The migration and accumulation of typical pollutants in the growing media layer of bioretention facilities

Abstract

A series of complex physical and chemical processes, such as interception, migration, accumulation, and transformation, can occur when pollutants in stormwater runoff enter the growing media layer of bioretention facilities, affecting the purification of stormwater runoff by bioretention facilities. The migration and accumulation of pollutants in the growing media layer need long-term monitoring in traditional experimental studies. In this study, we established the Hydrus-1D model of water and solution transport for the bioretention facilities. By analyzing the variation of cumulative fluxes of NO₃^--N and Pb with time and depth, we investigated pollutant migration and accumulation trends in the growing media layer of bioretention facilities. It can provide support for reducing runoff pollutants in bioretention facilities. The Hydrus-1D model was calibrated and verified with experimental data, and the input data (runoff pollutant concentration) for the pollutant concentration boundary was obtained from the SWMM model. The results demonstrated that the cumulative fluxes of NO₃^--N and Pb increased with the passage of simulation time and depth of the growing media layer overall. From the top to the bottom of the growing media layer, the change rates of the peak cumulative fluxes of NO₃^--N and Pb were strongly linked with their levels in the runoff. An increase in rainfall decreased the content of NO₃^--N and Pb in the growing media layer, and this phenomenon was more obvious in the lower part of the layer.

Graphical Abstract

Introduction

The global urbanization rate is expected to exceed 67.2% by 2050 (Yang et al. 2022). Unsurprisingly, the urban hydrological cycle will continue to face aggravating pressures from intensifying urbanization. Increases in impervious surface area have led to the doubling of surface runoff, the complexity of urban pollution sources has led to the exacerbation of nonpoint source pollution (Nargis et al. 2022), and issues related to urban water security and the water environment are becoming increasingly serious, posing a threat to urban infrastructure construction and social development (Li and Davis 2016). In recent years, many cities in China (such as Guangzhou, Beijing, Wuhan, and Zhengzhou) were unable to cope with the flood disasters brought by the heavy rainfall events, and about 60 − 80% of the water bodies in Shenzhen and many other towns in the Pearl River Delta were seriously polluted (Chan et al. 2018). More importantly, the polluted wastewater discharge will be used as a transmission medium in COVID-19 to further pollute the water environment and soil environment (Paital and Das 2022). To address these problems, stormwater management measures are being studied all over the world. As a low impact development (LID) facility, bioretention facilities can effectively reduce the peak flow of stormwater runoff and control the pollution in stormwater runoff. At present, a vast number of experimental studies have been undertaken by researchers on the effects of sewage interception and peak clipping (Yang and Chui 2018), the composition of the growing media layer (Rahman et al. 2020), the thickness of the filler (Vijayaraghavan et al. 2021), and the modification of the growing media (Jiang et al. 2018) in bioretention facilities. The bioretention technology applied to LID is gradually maturing.

In a growing number of studies, bioretention facilities have been shown to have a positive impact on emission reduction and pollution management (Goor et al. 2021; Jiang et al. 2017; Mai and Huang 2020). Shrestha et al. (2018) discovered that bioretention facilities may lower overall runoff and peak flow by 75% and 91%, respectively. Furthermore, bioretention facilities have shown a good removal impact on N, P, heavy metals, microplastics, and other contaminants in stormwater runoff (Smyth et al. 2021). For instance, the green bioretention facilities constructed by Caldelas et al. (2021) retained up to 73% Zn, 66% Cu, and > 99% Pb. However, several elements, particularly those in the growing media layer (Fan et al. 2019), influence the performance of bioretention facilities. On the one hand, the nature of the growing media layer itself has a certain influence on the movement of runoff pollutants after entering a facility. For example, the porosity between the particles in the growing media layer affects the infiltration of runoff and thus affects the treatment capacity of a bioretention facility in terms of pollutants (Skorobogatov et al. 2020); a growing media layer with a certain planting depth could remove more than 92% of heavy metal pollutants (Rycewicz-Borecki et al. 2016); the growing media layer of a single-layer medium has been modified into the growing media layer of a double-layer medium so that the removal rate of various forms of N increased to 76.8 − 98.3% (Luo et al. 2020). Compared with the traditional bioretention system, a biochar-pyrite bi-layer bioretention system has shown a high stability and efficiency for removing pollutants (Kong et al. 2021). On the other hand, as a medium related to the survival of plants and microorganisms, the growing media layer could control the treatment impact of bioretention facilities by regulating the development of plants and microorganisms (Jiang et al. 2019b).

