stations on the upper, middle, and lower reaches of Bai River, Duck river mouth, Nanyang penyao, and Wadian were selected as the model validation site, hydrological data for the whole year of 2018 is used as model validation data. There are 3 monitoring points, A1, A2, A3, and 2 sewage outlets (Sewerage outlets 1 and 2 are the industrial sources of Nanyang City) in the research section. The schematic diagram of the research area is shown in Figure 3.
Figure 3
Schematic diagram of the study area and hydrological station.
View largeDownload slide

Schematic diagram of the study area and hydrological station.

Figure 3
Schematic diagram of the study area and hydrological station.
View largeDownload slide

Schematic diagram of the study area and hydrological station.

Close modal

Data sources

Listen

The data of hydrological stations and corresponding water quality monitoring sections selected in this study are from China Hydrological Information Network (http://xxfb.mwr.cn/sq_dxsk.html). The locations of hydrological stations and water quality sections are shown in Figure 3. The elevation data of the study area were obtained from the China National Satellite Data Network (https://www.gscloud.cn).

Model building

Listen

Project settings

Listen

In this study, three calculation conditions were set for the water quality numerical simulation (Table 2), with the Duck river mouth section as the upstream boundary of the model and Wadian as the downstream boundary of the model.

Table 2

Calculation conditions

Working condition settingTypeTraffic at guaranteed rate m3/sInlet concentration
Working condition 1 P = 10% guaranteed Rate (high flow year) 59.24 sequentially 
Working condition 2 P = 50% Guaranteed Rate (average water year) 23.38 
Working condition 3 P = 90% Guaranteed Rate (dry year) 14.12 
Working condition settingTypeTraffic at guaranteed rate m3/sInlet concentration
Working condition 1 P = 10% guaranteed Rate (high flow year) 59.24 sequentially 
Working condition 2 P = 50% Guaranteed Rate (average water year) 23.38 
Working condition 3 P = 90% Guaranteed Rate (dry year) 14.12 
View Large

Mesh division

Listen
Mesh quality is one of the important parameters that determine model quality. The triangular unstructured meshing method is selected for the simulation. The grid generates a total of 16,354 nodes and 33,235 grids. It is generally believed that the mesh quality is better if the minimum angle of the unstructured mesh is not less than 30°. The mesh after processing is shown in Figure 4.
Figure 4
Mesh division.
View largeDownload slide

Mesh division.

Parameter setting

Listen

The simulation selects high flow year, average water year, and dry year as the simulation time, the total number of time steps is 730 steps, and the step size is set to 86,400 s. Through debugging, the dry water depth of the model is hdry = 0.006 m, the submerged water depth is hflood = 0.058 m, and the wet water depth is hwet = 0.13 m. The state variables of the water quality model are selected as dissolved oxygen (DO), biochemical oxygen demand (BOD), and ammonia nitrogen (NH3-N). The parameter settings of the model after calibration are shown in Table 3 below.

Table 3

Calibration results of water quality parameters

ParameterValueUnit
precipitation evaporation Measured sequence value mm 
wind field Measured sequence value m/s 
Coriolis force 7.92 × 10−5 s−1 
Eddy viscosity 0.3  
salinity 0  
water temperature 17 °C 
Diffusion coefficient 0.35  
COD degradation rate 0.25 /day 
BOD degradation rate 0.32 /day 
NH3-N nitrification rate 0.12 /day 
DO oxygen utilizaton rate 0.04 /day 
ParameterValueUnit
precipitation evaporation Measured sequence value mm 
wind field Measured sequence value m/s 
Coriolis force 7.92 × 10−5 s−1 
Eddy viscosity 0.3  
salinity 0  
water temperature 17 °C 
Diffusion coefficient 0.35  
COD degradation rate 0.25 /day 
BOD degradation rate 0.32 /day 
NH3-N nitrification rate 0.12 /day 
DO oxygen utilizaton rate 0.04 /day 
View Large

Model validation

Listen

Hydrodynamic model validation

Listen
The water level data of three hydrological stations were selected for water level verification (Figure 5). The overall trend of the simulated water level is basically consistent with the measured water level. However, the difference between the simulated water level and the measured data in the early stage is large, because the early stage of the simulation relies heavily on the initial water level conditions. As the simulation progresses, the simulated results gradually approach the measured data.
Figure 5
Water level verification diagram.
View largeDownload slide

Water level verification diagram.

