Assessing the characteristics of different-sized particles in the first flush of roof runoff

With rapid urbanization, polluted runoff from urban surfaces poses a severe threat to the environmental quality of surface water (Perera et al. 2019), which means that there is an urgent need to control urban runoff. The pollutant load in stormwater runoff is dominated by particulate pollutants (Sansalone & Cristina 2004; Zhang et al. 2021). Furthermore, polycyclic aromatic hydrocarbons, heavy metals, and other pollutants in urban runoff mostly exist in the particulate form and are attached to particulate matter of different sizes (Wang et al. 2020; Buyck et al. 2021; Qian et al. 2021).

Other researchers have examined the characteristics of particulate matter pollutants in urban runoff in previous studies (Chang et al. 2005; Mendez et al. 2011). Zhang et al. (2021) monitored runoff from an asphalt road and a cement roof in Beijing, and found that the suspended solids (SS) load in asphalt road runoff was significantly higher than that in cement roof runoff (p < 0.05), and that approximately 66.55 ± 9.44% of the phosphorus in the runoff was in the particulate form. Other researchers compared road runoff with roof runoff and found that, under the same rainfall conditions, roof runoff contained less pollutants than road runoff because of the influence of traffic and other factors (Hou & Zhang 2014; Zhang et al. 2021). Roof runoff is not affected by maintenance factors such as road sweeping, which represents the build-up and wash-off of pollutants from traffic.

The first flush is a commonly used term in urban runoff pollution that describes the initial part of a precipitation event when the majority of the accrued contaminants are transported in runoff (Deletic 1998). To assess the first flush strength, Geiger (1984) proposed the M(v) curve, which is a curve of the dimensionless cumulative pollutant mass M as a function of the dimensionless cumulative runoff volume V. Since then, several assessment methods based on the M(v) curve have been proposed, such as 20/80 (Stahre & Urbonas 1990), 20/40 (Deletic 1998), 30/80 (Bertrand-Krajewski et al. 1998), the mass first flush ratio (e.g., MFFR20, MFFR30, Wang & Li 2009; Jeung et al. 2019), and the b-value method (Saget et al. 1996).

Bach et al. (2010) proposed an innovative method for assessing the first flush and defined the first flush volume (V_FF) as the volume of runoff required to reduce catchment's stormwater pollutant concentrations to background levels. Applications of this method by Todeschini et al. (2019) and Zhang et al. (2021) have shown that it is efficient. Furthermore, Zhang et al. (2021) proposed an alternative method, based on the runoff depth vs. pollutant cumulative mass curve, for assessing the runoff pollution control efficiency of different V_FF.

The strength of the first flush of different pollutant types in urban runoff can vary significantly, especially for particulate matter (Zhang et al. 2012, 2021), and studies have shown that the particle size distribution (PSD) of particulate matter can vary by 2–3 orders of magnitude (Zhao et al. 2016; Morgan et al. 2017). Without considering the differences in the materials and particle shape, the strength of the first flush of particles in different size categories may be different. For example, Morgan et al. (2017) found that the strength of the first flushes of particles in five particle size categories (<10, 10–20, 20–45, 45–90, and >90 μm) differed significantly, and that the strength of the first flush increased as the particle size decreased. Further studies are needed to determine whether this trend applies to other types of urban runoff.

Urban stormwater management methods will generally be effective for particles within a specific particle size range (Fassman & Blackbourn 2011; Kayhanian et al. 2012). This means that to design, select, and evaluate appropriate stormwater management methods, information is needed about the size distribution of particles in urban runoff (Charters et al. 2015). Furthermore, there is concern about the first flush in many countries because local or national standards or specifications for pollution control are frequently based on the first flush (Sansalone & Cristina 2004; UK Highways Agency 2006; Todeschini et al. 2019; Zhang et al. 2021). Also, we need to improve our understanding of how V_FF is related to our ability to control particles in different size ranges.

