Pollutants discharged by roads may impact water bodies and soils. The best method to characterise road runoff is by monitoring, which is not always possible due to human or material constraints. Therefore, prediction tools can be a valuable method to manage road runoff discharges and protect the environment. The present work reviewed and evaluated international tools for road runoff quality prediction, in order to assess if an existing tool could be suitable for wide usage by stakeholders in Europe. Four tools from the USA and Europe were selected and tested at 22 road sites located in regions with annual precipitation values ranging from 500 to 1,000 mm, from seven European countries. The results for the site median concentration (SMC) of total suspended solids (TSS), Zn, Cu, Pb and Cd showed coefficients of determination (R2) from 0.0004 to 0.2890 for the different pollutants and tools. It was concluded that none of the tools could predict the road runoff pollutant concentrations, except for the country where it had been calibrated. The findings support practitioners and researchers all over the world, pointing out directions, and gaps to be filled, regarding the management of road runoff discharges and use of prediction tools.

  • Tools for road runoff quality prediction could support environmental management.

  • Evaluation of tools in seven European countries with monitoring data sets.

  • Site median concentrations of road runoff change over time and region.

  • It is not feasible that a tool accounts for the whole phenomena of road runoff.

  • Need for innovative approaches in water management under climate change.

Graphical Abstract

Graphical Abstract
Graphical Abstract

The accomplishment of the European Water Framework Directive in terms of a good ecological status for all water bodies, as well as overall environmental sustainability goals, requires a good understanding of the impacts of pollution sources and their control (European Commission 2000). Road runoff is a diffuse source of pollution that typically shows a huge variability of concentrations.

Several studies confirmed that road runoff can cause both cumulative impacts and acute effects on the chemical quality and ecological status of the receiving water bodies (e.g. Kayhanian et al. 2003; Vieira et al. 2013). According to Hvitved-Jacobsen et al. (2010), short-term impacts from stormwater discharge occur at time scales of less than 1 hour up to 1 day, and are related to hydraulic effects and discharge of biodegradable organic matter or of suspended solids.

The knowledge of the characteristics of road runoff pollution together with the evaluation of the vulnerability of the water bodies support the decision to implement mitigation measures such as the construction of treatment systems that are used widely (e.g. Barbosa & Fernandes 2009). Parameters such as the event mean concentration (EMC), defined as the concentration of one rainfall and runoff event in mass per volume, site median concentration (SMC), defined as the average concentration of several events for the same site and pollutant load are used to describe road runoff quality (e.g. Erickson et al. 2013). For a given site, the most precise method to characterise road runoff pollutant is to carry out road runoff monitoring studies, including automatic rainfall and flow measurements, and sequential runoff sampling. The laboratory analysis of the samples provides the concentrations for the selected parameters that are used to calculate the EMCs and SMCs, as well as the pollutants loads. It is generally accepted that at least 10 stormwater events from independent rainfall/runoff events, expressing seasonal variability, must be monitored to calculate a robust SMC for a given site/location (Hvitved-Jacobsen et al. 2010).

Monitoring work requires considerable time, human and material resources and it is subjected to risks, such as equipment damage or absence of appropriate rainfall events. Hence, tools for pollutant concentration or load predictions may be an important process to enable protection of the environment from road runoff pollution impacts. Prediction tools are commonly regression equations established based on large data sets from monitoring work.

It is noteworthy to mention that modelling road runoff quantity deals only with physical and hydraulic variables (e.g., Havryshchuk & Kaskiv 2020) whereas the quality of stormwater relates to physical, chemical and biological process.

Following an overview of road runoff characteristics, this study reviewed, selected, tested and discussed tools for road runoff quality prediction with the objective of evaluating if an established tool could be suitable for wide use by stakeholders in Europe.

The research included a thorough analysis of the pre-selected tools, including direct inquires to some of their authors. Although the objective was to test the tools for European use, the outcomes are applicable beyond this geographical context. Therefore, this paper provides new and applicable knowledge regarding the use of tools for road runoff prediction and water management. The conclusions of the work are clear and can support practitioners and researchers all over the world, pointing out directions, and gaps to be filled, regarding the management of road runoff discharges.

The composition and the concentration of pollutants in road runoff are affected by several factors such as the rainfall pattern, road material and construction, vehicles characteristics, ambient conditions and environmental aspects (e.g. Kayhanian et al. 2012; Huber et al. 2016; Bakr et al. 2020). The following constituents may be found in road runoff:

  • (1)

    Conventional water quality parameter: such as total suspended solids (TSS), total dissolved solids (TDS), dissolved organic carbon (DOC), total organic carbon (TOC), chemical oxygen demand (COD), biochemical oxygen demand (BOD), oil and grease (O&G), hardness as CaCO3, temperature and pH.

  • (2)

    Metal constituents: most frequently cadmium (Cd), chromium (Cr), copper (Cu), lead (Pb), nickel (Ni), and zinc (Zn) and, less frequently, aluminium (Al), the metalloid arsenic (As) and iron (Fe) (e.g.: Du et al. 2019).

  • (3)

    Nutrient constituents: nitrates (), nitrites (), ammonium (), total Kjeldahl nitrogen (TKN), total nitrogen (TN), phosphate () and total phosphorus (TP). It is acknowledged that the contribution of N and P from traffic-related sources in runoff is less significant than that from surrounding land uses, such as agriculture and farming (e.g. Kayhanian et al. 2012).

  • (4)

    Other less frequently measured water quality parameters are: faecal indicator bacteria (FIB), toxicity, polycyclic aromatic hydrocarbons (PAHs), herbicides (e.g. glyphosate) and de-icing salts (chloride).

Temperature is among the conventional water quality parameters of road runoff that is correlated with the local climate conditions (Hvitved-Jacobsen et al. 2010). Road runoff is warmed up by contact with the road pavement which has a higher temperature than the rainfall. This process may cause environmental impacts in specific cases of huge runoff volumes discharging into small water masses.

Markiewicz et al. (2017) listed sources of organic pollutants in road runoff, namely: tyre wear, brake lining, integrated vehicle components, car care products, fuels, oils and lubricants, road construction materials, concrete and road paint. The main sources of emitted PAHs were vehicle exhaust gases, followed by tyre wear, motor lubricant oils, road surface wear, and brake linings.

