Introduction

Metals from various sources are commonly found in stormwater (and to a lesser extent in wastewater) discharges and have long been in focus when stormwater impacts on receiving water bodies and/or water quality treatment demands are assessed and discussed. Early research evaluating stormwater quality has recognised metals to be of certain importance. Wilber and Hunter (1975) emphasise that ‘heavy metal concentrations (in stormwater runoff) were found to vary significantly throughout runoff events and from storm to storm’. This chapter will describe and discuss these variations and treatment technologies, which have been extensively evaluated with a focus on their metal treatment capacity.

Treatment Mechanisms

One main characteristic of metals which significantly affects removal processes is their distribution between the dissolved and particulate phase. A common way to estimate this distribution is by passing them through filters with a pore size of 0.45 µm and dividing them into the fractions as shown in Fig. 5.1. In a geochemical context, the dissolved fraction is commonly divided into colloidal and genuine dissolved fractions (Ingri 2012). Colloids, unlike particles, do not settle but remain in solution. The surface of colloids is often negatively charged, causing positively charged metal ions to bind to it. The ions and molecules present in free form without binding to colloids or particles are referred to as a true or authentic dissolved fraction. This is also the most bioavailable fraction since it can be taken up by aquatic plants and organisms, which also means it has increased toxicity (Ingri 2012; Campbell 1995).

Fig. 5.1
figure 1

Metal fractions in stormwater (simplified scheme: H. Österlund)

Full size image

Important factors that affect the solubility of metals and mobility are pH values and dissolved organic matter (DOM, such as humic and fulvic acids) and the access to particle surfaces for them to attach to. Generally, the solubility is higher at low pH values (Ingri 2012).

In stormwater quality studies, often both total and dissolved (i.e. <0.45 µm) fractions are analysed. This enables researchers to calculate the particulate fraction by subtracting the dissolved from the total concentration. The distribution between these fractions can vary substantially, not only between different metals but also within a runoff event and between different sites and seasons. For instance, investigations on the distribution of metals from runoff of five German highways by Dierkes (1999) revealed that

  • 51–90% of Cd (mean 70%),

  • 28–55% of Cu (mean 42%) and

  • 14–51% of Pb (mean 36%).

are in the dissolved phase (<0.45 µm). Boogaard et al. (2014) found even broader ranges and mean values of approximately 60% for Cd and for Cu (range ~20–90%), 55% for Ni (20–95%), 70% for Pb (10–99%) and 80% for Zn (10–90%). As already mentioned in Chap. 2, these particles are bound in large proportions to particles with a grain size of less than 90 µm or even 60 µm (Xanthopoulos 1990; Boogaard et al. 2014). Hence, treatment installations should ideally be at least capable of retaining fine suspended solids.

Metals in Stormwater from Separate Sewer Systems and Combined Sewer Overflows

One important source for metals in stormwater is vehicular traffic. Further, metals also leach from surfaces in the urban environment, such as roofs, lampposts, barriers, facades, etc. or are of geogenic origin. Their composition in runoff changes over time, e.g. substitutes used in industrial products, such as the replacement of lead in fuel over the last two decades.

As listed in Chap. 1, the concentrations found in stormwater often vary during single events (e.g. due to first flush effects or varying rain intensities during the event that transports different fractions), between different events (e.g. due to varying antecedent dry periods, seasonal variations and varying rain characteristics), seasons and between different catchments (due to different catchment characteristics).

Seasonal Variations

In a study in northern Sweden, significantly higher concentrations in snowmelt runoff have been observed in March and April (Cu: 37–199 mg/L, Pb: 16–80 mg/L; Zn 105–791 mg/L) compared to runoff from rain events in May and June (Cu: 30–45 mg/L, Pb: 14–19 mg/L, Zn 130–169 mg/L) (Westerlund et al. 2003). In this study, concentrations of both metals and suspended matter in stormwater are higher in snowmelt runoff than during rainy periods. In snowmelt runoff, relatively high concentrations of Cu, Pb and between 16 and 80 mg/L were measured. This can be explained by the long period of contaminant accumulation in the snow; these contaminants are then released during a relatively short period. In a study in Germany, similar results have been confirmed by Helmreich et al. (2010), who showed significantly higher metal and total suspended solids (TSS) concentrations in winter runoff compared to the summer season. Reasons given by the authors were the use of sand and gravel for anti-slip applications, which increases wear and tear on road surfaces and vehicles.

