Load quantification and effect evaluation of urban non-point source pollution in the Licun River based on SWAT model

In urban areas of China, the large percentage of impervious areas, the complexity of production activities, and the variety of pollution sources have led to serious pollution of urban rivers. Urban nonpoint source pollution is the second-largest source of nonpoint source pollution after agricultural nonpoint source pollution (Zhang et al. 2011; Cheng et al. 2017). In recent years, with the improvement of sewage collection, interception systems, and effluent standards of the sewage treatment plants in urban areas of China, point source pollution has been gradually controlled. On the other hand, the impact of urban nonpoint source pollution on river water quality is becoming more prominent. Therefore, it is urgent to study the load quantification and impact of urban river nonpoint source pollution, thus providing data support for improving urban river water quality.

At present, the load quantification of urban surface pollution is mostly estimated using empirical formulas, and there is a lack of a more accurate load quantification method. With the massive development of distributed hydrological models (e.g., SWAT, HSPF, AnnAGNPS models) (Radcliffe & Cabrera 2006; He et al. 2020; Marin et al. 2020), the high-accuracy simulation of agricultural nonpoint source pollution loads in large-scale basins has been achieved. However, there is a lack of research on applying models for load quantification and impact analysis of urban nonpoint source pollution in small-scale basins of urban rivers.

The Licun River basin in Qingdao is a typical urban river catchment area. At present, the nitrogen and phosphorus in the effluent of the wastewater treatment plant in the basin of the Licun River have reached the IV standard for surface water, and the sewage interception system in this region has become more comprehensive. Consequently, urban nonpoint source pollution has become the primary type of pollution affecting the water quality of the Licun River. Therefore, the Licun River basin in Qingdao was selected as the study object, and the SWAT model was used for simulation analysis. Through the field survey and surface accumulation sampling, the acquired land use data and the urban database of the model were revised, achieving the urban nonpoint source pollution load quantification and impact analysis of the Licun River basin. This study can provide a reference for analyzing urban nonpoint source pollution of similar urban rivers in China.

The SWAT model is commonly used for the analysis of nonpoint source pollution. It includes modules for rainfall simulation, sink production, and diffusion of river water quality (Ding & Zheng 2004; Radcliffe & Cabrera 2006). Because of its pesticide/insecticide module and the agricultural management module, the SWAT model is often used to simulate agricultural nonpoint source pollution in basins and is rarely used for urban nonpoint source pollution studies. According to the ArcSWAT Interface for SWAT2009: User's Guide (Winchell 2012) and the SWAT Input/Output File Documentation Version 2009 (Arnold et al. 2011), it was found that the SWAT model for versions 2009 and above already has an urban database. The basin land use data and urban database have been improved, enabling the simulation of urban nonpoint source pollution in the basin with the SWAT model. The key parameters of the SWAT model urban database (Arnold et al. 2011; Abdelwahab et al. 2018) are shown in Table 1.

In this study, DIRTMX, TNCONC, TPCONC, and TNO3CONC values were determined by monitoring and analyzing the surface deposits in the study area. Currently, two primary surface accumulation sampling methods are commonly used: the dry and wet sampling methods (Li et al. 2015). According to the tools used, the dry sampling method can be subdivided into the brush or broom sweeping method and the vacuum sampling method using a vacuum cleaner. In the wet sampling method, the ground is first cleaned using deionized water, and the mud-water mixture is absorbed through a vacuum cleaner (Chang et al. 2007). In this study, dry sampling of the study area was carried out using a wool brush and an 800 W German Kahl vacuum cleaner, and the sampling process was as follows:

• (1)

  In the selected sampling region, sampling points are laid along with the street side stones. According to the area of the sampling region, the number of sampling points was determined, with no less than five points in each region. A self-made sampling frame of 1.0 m × 1.0 m was placed at the sampling site to collect the samples in the frame.

• (2)

  During sampling, a vacuum cleaner was first used to collect dust from the frame horizontally and vertically. Then, the brush was used to sweep the surface of the frame back and forth, removing the particles attached to the surface. Finally, the vacuum cleaner was used again to collect the dust from the frame with three to five repetitions.