The use of tools for road runoff quality prediction requires a consistent understanding of road runoff characteristics. Some studies on road runoff are dedicated to a specific pollutant source or compound – for instance Klöckner et al. (2020) focused on tyre and road wear particles; or concern a specific road site (e.g. Barbosa & Fernandes 2012) or a given region in a country (e.g. Gan et al. 2008).

Moreover, several authors and national level organisations understood the relevance of addressing comprehensive and/or national level highway runoff characterisation.

The United States Geological Survey (USGS) developed a Highway Runoff Database, comprising more than 100 water quality constituents, monitored at 103 USA road sites (Granato & Cazenas 2009). This massive database has been filled since 1975, and includes a wide range of vehicle, precipitation events and operational conditions.

Crabtree et al. (2006) conducted a 5 year's duration research project in the UK, monitoring six highways, collecting at least 10 events at each site, and analysing 40 constituents in the runoff samples. The results showed that although pollutant concentrations in highway runoff were generally low, most constituents, and particularly metals, showed higher concentrations following winter salt applications. Based on the results, Crabtree et al. (2006) highlighted the relationship between runoff concentration and rainfall intensity.

Bruen et al. (2006), Desta et al. (2007) and Higgins et al. (2008) studied road runoff pollution in Ireland. They gathered monitoring data for approximately 200 individual storm events, over a 15-month period, at four different sites. The results showed clear relationships between storm event characteristics and pollutant concentrations and loadings. Rainfall intensity and volume, and antecedent dry period were pointed out as the principal driving factors of road runoff quality, together with traffic volume.

Van Duijnhoven et al. (2013) acknowledged the large differences among various road sites in The Netherlands and concluded that a nationwide typification of road runoff was not feasible. For instance, the authors observed higher pollutant concentrations in runoff from regular asphalt highways and secondary roads compared to porous asphalt roads. In the same country, Tromp (2005) and van den Berg et al. (2009) found out that the concentrations of PAHs, Cu and Zn in road runoff were higher than the national water quality standards, highlighting the relevance of controlling road runoff discharges.

Ranneklev (2016) showed that levels of metals and organic pollutants in Norwegian roads were high, and exceeding the environmental quality standards from the European Water Framework Directive, particularly for Cu, Zn, Pb, PAH, and suspended solids.

Gan et al. (2008) found that oils and grease, TSS and heavy metals were the dominant pollutants in highway runoff from a region in China. The authors compared road runoff from urban and rural sites and found that EMCs from the urban roads were up to 73% higher, except for pH and TOC. In this study, rainfall depth and the antecedent dry period (ADP) could explain approximately 30 to 70% of the variation in the EMCs except for TOC, TSS, TP and Cr.

In Southern Europe, Barbosa et al. (2011) based on monitoring data collected from 1996 to 2010, including more than 100 events at 10 Portuguese roads, found that concentrations of TSS and Fe exceeded the national limits for wastewater disposal in the environment. These standards were used as indicators due to the lack of specific national regulation for road runoff discharges. This study showed that the ADP was not particularly relevant to explain pollutants concentrations (Barbosa et al. 2011).

Nevertheless, the role of the ADP has been pointed out as relevant; for instance, by Hewitt & Rashed (1992). These authors found out that the behaviour of the particle-associated material (such as Pb) and total PAHs closely followed that of the TSS, allowing to establish a regression model based on the ADP and the discharge of Pb.

Table 1 presents a sample of site mean concentrations of five key pollutants for roads in different countries, illustrating the huge variability of results. It must be noted that the monitoring results were obtained in different dates and time frames. In the table, AADT stands for annual average daily traffic.

Table 1

Examples of site mean concentrations found in the literature

StudyAADT (vehicles)TSS (mg/l)Zn (μg/l)Cu (μg/l)Pb (μg/l)Cd (μg/l)Observations
Barbosa et al. (2011)  3,000−43,000 7–225 76–346 8–72 2–44 – Based on 10 highways in Portugal from 1996 to 2010 
Barrett et al. (1995)  9,000−60,000 19–131 22–208 7–34 7–50 – Based on three highways in Austin, Texas, USA, from 1993 to 1995 
Crabtree et al. (2006)  23,600–83,500 115 140 41 23 – Based on six UK highways and 10 events, from 1997 to 2002 
Drapper et al. (2000)  6,000–50,000 60–1,350 150–185 30–340 80–620 – Based on 21 sampling sites in southeast Queensland, Australia 
Driscoll et al. (1990)  >30,000 142 329 54 400 – Based on 993 highway runoff events, at 31 sites and 11 USA states, from 1975 to 1985 
<30,000 41 80 22 80 – 
Trafikverket (2011)  10,000–15,000 75 100 35 20 0.5 From several studies conducted in Sweden 
15,000–30,000 100 150 45 25 0.5 
>30,000 1,000 250 60 30 0.5 
StudyAADT (vehicles)TSS (mg/l)Zn (μg/l)Cu (μg/l)Pb (μg/l)Cd (μg/l)Observations
Barbosa et al. (2011)  3,000−43,000 7–225 76–346 8–72 2–44 – Based on 10 highways in Portugal from 1996 to 2010 
Barrett et al. (1995)  9,000−60,000 19–131 22–208 7–34 7–50 – Based on three highways in Austin, Texas, USA, from 1993 to 1995 
Crabtree et al. (2006)  23,600–83,500 115 140 41 23 – Based on six UK highways and 10 events, from 1997 to 2002 
Drapper et al. (2000)  6,000–50,000 60–1,350 150–185 30–340 80–620 – Based on 21 sampling sites in southeast Queensland, Australia 
Driscoll et al. (1990)  >30,000 142 329 54 400 – Based on 993 highway runoff events, at 31 sites and 11 USA states, from 1975 to 1985 
<30,000 41 80 22 80 – 
Trafikverket (2011)  10,000–15,000 75 100 35 20 0.5 From several studies conducted in Sweden 
15,000–30,000 100 150 45 25 0.5 
>30,000 1,000 250 60 30 0.5 

Distinct findings and conclusions have been obtained regarding how specific variables affect highway runoff characteristics (e.g. Huber et al. 2016). Even for a specific site it is possible to observe differences in the pollutant concentrations over time. Table 2 illustrates this statement, by comparing the concentrations of pollutants from A1 highway (Portugal) for years 2002 and 2009 (Barbosa et al. 2011; Barbosa & Fernandes 2012). Noteworthy is that the work was done by the same team, using the same equipment and methodology, and the total number of samples are comparable (73 and 93). In spite of that, as shown in Table 2, the differences observed in the range of values emphasise that the SMC is an estimation, based on a given set of data, and not a constant value.