The pollutants in stormwater runoff, after entering the growing media layer of bioretention facilities, take a long time to move through processes such as migration, transformation, and accumulation. Long-term on-site monitoring of operating bioretention facilities requires too much effort and is limited by objective factors such as equipment accuracy and objective environment. This has led to some difficulties in studying the long-term operation effects of bioretention facilities through experiments. However, choosing appropriate model software for numerical simulation can get accurate results based on limited measured data, which is rather convenient (Li et al. 2020). Thus, the experimental cycle can be effectively shortened, the objective variables can be controlled, and the migration and accumulation processes of pollutants can be quantitatively analyzed. However, the cumulative effects of pollutants and their ecological risks to soil and groundwater, as well as the simulation and optimization of structural parameters according to local conditions, are bottlenecks of the application for such facilities (Jiang et al. 2022). In recent years, for research on bioretention facilities based on numerical simulation, the RECARGA, DRAINMOD, SWMM, and Hydrus-1D models have usually been used for specific application designs (Lisenbee et al. 2021). Among them, the Hydrus-1D model was developed by the United States Saline Laboratory and other units and is used to simulate and analyze the water movement, solute transport process, and root water absorption of saturated and unsaturated porous media (Alam et al. 2022; Li et al. 2018). The model has many applications, such as studying the migration of water and salt in soil (Alam et al. 2022; Wang et al. 2018), the migration and transformation of nitrogen in soil (Pan et al. 2020), and the migration of heavy metal ions (Amiri and Nakhaei 2021). The Hydrus-1D model is mainly used for research on agricultural irrigation (Ventrella et al. 2019), pesticide leaching (Anlauf et al. 2018), soil moisture migration (Wang et al. 2018), and salinity leaching (Yang et al. 2019). Its advantage is that it can simulate the migration process of water and solutes in the soil environment under continuous rainfall or extreme stormwater. For permeable LID facilities, the Hydrus-1D model can simulate not only the effluent process of the facilities but also the vertical distribution of pollutants in the growing media layer. At present, research on this model in bioretention facilities is mainly aimed at the optimal design of the device under the key parameters (Li et al. 2018, 2021a) and improvements to the growing media layer (Jiang et al. 2019a; Li et al. 2021b). Pan et al. (2020) simulated the cumulative outflow volume and nitrate concentration of the subsurface infiltration system using the Hydrus-1D model and found that the fate of nitrate was directly controlled by the water temperature and hydraulic conditions. The Hydrus-1D model was used by Li et al. (2020) to simulate the regulatory effects of water quantity and water quality in bioretention facilities under different conditions; these results were compared with the experimental monitoring data, and then, optimization methods for the key parameters of the bioretention facilities were proposed by using the simulation results and response surface methodology. However, research on using the Hydrus-1D model to simulate the migration and accumulation of pollutants in the growing media layer of bioretention facilities is rare.

Therefore, based on our team’s previous experiments on a bioretention device (Gong et al. 2019), the Hydrus-1D model was used in this study to simulate the migration and transformation processes of typical pollutants in different growing media layers of bioretention facilities. On the premise of effectively saving time and controlling environmental variables, the migration and accumulation laws of pollutants in the growing media layer at both the temporal and spatial scales were revealed based on the model simulation results. They can provide support for the reduction of runoff pollutants in bioretention facilities, and provide basis and reference for the optimal design and construction of bioretention facilities. They can also further analyze the leaching potential in order to assist the ongoing operation and maintenance of the facilities as well as prepare for the evaluation of the bioretention facilities’ effectiveness after extensive use.

Materials and methods

Experimental devices and methods

Experiments and simulations were used in this study. Beginning in July 2018 and ending in November 2018, experimental studies were carried out. Actual monitoring data were collected for the calibration and validation of the subsequent model and to evaluate its applicability for the bioretention facilities. The bioretention devices were cuboids with a length of 800 mm, a width of 800 mm, and a height of 600 mm, as illustrated in Fig. 1. The devices’ structure was made up of the drainage layer, the filter layer, the growing media layer, the vegetation layer, and the aquifer. Based on the soil type of the growing media layer, the two devices were named SS-L and SSH-L, respectively. The growing media layer of SS-L was composed of the local fluvo-aquic soil (clay sand) and sandy soil with a volume ratio of 4:6, and SSH-L was composed of fluvo-aquic soil, sandy soil, and humus in a volume ratio of 3:6:1. The drainage layer was composed of gravel with a thickness of approximately 10 mm, and the filter layer consisted of geotextiles. The plants were Euonymus japonicus, Ophiopogon japonicus, Ligustrum vicaryi, Eleusine indica, Iris tectorum, Iris lactea, Berberis thunbergii, and Rosa davurica, which are all prevalent in Beijing.

Fig. 1.

Experimental devices for bioretention

In this study, NO₃^--N and Pb were chosen as the pollutants. On the one hand, the denitrification effect of traditional bioretention facilities is limited (Li et al. 2021c), and N removal efficiency fluctuates greatly (Dempster et al. 2012). Even though there is mutual conversion between different forms of N, organic nitrogen and ammonia are converted into nitrate by microorganisms in bioretention facilities (Jiang et al. 2018), and the form of NO₃^--N is the most stable. Therefore, NO₃^--N accounts for a large proportion of all existing forms of N, and this form is of great reference significance to the actual situation of studying the model simulation results. On the other hand, although the content of Pb in runoff is small, it is not similar to Cu and other heavy metals with double-sidedness; Pb has a small promoting effect on plants but a larger inhibiting effect. Pb is a hazardous heavy metal, and its long-term buildup in bioretention systems will impair microorganism breakdown and pollutant removal (Liu et al. 2020). Excessive Pb may impair plant development and agricultural productivity by reducing soil quality and fertility (Kabiri et al. 2019). Therefore, it is necessary to study the cumulative fluxes of Pb in the growing media layers of bioretention facilities.

During the experiment, according to the local climate characteristics and rainfall frequency in Beijing, the frequency of the experiments and the amount of water in a single experiment were determined. By simulating the concentration of pollutants in the rainwater runoff of typical roads in Beijing, the experimental influent water quality was obtained. A total of 16 times of water fed during the experiment (Gong et al. 2019). After each experiment, the effluent was sampled to measure the water quantity and water quality indices. Among the pollutants, NO₃^--N was measured by ultraviolet spectrophotometry (Lianhua Technology LH-3BA Multi-parameter Water Quality Analyzer, Lianhua Technology, Beijing, China), and Pb was measured by inductively coupled plasma-mass spectrometry (Agilent 7900 ICP-MS, Agilent Technologies, Santa Clara, CA, USA).

Model setup

For this investigation, the SWMM and Hydrus-1D models were chosen. Based on the effluent water quantity and pollutant concentration data of the bioretention experimental device, this simulation research first calibrated and validated the parameters of the bioretention Hydrus-1D mathematical model. Then, the migration and accumulation processes of pollutants in the growing media layer of bioretention facilities were simulated and analyzed using the Hydrus-1D model. The SWMM simulation provided the input pollutant concentration and water volume boundaries for the Hydrus-1D model.