Figure 5
Water level verification diagram.
View largeDownload slide

Water level verification diagram.

Close modal

It is generally considered that after the accuracy reaches 0.8, the model accuracy is acceptable. It can be seen from the table that R2 is greater than 0.80, indicating that the calculated value of the model can guarantee the accuracy of the state variable. The R2 for each model is shown in Table 4.

Water quality model validation

Listen
The actual monitoring water quality index values of the Nanyang penyao monitoring section were selected for verification (Figure 6). It can be seen from the Table 5 that the mean value of R2 reaches 0.94, indicating that the calculated value of the model can guarantee the accuracy of the prediction of state variables.
Figure 6
Water level and water quality verification chart.
View largeDownload slide

Water level and water quality verification chart.

Figure 6
Water level and water quality verification chart.
View largeDownload slide

Water level and water quality verification chart.

Close modal
Listen

Using the Ecolab water quality model that has been calibrated, the spatial and temporal distribution characteristics of each water quality index in a typical hydrological year are respectively explored. And set three monitoring points A1, A2, A3 in the model (Figure 3).

Result analysis of working condition I

Listen
Condition one in this study is the migration of each state variable under high flow year. There are multiple pollution source outbreak points in the region. The major impact on the river is the pollution from the industrial sources in Nanyang City, which caused obvious pollution in the river area after the confluence of the pollution sources. From Figure 7, it can be seen that the COD concentration of the section from Duck river mouth to Nanyang penyao is between 15 and 20 mg/L. The COD concentration between Nanyang penyao and the Wadian section is between 22 and 33.5 mg/L, and the COD concentration below the Wadian section is between 22 and 33.5 mg/L. The concentration is below 19.5 mg/L, and the overall COD concentration range is the standard of Class IV water. The NH3-N concentration between the Duck river mouth and the Nanyang penyao section is between 0.45 and 0.95 mg/L, meeting the Class III water standard. Under the condition of high flow year, the self-purification effect of ammonia nitrogen in the water body is more obvious, and the overall ammonia nitrogen concentration is between the standards of class III water and class IV water. The overall DO concentration in the study area is basically between 5.4 and 9 mg/L. This is because the DO concentration in the upstream inflow water is good, and the hydrodynamic conditions of the river in the high flow year are also good, so the reoxygenation rate also has a certain level. The concentration is maintained at about the standard of Class I water. The overall total phosphorus concentration in the water body is relatively low, and the total phosphorus concentration in the study area can reach the Class II water standard under the overall high flow year condition.
Figure 7
Time average distribution cloud chart of each state variable in wet year.
View largeDownload slide

Time average distribution cloud chart of each state variable in wet year.

Figure 7
Time average distribution cloud chart of each state variable in wet year.
View largeDownload slide

Time average distribution cloud chart of each state variable in wet year.

Close modal
There is a sewage treatment plant in the urban section of Bai River, so most of the concentration values of monitoring point A2 meet the Class IV water standard (Figure 8), and the concentration of COD is higher than other indicators. The concentration values of the state variables at the monitoring points A1 and A3 are mostly in the class III water standard during the wet season from May to August. This is because the water quality standards in the upper reaches of the Bai River are all above class II water quality, and only the COD concentration is higher than class III. The concentration values of other state variables were all good. The DO concentration in the wet season (July-September) was slightly higher than that in the dry season but above the Class II water standard. In December, the NH3-N concentration showed an upward trend, but the self-purification capacity of the river significantly reduced the overall pollution, and the overall water quality was above Class III water.
Figure 8
Change process of state variable concentration at each monitoring point in the wet year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in the wet year.

Figure 8
Change process of state variable concentration at each monitoring point in the wet year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in the wet year.