Table 2

Comparison of monitoring results (range of concentrations) for the years 2002 and 2009, for the A1 highway (Portugal) (Barbosa et al. 2011; Barbosa & Fernandes 2012)

A1 AADT = 30,299 Monit. 2002A1 AADT = 27,746 Monit. 2009
Total samples 5–93 37–73 
Total events 11 
pH 6.3–7.4 5.8–7.2 
Conductivity (μS/cm) 124–357 58.0–288.0 
TSS (mg/l) 10.0–872 0.3–350.0 
Fe (mg/l) 0.086–3.030 0.024–7.192 
Zn (μg/l) 62–736 13–834 
Cu (μg/l) 27–76 6–51 
Pb (μg/l) 2–58 2–32 
Cd (μg/l) <0.5 0.026–0.322 
Oil & Grease (mg/l) 3.2–40 0.04–16.00 
A1 AADT = 30,299 Monit. 2002A1 AADT = 27,746 Monit. 2009
Total samples 5–93 37–73 
Total events 11 
pH 6.3–7.4 5.8–7.2 
Conductivity (μS/cm) 124–357 58.0–288.0 
TSS (mg/l) 10.0–872 0.3–350.0 
Fe (mg/l) 0.086–3.030 0.024–7.192 
Zn (μg/l) 62–736 13–834 
Cu (μg/l) 27–76 6–51 
Pb (μg/l) 2–58 2–32 
Cd (μg/l) <0.5 0.026–0.322 
Oil & Grease (mg/l) 3.2–40 0.04–16.00 

As demonstrated by this simple overview of the literature, road runoff characteristics are highly variable and depend on several conditions. There are consensual observations that this diffuse pollution source impacts the environment, and should therefore be correctly evaluated and mitigated. The amount and quality of monitoring data available for each road site, as well as if the information is up to date, are relevant to take conclusions. Research also shows that there is room for continuous improvement on the knowledge of road runoff constituents (e.g. Klöckner et al. 2020).

Search of existing tools and of road runoff monitoring data sets

To carry out this work the first step was the identification of existing tools, as well as of data on road runoff monitoring studies that could later be used to test the selected tools. This was done by searching the international literature through digital research networks, databases, and consulting experts. The search was conducted in 10 languages. In order to ensure objectivity, a procedure for qualitative evaluation of each reference was implemented to assist the selection of the relevant information according to the following scale: (i) Poorly relevant; (ii) Relevant; (iii) Relevant with monitoring data; (iv) Relevant with monitoring data and modelling; and (v) Highly relevant. In the end, from the overall search, a set of 103 literature references with monitoring data and/or prediction tools were selected as relevant.

The characterisation of these 103 references regarding the type and the year of publication are depicted in Figures 1 and 2, respectively. As expected, the most common types of publication are scientific papers and reports, and are English written. They represent almost 80% of the total number of references. Most of the publication dates are after 2001; it should be noted that to set as criteria more recent studies would have limited the results.

Figure 1

Frequency of the references regarding the type of publication.

Figure 1

Frequency of the references regarding the type of publication.

Figure 2

Frequency of the references regarding the year of publication.

Figure 2

Frequency of the references regarding the year of publication.

Selection of the tools

The second step was to deepen the analysis of these 103 references. This allowed the identification of six references, among the 103, corresponding to six tools for road runoff prediction (Table 3).

Table 3

Summary of the six pre-selected tools

Tool/reference(s)Input dataApplicabilityOutput
PREQUALE
(Barbosa et al. 2011
Easily accessible data Equation Predicts SMC for TSS, Zn, Cu, Pb and COD 
HAWRAT
(Crabtree et al. 2008
Easily accessible data Excel spreadsheet EMCs for a large set of pollutants 
Kayhanian et al. (2007)  Accessible data Equation EMCs for a large set of pollutants 
SELDM
(Granato 2013
Some input data is automatically filled in, if a location in the USA is selected. For Europe the user must input several data Graphical User Interface application EMCs for a large set of pollutants 
Higgins (2007),
Higgins et al. (2008), Desta et al. (2007) and Bruen et al. (2006)  
Easily accessible data except for the flow Equation Note: Based on data from only four roads Only predicts EMCs for TSS 
RSS
(Gardiner et al. 2016
Large set of variables need to be imputed There is the need to use GIS to get data and Excel for calculations Only predicts the load for Cu and Zn 
Tool/reference(s)Input dataApplicabilityOutput
PREQUALE
(Barbosa et al. 2011
Easily accessible data Equation Predicts SMC for TSS, Zn, Cu, Pb and COD 
HAWRAT
(Crabtree et al. 2008
Easily accessible data Excel spreadsheet EMCs for a large set of pollutants 
Kayhanian et al. (2007)  Accessible data Equation EMCs for a large set of pollutants 
SELDM
(Granato 2013
Some input data is automatically filled in, if a location in the USA is selected. For Europe the user must input several data Graphical User Interface application EMCs for a large set of pollutants 
Higgins (2007),
Higgins et al. (2008), Desta et al. (2007) and Bruen et al. (2006)  
Easily accessible data except for the flow Equation Note: Based on data from only four roads Only predicts EMCs for TSS 
RSS
(Gardiner et al. 2016
Large set of variables need to be imputed There is the need to use GIS to get data and Excel for calculations Only predicts the load for Cu and Zn 

According to the objective of testing if a tool could be suitable for wide usage by stakeholders in Europe, three criteria were set to evaluate the pre-selected tools, namely: (i) the data requirement (should be easily obtained); (ii) the easiness of applicability (user friendly interface, methodology and/or calculations) and (ii) the consistency of the output results (preferably if SMC is directly calculated). To make the evaluation of the predicting models more objective, each criterion was assessed based on a score from 1 to 3, as presented in Figure 3. Table 4 shows the rating for each criterion and the global score, for each of the six tools.

Table 4

Rating each criterion for the six tools under appraisal

CriteriaPREQUALEHAWRATKayhanianSELDMHigginsRSS
Input data 
Applicability 
Output 
Global 7 8 7 6 5 5 
CriteriaPREQUALEHAWRATKayhanianSELDMHigginsRSS
Input data 
Applicability 
Output 
Global 7 8 7 6 5 5 
Figure 3

Score system used to evaluated the selected tools.

