Introduction

Increasing urbanisation worldwide has led to population explosion and changes in land use. Owing to the replacement of vegetation and permeable soil by impervious cover, a considerable amount of concrete floor constructed in urban areas causes stormwater runoff to easily collect pollutants (Al Mamoon et al., 2019; Chong et al., 2012; Muller et al., 2020; Silva & da Silva, 2020). Urban stormwater runoff containing large quantities of contaminants, such as organic matter, phosphorus, and nitrogen, is considered to be an important pathway for the delivery of contaminants to urban aquatic ecosystems (Al Mamoon et al., 2019). Studies have investigated stormwater runoff pollution from impervious surfaces since the 1980s (Ballo et al., 2009; Chow & Yusop, 2014; Gromaire-Mertz et al., 1999; Kim et al., 2007; Myers et al., 1982; Yang & Toor, 2016). Yan et al. (2019) indicated that stormwater runoff was regarded as the largest source of total nitrogen, chemical oxygen demand (COD), and ammonium nitrogen (NH4+-N) in the Taihu Basin, China. Silva et al. (2019) demonstrated that increasing impervious areas in the catchment enhanced the nutrient inputs from stormwater runoff, carrying total suspended solids (TSS), total phosphorus, and nitrate (NO3−) into Lake Pampulha, Brazil, at the beginning of the wet season.

Nitrogen pollution loads account for a large proportion of pollution in urban stormwater and contribute to the degradation of urban water quality, especially owing to algal blooms and eutrophication (Carey et al., 2013; Yang et al., 2020). For example, it was reported that stormwater from impervious surfaces contributed 80% of dissolved N to urban rivers in Melbourne, Australia, contributing to the risk of water eutrophication (Taylor et al., 2005). Further, in recent years, eutrophication in urban aquatic ecosystems has hindered the sustainable development of cities in China (Li et al., 2019a). Therefore, it is important to identify N sources and to study their transport to minimise the transport of N by urban stormwater runoff to urban aquatic ecosystems.

Stable nitrogen and oxygen isotopes of NO3− (δ15N-NO3− and δ18O-NO3−) have been widely used to identify NO3− sources and reveal N transport and transformation in aquatic ecosystems due to an unique isotopic signature of each NO3− source (Chen et al., 2021; Liu et al., 2014, 2021; Ma et al., 2015; Margalef-Marti et al., 2021; Peng et al., 2012; Yuan et al., 2019; Yue et al., 2020). Generally, δ15N-NO3− values range from 0.0‰ to + 25.0‰ for soil and organic N, -13.0‰ to + 13.0‰ for atmospheric deposition, -6.0‰ to + 6.0‰ for NH4+ fertiliser and NO3− fertiliser (Dong et al., 2021; Kendall et al., 2007; Xue et al., 2009). In recent years, with the development of technology, the nitrogen and oxygen isotopes combined with a Bayesian isotope mixing model (SIAR model) have been successfully applied to clarify the proportions of different N sources in surface water, groundwater, and surface runoff (Baral et al., 2018; Jani et al., 2020; Liu et al., 2021; Weitzman et al., 2021; Yang & Toor, 2016, 2017). Dong et al. (2021) investigated the values of δ15N-NO3− and δ18O-NO3− in a typical subtropical agricultural watershed and identified that nitrification of NH4+ mineralized from soil N and manure/sewage were the major sources of NO3− in stormwater runoff, accounting for 37–52% and 25–47% of the NO3− load, respectively. It was found that the δ15N-NO3− values ranged from -11.5‰ to + 4.9‰ and that atmospheric deposition contributed 43–71% of NO3−, followed by chemical fertilisers (< 1–49%) in urban residential stormwater runoff in Florida (Yang & Toor, 2016). On the other hand, water isotopes (δD-H2O and δ18O-H2O) can reveal the origin of water sources since different sources of water have distinct isotopic signatures (Rahal et al., 2021; Weitzman et al., 2021). Numerous studies have indicated that the combination of stable isotopes of NO3− with water isotopes can further enhance the ability to identify NO3− sources and its transformation processes in aquatic ecosystems (Gómez-Alday et al., 2022; Hu et al., 2019; Pastén-Zapata et al., 2014). For example, the δD-H2O and δ18O-H2O values of river water, groundwater, and rainfall in the Yongan watershed of eastern China suggesting a substantial portion of river water may originate from groundwater and subsurface water sources, and the δ15N-NO3− and δ18O-NO3− values with SIAR model indicated that groundwater (43 ± 17%), soil N (33 ± 8%), and wastewater (25 ± 15%) were the dominant river NO3− sources (Hu et al., 2019).