Close modal

Result analysis of working condition II

Listen
The second working condition in this study is the migration of each state variable in a flatwater year. It can be seen from Figure 9 that the water quality in downstream near the sewage outlet is worse than in other locations, and there are multiple pollution source outbreak points in the whole area. The COD concentration from Duck river mouth to Nanyang penyao section is between 16 and 22.5 mg/L, the COD concentration between Nanyang penyao and Wadian section is 24–34.5 mg/L, and the COD concentration below Wadian section is 13.5–21.5 mg/L. The COD concentration range of the study area in the overall level water year is basically within the V-class water standard. The NH3-N concentration between the Duck river mouth and the Nanyang penyao section was between 0.60 and 1.0 mg/L. The concentrations of NH3-N from the Nanyang penyao section to the Wadian section ranged from 0.85 to 1.40 mg/L. Compared with working condition 1, the flow velocity of the water body in the river is reduced, the self-purification effect of ammonia nitrogen in the water body is reduced, and the overall area of NH3-N concentration is at the level of IV water. The overall DO concentration in the study area is basically between 5.0 and 8.5 mg/L, which is lower than that of working condition 1, and the dissolved oxygen content in the water body is lower than that of working condition 1, but it is still basically maintained at about the standard of class I water. The content of total phosphorus is still low, and the concentration can still reach the standard of Class III water, which reflects that the content of total phosphorus in diffuse source and non-diffuse source pollutants is not high.
Figure 9
Time average distribution cloud chart of each state variable in normal flow year.
View largeDownload slide

Time average distribution cloud chart of each state variable in normal flow year.

Figure 9
Time average distribution cloud chart of each state variable in normal flow year.
View largeDownload slide

Time average distribution cloud chart of each state variable in normal flow year.

Close modal
It can be seen from Figure 10 that the concentration of COD at monitoring point A1 is an average low value, and the COD value at monitoring point A2 is relatively large, with an average of about 33 mg/L. Compared with working condition 1, the COD concentration has increased significantly, especially In March-April and June-August, the COD concentration in the urban section was relatively high, and it was basically near the Class V water. The change law of DO concentration in the water body is not very obvious, but basically, the DO content in the water body is the highest from May to September, and the DO concentration in December and January is the lowest. Pressure, temperature, and salt related. From the change of total phosphorus concentration, it can be seen that the total phosphorus concentration shows an upward trend as a whole, and the total phosphorus concentration value of the monitoring points is basically in the class IV water standard. The change law of NH3-N concentration showed a trend of first decline and then a small increase. The content of NH3-N in the water body was obviously low from March to September, which was the lowest period of the year. The NH3-N in the water body mainly came from industrial wastewater. The changes in concentration also fluctuated around the standard of Class IV water.
Figure 10
Change process of state variable concentration at each monitoring point in normal flow year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in normal flow year.

Figure 10
Change process of state variable concentration at each monitoring point in normal flow year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in normal flow year.

Close modal

Result analysis of working condition III

Listen
The third working condition in this study is the migration of each state variable in a dry year (90% guarantee rate). In dry years, the amount of upstream inflow is small, and it is a period of relatively less precipitation, which is relatively unfavorable to the diffusion of pollutants. From Figure 11, it can be seen that the COD concentration of the section from Duck river mouth to Nanyang penyao is between 16.5 and 25 mg/L, the COD concentration between Nanyang penyao and the Wadian section is between 30 and 35 mg/L, and the COD concentration below the Wadian section is between 30 and 35 mg/L. Above 31.5 mg/L, the COD concentration exceeds the standard limit of Class IV water. The main reason is that the amount of incoming water from the upstream is small in dry years, and the reduction of the flow rate will directly weaken the water body's carrying effect on pollutants. The overall range of COD concentration in the region is basically in the V water standard. The concentration of DO in the water is basically between 4.5 and 8.5 mg/L, which meets the standard of Class III water. The total phosphorus concentration in the section from Duck river mouth to Nanyang penyao is below 0.06 mg/L, and the total phosphorus concentration in the section from Nanyang penyao to Wadian is between 0.07 and 0.1 mg/L. The total phosphorus concentration continued to increase below the Wadian section, and the total phosphorus concentration in the whole area was above the class III standard.
Figure 11
Time average distribution cloud chart of each state variable in a dry year.
View largeDownload slide

Time average distribution cloud chart of each state variable in a dry year.

Figure 11
Time average distribution cloud chart of each state variable in a dry year.
View largeDownload slide

Time average distribution cloud chart of each state variable in a dry year.