Besides concentration variations, metal characteristics may also change. During winter in cold or temperate climates, de-icing salts are applied regularly, which, for instance, affects metal partitioning towards the dissolved phase (Marsalek et al. 2003). Higher percentages of dissolved pollutants can affect the performance of treatment technologies (Søberg et al. 2017).

Variation Between Catchments

The quality of stormwater depends on the surface characteristics of the catchment and the anthropogenic activities in or around the catchment (Eriksson et al. 2007). The contamination of stormwater with metals in urban catchments largely depends on the use of building materials, on the one hand, and the presence of frequently used roads, on the other hand. Studies have shown that runoff from metal roofs may have higher concentrations of, e.g. Cu and Zn than road runoff while other metals such as Cd, Pb, Ni and Cr are higher in road runoff (Göbel et al. 2007). In general, areas with direct connection to traffic and runoff from industrial and commercial areas often exhibit relatively high pollutant concentrations (Pitt et al. 1995; Czemiel Berndtsson 2014).

Although a correlation between the traffic density and the concentration of metals in road runoff is often assumed, Kayhanian et al. (2012) could not prove such a correlation in a literature review on road runoff worldwide. They showed, however, differences between the concentrations in North America, Europe and Asia, which prove that a local aspect has to be considered. The authors also mention the influence of preceding dry phases and the catchment area, as mentioned in Chap. 1.

Site variations may also vary for different pollutants. Gasperi et al. (2014) analysed pollutants in stormwater from three different areas. They found different Cu, Cr, Ni and Zn concentrations in these areas while they did not detect any differences in Cd and Pb concentrations, although the land use in the areas was different.

Variations Over Time

In general, it is quite difficult to compare the ranges of concentrations found in road runoff since the sampling points vary in the different studies, as can be found in literature reviews. However, many publications refer to investigations made in the middle of the 1990s or even earlier. The age of these studies is important, since metal concentrations have shifted over the decades; the ban on leaded gasoline in most countries has reduced lead concentrations in runoff significantly (Kayhanian et al. 2012; Ayrault et al. 2014). Future trends of stormwater quality changes depend on how treatment facilities perform during their lifespan. However, simulating these developments over time involves quite high uncertainties (Borris et al. 2016). Changes in climate, building materials and building design, environmental regulations and the use of unknown substances today may affect stormwater quality in the future.

Table 5.1 shows ranges of road runoff concentrations from different sources, which only include values published after 2005 given the fact that runoff composition has changed during the next last decades. In general, all values vary over two to three magnitudes. Due to the high variations, no clear overall trend can be derived, even if similar sampling locations are compared.

Table 5.1 Concentrations in urban runoff (µg/L), published after 2009
Full size table

In terms of CSO, to date, there are only few discharge measurements available before the flow volume enters the river. In most cases, researchers concentrate on an increase of the pollutants in the receiving surface water body by measuring upstream and downstream of a discharge point or in the river sediment. Table 5.1 presents results of some measurements taken in combined sewer systems or at their outlets. Since the sampling points and the catchment are not completely comparable, it is only possible to deduct general trends. Variations between minimum and maximum are within one magnitude. The values are in general also comparable to those from separate sewer systems and highway runoff despite the different composition of CSO.

Treatment Systems

Stormwater Ponds and Basins

As described in Chap. 2, the removal mechanisms of stormwater ponds rely mainly on sedimentation. Given that fine sediments show relatively higher metal concentrations (Sansalone and Buchberger 1997; Liebens 2002), sediment close to the inlet tends to have lower metal concentrations (Karlsson et al. 2010). As mentioned above, the coarser forebay sediment may show a lower toxicity.

Table 5.2 gives an overview of data on metal concentrations in stormwater pond sediments published between 2010 and 2017. As can be seen, similar to the metal concentrations in the stormwater itself, the range found in the dry matter (DM) is quite high.

Table 5.2 Overview of metal concentrations measured in sediment from stormwater ponds (mg/kg DM), published after 2009 (partly based on Søberg 2014)
Full size table

Dissolved substances are only reduced in shallow, planted areas, comparable to ponds with FTWs (see Chap. 2). For instance, a Swedish study demonstrated considerable removal of dissolved metals in a stormwater pond (Cd: 73%, Cu: 58%, Pb: 41% and Zn: 64%). However, this is still far lower than the removal of particulate metal [between 85 and 92%, (Al-Rubaei et al. 2016)]. In studies from USA and Sweden, stormwater pond influent and effluent concentrations of dissolved metals were in the same range (Stanley 1996; Pettersson 1998).

