Science of the Total Environment 334 – 335 (2004) 261 – 270 www.elsevier.com/locate/scitotenv

The performances of vegetative treatment systems for highway runoff during dry and wet conditions D.M. Revitt a,*, R.B.E. Shutes a, R.H. Jones a, M. Forshaw b, B. Winter b a

Urban Pollution Research Centre, Middlesex University, Queensway, Enfield, Middlesex, EN3 4SA, UK b Environment Agency, Howbery Park, Wallingford, OX10 8BD, UK Accepted 1 April 2004

Abstract The performances of two different highway runoff treatment systems, a horizontal subsurface flow-constructed wetland and a vegetated balancing pond, are described. Both systems have been assessed by collecting inlet and outlet grab samples during wet and dry weather conditions, and automatically controlled storm event samples have been obtained for the constructed wetland. Removal efficiencies are discussed for BOD (grab samples only), suspended solids, Cd, Cr, Cu, Ni, Pb, Zn, nitrate and sulphate, and explanations are offered for the trends observed under different weather conditions. The large variabilities in the removal efficiencies derived for both treatment systems, based on the analyses of grab samples, make accurate comparisons of the performances difficult and also raise concerns about using this type of sampling approach for this purpose. Treatment systems are required to function satisfactorily during the increased inlet loadings experienced during storm events, and this is shown to be the case for the constructed wetland for the majority of the monitored pollutants. The large removal efficiency ranges for five separate storm events, exhibited by Cu and Pb, are discussed and compared to the other monitored pollutants which showed positive median wet weather removal efficiencies of between 43% and 85%. Despite the existence of performance fluctuations, the generally low monitored inlet concentrations in the highway runoff indicated that the pond discharges did not threaten the environmental quality of the receiving waters. D 2004 Elsevier B.V. All rights reserved. Keywords: Constructed wetland; Metals; Nutrients; Pollutant removal efficiency; Vegetated balancing pond

1. Introduction Vegetative systems, which have been used for the treatment of highway runoff include filter strips, swales, detention basins, retention basins (balancing ponds) and constructed wetlands (Revitt and Ellis,

* Corresponding author. Tel.: +44-208-411-5308; fax: +44-208411-5440. E-mail address: [email protected] (D.M. Revitt). 0048-9697/$ - see front matter D 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.scitotenv.2004.04.046

2001). They can all be defined as sustainable drainage systems (SUDS) and have the potential to create wildlife habitats and to enhance the aesthetic aspects of the highway environment. Filter strips and swales are vegetated surface features, which can provide conveyance, storage and infiltration facilities for highway discharges. Basins and constructed wetlands can be categorised as systems which are designed to store and treat received stormwater prior to releasing it at an appropriate rate once the peak flow has passed. Extended detention basins specifically provide flow

262

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

attenuation and, by increasing the detention time by up to 24 h, allow removal of relatively fine suspended solids. Retention ponds are permanent water bodies which, as a result of increased runoff storage times, offer increased treatment through settlement of finer particles and also biodegradation processes. The incorporation of aquatic macrophytes enables additional treatment through biofiltration, adsorption and biological uptake. Purposefully designed constructed wetlands utilising various reed species are considered to be able to provide a similar level of treatment over shorter periods of time, although often at the expense of incorporating an equivalent storage capacity. This paper compares the performances of a vegetated balancing pond and a subsurface flow reed bed with respect to the treatment of runoff from the recently constructed Newbury Bypass in the south of England. The pollutant removal efficiencies of detention basins have been shown to be dependent on residence time with suspended solids removal decreasing from a maximum of 70% to 20% as containment time reduces from 48 to 2 h (Stahre and Urbonas, 1990). The removal efficiencies of hydrocarbons, BOD and metals (Zn and Pb) were reduced by similar factors. Hares and Ward (1999) have indicated removal efficiencies in excess of 84% for a range of 11 metals in a 500-m2 detention pond receiving runoff from a major motorway. For stormwater passing through a wet detention pond, Farm (2002) has reported average reduction rates of 26– 84% for total metal content, 67% for Ntot, 78% for Ptot and 92% for COD. In an extensive study of retention ponds in the Florida area, Yousef et al. (1996) have reported average sedimentary accumulation rates of 1.3, 13.8 and 6.9 kg/ha year for Cu, Pb and Zn, respectively. Similar metal accumulation rates have been observed in French studies of retention basins (Lee et al., 1997) highlighting the need for regular inspection and maintenance of these systems. Pontier et al. (2001) have tracked the changes in Zn, Fe and Cu sediment concentrations across a vegetated balancing pond and shown an increase between inlet and outlet with the metals being predominantly associated with size fraction below 63 Am. The use of constructed wetlands for the treatment of highway runoff, although well established in the United States (Kadlec and Knight, 1997), is a relatively new technology in the UK (Shutes et al., 2001).