Close modal
Figure 12 shows the time-varying process of the state variable concentration at each monitoring point in the model modeling area under dry year conditions. As can be seen from the Fig, it can be seen from the Fig that the concentration of COD is an average low value at monitoring point A1, and the COD value at monitoring point A2 is relatively large, with an average of about 35 mg/L. The concentration value of COD in the urban section of Bai river is higher than in other locations, it can be seen that the degradation effect of COD is significantly reduced in the dry season, and the COD concentration has increased significantly compared with other working conditions. From the changes in the concentration of total phosphorus can be seen, the overall trend of total phosphorus concentration is increasing, and the concentration values of water quality indicators at each monitoring point are above the limit value of class V water. The concentration value of water quality indicators at each monitoring point should be higher than the limit value of Class V water.
Table 4

Water level verification

Hydrological stationsDuck River mouthNanyang penyaoWadian
R2 0.89 0.92 0.91 
Hydrological stationsDuck River mouthNanyang penyaoWadian
R2 0.89 0.92 0.91 
View Large
Table 5

Water quality verification

State variablesCODDONH3-NTP
R2 0.90 0.95 0.96 0.94 
State variablesCODDONH3-NTP
R2 0.90 0.95 0.96 0.94 
View Large
Figure 12
Change process of state variable concentration at each monitoring point in dry year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in dry year.

Figure 12
Change process of state variable concentration at each monitoring point in dry year.
View largeDownload slide

Change process of state variable concentration at each monitoring point in dry year.

Close modal

BP network water quality evaluation

Listen

Based on the surface water environmental quality standard (GB3838-2002) Table 6, the rand function is used to generate the input layer sample data, and the upper limit of DO is set as 14.64 mg/L. Convert the decimal 1–6 into binary as a standard output sample: 001 means class I water, 010 means class II water, 011 means class III water, 100 means class IV water, 101 means class V water, 110 means water lower than class V. The tang excitation function is used between the input layer and the hidden layer, (Wen & Yuan 2020) and the pure excitation function is used between the hidden layer and the output layer. Because the data is too small to meet the training requirements, linear interpolation is used to expand the training samples to 5*170. In the sample data, 5*120 are selected as training data, and 5*50 are selected as test data.

Table 6

Basic project standard limit of surface water environmental quality standard (mg/L)

CODNH3-NDOTPWater quality grade
2 0.15 7.5 0.01 I 
4 0.5 6 0.025 II 
6 1 5 0.05 III 
10 1.5 3 0.1 IV 
15 2 2 0.2 V 
CODNH3-NDOTPWater quality grade
2 0.15 7.5 0.01 I 
4 0.5 6 0.025 II 
6 1 5 0.05 III 
10 1.5 3 0.1 IV 
15 2 2 0.2 V 
View Large

The 50 samples used for testing in this study include Class I water quality, Class II water quality, Class III water quality, Class IV water quality, and Class V water quality. The evaluation result of each sample is accurate, and the error is 1.54*10−2, which proves that the 120 samples used for learning and training can meet the learning accuracy requirements. The predicted versus expected values and the errors are shown in Figure 13.
Figure 13
Test data water quality evaluation effect.
View largeDownload slide

Test data water quality evaluation effect.

Figure 13
Test data water quality evaluation effect.
View largeDownload slide

Test data water quality evaluation effect.

Close modal
Figure 14 shows the change curve of the mean square error with the number of training steps during the model training process. It can be observed from the Fig that after 120 times of domestication, the mean square error reaches 10−6, and the network stops training.
Figure 14
The mean square error plot of the training set samples.
View largeDownload slide

The mean square error plot of the training set samples.

Figure 14
The mean square error plot of the training set samples.
View largeDownload slide

The mean square error plot of the training set samples.