Principle of the Hydrus-1D model

The Hydrus-1D model version 4.17 (Šimůnek et al. 2013) simulates the one-dimensional variably saturated water flow, heat movement, and solute transport in a partially saturated porous medium. The one-dimensional motion process of water in soil is described by Richards modified equation. The specific equations are expressed as follows:

$$\frac{\partial \theta }{\partial t}=\frac{\partial }{\partial x}\left[K,\left(\frac{\partial h}{\partial x},+,\cos ,\alpha \right)\right]-S$$ 1

$$K=K_s(x)K_r(h,x)$$ 2

where θ is the volumetric water content (cm³·cm⁻³); t is the time (s), x is the vertical coordinate; K is the unsaturated seepage coefficient function; h is the pressure head in the saturated region or matric potential in the unsaturated region (cm); α is the angle between the flow direction and X-axis; S is the source sink term (cm³·cm⁻³·s⁻¹); K_S is the saturated hydraulic conductivity (cm·day⁻¹); and K_r is the relative permeability coefficient (cm·day⁻¹).

In this study, the commonly used van Genuchten–Mualem (VG-M) model was selected as the soil moisture parameter model (Ghanbarian-Alavijeh and Hunt 2012) The specific equations are expressed as follows:

$$\theta \left(h\right)=\left\{\begin{array}{c}{\theta }_r+\frac{{\theta }_s-{\theta }_r}{{[1+|\alpha h|}^n{]}^m},h<0\\ {\theta }_s,h>0\end{array}\right)$$ 3

$$K\left(h\right)=K_s{S}_e^l{\left[1,-,\left(1,-,S_e^{1/m}\right)\right]}^2$$ 4

$$S_e=\frac{\theta -{\theta }_r}{{\theta }_s-{\theta }_r}$$ 5

where θ_s is the moisture content of the saturated region (cm³·cm⁻³); θ_r is the residual water content (cm³·cm⁻³); h is the pressure head in the saturated region or matric potential in the unsaturated region (cm); α is the reciprocal of inlet air suction (cm⁻¹); K_S is the saturated hydraulic conductivity (cm·day⁻¹); S_e is the available water capacity; $l$ is the pore connectivity parameter with a general value of 0.5; m is the parameter of the water characteristic curve; and n is the parameter of pore size distribution.

Regarding the solute transport process, the convection–dispersion equation is used in this study to describe the solute transport process and migration in saturated–unsaturated porous media. The model equation takes into account the processes of solute absorption and decomposition such as convection–dispersion reaction and nitrification–denitrification of pollutants in the soil (Li et al. 2020; Zhou et al. 2021).

Construction of the Hydrus-1D model

In this simulation, the Hydrus-1D model was established according to the growing media layer structure of the SS-L and SSH-L devices, and the model was calibrated and verified through experimental data. On this basis, the migration and accumulation processes of typical pollutants in a growing media layer were studied. The Hydrus-1D model built in this simulation study had two types of growing media layers, namely, sandy clay and sandy soil. The soil in the growing media layer was discretized at 1-cm equidistant distances.

According to the experimental period, the overall simulation duration was 134 days, the initial running time step was 0, the minimum running time step was 0.001 days, and the maximum running time step was 5 days. The upper boundary was the atmosphere because the upper boundary of the soil of the experimental device was in contact with the atmosphere. Since the influence of groundwater was not considered in this study model, the bottom boundary was free drainage. The initial volumetric water contents of six analysis points were 11.16%, 11.26%, 11.33%, 11.39%, 11.42%, and 11.43%, respectively, determined based on the preheating results in 2013.

Parameters of the Hydrus-1D model

The initial input parameters of this model mainly included hydraulic parameters and solute transport parameters. The hydraulic parameters included residual water content (θ_r), saturated water content (θ_s), reciprocal of intake suction (α), pore size distribution parameter (n), and saturated hydraulic conductivity (K_S); the solute transport parameters included bulk density (ρb), longitudinal dispersion (D_L), molecular diffusion coefficient in water (D_w) and air (D_a), and partition coefficient (K_d). The value range of each parameter is shown in Table 1. The hydraulic parameters of soil were determined by the type of local soil, and the solute transport parameters were determined by the nature of solute. In the specific modeling process, the soil hydraulic parameters were predicted using the pedotransfer function ROSETTA based on the dry bulk density, sand content, silt content, and clay content of the soil (Dou et al. 2022). Then, the specific values of each parameter can be determined through parameter sensitivity analysis, calibration, and validation process.

Table 1.

The value range of parameters

Parameter name	Unit	Dimension	Parameter name	Unit	Dimension
θr	cm3/cm3	0 − 0.1	KS2	cm/min	0.1 − 15
θs	cm3/cm3	0.1 − 0.5	ρb	g/cm3	0.9 − 1.5
α	1/cm3	0.02 − 0.04	DL	cm	1 − 30
n	——	1.10 − 2.90	Dw	cm2/d	4 − 20
KS1	cm/min	0.1 − 1.0	Kd	cm3/g	0 − 8

In terms of hydraulic parameters, there were 12 soil-related hydraulic parameters in the model. Generally, the residual water content of soil should be less than the saturated water content. The setting range of the reciprocal of intake suction and the pore size distribution parameter were specified by the model, and the relevant literature also provides reference values for the permeability coefficients of different media types (MEEPRC 2016). Given that molecular diffusivity was usually negligible (Zhang et al. 2021), the molecular diffusion coefficient was set to zero in this study. The molecular diffusion coefficients of solutes in water (D_w) are calculated as $\mathrm{D}\mathrm{w}=\frac{2.71·{10}^{-4}}{{\mathrm{M}}^{0.71}}$, where M is the molar mass of the solute (Li et al. 2020). The diffusion coefficient of NO₃^--N in soil was the empirical value, which was 1.64 cm²·d⁻¹, and the diffusion coefficient of heavy metals was also determined in the same way (Li et al. 2015). The longitudinal dispersion of soil is approximately 10% of the thickness of the vadose zone (Li et al. 2020). Because the depth of the growing media layer of the simulation object was shallow and the time difference of water inlet and outlet was short, the degradation degree was low. No degradation coefficient was set for this study.