In recent years, urban residential areas have been increasingly constructed due to the boom in the real estate industry in China, leading to an increase in impervious surfaces within residential areas, and accordingly, an increase in stormwater runoff carrying pollutants directly into urban rivers (Li et al., 2019b; Wen et al., 2019). Compared to China, more studies on stormwater runoff in residential areas have been conducted in other countries, whereas studies on stormwater runoff in China have focused more on non-point source pollution in agricultural areas (Sui et al., 2020). In residential areas, impervious surfaces mainly comprise roads and roofs, and the different types of human activities on roads and roofs lead to the corresponding differences in characteristic pollutants (Yang & Toor, 2017). For instance, Kojima et al. (2011) demonstrated that N in road dust mainly originated from fertilisers and soil, while the N source in roof dust originated from atmospheric deposition.

In this study, stormwater runoff was collected from roads and roofs in an urban residential area, and the concentration of different forms of N and stable isotopes of NO3− (δ15N-NO3− and δ18O-NO3−) were measured. The objectives were to evaluate the spatial and temporal distributions of the different forms of N and to quantify the major N sources in urban residential stormwater runoff. This study serves as a guideline for generating more effective mitigation strategies to reduce the concentration of N in urban aquatic ecosystems.

Materials and methods

Study site

The residential study area (30°17′N, 120°9′E) is located in Hangzhou, which is the political, economic, and cultural centre of Zhejiang Province, East China (Fig. 1). The continuously growing population in Hangzhou has increased from 8.89 million in 2014 to 10.36 million in 2019, and the urbanization rate of Hangzhou reached 78.5% in 2019 (HZSB, 2020). The residential area comprises approximately 12.7 ha, of which green land accounts for 27% and the rest are impervious surfaces (rooftops: 36%, driveways and sidewalks: 17%, and roads: 20%). The age of residential buildings in the study area is more than 30 years with 2525 households, and the area is served by a separate sewer system. We collected the roof runoff at the top of a five-storey residential building (approximately 15 m in height) which featured a flat and open-area cement roof without slope and any shelter. Rainwater on the roof is easily evacuated through the drainage ditch. Road runoff was collected on a residential road of bituminous concrete which featured tree, grass, and other plants planted on both sides of the roads. The overall study region is characterised by a subtropical monsoon climate with an annual average temperature of 15.7 ℃ and annual average rainfall of 1454 mm, of which 67% occurs during the wet season (March–September). The average annual rainfall was 1584.7 mm from 2019 to 2020, and monthly rainfall ranged from 37.7 to 374.7 mm in the wet season in Hangzhou (HZSB, 2020). The study period was from March to August in 2019 and 2020. And the urban runoff in this catcment finally enters into the Shangtang River, which is a tributary of the Grand Canal (Hangzhou).

Fig. 1
figure 1

Sampling sites in the study area

Full size image

Sampling and analysis

A weather application (Hangzhou weather network) showing the evolution of storm events was used to track storm events in the study area. Stormwater runoff samples were collected manually after identifying a major storm event. An acid-washed polyethylene container was used to gather runoff samples before the runoff flowed into the rain drainage systems. The roof and road runoff were collected at the top of the five-storey residential building (Site A) and on the road in the residential area (Site B), respectively (Table 1). At each catchment site, samples were collected at 5 min intervals within 1.5–2 h during each storm and were placed in a 500 mL plastic bottle. The stormwater runoff samples were collected during seven storm events, leading to a total of 321 samples (165 roof and 156 road runoff samples) (Fig. 1). After sampling, a portion of the samples was filtered through 0.45 μm membrane filters (Whatman) into 100 mL acid-washed polyethylene bottles on the sampling day and stored in a refrigerator at -20 ℃ until the analysis of dissolved total nitrogen (DTN), major ions (NH4+, NO2− and NO3−), and isotopes. The other portion was not filtered and stored at -4 ℃ for TN and COD analysis within 48 h.

Table 1 Information on runoff sampling in an urban residential area from 2019 to 2020
Full size table

TN and DTN were measured using the alkaline potassium persulfate digestion ultraviolet spectrophotometric method (HJ636-2012). Total suspended solids (TSS) and COD were analysed using the gravimetric method (GB11901-89) and the fast digestion-spectrophotometric method (HJ/T399-2007), respectively. The concentrations of NH4+, NO2−, and NO3− were measured using ion chromatography (Dionex ICS-900), and the detection limits of NH4+, NO2−, and NO3− were 0.03, 0.02, and 0.02 mg/L, respectively. Additionally, replicates were used for the determination of each sample, including TN, DTN, NH4+, NO2−, and NO3−, TSS, COD, δD-H2O, δ18O-H2O, δ15N-NO3− and δ18O-NO3−. The NO2− concentrations in most of the samples were below the detection limit (BDL); therefore, dissolved inorganic nitrogen (DIN) was defined as the sum of NH4+-N and nitrate nitrogen (NO3−-N). The dissolved organic nitrogen (DON) and particulate nitrogen (PN) were calculated using the following mass balance: [DON] = [DTN]—[DIN] and [PN] = [TN]—[DTN].