Close modal
Figure 15 shows the results of the water quality evaluation. The results of the water quality in dry years are Class II water quality, Class III water quality and Class IV water quality. It is mainly concentrated in the water quality of class III and class IV. The water quality of the river is class II from January to February, the water quality of class III from March to April, the water quality of class IV from May to August, and the water quality of class V from September to October. Due to the large discharge volume of the river sewage outlet, the flow velocity of the water flow is small, and the sewage holding capacity of the river is weak. From November to December, the river channel becomes Class III water quality through self-purification. From the water quality simulation results, it can be seen that from May to December, the main factors affecting water quality in the river are COD and NH3-N. Since the overall flow rate of the water body is relatively slow in dry years, the reduction of the flow rate will directly weaken the water body's impact on pollution. In addition, the urban section is disturbed by the amount of upstream water, and the industrial diffuse source discharges a large amount of pollutants. The discharge of COD and NH3-N concentration is high, which will bring great challenges to the water environment carrying capacity of the urban section of Bai River. Water quality assessment plays an important role in water resource management and water pollution control. Through water quality assessment, water quality status and water quality development trends can be predicted. Based on the BP neural network model, this paper makes an early warning assessment of the water environment in the urban section of Bai River, which can provide relevant institutions with an efficient water environment management plan.
Figure 15
Analysis of water quality evaluation results of BP neural network in dry years.
View largeDownload slide

Analysis of water quality evaluation results of BP neural network in dry years.

Figure 15
Analysis of water quality evaluation results of BP neural network in dry years.
View largeDownload slide

Analysis of water quality evaluation results of BP neural network in dry years.

Close modal
Listen

In view of the complexity of water quality analysis, this study uses the high-precision MIKE21 hydrodynamic and water quality coupling model to simulate the water environment in the urban section of the Bai river Basin, and combines the BP neural network to evaluate the regional water quality, and explores the calculation results and self-purification capacity.

  • (1)

    In terms of spatial distribution, the water quality in the urban section of the modeling area is worse than that in the upper reaches of the river. The main reason is that there are industrial sewage outlets in the modeling area, and the confluence of pollution sources has caused obvious pollution in the river area. The four water quality indicators showed the spatial distribution characteristics of upstream > urban segment > downstream.

  • (2)

    In terms of temporal distribution, the water quality of rivers during the year showed that the temporal distribution characteristics of the flood season were better than that of the non-flood season. The main reason is that the upstream inflow water volume is small in non-flood seasons, and the reduction of the flow rate will directly weaken the carrying effect of the water body on pollutants. The failure of water body to degrade in time has a great impact on water quality.

  • (3)

    Due to the limitation of verification data collection and time, the factors considered in the water environment model are relatively single, and only the main regional excess factors such as COD, NH3-N, total phosphorus, and DO are considered, and other pollution factors such as heavy metals are not included in the model. Therefore, the later research can expand and improve the model through supplementary data, so that the water environment simulation model can be more comprehensive. In this paper, a two-dimensional hydrodynamic model of the urban section of Bai River is established. Because the study area is too narrow and long, the vertical scale of the water body is not considered in this paper. But in fact, there is a pivot reservoir area in the upper reaches of the river, so the hydrodynamic and water quality characteristics of the river and the reservoir area can be further studied at a three-dimensional scale.

Listen

All the authors contributed to the conception and design of the study. Writing and editing: Xianqi Zhang and Xilong Wu; Preliminary data collection: Tao Wang; Chart editing: Guoyu Zhu Haiyang Chen; All authors have read and approved the final manuscript.

Listen

This work was supported by the Key Scientific Research Project of Colleges and Universities in Henan Province (CN) [grant number 17A570004].

Listen

All relevant data are included in the paper or its Supplementary Information.

Listen

The authors declare there is no conflict.

Ahn
J.
,
Na
Y.
 &
Park
S. W.
2019
Development of two-dimensional inundation modelling process using MIKE21 model
.
KSCE Journal of Civil Engineering
23
(
9
),
2
.
9
.
Google Scholar
Crossref
Search ADS
 
Aviation
E.
,
Pradiko
H.
,
Wahyuni
S.
 &
Pamungkas
O. F.
2021
Analysis of modification sedimentation process at communal scale water treatment with electrocoagulation technique
.
IOP Conference Series: Earth and Environmental Science
802
(
1
),
012044
.
(8pp)
.
Google Scholar
 
Kumar
P. M.
 &
Hong
C. S.
2021
Internet of things for security surveillance for sewage wastewater treatment systems
.
Environmental Research
203
(
4
),
111899
.
Google Scholar
PubMed
 