Parameters of the SWMM model

Since the accumulation and flushing process of pollutants is not considered in Hydrus-1D, with regard to the determination of NO₃^--N and Pb concentrations in the model, while SWMM can simulate that hydrological and hydraulic process of urban surface runoff under stormwater condition (Baek et al. 2020), the SWMM model was used in this study to simulate the accumulation and flushing process of pollutants in road runoff during the rainfall process. We used the NO₃^--N and Pb concentrations of the underlying surface runoff, which were output through the SWMM model, as the pollutant concentration boundary of the Hydrus-1D model.

This simulation constructed an SWMM model of the bioretention systems service drainage area. The catchment ratio of the bioretention pond was 1:10, area of subcatchment was 10 m², the percent imperviousness of subcatchment was 80%, and the underlying surface was composed of roads and green spaces. The setting of water quality module parameters is shown in Table 2 (Choi and Ball 2002). Based on the SWMM model manual (Rossman 2015), the Manning coefficient of the permeable area was 0.15, and the depth of depression storage was 4.5 mm; the Manning coefficient of the impervious area was 0.02, and the depth of depression storage was 1.9 mm; the percent of impervious area with no depression storage was 25%, and the confluence model was simulated by the nonlinear reservoir method.

Table 2.

Water quality module parameter settings in SWMM

Underlying surface type	Road		Bioretention area
Pollutants	NO3--N	Pb	NO3--N	Pb
Maximum cumulative amount (kg/ha)	2.2	0.02	1.2	0.005
Half-saturated accumulation constant (d)	4	4	4.5	4.5
Scour coefficient	0.004	0.002	0.004	0.002
Scour index	1.7	1.7	5	5
cleaning efficiency (%)	30	30	0	0

Methods for sensitivity analysis, calibration, and validation of parameters

Based on the above values, this study determined the initial settings of the parameters in the Hydrus-1D model, and further sensitivity analysis of the model parameters and model calibration and validation were carried out. In this study, the Morris screening method was used to analyze the sensitivity of the input parameters in the Hydrus-1D model. With a fixed step size of 10%, a single parameter was perturbed, and the perturbation range was − 30%, − 20%, − 10%, 10%, 20%, and 30% of the initial value. When sensitivity analysis is performed on one parameter, the other parameters remain unchanged. In this study, the sensitivity of the model parameters was also graded. The classification criteria were as follows: 0 ≤|S|< 0.05 was an insensitive parameter, 0.05 ≤|S|< 0.2 was a moderately sensitive parameter, 0.2 ≤|S|< 1 was a sensitive parameter, and |S|≥ 1 was a highly sensitive parameter (Li et al. 2021a).

The model uses the square of the correlation coefficient (R²) and the Nash–Sutcliffe efficiency coefficient (E_NS) to analyze the calibration and validation results. R² can quantify the fitting degree between the measured value and the simulated value; its value ranges from 0 to 1, and the closer to 1, the greater the fitting degree. E_NS is usually used to evaluate the accuracy of model results, and the value is between − ∞ and 1; the larger the value is, the better the simulation effect. When the E_NS is close to 1, it shows that the simulated value is in high agreement with the measured value, thus reflecting high reliability of the model (Nash and Sutcliffe 1970).

Simulation of pollutants cumulative fluxes in the growing media layer

The cumulative fluxes (CF) of pollutants refer to the mass of pollutants passing through the unit cross-sectional area of the growing media layer in unit time. The CF can be calculated as follows:

$$\mathrm{CF}=F_w·C$$ 6

where CF denotes the cumulative fluxes of pollutants (mg·m⁻²·d⁻¹); F_w indicates the amount of water passing through the unit cross-sectional area of the growing media layer per unit time, that is, the water fluxes (m·d⁻¹); and C represents the solute concentration at the analysis point (mg·m⁻³).

Excel and Origin 2018 software were used for data analysis and graph drawing. Statistical analysis was performed using SPSS (IBM Corporation. Chicago, IL, USA). To study the time-dependent trend of the cumulative fluxes of pollutants at different depths of the growing media layer, we chose 2013 to 2017 as the time scale of the model, and 2013 was taken as the preheating period simulated by the model. At the same time, at the spatial scale, six analysis points, N₁ − N₆, were used, each placed at 0 cm, 20 cm, 40 cm, 60 cm, 80 cm, and 100-cm deep in the growing media layer, and the one-dimensional vertical variation laws of the cumulative fluxes of pollutants from stormwater runoff in the growing media layer were investigated. Thus, the migration and accumulation laws of pollutants were simulated and analyzed at both the temporal and spatial scales.

Results and discussion

Sensitivity analysis results

Through the sensitivity analysis of model parameters to the simulation results, the key parameters for the simulation research can be identified. Sensitivity analysis can make the model calibration and validation convenient. The sensitivity analysis results of the model parameters to water volume, effluent NO₃^--N concentration, and Pb concentration are shown in Fig. 2, where subscript 1 is the upper fluvo-aquic soil and subscript 2 is the lower sandy soil.

Fig. 2.

Sensitivity analysis results

The input values of high-sensitivity parameters can greatly affect the numerical value of the output results, while the input values of low-sensitivity parameters have little effect on the output results. Within the specified parameter range, the parameters with high sensitivity were firstly adjusted, and the parameters with the second highest sensitivity were adjusted in sequence so that the model results were more consistent with the actual experimental results. In this study, the effluent quantity and quality of bioretention facilities were used as evaluation indicators for sensitivity analysis. It can be seen from Fig. 2 that the pore size distribution parameter n of the upper soil had a great influence on the water quantity and quality. In addition, for other parameters, the sensitivity of the upper soil parameters was generally higher than that of the lower soil parameters. Pei et al. (2021) showed that the mobility of solutes in soil was significantly affected by pore size and distribution, which was similar to the results of this study.