Figure 3

Score system used to evaluated the selected tools.

Based on the evaluation, the following four tools were selected to be tested: PREQUALE (Barbosa et al. 2011), Highways Agency Water Risk Assessment Tool (HAWRAT) (Crabtree et al. 2008), a multiple linear regression (Kayhanian et al. 2007) and Stochastic Empirical Loading and Dilution Model (SELDM) (Granato 2013). They are briefly described below, including the context in which they were established.

Prediction of road runoff quality for Portugal

In the scope of the research project G-Terra funded by the Portuguese Foundation for Science and Technology and coordinated by the National Laboratory for Civil Engineering (LNEC), the runoff from six Portuguese roads was monitored between 2002 and 2010 (Barbosa et al. 2011). In order to capture the variability of the annual mean precipitation in Portugal, the monitoring sites covered regions of the country with rainfall between 560 and 1,200 mm (Barbosa et al. 2011). The AADT of the roads ranged from 6,500 up to 30,300 vehicles per day (Barbosa et al. 2011). The monitored data were based on automatic and continuous sampling of road runoff, and each event was characterised by continuous flow and in situ precipitation data. The SMC of each road site was calculated by averaging 8 to 10 independent runoff events (Barbosa et al. 2011).

This dataset was used to establish the predicting tool PREQUALE that was validated for a data set of ten roads. The PREQUALE multiparametric equation estimates the SMC for key road runoff constituents, namely: TSS, COD, Fe, Zn and Cu (in mg/l):
formula
(1)
where ai, and to are regression coefficients that were calibrated based on the monitored data, DA is the Drainage Area (in km2), IF is the impervious fraction (in %), average rainfall (AR) is the average annual rainfall volume with the same duration as the time of concentration of the basin (in mm) and Pannual is the annual average precipitation (in mm).

The complete description of the tool, including the regression coefficients to be used may be found in Barbosa et al. (2011). This publication presents also Portuguese national guidelines to manage road runoff and protect the environment. Details for some roads can be found in Barbosa & Fernandes (2012) and Antunes (2014).

Highways agency water risk assessment tool for the UK

The HAWRAT was developed by the United Kingdom Highways Agency as a standalone application, aiming at assisting highway designers and operators in the decision if whether or not pollution mitigation measures are needed.

The tool has three steps, namely (i) road runoff pollution prediction; (ii) evaluation of the impacts on the receiving water bodies and (iii) mitigation measures. For the current work, only the first step was relevant.

HAWRAT was based on a dataset from 24 highway sites across England with AADT ranging from 11,000 to 159,000 vehicles (Crabtree et al. 2008). Statistical models were used to define the pollutant concentrations in the road runoff. HAWRAT calculates EMCs based on the following multiple linear regression:
formula
(2)
where, PC is a pollutant constant; CRC is a Climate Region Constant; AADTC is a constant related to the annual average daily traffic; MC is a constant related to the month of the event; MHI is the maximum hourly precipitation (mm); ADP is the antecedent dry period in hours, and γ1 and γ2 are regression coefficients. Further details regarding HAWRAT, such as the constants and the calibration coefficients, may be found in Crabtree et al. (2008).

Multiple linear regression for the USA

The multiple linear regression developed by Kayhanian et al. (2007) predicts EMC for several constituents, including TSS, TDS, Cu, Pb, Ni, Zn. The tool was based on monitoring data from 34 highway sites throughout California (USA), covering a wide range of AADT levels and environmental conditions. The monitoring data were collected from 2000 to 2003 and consisted of up to 8 storm events at each site. The authors found that the parameters with significant impact on the EMCs were total event rainfall (TER); cumulative seasonal rainfall (CSR); ADP; contributing DA, and AADT.

The regression equation established by Kayhanian et al. (2007) is the following:
formula
(3)

The equation was calibrated for 15 different water quality parameters including TSS and the total and dissolved fractions of Cu, Pb and Zn. More details may be found in Kayhanian et al. (2007).

Stochastic empirical loading and dilution model for the USA

The SELDM was developed by the US Federal Highway Administration. It predicts the flow discharge, the EMC and the loads of several pollutants in highway runoff. The input data comprise site and catchment characteristics, rainfall, stormflow and water quality. This tool generates statistical distributions of runoff quality in highway runoff and receiving water bodies, considering the performance of treatment systems. For the present work, only the component of SELDM regarding road runoff quality was considered.

SELDM is based on around 30 years of monitoring campaigns, gathering data for many variables. The tool utilises a Monte Carlo approach, where EMCs from more than a thousand storms are used to characterise highway runoff events at a given site (Granato 2013). The model is not deterministic but it uses a significant amount of monitoring data to give possible EMCs that may occur for a given site.

The tool was established for the USA and has defined regions for the country, called Ecoregions, this allows most parameters to be automatically filled from the database within SELDM. After experimenting to use the model and direct inquiries to the author, it was concluded that the tool can be used for other countries but information regarding weather conditions and road characteristics must be filled manually.

Characterisation of the monitoring data to test the tools

As mentioned in section 3.1, the literature review and experts’ consultation assisted the gathering of a representative European road runoff quality dataset, in order to assess the potential of the four predicting tools. It was possible to collect monitoring data for 22 case studies from seven different countries, namely: Portugal, The Netherlands, Norway, France, Ireland, England and Switzerland. The dataset comprised road runoff monitoring data – including quality parameters for several events – and hourly precipitation time series. The location of the case studies is presented in Figure 4 (red dots), over a map with the annual average precipitation regions.

Figure 4

Annual average precipitation and location of the case studies (red dots). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2021.427.

Figure 4

Annual average precipitation and location of the case studies (red dots). Please refer to the online version of this paper to see this figure in colour: http://dx.doi.org/10.2166/wst.2021.427.

The 22 case studies are located in regions with annual precipitation values ranging from approximately 500 to 1,000 mm, which represent most of the European territory. Note that data from France regard the same road (A11) but count as two cases, because the road pavement type has been changed and two runoff monitoring datasets were used, one regarding porous pavement and other for conventional asphalt. Consequently, in Figure 4, this location stands for two case studies.

