Removal of N, P from seepage and runoff by different vegetated and slope buffer strips

Since the 1980s, problems such as wetlands loss, biodiversity decrease, and agricultural nonpoint pollution have posed serious threats to humans and wildlife. Hazlett et al. (2008) and Tomer et al. (2008) discovered that one of reasons for eutrophication of rivers and lakes was the degradation of the riparian ecosystem, which deprived its function of trapping sediment and debris, stabilizing stream banks, reducing erosion, and improving the infiltration of runoff. As an ecological link between terrestrial and aquatic ecosystems, the riparian buffer zone has a significant effect on restoring the degraded riparian ecosystem. Lowrance et al. (1985) considered that riparian buffer strips had a number of ecological functions, such as protecting the river water environment, maintaining water quality and improving biodiversity. Smith et al. (2014) regarded riparian buffer strips as increasing the biodiversity of ecosystems, retarding sediment and increasing soil productivity.

In China, the utilization of pesticide and fertilizer in agriculture caused significant pollution to surface waters. Through surface runoff and soil seepage (infiltration), the pollutants ultimately flow into surface water bodies, such as rivers and lakes. In particular, N and P from fertilizer and sewage are well known to cause negative environmental effects including eutrophication (Cao et al. 2018; Nummer et al. 2018; Xu et al. 2019). In recent years, some researchers also conducted research and developed the principles of using riparian buffer strips to control pollution. Pan & Shangguan (2007) concluded that the interception rates of grasslands with different slopes for runoff sediment were from 19.5% to 43.6%, and slope gradient had a negative correlation with the interception rates. Qi & Altinakar (2011) studied the purification effect of vegetated buffer strips on agricultural non-point N, P pollution on a small scale. The interceptions of N and P in farmland runoff in the buffer zone are mainly in precipitation, infiltration, and adsorption, and the processes of pollutants migration and transformation in the buffer zone are very complex, with many influencing factors, including hydraulic conductivity, soil type and initial water quality, plant type and root state, etc. (Balestrini et al. 2011).

For areas with similar soil types, vegetation characteristics and slope are two important factors of riparian buffer strips to intercept runoff pollutants (Huang et al. 2008; Wang et al. 2008). Vegetation can make a significant contribution to the control of nonpoint pollution and ecosystem recovery by absorption, increasing dissolved oxygen, providing biological habitats, loosening soil, retarding runoff, and adjusting microclimate (Milan et al. 2013; Nash et al. 2015; Haukos et al. 2016). The interception of suspended solids (SS) and reduction of other pollutants by traditional riparian buffer strips mainly happens at the front end of turf buffer strips. This is because herbaceous plants that grow intensively and cover the surface can retard runoff, intercept runoff pollutants, degrade, absorb, and deposit pollutants effectively (Yan et al. 2019; Zhu et al. 2019). At the same time, the slope of river banks is another important factor to determine the capacity of buffer strips in intercepting sediments and retaining nutrients. The gentler the slope is, the slower the runoff flows, providing longer time of contact and buffering. Thus, pollutant interception and degradation efficiency are higher (Huang et al. 2008; Wang et al. 2008).

Research into riparian buffer strips was developed and progressed rapidly in America, Europe and Japan with studies on using it to control contaminants in agricultural runoff (Weigelhofer et al. 2012; Cardinali et al. 2014; Habibiandehkordi et al. 2014). Noij et al. (2013) found that a riparian ecosystem was able to absorb N, P, and a large amount of sediment particles, including heavy metals and organic pollutants from runoff. It also reduced persistent pollutants in groundwater by complexation and chelation. A filter strip covering 10–50% of the subbasin area with a perennial vegetation of switchgrass could potentially reduce NO₃^–N by 55–90% in outflows from the subbasin under average rainfall conditions (Sahu & Gu 2009; Nummer et al. 2018). Shan et al. (2014) and Collins et al. (2013) showed that riparian buffer strips could effectively reduce the runoff rate, improve sediment deposition, and intercept and degrade most N, P nutrients. However, most research focused on qualitative analysis of the effect of buffer strips for retarding runoff and reducing pollutant concentrations (Guo et al. 2015; Haukos et al. 2016; Li et al. 2016; Cao et al. 2018; Zhou et al. 2019).