Model calibration and validation results

Based on the sensitivity analysis results, we calibrated and verified the water quantity parameters and water quality parameters of the Hydrus-1D model by using the 9 experimental data parameters of the bioretention experiment. All bioretention device parameters in the model were calibrated using the experimental data of the SS-L device and validated using the experimental data of the SSH-L device. Within the specified parameter range, the parameters with high sensitivity were adjusted first, and the parameters with the second highest sensitivity were adjusted in sequence.

The calibration and validation results are shown in Fig. 3. It could be seen from the above analysis that the fitting effect between the simulation results and the measured values of the bioretention experimental device were acceptable and the constructed model could be used for subsequent analysis and research.

Fig. 3.

Calibration and validation results of the model parameters: a calibration results of water quantity; b calibration results of NO₃^--N; c calibration results of Pb; d validation results of water quantity; e validation results of NO₃^--N; f validation results of Pb

Simulation results of pollutant concentration

In this study, the rainfall monitoring data from 2013 to 2017 in Beijing were chosen and applied as the input rainfall value of the SWMM.

As shown in Fig. 4, the peak rainfall during the simulation period usually occurred in July and August, but the peak concentrations of NO₃^--N and Pb in the runoff generally occurred in May and June. This scenario occurred because pollutants on the urban surface accumulate in the dry season or in areas where no runoff is formed, and the concentration of pollutants in runoff is related to rainfall, runoff, and surface pollution load accumulation. Generally, November to April is the dry season in northern cities of China. During this period, there is little rainfall and almost no runoff on the surface, so a large amount of pollutants have accumulated on the surface. When the rainfall increases in May and June, the runoff washes off most of the pollutants on the surface, making the initial runoff more polluted. When peak rainfall occurs in July and August, the residual pollutants on the surface are reduced, resulting in a lower concentration. Previous studies have also shown that the concentration of pollutants is related to rainfall duration (Wang et al. 2013), rainfall intensity (Ma et al. 2018), seasonal changes (Smith et al. 2020), and the accumulation of surface pollutants (Zhang et al. 2017). In addition, this study calculated the rainfall from 2014 to 2017 and calculated the loads of NO₃^--N and Pb in the runoff in the corresponding years by using the product of the daily average concentration of the pollutants per unit area and daily runoff as the pollutant load. As shown in Fig. 5, the linear regression analysis results of NO₃^--N was R² = 0.849; the linear regression analysis results of Pb was R² = 0.947. Therefore, for the simulation conditions of multiyear rainfall, when the surface runoff coefficient is fixed, the pollutant load of runoff was positively correlated with the rainfall.

Fig. 4.

NO₃^--N and Pb concentrations in runoff from 2014 to 2017

Fig. 5.

Linear regression analysis of the pollutant load and the rainfall

Variation in the cumulative fluxes of pollutants in the growing media layer with time

Through the changes in cumulative fluxes at different depths (N₁ − N₆) of the growing media layer under multiyear simulation conditions, the migration and accumulation of pollutants in the growing media layer were analyzed in this study, and the results are shown in Fig. 6.

Fig. 6.

Trend of the cumulative fluxes at different depths of the growing media: a NO₃^--N; b Pb

The pollutants NO₃^--N and Pb in the runoff gradually migrated to the lower layer and accumulated internally after infiltrating into the growing media layer of the bioretention, and the accumulation degree varied from year to year. However, the annual accumulation processes of NO₃^--N and Pb in the growing media were the same, both of which mainly occurred during the runoff of rainfall, specifically manifested as NO₃^--N and Pb began to accumulate from the beginning of the annual rainfall, and then entered the release phase after accumulating to a certain amount. Following that activity, according to the two scales of the entire simulation period and the rainy season, the yearly simulation time was utilized to further evaluate the fluctuation laws of the cumulative fluxes of the pollutants in the growing media layer with time.

Variation in the cumulative fluxes of NO₃^--N and Pb over the whole simulation period

The cumulative fluxes of NO₃^--N and Pb over the whole simulation period are shown in Tables 3 and 4. The results of multifactorial ANOVA using SPSS for all outcome data are presented in Table 5. The cumulative fluxes of NO₃^--N and Pb were taken as dependent variables, and time and depth were taken as influencing factors. Then, the main effects of different influencing factors were analyzed. The results showed that the time had significant effects on the cumulative fluxes of NO₃^--N (p = 0.000 < 0.05) and Pb (p = 0.000 < 0.05), the depth also had significant effects on the cumulative fluxes of NO₃^--N (p = 0.000 < 0.05) and Pb (p = 0.000 < 0.05). There was a main effect in both the time and the depth, indicating that the cumulative fluxes of pollutants would be significantly changed with the increase of the time and the depth.

Table 3.

Statistics of the cumulative fluxes of NO₃^--N at different depths of the growing media (mg·m⁻²·d⁻¹)

Depth	2014	2015	2016	2017
N1	1.73 ± 0.70	2.04 ± 1.51	1.93 ± 0.82	2.79 ± 2.11
N2	2.29 ± 1.04	2.61 ± 1.77	2.45 ± 1.03	3.77 ± 2.88
N3	2.55 ± 1.33	2.88 ± 1.98	2.68 ± 1.20	3.97 ± 2.89
N4	2.70 ± 1.47	3.13 ± 2.19	2.92 ± 1.57	4.01 ± 2.86
N5	2.79 ± 1.52	3.34 ± 2.35	3.21 ± 2.21	3.99 ± 2.82
N6	2.83 ± 1.55	3.46 ± 2.41	3.39 ± 2.66	3.96 ± 2.79

The values refer to mean ± standard deviation

Table 4.