Twenty-two road runoff samples were analysed for hydrogen and oxygen isotopes of water (δD-H2O and δ18O-H2O) using an isotopic water analyser (Picarro L2140-i). The precisions of δD-H2O and δ18O-H2O were ± 0.5‰ and ± 0.1‰, respectively. Thirty-two roof runoff and thirty-five surface runoff samples were analysed for stable nitrogen and oxygen isotopes of NO3− (δ15N-NO3− and δ18O-NO3−) which were measured following the bacterial denitrification method (Kaiser et al., 2007; McIlvin & Casciotti, 2011). Nitrite (NO2−) in the samples was reduced to nitric oxide (NO) using ascorbic acid (pH < 3.5), and then, the NO produced was continually degassed with an inert gas (helium) during the reaction. In brief, denitrifying bacteria (Pseudomonas aureofaciens) lacking gaseous nitrous oxide (N2O), converted NO3− from the water samples to N2O through reductase activity. Then, the N2O was stripped by helium carrier gas and thermally decomposed to N2 and O2. The isotopic ratios of N2 (15 N/14 N) and O2 (18O/16O) were measured by a mass spectrometer (Thermo Delta V Advantage). The analytical errors for δ15N-NO3− and δ18O-NO3− were ± 0.3‰ and ± 0.5‰, respectively. In this paper, isotopic results are expressed as δ values (per mil unit), such that

$$\delta(^0/_0{_0})=1000 \times \left(\frac{{\mathrm{R}}_{\mathrm{sample}}}{{\mathrm{R}}_{\mathrm{standard}}}-1\right),$$

where R is the isotopic ratios (15 N/14 N, 18O/16O, and D/H). The standard of 15 N/14 N was atmospheric air (AIR), and the collective standard of 18O/16O and D/H was the Vienna Standard Mean Ocean Water (VSMOW).

SIAR model

The proportions of the different NO3− source contributions were evaluated by the SIAR model. The SIAR model can be expressed as follows (Parnell et al., 2010):

$${X}_{ij}={\sum }_{k=1}^{K}{p}_{k}\left({S}_{jk}+{C}_{jk}\right)+{\varepsilon }_{ij};$$
$${S}_{ij}\sim N\left({\mu }_{jk},{\omega }_{jk}^{2}\right);$$
$${C}_{ij}\sim N\left({\lambda }_{jk},{\tau }_{jk}^{2}\right);$$
$${\varepsilon }_{ij}\sim N\left(0,{\sigma }_{j}^{2}\right);$$

where Xij is the isotope value j of the mixture i (i = 1, 2, 3,…, N, and j = 1, 2, 3,…, J); Sjk is the isotope value j in source k (k = 1, 2, 3,…, K), which conforms to a normal distribution with a mean of μjk and a standard variance of ωjk2; Cjk is the fractionation coefficient for isotope j for source k, which is a standard distribution with a mean value of λjk and a variance of τjk2; and εij is the residual error, which is used to characterise the remaining unquantified variation between the individual mixed samples and is a normal distribution with a mean value of 0 and a variance of σj2. The variable pk is the proportional contribution of source k, estimated using the SIAR model.

Results and discussion

N concentration and N forms in stormwater runoff

The concentrations of TN, DTN, NH4+-N, NO3−-N, PN and DON in roof runoff and road runoff for 2019–2020 are shown in Table 2. The concentrations of TN, NH4+-N, and NO3−-N in the roof runoff of seven storm events ranged from 0.26–6.13 (mean 1.23 mg/L), 0.03–1.96 (mean 0.42 mg/L), BDL–1.69 (mean 0.26 mg/L), respectively. The concentrations of TN, NH4+-N, and NO3−-N in the road runoff of the same seven storm events were higher than those in the roof runoff (TN: 0.65–12.7 (mean 3.76 mg/L), NH4+-N: BDL–5.28 (mean 0.50 mg/L), and NO3−-N: BDL–3.05 (mean 0.50 mg/L)). The concentrations of PN and DON in roof and road runoff also showed the same trend as those of TN, NH4+-N, and NO3−-N. In addition, the concentrations of TN, PN, and DON in road runoff were more than three times higher than as those in the roof runoff in the same storm event, proving the important influence of surrounding land use on N in stormwater runoff (Table 2). Road runoff is an important source of nonpoint pollution in urban aquatic ecosystems because the N compounds within the road dust, chemical fertiliser, soil materials, pet waste, and leaf litter on road surfaces are eventually washed into stormwater runoff (Janke et al., 2017; Lusk & Toor, 2016; Yang & Toor, 2017). In contrast, human activities had relatively little impact on the roof runoff, while atmospheric deposition and organic N, such as bird and rodent droppings, were the main N sources on the cement roof surfaces (Song et al., 2019). Moreover, the TN concentration was dominated by PN and DON in road runoff, while NH4+-N and NO3−-N were the major N compounds in roof runoff (Table 2). The results were consistent with those of Vaze and Chiew (2004), who found that the proportion of DTN in TN ranged between 20 and 50%, and proportion of PN in TN ranged from 50 to 80% in road runoff in Australia. It was suggested that roads were more likely to accumulate particulate matter than roofs, and that the particulate matter in the soil on the sides of the roads was easily mobilised and transported to the impervious surface through stormwater runoff. Previous studies have also documented that climatic parameters such as the frequency and intensity of storms and antecedent rainfall conditions have significant impacts on the concentrations of N in stormwater runoff, leading to a wide range of N concentrations in stormwater runoff (Yang & Lusk, 2018). The event mean concentration of TN average value in road runoff in this study (3.61 mg/L) was higher than that in road runoff in humid subtropical urban residential areas in Tampa, Florida, (TN: 0.42 mg/L) and much lower than that in a study conducted on road runoff in a semi-arid urban residential area in the Aliso Creek watershed, California (TN: 10.85 mg/L) (Toor et al., 2017; Yang & Toor, 2017).