There is a lack of information about the first flush of particles of different sizes in runoff or the variation in the ability to control particles of different sizes. Here, the runoff was collected from three different roof types, asphalt (AR), metal (MR), and cement (CR), and analyzed to assess the strength of the first flush of different-sized particles. The objectives of this study were to (a) evaluate the strength of the first flush of different-sized particles, (b) quantify the ability to control runoff that contained particles of different sizes under a specific V_FF, and (c) propose a new method to quantify the relationship between the runoff depth and the particle cumulative mass.

The roof platforms were made of potable-quality polyvinyl chloride (PVC). They each measured 1 m long and 2 m high, and had a longitudinal slope of 5%. The roof surfaces were positioned 0.5 m above the ground to prevent rainwater splashing in around the ground and a 20 cm fence was set up to prevent rainwater splashing out of the platforms. An outflow pipe was installed at the lower end of the device (D = 50 mm). The flow rate was monitored and the runoff was collected at this outflow pipe. The roof materials were supplied by a local building materials market. The asphalt, metal, and cement roofs were labeled AR, MR, and CR, respectively.

A tipping bucket rain gauge that was accurate to 0.2 mm (HOBO U30 Station, Onset) was installed on the roof of a building that was at a distance of 50 m from the roof platforms to monitor rainfall at a time step of 1 min. A total of 41 rainfall events were recorded during the monitoring period from June to September 2021. The average duration, average rainfall depth, average maximum intensity, average intensity, and average antecedent dry period (ADP) for the rainfall events were 309 ± 555 min, 15.2 ± 16.9 mm, 0.8 ± 0.7 mm/min, 0.13 ± 0.17 mm/min, and 2.62 ± 2.50 days, respectively. The statistical summary of the rainfall characteristics of the 41 rainfall events is provided in Supplementary Materials (Table S1).

The runoff flow rate monitoring and water sampling were implemented synchronously. We did not use the data for all the rainfall events for various reasons; for example, some events occurred at night, the data were incomplete, or the equipment failed. The results and discussion are based on an analysis of the water quantity and water quality data from eight rainfall events (Table 1).

Runoff from the AR, MR, and CR was collected manually according to the protocols outlined in Burton & Pitt (2002). The runoff samples were recovered immediately after the rainfall events and transported to the laboratory, where they were analyzed for SS, chemical oxygen demand (COD), total nitrogen (TN), and total phosphorus (TP) using standard methods (APHA 2012). The PSDs of the samples were determined within 6 h of collection using a Malvern Mastersizer 3000 (Malvern Instruments Ltd, UK) with a particle size resolution from 0.01 to 3,500 μm (Zhang & Li 2015). Additional details of the analysis are provided in Supplementary Materials.

The pollutant concentration of a rainfall event is usually expressed by the event mean concentration (EMC) (Lee et al. 2002), and the EMCs of the pollutants in the AR, MR, and CR are provided in Supplementary Materials. The M(v) curve, mass first flush ratio (MFFR), and b-value method were used to assess the strength of the first flush (Saget et al. 1996; Deletic 1998), and the analytical procedures are detailed in Supplementary Materials.

The particle first flush (PFF) was adopted to assess the strength of the first flush of particles in five different size categories, namely <10, 10–20, 20–45, 45–75, and >75 μm. These categories were selected as they were used in similar previous studies (Morgan et al. 2017; Zhang et al. 2021). The particle concentrations in the different categories can be obtained from the PSD and SS in the runoff. The first flush of a specific size category of particles can be assessed by the M(v) curve and the MFFR method. The PFF20, defined as the proportion of the total pollutant event load for the specific range of particles transported by the first 20% of the total event volume, was also calculated.

The runoff depth–cumulative mass (RD–CM) curve proposed in previous research (Zhang et al. 2021) was used to analyze the ability to control different-sized particles in the runoff for a specific V_FF. One-way analysis of variance (ANOVA) was used to determine significant differences (with a significance level of p > 0.05).

The EMCs of COD were also affected by the roof material. The average EMCs of COD were significantly higher in the AR runoff than those in the MR and CR runoff. The AR may have been a source of organic matter (Hou & Zhang 2014). There were no significant differences in the EMCs for TN and TP in the runoff for the three roofs (p > 0.05). Wet and dry deposition may be the main source of nitrogen and phosphorus in roof runoff rather than the roof materials (Hou et al. 2012; Wang et al. 2021).