This study intends to focus on the distribution of runoff/seepage volume to derive detailed quantitative analysis on retarding runoff and the removal of N, P pollutants in runoff and seepage flow. To explore better design of buffer strips using typical vegetation and slopes, a research project was designed in Qingpu district, Shanghai. The test facility was constructed including buffer strip test bases and runoff measurement devices. This research intends to compare and analyze the pollutant removal effect of a typical vegetation buffer, in order to find the removal capacities of different vegetation and slopes for non-point source pollutants, conduct the correlation of runoff/seepage volume ratios and removal loads per area, and determine the best buffer vegetation types and slope. On one hand, the research can provide a technical basis for the prevention and control of non-point source pollution in Shanghai or similar plain river network areas. On the other hand, it can also provide technical support for the design and construction of a riparian buffer zone.

The test base is located at Dongfeng River, Qingpu district, in the west part of Shanghai, covering an area of about 8,000 m². It is around 100 meters long from north to south, and 80 meters wide from east to west. The area is at the upstream of Suzhou River, 121°236′ E and 31°257′ N (Figure 1). The whole length of Dongfeng River is 3.1 kilometers and the width of the water body is about 20 meters. The river is affected by the tide, with a normal water level of 2.7–2.8 meters. Under the north subtropical monsoon climate, the annual average temperature is about 18 degrees Celsius, with around 1,929 hours sunshine. Annual average precipitation is 1,104.4 millimeters and the annual average rainy days is about 130, mainly during May to October. The local soil is mainly purple clay and groove clay, and the quality is slightly sticky.

A 10 m³ service reservoir was constructed in the test base to simulate runoff water. Seven test strips were used to construct buffer strips with different vegetation and gradients. Most studies adopted artificial runoff simulation in accordance with the rainfall characteristics of the subject area, using grass as the plant at a width of approximately 4.6–27.4 meters through which runoff flows to the river. Most of them applied a gentle slope in the range of 10% (Schwer & Clausen 1989; Mander et al. 1997; Lee et al. 2004; Keffala & Ghrabi 2005). In our study, every test strip was 19 meters long and 2 meters wide. Every two strips were divided by anti-seepage membranes to avoid any interaction. Imitating the characteristics of agricultural nonpoint pollution and rainfall runoff in Shanghai, the concocted runoff ran artificially from the service reservoir, and ran back to the Dongfeng River after treatment by buffer strips. The flow chart of the buffer strips field test is shown as Figure 2. Water sampling devices were located at the weir mouth of the right triangle weir at the end of each test strip. u-PVC (Unplasticized Polyvinyl Chloride) pipes were buried in each strip for water sampling.

In eastern China, coastal cities with abundant river networks are usually on low and flat plains, where river banks only have gentle slopes with a 2%–5% gradient. Cynodon dactylon, Festuca arundinacea, Trifolium repens Phragmites australis and Typha angustifolia are native and typical herbaceous plants in Shanghai (Wu et al. 2018). Cynodon dactylon, Festuca arundinacea and Trifolium repens were chosen to be planted on the test buffer strips with 2% gradient. Cynodon dactylon is a perennial warm-season herbaceous plant with 270 days' green period. Festuca arundinacea and Trifolium repens are both perennial and with good resistance to cold and heat. Festuca arundinacea is evergreen and suitable for growing in a warm and wet area. Trifolium repens is easy to grow in most areas of China and evergreen in all seasons in Shanghai. When comparing the runoff at different slopes of 2%, 3%, 4% and 5% gradients four buffer strips, Cynodon dactyl were chosen for the test strips (Figure 3).

The test was designed to imitate confluence with the typical single rainfall of 40 millimeters for 60 minutes during May to October. Confluence water volume was calculated by confluence area and test strips, which was 4 m³. According to the characteristics of agricultural nonpoint pollution in Shanghai, discharged loads of N, P pollutants were simulated by using ammonium bicarbonate, superphosphate, and soil particles as agricultural runoff water samples by artificial configuration (Huang et al. 2007). Main concentrations of N, P pollutants are shown in Table 1.

The test base was completed in March 2012 and all kinds of vegetation were planted by the end of the year. In order to inspect the agricultural runoff control effect of buffer strips during the rainy season, the majority of the experiment was conducted during May to October of 2016, 2017 and 2018. All seven test strips ran the experiment twice every year.

The mean data from the right triangle weir device during the experiment processes are shown in Table 2. The mean time of surface flow from the start to the end (initial outflow time of runoff) for 2% gradient buffer strips with three different vegetation was 13.8 minutes, while the control group duration time was 6.7 minutes. This showed that the runoff volume of three kinds of buffer strips with different herbaceous plants was significantly less than that of the control groups. The mean runoff volume of three kinds of vegetated buffer strips was 2.71 m³, while the control group was 3.5 m³. The ratio of runoff to that of seepage in the control group was 6.84, which was 2.6–4.2 times higher than all the other three vegetated buffer strips. This indicated that vegetated buffer strips not only slowed down the runoff effectively, but also improved the seepage capacity remarkably by the diversion effect of vegetation roots. The result was consistent with the other research that found vegetation slowed down runoff effectively (Haukos et al. 2016; Li et al. 2016; Stutter et al. 2019).