Table 2 Statistical parameters of different N forms, TSS, COD, and the NO3−-N/NH4+-N ratios in roof runoff and road runoff in an urban residential area from 2019 to 2020
Full size table

Significantly positive correlations (P < 0.01) were found between antecedent dry weather period and TN, DTN, NH4+-N, and DON concentrations, both in roof runoff and road runoff, revealing that N concentrations in stormwater runoff are linked to the duration of dry periods before storm events (Table S1). Lewis and Grimm (2007) also reported high NH4+-N concentration in stormwater runoff in arid urban catchments after a longer antecedent dry period. Higher concentrations of TN, DTN, NH4+-N, and DON were observed in roof and road runoff during storm event 2, as compared to those in other storm events. This is attributed to the longer antecedent dry weather period (13 days) preceding 13 May 2019. As pollutants accumulate on impervious surfaces during antecedent dry periods, the stormwater runoff after a longer antecedent dry weather period carries more pollutants (Lewis & Grimm, 2007; Li et al., 2007a; Zhi et al., 2018). The PN and DON concentrations in the road runoff were higher in May than those in the other months (Table 2). Furthermore, higher concentrations of PN corresponded to the higher concentrations of TSS in road runoff (Table 2). Similarly, higher concentrations of DON were consistent with the higher concentrations of COD in road runoff (Table 2). Significantly positive correlations (P < 0.01) were found between TSS and PN, COD, and DON concentrations, in road runoff (Table S1). This may be attributed to extensive plant growth during May, which is a warm and humid month in late spring. During this time, organic matter, including leaf litter, flower debris, pollen, and seeds, is expected to reach the ground and eventually enter into stormwater runoff, increasing the organic N load of the road runoff in the urban residential area. These results are in agreement with Janke et al. (2017), who found that seasonal peaks of N in urban stormwater runoff coincided with spring leaf-out and flowering.

The mean concentrations of NO3−-N were lower than the mean concentrations of NH4+-N in roof runoff, while the opposite was true for in road runoff (Table 2). NO3−-N concentrations in rainwater of Hangzhou were lower than NH4+-N concentrations, and the mean NO3−-N/NH4+-N ratio was found to be 0.87 from 2015 to 2017 (Jin et al., 2019). In this study, all mean NO3−-N/NH4+-N ratios in the roof runoff of the seven storm events were lower than 1.0, indicating that atmospheric deposition was the dominant N source in the roof runoff. The NO3−-N/NH4+-N ratios in road runoff varied widely (0.37–2.33). In addition to atmospheric deposition, it was found that anthropogenic inputs, such as chemical fertilisers, soil materials, pet waste, and leaf litter, also influenced the NO3−-N and NH4+-N concentrations in road runoff. NO3−-N may have been preferentially washed out from surface deposits, and NH4+-N could have been easily absorbed by surface deposits (Kojima et al., 2011; Wang et al., 2019). However, the NO3−-N/NH4+-N ratios in road runoff in May (storm events 2 and 5) were much lower than those in other months (Table 2). On the one hand, due to the increased air temperature and humidity in May, the activities of microbes were likely enhanced; therefore, the mineralisation of tree branches, leaf litter, flower debris, and soil organic N was likely a significant contributor to the NH4+-N in road runoff. On the other hand, low rainfall in May would not have been conducive to NO3−-N export in stormwater runoff (Kaushal et al., 2014). The above two reasons could explain why the NO3−-N concentrations were lower and the NH4+-N concentrations were higher in road runoff in May than in other months.