Also, when the metal contamination is assessed in sediment accumulated in ponds, only looking at the total metal content can be misleading since metals may be present in different fractions and, thus, potentially available to different degrees. Sequential extraction procedures reveal the metal fractionation by distinguishing between the five fractions: exchangeable (I), carbonate-associated (II), Fe–Mn oxide-associated (III), organic matter/sulphide-associated (IV) and residual (V). Metals in fractions I to IV are potentially bioavailable since they can be released from the sediment if the ambient conditions change (e.g. after excavation during maintenance, see the following text). For instance, a recent study from Sweden (Karlsson et al. 2016) shows—for sedimentation ponds and tanks as well as for storm drain sediment—that the majority of Cd, Cu, Pb and Zn and a significant amount of Ni were in potentially mobile forms. This fact must be considered during pond maintenance (sediment removal, drying/de-watering/disposal) to prevent metal from being released. Similar results were reported in various studies (Marsalek and Marsalek 1997; Camponelli et al. 2010; Lee et al. 1997).

In winter, an ice cover on the sediment pond reduces oxygenation of the pond water (e.g. by wind) (German et al. 2003), which can affect metal partitioning. Additionally, road salt used in cold climates affects the metal partitioning between particulate and is dissolved: if road salt is present in stormwater, a higher percentage of the metals is in the dissolved phase (Søberg 2014). Since ponds mainly remove metals in particulate form, the overall metal treatment performance may decrease.

Constructed Wetlands

Surface-Flow Constructed Wetlands

Since metals are often bound to particles and since wetlands capture such particles to a great extent (Sansalone and Buchberger 1997), wetlands remove a significant reaction of total metals thanks to their sedimentation process. Resuspension of the captured metals has to be avoided (Zhang et al. 2012). In comparison to ponds, wetlands provide more heterogeneous morphology including dense vegetation; therefore, treatment of fine particles and/or dissolved metals is potentially more effective than in ponds.

In an extensive literature study and meta-analysis, Carleton et al. (2001) investigated factors affecting the stormwater quality treatment performance of constructed wetlands (CWs). The review published data from 49 wetlands in 35 studies. In combination with results of other studies, the removal rates achieved are illustrated in Fig. 5.2.

Fig. 5.2
figure 2

Interval plot (95% confidence interval bar) of removal percentages achieved in constructed stormwater wetlands

Full size image

The figure underlines that, similarly as for ponds, the metal treatment efficiency reported in different studies varies significantly depending on a wide range of factors. In general, however, the figure corroborates the assumption of Birch et al. (2004), who conclude that a mean removal of Cd, Cu, Pb and Zn of approximately 60% can be achieved. In studies done in 1997 and 2012/13 in Bäckaslöv, Växjö (Sweden), metal removal exceeding 80% was observed (Semadeni-Davies 2006; Al-Rubaei et al. 2016), which is in the upper range of the data included in the meta-analysis done by Carleton et al. (2001). The study of Al-Rubaei et al. (2016) also included dissolved metals. Their outflow concentrations were significantly below the inflow concentrations. Removal rates were between 55 and 80%. In the combined pond–wetland system evaluated in this study, the wetland increased the removal of the dissolved metals significantly compared to the removal in the pond only, underlining the importance of more advanced treatment processes for dissolved contamination removal.

Floating Treatment Wetlands

Since the treatment in stormwater ponds relies on sedimentation to large extent, the treatment of dissolved metals (and other compounds) in stormwater ponds may be insufficient. Accordingly, retrofitted Floating Treatment Wetlands should improve their treatment performance. After such pond retrofitting, the metal and sediment removal significantly increased [TSS, particulate Cu, and particulate Zn by 40% and dissolved Cu by 16% (Borne et al. 2013)]. Reasons for that are an increased direct plant uptake (Borne et al. 2013; Ladislas et al. 2013, 2015), bacterial/biofilm uptake (Borne et al. 2014), increased sorption [e.g. to organic matter (Borne et al. 2014)] and precipitation processes due to higher humic content, lower dissolved oxygen and more neutral pH value (Borne et al. 2013). In a study in New Zealand, some release of metals was observed in the spring, especially of Cu, due to organic matter degradation and, and thus the export of dissolved organic matter from the pond (Borne et al. 2014).