Statistics of the cumulative fluxes of Pb at different depths of the growing media (10⁻² mg·m⁻²·d⁻¹)

Depth	2014	2015	2016	2017
N1	2.77 ± 0.81	3.24 ± 1.15	3.43 ± 1.17	3.73 ± 1.56
N2	3.51 ± 0.90	3.92 ± 1.20	3.97 ± 1.06	4.54 ± 1.82
N3	3.69 ± 1.03	4.11 ± 1.32	4.12 ± 1.17	4.65 ± 1.87
N4	3.80 ± 1.13	4.28 ± 1.44	4.26 ± 1.42	4.68 ± 1.88
N5	3.86 ± 1.18	4.42 ± 1.51	4.41 ± 1.78	4.66 ± 1.87
N6	3.89 ± 1.21	4.49 ± 1.54	4.51 ± 2.03	4.65 ± 1.86

The values refer to mean ± standard deviation

Table 5.

Analysis of the significance level

Influencing factors	The cumulative fluxes of NO3--N		The cumulative fluxes of Pb
	F	p	F	p
Time	10.957	0.000	9.842	0.000
Depth	5.474	0.000	7.682	0.000

F and p indicate F statistics and p values respectively; p value < 0.05 indicates significant difference

The cumulative fluxes of pollutants at the same depth of the growing media layer changed in a generally positive trend. When runoff occurred at the beginning of each year's rainfall, the cumulative fluxes of NO₃^--N and Pb exhibited increasing trends with varying amplitudes at each analysis point. During the rainy season, the cumulative fluxes increased until they reached the saturation point, at which time they began to fall. The cumulative fluxes of NO₃^--N and Pb in the growing media layer varied slowly over the runoff-free period, and the change amplitudes at different analysis points tended to be consistent.

With the least rainfall and pollution load in 2014, the cumulative fluxes of pollutants in the growing media layer were low. There would be some adsorption and accumulation of pollutants in the growing media layer as simulation time passed, resulting in an increase in the cumulative fluxes of pollutants at the same depth. However, pollutants in the upper part of the growing media layer in 2016, with high rainfall and pollution load, migrated to the lower part to a greater extent. In addition, the peak cumulative fluxes at the bottom of the growing media layer were the highest, and the mean cumulative fluxes were lower. The peak cumulative fluxes of the growing media layer occurred during the rainy season as a result of the heavy rainfall event that scoured the growing media layer and caused a significant amount of pollutants to be leached out. The mean cumulative fluxes at the bottom of the growing media layer were reduced as a result of the reduction in the pollutant load in the growing media layer and the subsequent reduction in cumulative fluxes as a whole. It indicated that the cumulative fluxes were related to the simulated annual rainfall, pollution load and pollutant accumulation. Sun et al. (2022) simulated the heavy metal leaching characteristics of Pb–Zn tailings and found that the concentration of heavy metals in leachate increased with increasing rainfall. Most soils or filter media have poor N and P adsorption (Tirpak et al. 2021), according to studies, and soil moisture can impact solute transport (Fan et al. 2014) and contribute to nitrate leaching in future rainfall events, thereby affecting the accumulation of nitrate in soil profiles (Liu et al. 2019). As a result, when rainfall was heavy and the runoff pollution load was significant, it provided power for the migration of pollutants, allowing pollutants to migrate more easily in the growing media layer.

Cumulative fluxes of NO₃^--N and Pb in the rainy season

Peak cumulative fluxes refer to the maximum mass of pollutants passing through a unit cross-sectional area of a growing media layer in a unit of time. By taking the maximum value of the cumulative flux variation range of the analysis points at different depths, the influence of different factors on the achievable saturation cumulative pollutants at different depths of the growing media layer was reflected. Figure 6 shows that during the whole process, the cumulative flux curves fluctuated with the variation of the pollutant concentrations in the runoff; the cumulative fluxes of NO₃^--N and Pb began to rise slowly with pollutant concentrations in the initial stage of runoff, and the change range was small. By analyzing the changes in the cumulative fluxes of NO₃^--N and Pb from 2014 to 2017, it was found that the peak cumulative fluxes of both occurred in the rainy season every year. The peak cumulative fluxes of NO₃^--N and Pb at different depths of the growing media layer were calculated, as shown in Table 6.

Table 6.

Statistics of the peak cumulative fluxes of NO₃^--N and Pb at different depths of the growing media

Depth	NO3--N peak cumulative fluxes (mg·m−2·d−1)				Pb peak cumulative fluxes (10−2 mg·m−2·d−1)
	2014	2015	2016	2017	2014	2015	2016	2017
N1	4.11	7.13	3.87	7.72	5.89	6.38	6.81	7.06
N2	5.12	8.83	4.77	9.48	6.14	7.55	7.07	8.37
N3	5.69	9.04	5.53	9.17	6.64	7.73	7.94	8.58
N4	6.51	9.18	8.27	9.15	7.11	7.87	9.65	8.73
N5	7.31	9.21	12.22	9.00	7.61	7.92	11.79	8.76
N6	8.01	9.16	14.68	8.94	8.01	7.91	13.12	8.71

As shown in Table 6, the peak cumulative fluxes of NO₃^--N in the upper part of the growing media layer showed a trend of increasing with simulation time except in 2016, and the peak cumulative fluxes in the lower part of the growing media layer gradually increased from 2014 to 2016 but decreased in 2017. The peak cumulative fluxes of Pb at the same depth in the growing media layer showed a trend of increasing with the passage of simulation time except in 2016. Compared with those of NO₃^--N, the changes in the peak cumulative fluxes of Pb in the growing media layer were slightly different, which was related to the low content of Pb in the runoff and the growing media layer, and the slight changes led to the nonsignificant difference.