The temporal variation of N forms in roof and road runoff during the seven storm events is shown in Fig. S1. The concentrations of TN, NH4+-N, and NO3−-N in roof and road runoff were higher in the beginning of the storm event and then gradually stabilized with an increase in rainfall duration. Climate variables such as the frequency and intensity of storms are important factors that influence N transport in stormwater runoff, and NO3−-N exports in stormwater runoff increased during strong storms (Kaushal et al., 2014; Li et al., 2015). For example, the comprehensive wind and rain intensity index of Super Typhoon Lekima was 158.6, and the rainfall amount during the sampling period was 19.1 mm on 9 August 2019 (NMC, 2019). The onrush of water that accompanied strong winds during Super Typhoon Lekima caused a strong scouring effect on the land. As a result, the temporal variations in NO3−-N concentrations in the road runoff of storm event 4 (9 August 2019) showed significant fluctuations. Owing to the low rainfall, the temporal variations of NO3−-N concentrations in road runoff during storm events 2 and 5 were maintained at a low level. The antecedent conditions could have also affected N transport in the stormwater runoff (Lee et al., 2002; Taebi & Droste, 2004). Thus, it is considered that the temporal variations of TN, NO3−-N, and NH4+-N concentrations in the roof runoff of event 7 were more stable owing to a lower rainfall amount and relatively short antecedent dry weather period.

Water sources of stormwater runoff

Unlike roof runoff, road runoff can originate from a combination of various water sources, such as rainfall and the overflow of sewer system in urban residential areas. Therefore, water isotopes were used to identify the water sources in the road runoff. The values of δD-H2O ranged from -44.3‰ to -26.1‰ (mean -34.7 ± 7.6‰), and the δ18O-H2O values ranged from -6.5‰ to -4.3‰ (mean -5.4 ± 0.9‰) in road runoff in storm events 3 (n = 11) (Fig. 2 and Table 3). The values of δD-H2O ranged from -72.2‰ to -67.8‰ (mean -70.4 ± 1.4‰), and the δ18O-H2O values ranged from -10.5‰ to -9.6‰ (mean -10.1 ± 0.3‰) in road runoff in storm events 7 (n = 11) (Fig. 2 and Table 3). The relationship between δ18O-H2O and δD-H2O in road runoff could be described as δD-H2O = 7.6 δ18O-H2O + 6.2 (R2 = 0.99), which was remarkably close to the local meteoric water line (LMWL: δD-H2O = 8.4 δ18O-H2O + 17.5) and the global meteoric water line (GMWL: δD-H2O = 8.0 δ18O-H2O + 10.0) (Jin et al., 2021a; Craig, 1961). It was suggested that water in the road runoff mainly originated from local rainwater in the study area. The slope and intercept of this isotopic line for the road runoff samples were lower than those of the LMWL and GMWL, implying that slight evaporation occurred during runoff generation. Owing to the high temperature during the study period and the impervious surface in the study area (73%), evaporation was likely to take place as runoff travelled over the impervious surface. (see Tables 4 and 5)

Fig. 2
figure 2

Relationship between δ18O-H2O and δD-H2O for road runoff (n = 22)

Full size image
Table 3 The δ15N-NO3− and δ18O-NO3− values and the δD-H2O and δ18O-H2O values in the roof runoff and road runoff and the δ15N-NO3− and δ18O-NO3− values of NO3− sources used in SIAR
Full size table
Table 4 The concentrations of N, TSS, COD, Major ions (NH4+, NO2−, NO3−) in individual 7 stormwater runoff events
Full size table
Table 5 The isotope values (δ18O-NO3−, δ15N-NO3−, δ18O-H2O, δD-H2O) of runoff in a residential catchment of Hangzhou, Zhejiang
Full size table

NO3 − sources in stormwater runoff

Identifying NO3 − sources and NO3 − transport in stormwater runoff

The δ18O-NO3− values in roof runoff varied from + 49.1‰ to + 68.6‰ (mean: + 57.5 ± 6.3‰) in storm events 3 (n = 10), from + 39.9‰ to + 53.1‰ (mean: + 46.5 ± 3.5‰) in storm events 4 (n = 11), and from + 72.3‰ to + 79.8‰ (mean: + 75.7 ± 2.4‰) in storm events 7 (n = 11) (Fig. 3 and Table 3). The δ18O-NO3− values in road runoff varied from + 21.9‰ to + 50.2‰ (mean: + 31.2 ± 10.1‰) in storm events 3 (n = 13), from + 15.6‰ to + 36.6‰ (mean: + 22.2 ± 5.8‰) in storm events 4 (n = 11), and from + 65.4‰ to + 74.3‰ (mean: + 70.0 ± 2.7‰) in storm events 7 (n = 11) (Fig. 3 and Table 3). Temporal variations of δ18O-NO3− in stormwater runoff during storm events 3, 4, and 7 are shown in Fig. 4. The temporal variations of δ18O-NO3− in roof runoff were limited, and the δ18O-NO3− values in the roof runoff were higher than those in road runoff. In general, δ18O-NO3− values range from + 25.0‰ to + 75.0‰ for atmospheric deposition, + 17.0‰ to + 25.0‰ for NO3− fertilisers, -5.0‰ to + 15.0‰ for NO3− derived from nitrification (Kendall et al., 2007; Xue et al., 2009; Yue et al., 2020). The δ18O-NO3− values of rainwater and dry deposition in Hangzhou in wet season were + 31.5– + 71.6‰ (mean: + 57.4‰), and + 24.5– + 79.2‰ (mean: + 62.5‰), respectively (Jin et al., 2021b; Jin et al., 2019). The high δ18O-NO3− values in roof runoff were in the range of δ18O-NO3− values of global atmospheric deposition and similar to the δ18O-NO3− values of rainwater and dry deposition in Hangzhou, implying that the atmospheric deposition (both dry and wet) was the dominant NO3−-N source in roof runoff (Fig. 3). The higher values of δ18O-NO3− in roof and road runoff for storm event 7 than those in other events resulted from the influence of the high δ18O-NO3− values of atmospheric deposition (Fig. 4). The δ18O-NO3− values in road runoff were lower than those in roof runoff and were further from the δ18O-NO3− values in rainwater and dry deposits in Hangzhou, indicating that NO3−-N from other sources was carried with the urban stormwater runoff.