Subsurface-Flow Constructed Wetlands

Since most metals entering media-based systems are particle-bound, mechanical filtration of the incoming stormwater sediment also removes substantial loads of metals (and other particle-bound pollutants). Thus, the efficiency of TSS and particle-bound metal removal is correlated which was shown by Hatt et al. (2008) for vertical-flow stormwater wetlands (see section Bioretention filters). Studies on systems for horizontal-flow wetlands used for stormwater or CSO treatment are missing; however, the general processes in media-based systems are the same. Dissolved metal removal varies more since it is affected by diverse factors that influence soil and metal interactions. The main metal retention processes in soil are adsorption (including metal-OM complexation and cation exchange), surface precipitation and fixation (mainly to clay minerals) (Alloway 1995). Key soil properties controlling these processes are, among others, pH, OM content, clay mineral content and oxidation reduction potential (Bradl 2004). Besides these geochemical processes, plant metal uptake plays a less significant role (Read et al. 2008; Søberg et al. 2014a; Muthanna et al. 2007a) and is less important since the plants are not usually harvested. However, vegetation in media-based systems plays an important role in maintaining the infiltration capacity, facilitating treatment indirectly (e.g. by effects on microbial communities in the filter) and providing aesthetical values and/or (urban) biodiversity.

Bioretention Filters

From approximately 2000 onwards, numerous studies have been published on how well stormwater bioretention filters remove pollutants. A summary of inflow and outflow metal concentrations reported in selected studies is given in Table 5.3. Cadmium concentrations were only investigated in two studies with inflow values between 4.6 and 5.6 mg/L and removal efficiencies between 66 and >99.5%.

Table 5.3 Biofilter inflow and outflow concentrations of metals (µg/L) from selected studies. (diss. dissolved concentrations; non-veg. non-vegetated; In. inflow concentrations; Out.: outflow concentrations) (partly based on Søberg 2014)
Full size table

The total metal removed by bioretention filters often exceeds 80–90% (Hatt et al. 2009; Muthanna et al. 2007b; Read et al. 2008; Sun and Davis 2007).

As for most compounds removed by bioretention filters (see, e.g. Chap. 2) the processes and properties are, to varying degrees, affected by ambient conditions, e.g. the drying/wetting pattern, ambient temperatures, road salt in the runoff, the pollutant concentrations in the runoff and the runoff intensity (Hatt et al. 2007b; Blecken et al. 2009; Søberg et al. 2014b; Muthanna et al. 2007a; Denich et al. 2013; Bratieres et al. 2008), the filter design (e.g. water saturated zone, different filter materials) (Dietz and Clausen 2006; Davis et al. 2009; Hatt et al. 2006; Fassman et al. 2013) and, with minor effect, the plant species utilised (Read et al. 2008).

Although dissolved metal removal has been shown to vary far more than the quite stable total metal removal, only the total metal removal has been investigated in most biofilter studies (see Table 5.1). Dissolved metal removal has been considered in fewer of the investigations (Muthanna et al. 2007b; Read et al. 2008; Hatt et al. 2007a; Sun and Davis 2007; Søberg et al. 2014b). In pilot-scale stormwater biofilters, Muthanna et al. (2007b) found removal rates of dissolved Zn up to 70%, whereas leaching was observed for both dissolved Cu and Pb. In a laboratory study investigating biofilter columns at three different temperatures, Blecken et al. (2011) found lower removal efficiencies (24–66%) for dissolved Cu and Pb compared to Zn and Cd (99%), and a negative correlation between temperature increase and removal of dissolved Cu and Pb. In a study about temperature and salt influence on metal removal in laboratory pilot-scale bioretention filters, Søberg et al. (2014a) found high removal of dissolved Zn and Cd (>90%), whereas removal of dissolved Cu and Pb was less efficient, ranging from −1345 to 71% being deteriorated by the presence of salt, particularly in connection with high temperature.

Although some findings indicate that dissolved metal removal is significantly lower than total metal removal and, in particular, Cu leaching was observed (Hatt et al. 2007a; Chapman and Horner 2010; Muthanna et al. 2007b), biofilters seem to have potential to provide adequate dissolved metal treatment if filter material with specific sorption properties is used (Sun and Davis 2007; Hsieh and Davis 2005). An efficient removal of dissolved metals has also been reported for bioretention filters where sandy soils with only little organic content are used as filter material (e.g. Blecken et al. (2011) reported removal rates of >99% for dissolved Zn and Cd and >60% for dissolved Cu when using filter material with 90% sand). Numerous studies have further tested various filter materials to enhance metal treatment. Examples are zeolites and peat (Färm 2003), blast furnace slag, chitosan, crab shell, peat, sawdust and sugar cane (Vijayaraghavan et al. 2010), limestone, shell sand, zeolite, and olivine (Wium-Andersen et al. 2012). Many of these results are derived from short-term laboratory studies; when these results are transferred to praxis, it is important to consider long-term behaviour of the material (e.g. breakdown and release of associated pollutants over time). When choosing filter materials for bioretention systems, one thus has to compromise between infiltration rate, adsorption capacity and support of plant growth.