Among different vegetation, the difference in the runoff volumes between every two buffer strips was significant (P < 0.05) except between Trifolium repens and Festuca arundinacea (Figure 5). Combining the growth condition of different vegetation in the field and the research results of Wu et al. (2008), Cynodon dactylon grew well during May to October because of abundant rainfall and suitable temperature (17.0–28.0 °C in Shanghai) with crowded stolons covering the whole surface of the buffer strips, providing the highest stagnation capacity of runoff. Its mean initial outflow time of runoff was 16.4 minutes and mean runoff volume was the least, just 2.47 m³. While Festuca arundinacea was in aestivation during summer (when temperature is more than 26 °C), it did not grow well, resulting in lower stagnation ability of runoff than the other two vegetated buffer strips, which was the largest among these three, 2.90 m³. However, Trifolium repens doesn't have stolon that can cover the whole surface of the ground as Cynodon dactylon does (Kou et al. 2013; Wu et al. 2018). Its ability for runoff stagnation was between the other two.

With the increase of buffer strip gradient, the initial outflow time of runoff was shortened. Among the buffer strips with different gradients, the initial outflow time of runoff was 16.4 minutes for 2% gradient slope, while it was only 9.1 minutes for 5% gradient slope buffer strips. The initial outflow time of runoff through 2% gradient slope buffer strips was 1.8 times the 5% gradient one. This showed that with the same vegetation, the gentler the slope was, the larger the runoff stagnation ability it has. In the case of ignoring the effects of evaporation and vegetation absorption, for the four buffer strips with different gradients, the records of water level in the outlet with gentler slopes were less, and the peak of water level was also smaller, so the calculated runoff volume was smaller, and the corresponding seepage volume was larger. Without vegetation, the runoff volume of control tests was more stable than other buffer strips, at between 3.35 m³ and 3.60 m³. The mean runoff volume of the 2% gradient slope was 2.47 m³, and the ratio of runoff volume to that of seepage volume was 1.61. While the runoff volume of the 5% gradient slope was 3.12 m³, and its distribution ratio of runoff and seepage volume was 3.54. Meanwhile, the difference in runoff volume between every two gradient buffer strips was significant (P < 0.05), except between the 4% and 5% gradient ones (Figure 5). Young et al. (1980) also investigated that vegetated buffer strips (21.3 meters and 27.4 meters width) on a 4% gradient slope reduced runoff and total solids transport by 67% and 79% from a feedlot. This showed gentler gradient buffer strips can improve the hydraulic permeability of soil, while slowing down the runoff substantially.

The removal loads of TN, NH₄⁺-N and TP through different buffer strips revealed that vegetated buffer strips had better N, P removal efficiency for agricultural runoff than control strips (Figure 6). N, P removal efficiency is calculated according to the removal loads. The total TN, NH₄⁺-N and TP removal ratios of the three vegetated buffer strips increased by 237%, 268% and 274% compared with control ones. Young et al. (1980) found that vegetated buffer strips on a 4% gradient slope reduced total N and P by an average of 84% and 83%, respectively. Schwer & Clausen (1989) indicated that the effectiveness of filter strip treatment is governed by hydraulic loading rate, and nutrient uptake rate by the plants were almost 2.5%, 15% in P, and N respectively. In our study, buffer strips planted with Cynodon dactylon had the highest TN, NH₄⁺-N and TP removal ratios, at 45.7%, 50.0% and 43.4%, respectively. This was mainly because of its high stagnation capacity for runoff and seepage improvement (Bai et al. 2016). With outstanding ability for nitrogen fixation, Trifolium repens also had higher TN and NH₄⁺-N removal efficiency than the strips with the other two plants and the strip without any vegetation, with TN and NH₄⁺-N removal ratios of over 40% respectively.