Fig. 3
figure 3

δ15N-NO3− and δ18O-NO3− values of stormwater runoff in the urban residential catchment for storm events 3, 4, and 7

Full size image
Fig. 4
figure 4

Temporal variations of δ15N-NO3− and δ18O-NO3− in (a)–(b) roof runoff and (c)–(d) road runoff

Full size image

The δ15N-NO3− values ranged from -10.7‰ to + 1.4‰ (mean: -4.0 ± 3.3‰) in roof runoff, and from -8.6‰ to + 8.4‰ (mean: -2.9 ± 3.1‰) in road runoff (Fig. 3 and Table 3). The δ15N-NO3− values of rainwater (-4.4‰ to + 3.6‰) and dry deposition (-1.0‰ to + 5.8‰) in Hangzhou during the wet season were in the range of the δ15N-NO3− values in road runoff, suggesting that other NO3− sources contributed to road runoff in addition to the atmospheric NO3− (Fig. 3) (Jin et al., 2021b; Jin et al., 2019). The δ15N-NO3− values in roof and road runoff in the urban residential area in Hangzhou were higher than the δ15N-NO3− values in forest runoff (mean: -6.07‰) reported by Zhang et al. (2019) because chemical fertilisers with low δ15N-NO3− values were identified as the main NO3− source in forest runoff. The δ15N-NO3− and δ18O-NO3− values of roof and road runoff portrayed in Fig. 3 demonstrate that atmospheric deposition was the only NO3− source in roof runoff, while NO3− sources of road runoff mainly reflected a mixture of atmospheric deposition and chemical fertilisers during the study period. Previous studies have pointed out that chemical fertilisers contributed an average of 16–64% of NO3−-N in road runoff, and a large proportion of N inputs were from chemical fertilisers application for residential lawns and plants in urban residential areas (Muller et al., 2020; Riha et al., 2014; Yang & Toor, 2017). The estimated annual NPK (nitrogen–phosphorus–potassium) compound fertiliser application (containing NO3− fertiliser and NH4+ fertiliser) in urban green land was found to be 75–150 kg N/ha (two applications on average) in Hangzhou (Teaching Material Office of the Ministry of Labor & Social Security, 2005). Further research has suggested that NO3− derived from chemical fertilisers (NO3− and NH4+ fertiliser) had typical δ15N-NO3− values from -6‰ to + 6‰, and the δ18O-NO3− values in NH4+ fertiliser and in NO3− fertiliser were from -5.0 to + 15.0‰ and + 17.0 to + 25.0‰, respectively (Bateman & Kelly, 2007; Xue et al., 2009). Therefore, chemical fertilisers were one of the main NO3− sources, as the green coverage rate (27%) in the study area was high.

The temporal variations of δ15N-NO3− values in road runoff in the urban residential area in Hangzhou followed a consistent decreasing trend, whereas the temporal variations of δ15N-NO3− in roof runoff fluctuated up and down smoothly (Fig. 4). It was implied that NO3− sources in road runoff were more varied, which were not only derived from atmospheric deposition but also from chemical fertilisers, soil particles containing N, and organic N sources (pet waste, leaf litter, etc.), in comparison with those in roof runoff. Based on the variations of TN and NO3− concentrations and the δ15N-NO3− and δ18O-NO3− values in stormwater runoff, the first 10 min of the sampling period of each storm event was assumed to be the initial period of stormwater runoff. There are two possible reasons for the high δ15N-NO3− values in road runoff in the beginning (initial period of stormwater runoff, i.e. the first 10 min of sampling time). First, the elderly population (over 60 years old), accounting for 41.7% of the study area, led to high per-area rates of pet ownership (HZSB, 2020); therefore, pets such as dogs that were kept by retired persons may have excreted faeces on roads or green land in residential areas. For example, dog waste in one urban area of Minnesota has been found to contribute up to 28% of TN inputs (Hobbie et al., 2017). Second, the mineralisation of organic matter including leaf litter, flower debris, pollen, and seeds from trees, and other plants on the roadsides was also a NO3− source in stormwater runoff. Similarly, the highest δ15N-NO3− value of the first sample in roof runoff was also ascribed to organic N (bird and rodent droppings) on the roof surface during storm events 4 and 7. After the surface pollutants were washed away, the soil moisture gradually become saturated. Then, nitrified soil and chemical fertilisers were washed into the road runoff as a result of continuous rainstorms. Thus, low δ15N-NO3− values in road runoff were observed in the middle and late period of stormwater runoff (i.e. after the first 10 min of sampling time). As reported by Baral et al. (2018), the NO3− in stormwater runoff during smaller storms mainly originates from atmospheric deposition. In contrast, the NO3− contribution from atmospheric deposition may be lower than that from nitrified soil and fertiliser washed into the stormwater runoff during larger storms. Therefore, in this study, atmospheric deposition is thought to be the main NO3− source in road runoff of storm event 7 due to the small rainfall amount and relatively short antecedent dry weather period during the sampling period, which was confirmed by the lower δ15N-NO3− values and higher δ18O-NO3− values during storm event 7.