These 22 roads cover a wide range of conditions and characteristics as illustrated in Table 5. The following literature references were the sources of most of the characterisation of the 22 sites: Moy & Crabtree 2002a, 2002b, 2002c, 2002d, 2002e, 2002f; Leitão et al. 2005; Higgins 2007; Vollertsen et al. 2007; Barbosa et al. 2011; Brongers 2011; Barbosa & Fernandes 2012; Antunes 2014. Nevertheless, as mentioned above, some of the data summarised in Table 5 was provided by direct contacts with the persons responsible for the monitoring studies (non-published data).

Table 5

Characterisation of the case studies

CodeCountryRoad IDDrainage area (DA) (m2)Impervious fraction (IF) (0-1)Annual precipitation (Pannual)a (mm)Annual average daily traffic (AADT) (no. vehicles)Drainage length (m)Slope (%)AR (mm)
P1 Portugal A 1 22,800 0.41 646 27,746 814 2.95 7.80 
P2 A 2 1,287 528 16,344 117 7.70a 6.00 
P3 A 6 5,580 744 2,918 465 3.00a 5.50 
P4 A 22 15,422 0.85 518 24,000 612 3.40a 7.00 
P5 A 25 287.5 1,014 15,673 25 2.50 6.00 
P6 IP 6 7,280 709 6,539 520 3.30a 6.00 
N1 Netherlands A 27–pervious 48,590 0.5 776 63,000 1,600 0.20a 3.67 
N2 A 27–impervious 30,510 776 63,000 2,700 0.20a 6.00 
N3 Norway E 6 22,000 834 42,000 1,630a 3.40a 2.50 
F1 France A 11–pervious 3,200 0.5 786 24,103 275 2.50 9.00 
F2 A 11–impervious 3,200 786 24,103 275 2.50 9.00 
I1 Ireland M 7–Kildare 14,184 731 27,500 1,200 0.94 3.80 
I2 M 7–Monasterevin 11,368 731 27,500 480 0.50 3.80 
I3 M 7–Portlaoise 9,600 731 27,500 800 0.50 3.80 
E1 England M 4–Brinkworth 8,755 745 70,000 724 1.10a 2.08 
E2 M 4–River Ray 4,348 745 35,000 303 0.66a 1.48 
E3 M 40 58,680 615 78,000 1,800 2.40a 3.27 
E4 A 417 20,232 843 24,000 735 3.10a 1.55 
E5 A 34–Gallos Brook 2,760 660 64,000 250 0.80a 1.19 
E6 A 34–River Enborne 19,425 0.5 635 36,000 1,050 0.19a 5.90 
S1 Switzerland A12–Bümplizstrasse 42,084 986 38,985 1,625 0.43a 2.15 
S2 A1–Gabelbach 12,200 986 39,500 4,300 1.67a 2.60 
CodeCountryRoad IDDrainage area (DA) (m2)Impervious fraction (IF) (0-1)Annual precipitation (Pannual)a (mm)Annual average daily traffic (AADT) (no. vehicles)Drainage length (m)Slope (%)AR (mm)
P1 Portugal A 1 22,800 0.41 646 27,746 814 2.95 7.80 
P2 A 2 1,287 528 16,344 117 7.70a 6.00 
P3 A 6 5,580 744 2,918 465 3.00a 5.50 
P4 A 22 15,422 0.85 518 24,000 612 3.40a 7.00 
P5 A 25 287.5 1,014 15,673 25 2.50 6.00 
P6 IP 6 7,280 709 6,539 520 3.30a 6.00 
N1 Netherlands A 27–pervious 48,590 0.5 776 63,000 1,600 0.20a 3.67 
N2 A 27–impervious 30,510 776 63,000 2,700 0.20a 6.00 
N3 Norway E 6 22,000 834 42,000 1,630a 3.40a 2.50 
F1 France A 11–pervious 3,200 0.5 786 24,103 275 2.50 9.00 
F2 A 11–impervious 3,200 786 24,103 275 2.50 9.00 
I1 Ireland M 7–Kildare 14,184 731 27,500 1,200 0.94 3.80 
I2 M 7–Monasterevin 11,368 731 27,500 480 0.50 3.80 
I3 M 7–Portlaoise 9,600 731 27,500 800 0.50 3.80 
E1 England M 4–Brinkworth 8,755 745 70,000 724 1.10a 2.08 
E2 M 4–River Ray 4,348 745 35,000 303 0.66a 1.48 
E3 M 40 58,680 615 78,000 1,800 2.40a 3.27 
E4 A 417 20,232 843 24,000 735 3.10a 1.55 
E5 A 34–Gallos Brook 2,760 660 64,000 250 0.80a 1.19 
E6 A 34–River Enborne 19,425 0.5 635 36,000 1,050 0.19a 5.90 
S1 Switzerland A12–Bümplizstrasse 42,084 986 38,985 1,625 0.43a 2.15 
S2 A1–Gabelbach 12,200 986 39,500 4,300 1.67a 2.60 

aEstimated through the Google Earth Pro function.

The Intensity–Duration–Frequency (IDF) curves used to calculate AR are available at: Brandão et al. (2001); Korving et al. (2009); http://eklima.met.no; EDF-DTG & Cemagref 1993; and https://www.met.ie.

The selected tools and the literature review point out that the most relevant variables affecting the road runoff quality are: DA, impervious fraction, annual precipitation, annual average daily traffic (AADT), drainage length (DL), slope (S) and AR event.

Analysing Table 5, it is observed that the DA ranged from 290 m2 to 58,680 m2; the annual precipitation (Pannual) from 518 to 1,014 mm and the AADT from 2,918 up to 78,000. For these three variables, the lowest values concern to Portuguese roads and the highest to English roads. The lowest impervious fraction of a monitored road catchment regards a Portuguese road (A1), with 41% of impervious area. The reason for this observation is the fact that the monitoring took place at the entrance of a treatment pond and the road drainage system was prepared to convey all runoff to this point (Barbosa et al. 2011). The two countries providing more road sites are Portugal and England. In this sample, Portugal six sites overall include the widest variety of DA, IF, Pannual and AADT.

Testing the tools

The information available for some of the 22 roads was not enough to fulfil all the input requirements of the tools. Therefore, some assumptions were made, in order to allow the testing of the tools, as explained below. The climate regions established in HAWRAT refer to the division of the UK in four areas according to temperature (hot or cold) and humidity conditions (wet or dry). Following HAWRAT specifications, a climate region was assigned to each of the 22 case studies by extending the lines defined (by the authors of the tool) for the UK.