The particle volume proportions and particle concentrations of the different size ranges in the AR, MR, and CR runoff are shown in Figure 2(c) and 2(d). Particles <10 μm dominated the particles in AR, MR, and CR runoff, and accounted for 60.14 ± 39.75, 60.64 ± 36.60, and 63.58 ± 35.73% of the total particle volumes. The proportions of particles <10 μm did not vary significantly among the three types of roof runoff (p > 0.05), but the concentrations of the particles <10 μm in the three types of roof runoff were different because the SS concentrations (particle mass) varied among the three runoff types (Figure 2(d)).

The mean b values of SS in the runoff from the AR, MR, and CR were 0.71 ± 0.17, 0.73 ± 0.36, and 0.64 ± 0.13, respectively. These values represent a moderate first flush, and the value did not vary significantly among the three runoff types (p > 0.05).

Although there were no significant differences in the b values for the first flush of SS in the runoff for the three roof types, the trends were different for the different size categories (<10, 10–20, 20–45, 45–75, and >75 μm). This is because the particle wash-off process is closely related to the particle size (Morgan et al. 2017).

For the same roof runoff, the first flush strength of a specific particle size category varied significantly in different rainfall events. For example, the 20–45 μm particles in the AR runoff presented a moderate first flush in the rainfall events of 06/09 and 07/29 but were strongly and moderately diluted in the rainfall events of 07/05 and 08/23, respectively. For the same rainfall event, the first flush strength of a specific particle size category varied significantly for the runoff from different roofs (p < 0.05). For example, for the 06/09 rainfall event, 45–75 μm particles presented a moderate first flush in runoff from the AR but were weakly diluted in the runoff from the MR and CR.

In the rainfall event of 06/09, the particles in the runoff from the AR were mainly distributed between 1 and 100 μm during the initial period, but large-sized particles (>100 μm) appeared after 10 min as the flow rate increased. When the flow rate peaked for the second time, the PSD peak around 100 μm was enhanced. Meanwhile, larger particles (>1,000 μm) appeared, which was consistent with the occurrence of the peak flow rate. It is also worth noting that, for the rainfall event of 06/09, the patterns in the flow rate for the MR and CR runoff differed from those in the AR runoff because of the roof material characteristics. Furthermore, the physical properties of the particles differed among the three types of roof runoff. These factors help to explain the significant differences in the PSD variations in the runoff from the AR, MR, and CR.

The variability in the PSD through the rainfall events of 06/09 and 07/29 shows that the rainfall characteristics can significantly influence the PSD for the same roof material. Furthermore, for the rainfall event of 07/29, the PSD in the three types of roof runoff was significantly different (p < 0.05), which indicates that the roof material influences the variation in the PSD in the runoff. As well as these two factors, the ADP determined the mass of particles that accumulated, and also influenced the PSD characteristics in the runoff.

An analysis of the PSD variation in the different rainfall events indicated that the variability of the PSD in the runoff was complicated, and that the PSD was affected by several factors. The first flush of particles in different size categories should be analyzed separately to provide useful information about how to control particles in the runoff.

The national standards of China require a V_FF of 3 mm (GB50400-2016). This means that a mean SS cumulative mass of 82.34 ± 15.44% could be effectively controlled from the MR runoff, but that only 67.54 ± 17.12 and 64.18 ± 21.39% of the SS mass could be controlled from the runoff from the AR and the CR, respectively. This result suggests that it was easier to control the SS in runoff from the MR than from the other roof types. However, the SS cumulative mass in the MR runoff varied significantly in different rainfall events (p < 0.05), and the results for the MR, and also for the AR and CR, may be influenced by rainfall characteristics (Zhang et al. 2021).