The δ18O-NO3− values are a useful indicator for identifying whether nitrification occurred during runoff. Theoretically, the δ18O-NO3− of nitrification is generated by one oxygen atom from oxygen in the atmosphere, and two oxygen atoms from water. The equation can be expressed as δ18O-NO3− = 2/3 (δ18O-H2O) + 1/3 (δ18O-O2) (Kendall et al., 2007). According to this equation, the observed δ18O-H2O values of water samples and δ18O-O2 (+ 23.5‰), it was expected that the theoretical δ18O-NO3− values in the road runoff from nitrification could range from + 0.8‰ to + 5.0‰, which was considerably lower than the obsreved δ18O-NO3− values in the road runoff in storm events 3 and 7 (Fig. 5). Therefore, the possibility of an instant nitrate production within the runoff water was eliminated. Generally, denitrification occurs when oxygen is limited and organic carbon is available in an aquatic ecosystem, where bacteria reduce NO3− to N2 or N2O. Heterotrophic microorganisms metabolise light isotopes (i.e. 14 N and 16O) in preference to heavy isotopes (i.e. 15 N and 18O) during denitrification (Kendall et al., 2007). Denitrification causes the δ15N-NO3− and δ18O-NO3− values of the residual NO3− to increase with a δ15N-NO3−/δ18O-NO3− ratio from 1:1 to 2:1 (Xue et al., 2009). In this study, no linear relationships between δ15N-NO3− and δ18O-NO3− were observed, suggesting that no instant denitrification occurred in the roof or road runoff in the urban residential area of Hangzhou.

Fig. 5
figure 5

Relationship between δ18O-H2O and δ18O-NO3− in the road runoff of storm events 3 and 7

Full size image

Estimating the contribution of NO3 − sources

According to the above analysis, two NO3− sources (atmospheric deposition and soil and organic N such as bird and rodent droppings) in roof runoff and four NO3− sources (atmospheric deposition, NO3− fertiliser, NH4+ fertiliser, and soil and organic N such as pet waste, leaf litter and soil N) were identified in road runoff in the urban residential area of Hangzhou. The contributions of NO3− in the urban residential stormwater runoff were estimated using the SIAR model. The δ15N-NO3− and δ18O-NO3− values of the NO3− sources were based on relevant literatures (Bedard-Haughn et al., 2003; Curt et al., 2004; Divers et al., 2014; Jin et al., 2019; Kendall et al., 2007; Li et al., 2007b; Widory et al., 2004; Yang & Toor, 2016), as shown in Table 3. We assumed Cjk = 0 in the SIAR model because of the absence of denitrification in the roof and road runoff in the study area. The contributions of NO3− sources to roof and road runoff are shown in Fig. 6 and Table S2. The NO3− contributions from atmospheric deposition (84–98%) were predominant, and the contributions from organic N were only 2.0–16% in roof runoff. In road runoff, atmospheric deposition (41% in storm event 3; 34% in storm event 4) contributed the most, while soil and organic N (6.0% in storm event 3, 12% in storm event 4) contributed the least, and NH4+ fertiliser (31% in storm event 3; 30% in storm event 4), and NO3− fertiliser (22% in storm event 3; 24% in storm event 4) were intermediate. The contribution of atmospheric deposition (92%) was dominant, followed by that of NO3− fertiliser (3.7%), NH4+ fertiliser (2.5%), and soil and organic N (1.8%) in road runoff during storm event 7. In this case, atmospheric deposition was an important contributor to stormwater runoff N in the urban residential area, which is similar to the findings of urban stormwater runoff in Florida, where 30–88% of NO3− was found to be from atmospheric deposition (Krimsky et al., 2021; Yang & Toor, 2016). Compared with the values in storm events 3 and 4 in this study, a significant increase in the contribution of atmospheric deposition occurred for roof and road runoff during storm event 7, reflecting the short antecedent dry weather period (1 day). Rain from the previous day would have washed away the N pollutants on the road surface. Accordingly, the NO3− contributions in road runoff from chemical fertiliser (NH4+ fertiliser and NO3− fertiliser), and soil and organic N were relatively low for storm event 7. Our results highlighted that chemical fertiliser (NH4+ and NO3− fertiliser) were the main NO3− source in road runoff (an average contribution of more than 50% in road runoff in storm events 3 and 4), owing to the application of chemical fertilisers for plant growth in urban residential areas. This is in agreement with the investigations in urban areas by Hale et al. (2014) and Krimsky et al. (2021). For example, chemical fertilisers contributed 44% of NO3− in stormwater runoff in the urban areas of Phoenix, Arizona (Hale et al., 2014). The NO3− contribution from soil and organic N was lower than other NO3− sources in road runoff in the urban residential area of Hangzhou. Soil erosion was mitigated by 27% of the green land and 73% of the impervious surface, and road sweeping was carried out on alternate days. Although soil and organic N generally have diverse origins, the quantities of soil particles, leaf litter, and pet waste, etc., that were washed into runoff were likely small.