Seepage has a better removal efficiency than that of runoff for agricultural N, P pollutants. Seepage total removal ratios were 73.6%, 66.9%, and 73.9% of TN, NH₄⁺-N and TP, respectively. The ratios of average seepage removal load to runoff were 2.79, 2.02 and 2.83 for TN, NH₄⁺-N and TP, which indicated that seepage had higher degradation capacity for TN and TP than NH₄⁺-N. For N, P removal loads of different buffer strips, the contrast average ratios of seepage and runoff were also different, and were 2.32, 2.11, 1.79 and 3.96. The N and P of runoff adsorbed in the soil particles. As the runoff resistance increases, the flow velocity decreases, the particles in the runoff precipitate, and the granular nitrogen is then fixed in the buffer zone (Messer et al. 2012). The deep decaying layer and loose soil structure in the buffer zone are conducive to the infiltration of runoff, promote the vertical migration of dissolved nitrogen with water infiltration, and reduce the transport capacity of runoff for soluble nitrogen (Schoonover et al. 2005). When N and P in the soil diffuse to the root region of the plant, the root system will assimilate N and P for its growth (Lowrance et al. 2000). When farmland runoff flows through the buffer zone, soil particles and colloids can absorb most ammonia nitrogen. Ammonia nitrogen becomes binding nitrogen and is preserved in the soil, while nitrate easily leaches into groundwater (Kaushal et al. 2011). On the contrary, the control group, without vegetation interception, had a high pollutant removal ratio for seepage to that of runoff, which can reach as high as 4.01–5.67 (5.67 for TN, 4.01 for NH₄⁺-N and 4.22 for TP). Compared to the control group, vegetated buffer strips had a higher pollutant removal ratio due to the interception of pollutants when runoff flowed through plants (Salazar et al. 2015), while seepage removed pollutants by holding up and absorbing them through soil particles and colloids (Kaushal et al. 2011). Therefore, the control group still had the function of interception and absorption. However, due to the absence of plants, its pollutant removal ratio by seepage was much higher than that by runoff. The total removal loads of TN, NH₄⁺-N and TP per unit area were calculated on different buffer strips. On the same gradient of 2%, the buffer strips with Cynodon dactylon had the highest amount of 0.661 g/m², 0.672 g/m² and 0.044 g/m² for TN, NH₄⁺-N and TP respectively. The buffer strips with Trifolium repens followed, with removal loads per unit area of TN and NH₄⁺-N both over 0.6 g/m² while the load of TP was 0.035 g/m² due to nitrogen fixation.

Among the vegetated buffer strips in the study, because of its thick stolons covering the whole strip, its runoff stagnation ability as well as the effective increase of seepage volume, Cynodon dactylon performed the best at N, P removal efficiency and pollutant load interception per unit area. The result matched the conclusions that the advantages of vegetated buffer strips are to stagnate runoff sufficiently, increase the soil hydraulic permeability, intercept, store, absorb and transform non-point pollutants of the surface runoff (Huang et al. 2009; Miao et al. 2013; Hille et al. 2018; Stutter et al. 2019). Although the dry weight of Cynodon dactylon was heavier than that of Trifolium repens, the latter had more nodules but fewer stolons in the root than the other two plants. Through abundant rhizobium that can transform nitrogen molecules into inorganic nitrogen absorbed by the plants, the root gave Trifolium repens an outstanding nitrogen fixation ability and efficiency as well (Stewart & Siciliano 2015), so the buffer strips containing this also had a good removal effect for nitrogen.

The contrast relation between removal load of TN, NH₄⁺-N and TP by runoff and seepage along slopes with different gradients was analyzed (Figure 7). In terms of load reduction of pollutants, seepage performed much better than runoff for all buffer strips. Among the four slopes with different gradients, the average contrasts for TN, NH₄⁺-N and TP removal load of seepage to that of runoff were 2.19 1.66 and 2.10. Gradients also made a difference. On the gradients of 2%, 3%, 4% and 5%, the ratios were 2.32, 2.15, 1.82 and 1.64. The smaller the gradient was, the larger the ratio was, which was inverse to the volume ratio of runoff and seepage. Gentler slopes enabled runoff to achieve better seepage, and seepage had a better N, P removal capacity than that of runoff.

The lower the gradient, the higher the total removal ratio of pollutants of buffer strips (Wu 2011; Hille et al. 2018). On the 2% gradient slope, TN, NH₄⁺-N, TP removal ratios were the highest, which were 45.65%, 50.05% and 43.43%, while buffer strip with 5% gradient had the lowest ratios, which were 29.04%, 32.83% and 27.13%, respectively. After comparison and analysis of the pollutant removal efficiency, it was found that the average removal ratios of TN, NH₄⁺-N and TP by seepage on 2%, 3%, 4% and 5% gradient slope buffer strips accounted for 71.66%, 68.14%, 64.39% and 61.93% of the total removal load.