The DA and DL were known for most of the cases. For one case (E2, Table 5), the value was estimated by multiplying the length of the road by its width. When the DL and the road slope were also missing, they were calculated through Google Earth; the cases where this procedure was used are identified with ‘*’ in Table 5.

The impermeable fraction for roads N1, F2 and E6 was assumed as being 0.5, since the reports of the study only inform that the highways have permeable asphalt (Moy & Crabtree 2002b; Higgins 2007; Brongers 2011).

The annual precipitation was based on the average of annual precipitation records for each site. Intensity–Duration–Frequency (IDF) curves were used to calculate the AR for PREQUALE. For some cases, these curves were not available and AR was estimated as the average of the precipitation volume for the events identified in the precipitation time series.

The implementation of the predicting tools demanded different methods and approaches. SELDM has a graphical interface and all calculations have been done in the software itself. PREQUALE is a multiparametric equation that directly calculates the SMC and it was easily implemented in an Excel spreadsheet. The same was done for Kayhanian et al. (2007) equations although, for this case, the outputs are estimations of EMCs. HAWRAT is itself a spreadsheet established for the UK context, and it was used directly to the six cases that are located in England (E1 to E6). In order to use this tool for the remaining other 16 sites, the HAWRAT equations were implemented in an Excel spreadsheet established for this purpose.

For each case study presented in Table 5, SMCs were calculated by averaging the EMCs, both monitored and predicted, except for the PREQUALE tool that gives directly the SMC. After comparing the available monitoring water quality parameters for the 22 cases and the list of pollutants predicted by the four tools, five constituents were selected for the assessment of the tools namely TSS, Cu, Zn, Pb and Cd. These pollutants are relevant and included as key pollutants in road runoff mentioned in most of the relevant literature references.

Comparison of predicted and real SMCs

The comparison between the SMC predicted by the tools and the real SMC measured at the road sites is presented in Figure 5, for the 22 roads and the five selected pollutants. Some tools do not predict all these five pollutants. Only SELDM and HAWRAT predict cadmium, and lead is predicted by Kayhanian et al. and SELDM.

Figure 5

Comparison of real and predicted SMC for the 22 case studies. (a) TSS; (b) copper; (c) zinc; (d) lead; and (e) cadmium.

Figure 5

Comparison of real and predicted SMC for the 22 case studies. (a) TSS; (b) copper; (c) zinc; (d) lead; and (e) cadmium.

The comparison between monitored and predicted SMC clearly shows that none of the four tools is able to predict the road runoff concentration of pollutants or to capture the trend observed in the monitoring data. The overall predicted values for the 22 case studies present little variability, whereas monitored data show this characteristic variability. This means that the tools under evaluation do not have much sensitivity to the input values. The qualitative analysis based in Figure 5 was supported by a quantitative assessment through the coefficients of determination (R2, Table 6). The R2 are low (ranging between 0.0004 and 0.2890), confirming the empirical observation that no linear relationship exists between the monitored and the predicted values.

Table 6

Coefficient of determination (R2) for the different pollutants and tools

ToolTSSCopperZincLeadCadmium
PREQUALE 0.0116 0.1219 0.0087 – – 
HAWRAT 0.1682 0.0209 0.0111 – 0.0127 
Kayhanian 0.1468 0.0019 0.0072 0.2890 – 
SELDM 0.1803 0.0004 0.0685 0.0575 0.0039 
ToolTSSCopperZincLeadCadmium
PREQUALE 0.0116 0.1219 0.0087 – – 
HAWRAT 0.1682 0.0209 0.0111 – 0.0127 
Kayhanian 0.1468 0.0019 0.0072 0.2890 – 
SELDM 0.1803 0.0004 0.0685 0.0575 0.0039 

A relevant issue to discuss is the ease of using each of the tools. In fact, using a tool in the framework of a research work, as this one, is distinct from routine usage by practitioners from road administration boards, road operators, water managers, etc.

Regarding the PREQUALE, almost all data are easily available. In this case, only the AR needs further calculation and a calibrated IDF curve for the site. The application of HAWRAT is rather straightforward for sites located in the UK because a spreadsheet was developed and ready to use. For countries outside the UK, HAWRAT requires the implementation of new calculations and extra work, such as getting hourly precipitation time series. It was also observed during this study that HAWRAT is quite sensitive to the month of the rainfall event, which intends to represent the seasonality of the rainfall pattern.

Similarly, the implementation of SELDM outside USA is complicated due to the need to collect the climate variables data and input them manually. A recent review study was successful in using SELDM within the USA territory (Bakr et al. 2020), corroborating this fact.

The second criteria to evaluate the tools was the consistency of the output compared to the real monitoring data. The comparison of these data is presented in Figure 4 and validated by Table 6. Both support the deduction that it is not likely that a tool is suitable to predict road runoff pollutant concentrations for wide use by practitioners in Europe. The overview of road runoff characteristics presented in section 2 confirms this conclusion. As precipitation is the driving mechanism for road runoff, it is reasonable to think that this, as well as other climate variables, would be so different across Europe such that determination of differences could not be easily tackled by a single tool.

When discussing the estimation of pollutant concentrations, it must be clear that predicted SMC or EMC are correlated to the dataset that supported the establishment of the tool and subject to temporal variability. The four tools appraised have been established on different dates, and with diverse amounts of data. SELDM is the oldest tool, based on more than 30 years of monitoring data, and on 103 road sites, whereas the other three tools are more recent and based on 34 (Kayhanian et al. 2007), 24 (HAWRAT, Crabtree et al. 2008) or 10 road sites (PREQUALE, Barbosa et al. 2011). Nevertheless, this fact does not seem to have had any influence on the results and the goodness of prediction.

Published work demonstrates that there are variations over time even in the monitoring results for single road. The example from Table 2, a highway in Portugal, corroborates that the SMC is an estimation and should not be taken as an exact value. Therefore, to pursue the task of establishing a ‘constant’ site mean concentration for a specific road site is not reasonable, due to the variability of rainfall events, as well as of other variables, such as traffic volume, surrounding soil use, or wind. Moreover, pollutant concentrations are also influenced by changes caused by the development of new engines, road materials, road furniture, and maintenance activities, among others. Highway and vehicle changes over decades – for instance, the reduction of lead content in fuel – result in alterations in the road runoff quality characteristics. The accurate knowledge of SMC can only be achieved by periodical monitoring work, and it is not expectable that a deterministic tool that has been calibrated and validated for a given site, region or country can be used widely with success.