• (3)

  The surface dust collected in the dust collection bag of the vacuum cleaner was moved into a sterile sampling bag and labeled.

According to Cheng et al. (2017), urban rainfall-runoff pollution is mainly caused by fine dust with particles ranging from 50 to 1,000 μm. Therefore, the collected surface accumulations were sieved using metal sieves in this study. Samples of particles smaller than 1 mm are weighed, recorded, and then repeatedly bumped to achieve a uniform distribution of different-sized particles. Each sample was dissolved in 1 L of deionized water with 5 g, and the total nitrogen (TN), nitrate-nitrogen, and total phosphorus (TP) of the mixed water samples were measured. In addition, three sets of parallel samples were set up for each group of samples.

This study uses a combination of basic data collection and on-site monitoring analysis to obtain the relevant data required for SWAT model construction. Data types can be divided into spatial data and surface observation data. Unlike previous studies, there is less agricultural land and larger urban land in the Licun River basin. Therefore, agricultural management data were not collected during data collection. The data and basic information required for SWAT model construction are shown in Table 2.

The relative significance of each parameter b_i is determined by t-test; the sensitivity of each parameter is measured by the algorithm output parameter t_stat, with larger absolute values indicating high sensitivities.

The sampling study of the surface accumulation of the urban land types in the Licun River basin was conducted. Afterward, based on the relevant parameter ranges provided in the SWAT Input/Output File Documentation Version 2009 (Winchell 2012) and the existing research results (Zhang et al. 2021), a SWAT model urban database of the Licun River basin was constructed. Among the parameters, FIMP, FCIMP, and CURBDEN were determined from field investigations, DIRTMX, TNCONC, TPCONC, and TNO3CONC were determined from surface sediment examinations, and URBCOEF, THALF, and URBCN2 were determined by referring to the SWAT Input/Output File Documentation Version 2009 (Winchell 2012) with the calibration studies of Feng-Jiao & Tian-Hong (2012) and Zhang et al. (2022) on CN values in the Guanlan River in Shenzhen and the Licun River basin in Qingdao.

By sampling and monitoring the surface sediments of seven types of urban land in the Licun River basin, the SWAT model urban database was constructed, and the parameters were calibrated in a localized manner. Sensitivity analysis of the parameters showed that: The parameters RCN_SUB_BSN, FIXCO, and BC3 are most sensitive to the simulation of monthly TN load, and the parameters LAT_ORGP, ERORGP, and PSP are most sensitive to the simulation of monthly TP load. In addition, the E_NS and the R² of the constructed SWAT model for the simulation validation period of runoff, TP load, and TN load of the control section suggest an excellent fit to the actual values and high confidence of the simulation results.

By using the corrected SWAT model, the spatial and temporal distribution characteristic analysis of TN and TP in the Licun River basin in 2021 revealed that the highest spatial distribution of TN and TP unit area annual output loads are 58,871.16 and 54,173.85 kg/km², respectively. Among seven urban land types, the contribution of TN and TP load from UTRN is the highest, accounting for 34.61 and 35.20%, respectively; the contribution of TN load from UINS is the lowest, accounting for 0.57%; the contribution of TP load from URMD is the lowest, accounting for 0.15%. The TN load and TP load from rainfall-runoff pollution from April to September are higher than those in the rest of the months and the most polluted rainfall-runoff is in July.

The authors would like to thank Qingdao Municipal Bureau of Ecology and Environment for providing the historical data.

Minghui Zhang developed the methodology and software, validated the article, conducted a formal analysis and data curation, and wrote the original draft. Lin Wang conceptualized the whole article, wrote the review, edited the article, supervised the work, and conducted funding acquisition. Xuda Huang conducted investigation, brought resources, reviewed & edited the article, and administered the project. Zhonghua Yu brought resources, reviewed & edited the article, and administered the project.

This work was supported by the National Key Research and Development Program of China [NO.2018YFC0408000,2018YFC0408004] and the Jinan Water Science and Technology Project [JNSWKJ202103].

Data cannot be made publicly available; readers should contact the corresponding author for details.

The authors declare there is no conflict.