Fig. 6
figure 6

Contributions of different nitrate sources in roof runoff and road runoff for storm events 3, 4, and 7

Full size image

The SIAR outputs revealed that NO3− contributions varied significantly between the initial period and middle and late periods in road runoff during a storm event (Fig. 7 and Table S2). The NO3− contributions were similar during the same period in storm events 3 and 4. Atmospheric deposition and chemical fertiliser were the primary N sources in both the initial period and middle and late periods in road runoff in storm events 3 and 4. Coupled with the continuing storm, the combination of atmospheric deposition and chemical fertilisers became more important, suggesting that longer duration storms were more likely to transfer N pollutants from urban green land or urban soils to stormwater runoff. Therefore, we find that during the initial period of stormwater runoff, the storm runoff generated from impervious surfaces was able to quickly wash off the soil, organic matter, and atmospheric dry-deposited NO3− on the impervious surface. Thus, the NO3− contributions from atmospheric deposition, soil and organic N were higher in the initial period than those in the middle and late periods in road runoff during storm events 3 and 4. Similarly, Lewis and Grimm (2007) revealed that frequent N transport by rain is easier in urban environments. The short antecedent dry weather period (1 day) in storm event 7 was therefore likely responsible for the higher NO3− contribution from chemical fertilisers in road runoff during the initial period as compared to the middle and late periods. Moreover, with a decrease in NO3− contribution from chemical fertilisers, and soil and organic N, the NO3− contribution from atmospheric deposition in road runoff increased dramatically during the middle and late periods for storm event 7.

Fig. 7
figure 7

Contributions of different nitrate sources in road runoff during different periods of the storm events. Ini.: initial period of stormwater runoff (first 10 min of sampling time); M.&L.: middle and late periods of stormwater runoff (after the first 10 min of sampling time)

Full size image

However, the overlapping in isotope values of NO3− sources and the isotopic fractionation effect in N transformation processes might affect the NO3− source apportionment by SIAR (Hu et al., 2021; Liu et al., 2018; Yu et al., 2020). In order to reduce the uncertainties and to improve the accuracy of SIAR outputs, the actual values of NO3− sources in the study area will be measured and the isotopic fractionation effect in N transformation processes in urban runoff will be considered in the future studies. In addition, it was found that our SIAR outputs about the NO3− contributions of the different sources at the different sampling time in the same sampling site have relatively large variations. Therefore, future studies can be paid attention to the temporal variability of NO3− isotopes in sampling sites.

Conclusions

The different forms of N and multiple isotopes (δD-H2O, δ18O-H2O, δ15N-NO3−, and δ18O-NO3−) in stormwater runoff were measured from 2019 to 2020 in a typical urban residential area in Hangzhou, East China. Based on the findings, N concentrations in road runoff were higher than those in roof runoff. The SIAR model showed that atmospheric deposition was the dominant NO3− source, contributing 84–98% of the NO3− in roof runoff in 3 storm events. Atmospheric deposition and chemical fertilisers were the major NO3− sources in road runoff in 3 storm events, with NO3− contributions from atmospheric deposition, NH4+-N fertiliser and NO3−-N fertiliser accounted for 34–92%, 2.5–31%, and 3.7–24%, respectively. The contributions of soil and organic N to NO3− in roof and road runoff were relatively low (1.8–16%). The antecedent dry weather period before storm event had a significant impact on NO3− in road runoff, and with the increased antecedent dry weather period the NO3− contribution of chemical fertilisers was dramatically increased. It was demonstrated that much of the NO3− in road runoff originated from impervious areas (soil and organic N) during the initial period of stormwater runoff. The results of this study suggest that it is necessary to take effective measures to optimise chemical fertilisers application and control its loss from urban green land. Frequent road sweeping and cleaning are useful in preventing soil and organic N from entering urban ecosystems. Reducing the amount of impervious areas is also essential to reducing the overall N load in urban ecosystems.