The highest rates of change in the growing media layer in 2016 were for the peak cumulative fluxes of NO₃^--N and Pb, which were connected to the concentration of pollutants in the runoff, when the rainfall was unusually heavy, resulting in high levels of NO₃^--N and Pb in the runoff. As a result, peak cumulative fluxes at various depths of the growing media layer rose during the downward migration. However, owing to very heavy rainfall, NO₃^--N and Pb were absorbed less in the surface layer of the growing media layer over a short amount of time, and the relatively high water content between the particle pores promoted the saturation of the upper part of the growing media layer. According to other studies, the input of NO₃^--N has a significant positive correlation with the accumulation of nitrate in the soil layer (Ascott et al. 2017; Gao et al. 2019), with excess nitrate migrating vertically to a deeper soil depth (Zhou et al. 2016). The effect of heavy metal concentrations on their migration ability has been studied, and it has been found that heavy metal leaching is proportional to their overall content (Sun et al. 2022); in addition, salt content in runoff also affects heavy metal adsorption and migration accumulation in the growing media layer of bioretention facilities (Costello et al. 2020). It was shown that the amount of NO₃^--N in the growing media layer may have had an impact on the migration and accumulation of Pb.

Variation in the cumulative fluxes of the pollutants in the growing media layer with depth

The cumulative flux laws of NO₃^--N and Pb over the whole simulation period are shown in Tables 3 and 4, demonstrating that cumulative fluxes in the bottom half of the growing media layer were overall larger than those in the top section. As shown about the multifactorial ANOVA results in Table 5, the depth had a significant effect on the cumulative fluxes of NO₃^--N (p = 0.000 < 0.05) and Pb (p = 0.000 < 0.05). The variation of the peak and net cumulative fluxes with depth in the growing media layer was discussed further below.

Variation in the peak cumulative fluxes

By comparing the peak cumulative fluxes at each analysis point in Table 6, it was found that the peak cumulative fluxes of NO₃^--N increased by 95%, 28%, 280%, and 16%, and the peak cumulative fluxes of Pb increased by 36%, 24%, 93%, and 23% from the surface to the bottom of the growing media layer from 2014 to 2017, respectively. The changes in the peak cumulative fluxes of Pb in the growing media layer were slightly different from those of NO₃^--N because the background cumulative fluxes of Pb in the runoff and the growing media layer were lower, which resulted in smaller change rates of the peak cumulative fluxes of Pb with depth.

The peak cumulative fluxes of NO₃^--N and Pb in the growing media layer with runoff infiltration increased with the depth, and except in 2016, the peak cumulative fluxes of NO₃^--N and Pb in the growing media layer increased slowly with depth. The peak cumulative fluxes in the upper part of the growing media layer were lower in 2016 than in other years and in the lower part of the growing media layer were the highest. This was because the amount of rainfall in 2016 was relatively large, resulting in a high inflow of NO₃^--N and Pb into the surface layer, the saturation phenomena in the upper part of the growing media layer led the surface layer to fail to absorb pollutants in a timely manner and enhanced the infiltration of NO₃^--N and Pb in the growing media layer, thus resulting in the maximum increase in the peak cumulative fluxes from the surface to the bottom of the growing media layer, and the growth trend was more obvious. The study by Zhang et al. (2022) on the downward movement of nitrate in soil found that soil depth had a significant indirect effect on the accumulation of NO₃^--N, and a peak value of NO₃^--N accumulation appeared with the downward movement in soil. In addition, since soil moisture was the main carrier of NO₃^--N, soil water storage had a direct impact on the accumulation of NO₃^--N. The downward migration of NO₃^--N in the soil was related to the infiltration of water in the soil during rainfall, and the degree of migration was affected by rainfall intensity (Wang et al. 2022). The storage and movement of soil moisture promoted the migration and accumulation of NO₃^--N in the soil. Zeng et al. (2020) also demonstrated that rainfall drives the migration of NO₃^--N. Hermawan et al. (2021) found that the deeper filter media can improve the retention of contaminants, but may result in a higher percentage of heavy metals accumulating in the bottom layer. The soil–water characteristic curve constructed by Bai et al. (2020) showed that the heavy metal Pb would migrate with water migration and its migration degree increased with the increase in the initial water content.

Furthermore, the change rates of the peak cumulative fluxes of NO₃^--N and Pb in the growing media layer were the lowest in 2017 because the rainfall and pollutant load in the runoff that year experienced a migration and accumulation process of 4 years over the course of the simulation cycle, resulting in higher initial cumulative fluxes of pollutants in the growing media layer in 2017 than in other years, which was the reason why the peak cumulative fluxes were higher but the growth rates were lower. An et al. (2022) found that long-term N source input to soil has been demonstrated to reduce NO₃^--N leaching and induce N₂O generation. Afrooz and Boehm (2017) accepted that frequent exposure of the media to stormwater will gradually result in saturation of sorption sites in the surface layer, which may affect the overall performance of the bioretention system in retaining heavy metal. In this study, the migration and accumulation of NO₃^--N and Pb in the growing media layer mainly occurred in the rainy season, which were driven by the rainfall and greatly affected by the runoff pollution load. Therefore, NO₃^--N and Pb migrated to the lower part of the growing media layer with soil moisture during rainfall so that their peak cumulative fluxes were increased with the growing media layer depth, and the change rates of their peak cumulative fluxes with depth were correlated with the simulated annual rainfall and the initial cumulative fluxes of the simulated year.

Variation in the net cumulative fluxes

The mass per unit cross-sectional area of pollutants moving through the growing media layer on the last day after a year of migration accumulation is known as the net cumulative fluxes. As indicated in Table 7, the net cumulative fluxes of NO₃^--N and Pb in the growing media layer for each year were calculated. With the deepening of the growing media layer, the net cumulative fluxes of NO₃^--N and Pb increased. The results revealed that the net cumulative flux growth rate of pollutants from the surface to the bottom of the growing media layer was lowest in 2016, when the change rates of NO₃^--N and Pb in the growing media layer were calculated for each year.

Table 7.