The distribution of the RD–CM curves of particles in the five size categories can be illustrated by the PFF strength of the different particle size categories. For example, for the rainfall event of 06/09, the five RD–CM curves for the runoff for the three roof types were relatively centralized, which indicates that the particles in the five size categories presented a relatively consistent first flush. It is worth noting that the variation in the cumulative mass of the different-sized particles over the runoff depth was generally consistent with the SS. But for the other three rainfall events, the five RD–CM curves for the three types of roof runoff were more dispersed and highlight the differences in the strength of the first flushes of particles in the different size categories. When a RD–CM curve of particles in a specific size category is above the RD–CM curve of SS, it means that the first flush of particles in this size category is strong, and that the particle cumulative mass would be high for a specific runoff depth. Conversely, when a RD–CM curve of particles in a specific size category is below the RD–CM curve of SS, the first flush would be weak and the cumulative mass of particles would be low.

For the rainfall events in this study, the strength of the first flushes of particles in the different size categories differed significantly. For example, for the runoff from the AR, the <10 μm particles presented a first flush that was significantly stronger than for SS. The first flush of the 10–45 μm particles was strong in runoff from the MR, while the first flush of the <45 μm particles in CR was relatively strong. The differences may be related to the roof material characteristics and the build-up of surface sediments during the dry period. Fine particles (<45 μm) easily peel off the CR as the materials age, and these fine particles can easily be washed off into runoff from the CR. The MR and AR have relatively slow aging processes, so fewer particles peel from these surfaces than from the CR. Particles on the surfaces of the AR and MR mainly come from atmospheric dry deposition. The difference between the AR and MR was related to the roughness of the AR and MR. The AR has a rough surface, which can delay the wash-off of 10–45 μm particles. In contrast, the MR has a smooth surface and the roughness coefficient is low, so the first flush for 10–45 μm particles in the runoff was strong.

Generally, the differences in the runoff for the three roof types were mainly reflected in the <45 μm particles. Large particles >45 μm are mainly affected by the rainfall intensity and the runoff kinetic energy. Hence, the first flushes of particles >45 μm were relatively weak for the runoff from the three types of roof and did not differ significantly (p > 0.05). So, the control efficiency of the particles <45 μm will be higher, but the control efficiency of the particles >45 μm will be lower than the SS control design expectation (based on the requirements for runoff pollution control in the Chinese national standard (V_FF = 3 mm)); it is worth noting that there were also differences in <45 μm particles as they were influenced by the rainfall and roof materials.

The PSD characteristics and PFFs in runoff from an AR, MR, and CR were assessed. The main findings were as follows:

• (1)

  The EMCs of SS in runoff from the CR were significantly higher than those in the runoff from the AR and MR, and the EMCs of COD in runoff from the AR were significantly higher than those in runoff from the MR and CR. The EMCs for TN and TP did not differ significantly in the runoff from the three roof types. The variation in the EMCs can mainly be explained by the source of the pollutants.

• (2)

  The first flushes of pollutants (SS, COD, TN, and TP) were relatively stronger in the runoff from the AR, followed by the MR, and were relatively weak for the runoff in the CR. Note, however, that there were slight variations.

• (3)

  The patterns of the PSD in the runoff were complicated and were affected by several factors. The strength of the first flushes differed significantly for particles in different size categories in the roof runoff and were very different from SS.

• (4)

  The results from this study showed that the control of particles >45 μm would be lower than, but the control of particles <45 μm would be higher than, the SS control design expectation, which are outlined in the Chinese national standard for runoff pollution control (V_FF = 3 mm). Note that the efficiency of the control of the <45 μm particles may vary under the influence of rainfall and roof materials.

Statistically significant results were obtained from monitoring rainfall events in this study, and the results were affected by the sampling interval, sampling frequency, and the number of water samples. The differences in the results for the different roof materials and rainfall characteristics highlight the need to carry out further monitoring and verify the results on different land use types and various rainfall types. Also, because the PSD results were closely related to the sampling and testing conditions, it would be useful to collect more PSD monitoring data to verify the results. In the meantime, the results show that information about the first flush of particles can be used as an alternative approach for evaluating the effectiveness of pollution control efforts.

This study was financially supported by the National Key R&D Program of China (Grant No. 2022YFC3800500). The authors thank the research team members for their enthusiastic support.

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

The authors declare there is no conflict.