Typically, metals do not ingress far into the filter material, but are trapped on or near the top of the filter due to both mechanical removal and sorption processes (e.g. Davis et al. 2001; Grotehusmann et al. 2017). Grotehusmann et al. (2017) found that metals accumulate on the filter surface and in the first 10–15 cm of the filter layer in correlation with how much calcium carbonate (CaCO3) is available, which is often added as additional layer on top of the filter surface at the large-scale sites investigated in Germany. Although in general, high inflow values of CaCO3 onto the filter could also lead to building up a carbonate layer, due to the hydraulic conditions on the filter surfaces, it is usually limited to areas close to the inflow and did not result in overall clogging of the filter surface. However, when CaCO3 is added as additional surface layer or mixed into the filter material, the additive itself may not be contaminated with heavy metals, e.g. lead (Grotehusmann et al. 2017). The high metal removal in the upper layer facilitates filter maintenance since merely scraping off the top layer may remove a high proportion of accumulated metals from the system, and thus postpone the need to replace the whole filter media (Hatt et al. 2008).

Some field investigations predicted that the accumulation of fine stormwater sediment on top of the filter material and in the upper layers reduces the hydraulic conductivity relatively quickly, sometimes even within several months, and leads to clogging (Li and Davis 2008). However, Grotehusmann et al. (2017) could not confirm this in large-scale investigations on filters designed according to German standards. The main reason for this finding was oversized filter layers which led to low long-term loads of fine sediments.

During winter in cold or temperate climates, pollutant concentrations are particularly high, and de-icing salt often affects metal partitioning towards the dissolved phase (Marsalek et al. 2003; Oberts 2003). The presence of salt has been shown to substantially influence the ability of stormwater biofilters to remove metals. The latter is particularly pronounced for dissolved metals (Søberg et al. 2014b). Søberg et al. (2014b) found that ion exchange by Na+ was probably entirely responsible for the leaching of dissolved Pb from the filter material.

In winter, plant metal uptake is generally inhibited by salt in stormwater runoff (Fritioff et al. 2004) and low temperatures generally reduce biological activities. Søberg et al. (2014b) examined the impact of temperature, salt and a submerged zone on metal uptake in three native (Northern Sweden) wet/drought tolerant plant types: Juncus conglomeratus, Phalaris arundinacea and Carex panacea. They found a generally higher metal uptake at low temperature. Their results suggested that the three plant species were not particularly affected by different temperatures and/or the presence/absence of a submerged zone in the filter and/or salt in stormwater. This indicates the potential to use the investigated plant species for targeted cold climate biofilter design. Additionally, Denich et al. (2013) found that biofilter vegetation was capable of withstanding high salt exposure. Despite the reduced biological activity in cold seasons as described in Chap. 2, metal retention was good for both seasons with mass reductions of 90, 82 and 72% of Zn, Pb and Cu, respectively (Muthanna et al. 2007b). The latter is supported by findings of a study evaluating seasonal performance variations (Roseen et al. 2009), where seasonal contaminant removal performance was found to vary little for stormwater biofilters.

Swales and Buffer Strips

Grotehusmann et al. (2017) found out that a major part of the metals is already captured within the first 10 cm of the buffer strip leading to the swale. Since the buffer strips contain rather coarse media, the metal accumulation was also found in deeper layers (25–30 cm). The authors revealed that a major part of the retention was, thus, already provided by the shoulder, and concluded that the treatment of swale effluent, as often practiced in Germany, is not necessary.

Reported removal percentages of metals in swales vary as follows: Bäckström et al. (2006) report about 20% metal removal while Stagge et al. (2012) and Knight et al. (2013) report very efficient metal removal rates. Bäckström et al. (2006) found that the particle size distribution influences the removal efficiency: only large particles >250 µm settle in swales. In general, the pollution removal capacities for dissolved pollutants and small particles are low. Thus, Bäckström et al. (2006) conclude that, while efficient for flow retention, swales cannot produce consistently high pollutant removal.

Although swales commonly tend to be comprised of grass, they can have particular design modifications (such as wetland planting) to improve nutrient reduction (Winston et al. 2012). Metal uptake by plants can be significant. This uptake is specific to metal and plant species (Zhang et al. 2012). Most plants accumulate the metals in their roots, but also transport to the leaves occurs (Weis and Weis 2004). It is, thus, important that swales and buffer strips be harvested regularly.