The goal of this study was to evaluate if existing tools for road runoff quality prediction could be widely used by practitioners across Europe. None of the four evaluated tools could predict the road runoff pollutant concentrations from the 22 road sites, located in seven European countries. As expected, PREQUALE and HAWRAT could predict the concentrations, respectively, from Portugal and the UK where each of the tools have been calibrated. Noteworthy that some authors have previously concluded that even a nationwide standardisation of road runoff is not attainable (e.g. Van Duijnhoven et al. 2013).

Other relevant outcomes from this work are:

  • If the SMC from a given location needs to be accurately known, monitoring work must be done periodically to keep it updated.

  • Decision making concerning road runoff pollution management should not be based on prediction tools that were not established for that same region/country, and/or that are outdated.

  • For most of the cases, sound decisions regarding the control of diffuse discharges from roads can be taken without knowing the precise SMC.

The work presented herein supports placing in the right context the value of tools for road runoff quality prediction. They are not likely to be useful beyond the site, region or country where they have been developed. For the purpose of water resources management and mitigation of impacts from road runoff discharges – also in soils and related environmental compartments – different innovative approaches to manage road runoff pollution should be discussed.

The authors express their gratitude to the partners from the PROPER Project consortium that provided information, including non-published road runoff monitoring data. The authors also thank Masters student Duarte Galhardo who collaborated in the test of the tools.

This work was based in results from two research projects, namely G-Terra (funded by the Portuguese National Foundation for Science and Technology) and PROPER Project (funded by the CEDR – Conference of European Directors of Roads, through the CEDR Call 2016 Water Quality). The work was also supported by LNEC through the Program E3T1 (Integrated Resources Management – Sustainability and Climate Changes).