The total pollutants removal loads per unit area were calculated by the buffer strips areas. 2% gradient buffer strip had the highest for TN, NH₄⁺-N and TP, which were 0.66 g/m², 0.67 g/m² and 0.04 g/m², while 5% slope had the lowest, which were 0.42 g/m², 0.38 g/m² and 0.03 g/m².

The study showed that steeper buffer slope can cause lower total removal efficiency and removal loads. On the four typical slope buffer strips, 2% gradient slope buffer strip had the highest TN, NH₄⁺-N and TP removal ratios, which were 15% higher than those of 5% gradient. And the average pollutants removal loads per unit area of 2% gradient slope buffer strip was also 68.42% higher than that of 5% gradient. The results aligned with the two conclusions that the lower the gradient was, the larger the seepage water volume was and the pollutants removal capacity of seepage was significantly higher than that of the runoff (Wu 2011; Hille et al. 2018).

In terms of the results from buffer strips with Cynodon dactylon on different slopes, the volume ratio of the runoff to that of seepage and three pollutants' removal loads per unit area all had significant exponential correlation under 0.99 confidences by SPSS analysis. The function curve was shown in Figure 8. The result indicated that N, P removal loads of seepage were higher than that of runoff in spite of the seepage volume being less than the runoff volume. The main reason was that seepage pollutants passed through soil filtration, root absorption, microbial degradation, etc. The ability of the riparian zone and buffer zone to intercept runoff pollutants and improve surface water quality has been recognized by many previous studies. Mankin et al. (2007) found that nitrogen removal efficiency in surface runoff was closely related to surface runoff permeability. Yin & Lan (1995) found that the formation of root pores in soil in the buffer zone was beneficial to the enhancement of nitrogen filtration and the expansion of adsorption capacity. Bu et al. (2015) concluded that poplar-grass composite buffer zone in Taihu Lake area (China) can significantly reduce the amount of sediment and nitrogen loss in farmland surface runoff, and the loss flux decreased with the increase of planting density in the buffer zone. Moreover, the types and forms of pollutants from surface runoff had a great influence on the interception effect, and the interception effect of particulate nitrogen was greater than that of soluble nitrogen. It showed that higher soil hydraulic permeability of buffer strips could effectively improve the removal capacity for non-point pollutants (Wu 2011; Haukos et al. 2016; Li et al. 2016; Hille et al. 2018; Stutter et al. 2019). The N, P pollutants removal load of seepage of buffer strips was larger than that of runoff, due to soil interception, soil filtration, microbial degradation and root absorption. Therefore, vegetation and gentler slope buffer strips stagnated runoff and improved soil hydraulic permeability effectively, which can improve the overall removal capacity of the N, P pollutants from agricultural runoff water.

Vegetated buffer strips had a better stagnation capacity for runoff than strips without any cover and could also enhance soil hydraulic permeability effectively. For a 2% gradient slope, the mean initial outflow time of runoff in the buffer strips planted with Cynodon dactylon, Trifolium repens and Festuca arundinacea was 7.1 minutes longer than the control group, while the average seepage volume was 2.12 times larger. Runoff stagnation capacity of buffer strips varied with the gradient. Comparing 2% and 5% gradient slopes planted with the same Cynodon dactylon, the initial outflow time of runoff along the 2% slope delayed 6.3 minutes while the seepage volume was 1.74 times higher than that of the 5% one. Buffer strips with 2% gradient and planted with Cynodon dactylon had the highest TN, NH₄⁺-N, TP removal loads per unit area, which were 0.661 g/m², 0.672 g/m² and 0.044 g/m² respectively.

In the research and practice of buffer zone for controlling non-point pollution, due to the limitation of land resources and slope structure, it can't build wide enough buffer strips to improve the purification effect on nitrogen and phosphorus pollutants. However, according to the difference of purification effect of runoff and seepage on pollutants, the soil infiltration capacity of buffer zones can be improved from the aspects of vegetation selection, soil improvement and soil animal cultivation, so as to improve the removal capacity of pollution in buffer zone for nonpoint pollution form farmland.

This study was supported by funding from Natural Science Foundation of China (No. 51679141 and No. 51979168).

J.W. conceived and designed the experiments. J.W. and C.S. performed the experiments. L.X. analyzed the data. J.W., C.S. and L.X. wrote the manuscript, and all authors provided editorial advice and reviewed the manuscript.

Competing financial interests: The authors declare no financial and non-financial competing financial interests.