Li
D.
,
Bu
S.
,
Li
Q.
,
Chen
S.
,
Zhen
Z.
 &
Fu
C.
2021a
Water environment capacity estimation and pollutant reduction of Yanghe Reservoir Basin in Hebei Province, China, via 0-D water quality model
.
Environmental Earth Sciences
80
(
16
),
1
–
13
.
Google Scholar
 
Li
Z. X.
,
Xin
S. K.
 &
Liu
R. F.
2021b
MIKE SHE model-based simulation of the water environment in the Yangtze River basin
.
People's Yellow River
23
,
1
–
6
.
Google Scholar
 
Lin
Q. Y.
 &
Li
Z. R.
2015
Design of a watershed water environment management system and its prototype implementation
.
Hydroelectric Energy Science
33
,
159
–
163
.
Google Scholar
 
Qiang
L.
,
Xiaoqun
X.
,
Jiangning
Z.
, Huang, W., Xu, X., Shou, L. & Chen, Q.
2020
Development of marine water quality criteria for inorganic mercury in China based on the retrievable toxicity data and a comparison with relevant criteria or guidelines
.
Ecotoxicology
28
(
4
),
412
–
421
.
Google Scholar
 
Shahady
T. D.
 &
Cleary
W. C.
2021
Influence of a low-head dam on water quality of an urban river system
.
Journal of Environmental Management
297
(
6500
),
113334
.
Google Scholar
PubMed
 
Sun
T.
,
Pan
S.
 &
Li Y
J.
2004
Application of artificial neural network model in groundwater quality evaluation and classification
.
Hydrogeology and Engineering Geology
31
(
3
),
58
–
61
.
Google Scholar
 
Wang
T. L.
,
Su
F. L.
,
Sun
D.
,
Sun
Y. M.
 &
Li
H. F.
2020
Water quality evaluation of Liaohe estuary wetlands based on fuzzy BP neural network
.
Journal of Northwest Forestry Academy
35
,
195
–
200
.
Google Scholar
 
Xue
B. L.
,
Zhang
H. W.
,
Wang
Y. T.
,
Tan
Z. X.
,
Zhu
Y.
 &
Shrestha
S.
2021
Modeling water quantity and quality for a typical agricultural plain basin of northern China by a coupled model
.
Science of The Total Environment
790
,
148139
.
0048-9697
.
Google Scholar
Crossref
Search ADS
PubMed
 
Yang
H. Y.
,
Wang
J. Q.
,
Li
J. H.
,
Zhou
H. L.
 &
Liu
Z. H.
2021
Modelling impacts of water diversion on water quality in an urban artificial lake
.
Environmental Pollution
276
,
116694
.
0269-7491
.
Google Scholar
Crossref
Search ADS
PubMed
 
Yazdi
M. N.
,
Ketabchy
M.
,
Sample
D. J.
,
Scott
D.
 &
Liao
H. H.
2019
An evaluation of HSPF and SWMM for simulating streamflow regimes in an urban watershed
.
Environmental Modelling and Software
118
,
211
–
225
.
Google Scholar
Crossref
Search ADS
 
Zhang
Y.
,
Meng
D. J.
,
Yu
Z. C.
,
Zhao
Z. Y.
,
Zhao
J. Y.
,
Peng
W. Q.
 &
Zhang
J.
2020
Analysis of urban river water quality improvement and compliance based on MIKE21
.
Hydroelectric Energy Science
38
(
09
),
48
–
52
.
Google Scholar
 
Zhu
Y. T.
,
Chen
Q. H.
,
Zhang
Y. T.
,
Li
Q. F.
 &
Chen
J.
2021
Simulation analysis of water quality improvement effect of Yangtze River diversion on the frontal receiving area of South-North Water Transfer
.
Hydrology
41
(
3
),
7
.
Google Scholar
 
Zuo
J.
,
Zhang
H. X.
,
Deng
S.
,
Gu
P.
 &
Lian
L. L.
2020
Analysis of the influence of mesh partition on mike21 calculation in flood impact assessment
.
IOP Conference Series: Earth and Environmental Science
612
(
1
),
4
.
Google Scholar
 
© 2022 The Authors
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).