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

Antunes
P.
2014
The Effect of Saline Deposition on the Characteristics of Road Runoff in Coastal Zones
.
PhD Thesis
,
University of Minho
.
Bakr
A. R.
,
Fua
G. Y.
&
Hedeenb
D.
2020
Water quality impacts of bridge stormwater runoff from scupper drains on receiving waters: a review
.
Science of the Total Environment
726
,
138068
.
https://doi.org/10.1016/j.scitotenv.2020.138068
.
Barbosa
A. E.
&
Fernandes
J. N.
2009
Assessment of treatment systems for highway runoff pollution control in Portugal
.
Water Science and Technology
59
(
9
),
1733
1742
.
Barbosa
A. E.
&
Fernandes
J. N.
2012
Comparison of the pollutant potential of two Portuguese highways located in different climatic regions, Urban Environment Series
. In:
Proceedings of the 10th Urban Environment Symposium
.
Springer
.
Barbosa
A.
,
Telhado
A.
,
Caliço
J.
,
Fernandes
J.
,
Vieira
J.
,
Almeida
L.
,
Whitehead
M.
,
Ramísio
P.
,
Antunes
P.
&
Baguinho
R.
2011
Guidelines for Integrated Road Runoff Pollution Management in Portugal.
Final report of the G-Terra Project
,
Portugal
.
Barrett
M.
,
Malina
F.
,
Charbeneau
R.
&
Ward
G.
1995
Characterization of Highway Runoff in the Austin Texas Area
.
Technical report CRWR 263
,
University of Texas at Austin
.
Brandão
C.
,
Rodrigues
R.
&
Costa
J. P.
2001
Analysis of Extreme Intense Precipitations in Continental Portugal
.
Portuguese Environmental Agency
,
Lisbon
,
Portugal
.
Brongers
I.
2011
Jaarverslag 2010 Monitoring WVO-Vergunning A27. Report
.
Ministry of the Infrastructures and the Environment. Government of The Netherlands
.
Bruen
M.
,
Johnston
P.
,
Quinn
M.
,
Desta
M.
,
Higgins
N.
,
Bradley
C.
&
Burns
S.
2006
Impact Assessment of Highway Drainage on Surface Water Quality
.
Report 2000-MS-13-M2
,
Environment Protection Agency
,
Dublin
,
Ireland
.
Cemagref
1993
Etude des Courbes Intensite-Duree-Frequence de Precipitations dans les Alpes
.
EDF-DTG
,
Rhône-Alpes
.
Crabtree
B.
,
Moy
F.
,
Whitehead
M.
&
Roe
A.
2006
Monitoring pollutants in highway runoff
.
Water and Environment Journal
20
(
4
),
287
294
.
Crabtree
B.
,
Dempsey
P.
,
Johnson
I.
&
Whitehead
M.
2008
The development of a risk-based approach to managing the ecological impact of pollutants in highway runoff
.
Water Science and Technology
57
(
10
),
1595
1600
.
Desta
M.
,
Bruen
M.
,
Higgins
N.
&
Johnston
P.
2007
Highway runoff quality in Ireland
.
Journal of Environmental Monitoring
9
(
4
),
366
371
.
Drapper
D.
,
Tomlinson
R.
&
Williams
P.
2000
Pollutant concentrations in road runoff: southeast Queensland case study
.
Journal of Environmental Engineering
126
(
4
),
313
320
.
Driscoll
E. G.
,
Shelley
P.
&
Strecker
E.
1990
Pollutant Loadings and Impacts From Highway Stormwater Runoff
.
Volume I: Design Procedure. Report No. FHWA-RD-88-006, April 1990
.
Du
X.
,
Zhu
Y.
,
Han
Q.
&
Yu
Z.
2019
The influence of traffic density on heavy metals distribution in urban road runoff in Beijing, China
.
Environmental Science and Pollution Research
26
,
886
895
.
https://doi.org/10.1007/s11356-018-3685-4
.
Erickson
A. J.
,
Weiss
P. T.
&
Gulliver
J. S.
2013
Optimizing Stormwater Treatment Practices: A Handbook of Assessment and Maintenance
.
Springer
.
doi:10.1007/978-1-4614-4624-8
.
European Commission
2000
Directive 2000/60/EC of the European Parliament and of the Council of 23 October 2000 Establishing A Framework for Community Action in the Field of Water Policy
.
Off. J. Eur. Communities 2000
.
Gan
H.
,
Zhuo
M.
,
Li
D.
&
Zhou
Y.
2008
Quality characterization and impact assessment of highway runoff in urban and rural area of Guangzhou, China
.
Environmental Monitoring and Assessment
140
(
1-3
),
147
159
.
Gardiner
L. R.
,
Moores
J.
,
Osborne
A.
&
Semadeni-Davies
A.
2016
Risk Assessment of Road Stormwater Runoff
.
New Zealand Transport Agency research report 585
,
146
pp.
Granato
G. E.
2013
Stochastic empirical loading and dilution model (SELDM) version 1.0.0, U.S. Geological Survey Techniques and Methods, book 4, chap. C3, 112 pages
.
Granato
G. E.
&
Cazenas
P. A.
2009
Highway-Runoff Database (HRDB Version 1.0) A data warehouse and preprocessor for the stochastic empirical loading and dilution model. Washington, DC, USA. Department of Transportation, Federal Highway Administration, FHWA-HEP-09-004, 57 pp
.
Havryshchuk
V.
&
Kaskiv
V.
2020
Mathematical model for the duration of runoff formation determined from the road surface
.
Heliyon
6
.
https://doi.org/10.1016/j.heliyon.2020.e05687
.
Higgins
N.
2007
Analysis of highway runoff in Ireland. Trinity College, Department of Civil
,
Structural and Environmental Engineering
2007
,
443
pp.
Higgins
N.
,
Johnson
P.
,
Gill
L.
,
Bruen
M.
&
Desta
M.
2008
Highway Runoff in Ireland and Management with a French Drain System
. In:
Proceedings of the 11th International Conference Urban Drainage
,
Edinburgh, Scotland
.
Hvitved-Jacobsen
T.
,
Vollertsen
J.
&
Nielsen
A. H.
2010
Urban and Highway Stormwater Pollution. Concepts and Engineering
.
CRC Press. Taylor and Francis Group
,
Boca Raton, Florida, USA
,
347
pp.
Kayhanian
M.
,
Singh
A.
,
Suverkropp
C.
&
Borroum
S.
2003
Impact of annual average daily traffic on highway runoff pollutant concentrations
.
Journal of Environmental Engineering
129
(
11
),
975
990
.
Kayhanian
M.
,
Suverkropp
C.
,
Ruby
A.
&
Tsay
K.
2007
Characterization and prediction of highway runoff constituent event mean concentration
.
Journal of Environmental Management
85
(
2
),
279
295
.
Kayhanian
M.
,
Fruchtman
B. D.
,
Gulliver
J. S.
,
Montanaro
C.
,
Ranieri
E.
&
Wuertz
S.
2012
Review of highway runoff characteristics: comparative analysis and universal implications
.
Water Research
46
(
20
),
6609
6624
.
Klöckner
P.
,
Seiwert
B.
,
Eisentraut
P.
,
Braun
U.
,
Reemtsma
T.
&
Wagner
S.
2020
Characterization of tire and road wear particles from road runoff indicates highly dynamic particle properties
.
Water Research
185
,
116262
.
Korving
H.
,
Noortwijk
J.
,
Van Gelder
P.
&
Clemens
F.
2009
Risk-based design of sewer system rehabilitation
.
Structure and Infrastructure Engineering
5
(
3
),
215
227
.
https://doi.org/10.1080/15732470601114299
.
Leitão
T.
,
Barbosa
A. E.
,
Henriques
M. J.
,
Ikavalko
V. M.
&
Menezes
J.
2005
Evaluation and Environmental Management of Road Runoff. Final Report
.
Report 109/05 – NAS, Laboratório Nacional de Engenharia Civil
,
April
,
243
pp.
Markiewicz
A.
,
Björklund
K.
,
Eriksson
E.
,
Kalmykova
Y.
,
Strömvall
A.
&
Siopi
A.
2017
Emissions of organic pollutants from traffic and roads: priority pollutants selection and substance flow analysis
.
Science of The Total Environment
580
,
1162
1174
.
Moy
F.
&
Crabtree
R.
2002a
Monitoring of Pollution From Highway Runoff. A34-Gallos Brook
.
Environment Agency R&D Report
.
Moy
F.
&
Crabtree
R.
2002b
Monitoring of Pollution From Highway Runoff. A34-River Enborne
.
Environment Agency R&D Report
.
Moy
F.
&
Crabtree
R.
2002c
Monitoring of Pollution From Highway Runoff. A 417-River Frome
.
Environment Agency R&D Report
.
Moy
F.
&
Crabtree
R.
2002d
Monitoring of Pollution From Highway Runoff. M4-Brinkworth Brook
.
Environment Agency R&D Report
.
Moy
F.
&
Crabtree
R.
2002e
Monitoring of Pollution From Highway Runoff. M4-River Ray
.
Environment Agency R&D Report
.
Moy
F.
&
Crabtree
R.
2002f
Monitoring of Pollution From Highway Runoff. M40-Souldern Brook
.
Environment Agency R&D Report
.
Ranneklev
S.
2016
Et litteraturstudium over forurenset snø fra bynære områder: stoffer, kilder, effekter og håndtering
.
NIVA Report 27 pp
.
Trafikverket
2011
Vägdagvatten – Råd och rekommendationer för val av miljöåtgärd
.
Trafikverket
2011
,
112
.
Tromp
2005
Helofyteninfiltratiesystemen voor zuivering van wegwater. Onderzoek naar het milieurendement van een Helofyteninfiltratiesloot langs de A1 in 't Gooi
.
Utrecht University
,
July, 2005
, Utrecht.
van den Berg
G. A.
,
Hunneman
H.
&
Langemeijer
H. D.
2009
Emissie van verontreinigingen door run-off en verwaaiing van dunne deklagen. Pilot Noordoostpolder. KWR 09.072. December 2009
.
van Duijnhoven
N.
,
Klein
J.
&
den Hamer
D.
2013
Update Verontreinigingsbeeld Afstromend Wegwater
.
Deltares rapport 1208038-000-ZWS-0003
.
Vieira
R.
,
Fernandes
J. N.
&
Barbosa
A. E.
2013
Evaluation of the impacts of road runoff in a Mediterranean reservoir in Portugal
.
Environmental Monitoring and Assessment
185
(
9
),
7659
7673
.
Vollertsen
J.
,
Åstebøl
S.
,
Coward
J.
,
Fageraas
T.
,
Madsen
H.
,
Nielsen
A.
,
Hvitved-Jacobsen
T.
2007
Highway and urban environment
. In:
Proceedings of the 8th Highway and Urban Environment Symposium
(
Morrison
G.
&
Rauch
S.
eds),
Springer, Dordrecht
.
This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/).