Statistics of the net cumulative fluxes of NO₃^--N and Pb at different depths of the growing media

Depth	NO3--N net cumulative fluxes (mg·m−2·d−1)				Pb net cumulative fluxes (10−2 mg·m−2·d−1)
	2014	2015	2016	2017	2014	2015	2016	2017
N1	0.90	1.05	1.25	0.98	2.04	2.59	2.61	2.52
N2	1.20	1.16	1.28	1.30	2.79	2.66	2.68	2.72
N3	1.29	1.27	1.26	1.50	2.85	2.74	2.69	2.84
N4	1.34	1.40	1.25	1.57	2.88	2.81	2.69	2.89
N5	1.37	1.52	1.26	1.60	2.90	2.87	2.71	2.91
N6	1.39	1.58	1.27	1.61	2.91	2.90	2.71	2.91

Throughout the simulation period, the net cumulative fluxes of NO₃^--N and Pb in this study followed the rule that the net cumulative fluxes in the lower section were typically larger than those in the upper section. This result was due to the partial migration of pollutants in the runoff and pollutants deposited by the growing media layer to the bottom section of the growing media layer. It has been found that leached nitrates may accumulate in deep seepage areas (Bai et al. 2021). The accumulation of heavy metals also increases with the depth of soil (Ke et al. 2020). According to Wang et al. (2019), leaching of Pb in the shallow layer and redeposition of Pb in the deep layer at 40 − 100 cm dominate the vertical movement of Pb. Moreover, Pb has a significant migratory capacity that is enhanced as soil depth increases (Zhang et al. 2018).

The net cumulative fluxes were expressed by CF_N. The net cumulative fluxes of NO₃^--N and Pb at the surface were ranked as CF_N 2016 > CF_N 2015 > CF_N 2017 > CF_N 2014. As the depth of the growing media layer increased, the net cumulative fluxes of NO₃^--N at the bottom were ranked as CF_N 2017 > CF_N 2015 > CF_N 2014 > CF_N 2016, and the net cumulative fluxes of Pb at the bottom were ranked as CF_N 2017 > CF_N 2014 > CF_N 2015 > CF_N 2016. It was found that the net cumulative fluxes of NO₃^--N and Pb in the surface layer had a large correlation with their contents in the runoff. Due to heavy rainfall in 2016, the migration activities of the pollutants originally existing in the growing media layer were ongoing, leading to the migration of NO₃^--N and Pb in the lower part of the growing media layer to the filter layer. That is, the migration speed of the pollutants was fixed, and the accumulated pollutant content at the bottom of the growing media layer was lowered over time. According to Lu et al. (2019), the accumulation of NO₃^--N in soil increases as rainfall increases, but as rainfall reaches its peak level in the current year, the accumulation declines as rainfall increases further. Furthermore, in this study, the migration process was ongoing, with certain pollutants continuing to move downhill over time. As a result, although rainfall in 2017 was not the highest, the net cumulative fluxes of pollutants at the growing media layer’s bottom were the highest. Speak et al. (2014) and Mitchell et al. (2021) demonstrated that the growing media of LID facilities after long-term operation could accumulate N and the heavy metal Pb as potential sources of runoff pollution. As a result, the net cumulative fluxes of pollutants in bioretention facilities grew with downward movement, with runoff in the upper half of the growing media layer having a significant effect. However, with downward movement, time became one of the elements affecting accumulation.

Conclusion

1. The cumulative fluxes of NO₃^--N and Pb increased with the passage of simulation time and growing media layer depth overall, and the growth amplitude increased slowly with the increase in the growing media layer depth. During the rainfall runoff period, the cumulative fluxes of NO₃^--N and Pb at different depths of the growing media layer increased with different amplitudes and began to decline after reaching the saturation point; during the runoff-free period, the cumulative fluxes of NO₃^--N and Pb in the growing media layer changed gradually, and the variation amplitudes at different depths tended to be consistent.

2. The peak and net cumulative fluxes of NO₃^--N and Pb of the growing media layer were related to rainfall, growing media layer depth, pollutant content in runoff, and the initial content of pollutants in the growing media layer. The change rates of the peak cumulative fluxes decreased as the simulated annual background initial cumulative fluxes increased. Runoff had a considerable impact on the cumulative fluxes of NO₃^--N and Pb at the surface of the growing media layer.

3. The accumulation of runoff pollutants in the growing media layer was related to the interception and adsorption of pollutants by the growing media of the bioretention facility and the migration of the pollutants themselves. Interception and adsorption were mainly affected by growing media materials. Runoff, pollutant load, initial pollutant content, and initial water content of the growing media layer all affected the migration of pollutants themselves. With the continuous rainfall, the conditions at various depths of the growing media layer were changed, resulting in different cumulative fluxes of pollutants at different depths.

Overall, in order to accurately predict the accumulation degree of runoff pollutants in the bioretention facilities, the effect of the year of stormwater on pollutant distribution in the growing media layer at different depths of bioretention facilities after long-term operation should be accounted for. The research findings can be used to determine whether the effluent quality of the growing media layer at different depths meets the design requirements and whether the cumulative amount of pollutants exceeds the ecological safety threshold in actual engineering. The results will also serve as a theoretical reference for the repair and replacement of the growing media layer during actual engineering operations and maintenance.

Author contribution

Yongwei Gong: conceptualization, methodology, project administration, writing-review and editing, funding acquisition. Xia Li: methodology, data curation, formal analysis, writing-original draft. Peng Xie: methodology, data curation, formal analysis. Hongyan Fu: formal analysis, writing-original draft. Linmei Nie: review and editing. Junqi Li: methodology, review and editing, supervision. Yanhong Li: review and editing.

Funding

This research was supported by the National Natural Science Foundation of China (51879004), the Major Science and Technology Program for Water Pollution Control and Treatment (2017ZX07403001), and the Basic Research Fund of Beijing University of Civil Engineering and Architecture (X20145 and X20160).

Data availability

All data and materials generated or analyzed during this study are included in this published article.

Declarations

Ethical approval

Not applicable.

Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.