More extensive data sets have been reported for urban storm-water treatment with removal efficiency ranges for subsurface flow systems of 67 –97% for TSS, 25– 98% for Ntot, 5 –94% for Pbtot and 10– 82% for Zntot (Strecker et al., 1992). The variability in performance was attributed to a number of factors including shortcircuiting, short detention and contact times, pollutant remobilisation and seasonal vegetation effects. There is a need to design constructed wetlands for the treatment of highway runoff to address these and other factors, and Shutes et al. (1999) have commenced this process. This paper contributes further to this approach and therefore also addresses the requirements of the Environment Agency for England and Wales, which include assessing methods for improving surface water management with an emphasis on sustainable drainage systems.

2. Methodology 2.1. Site description The A34 Newbury Bypass is a 13.5-km porous asphalt-surfaced dual carriageway which opened in November 1998. The drainage system includes a series of nine vegetated balancing ponds located adjacent to the highway. Each balancing pond incorporates a front – end oil interceptor and rectangular concrete sediment trap followed by a grassed slope to deliver the highway runoff to the treatment system. Two ponds have been studied by Middlesex University. Pond B exists as originally designed with a sloping profile, which is able to support a variety of fringing macrophytes in the shallows with the predominant species in the main water body (depth; 0.05 – 1.0 m) being Phragmites australis. The original design of Pond F/G was amended by retrofitting to produce a subsurface flow-constructed wetland containing a gravel substrate preceded by a small settlement pond. The constructed wetland within Pond F/G was planted with both P. australis (front half) and Typha latifolia (final half). 2.2. Sampling and analysis Water samples were collected at the inlets and outlets to each pond to enable the assessment of

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

pollutant attenuation/enhancement in both treatment systems. The inlet sampling points were located in the silt traps adjacent to the inflow pipe, and the outlet samples were taken from the exit pipe prior to discharge into a receiving stream. Samples were collected over a 33-month period from December 1998 to August 2001. Grab samples were obtained from Pond F/G on 24 occasions, and from Pond B on 16 occasions. In addition to this routine monitoring at both ponds, five storm events were sampled at Pond F/G with concurrent measurement of inlet flow rates obtained via a V-notch weir located on the sill of the silt trap and a pressure transducer probe/data logger positioned above the silt trap. An automatic water sampler, triggered by the rising water level in the silt trap, was used to collect 24 inlet samples over the 4-h period of a storm event. Outflow samples were collected automatically at 2-h intervals following an initial delay, after the commencement of the storm event, of 6 h. This provided a comprehensive sample coverage of the period following the completion of the storm event consistent with tracer studies, which showed that a typical time taken for a storm to pass through Pond F/G was between 24 and 54 h. Inlet and outlet pollutant loadings were calculated for each storm event from the relevant pollutographs. Dissolved oxygen, temperature and pH were measured in situ during each routine-monitoring visit. The grab samples collected at the inlets and outlets to both ponds were analysed in duplicate for suspended solids, BOD, conductivity, nutrients (nitrate, sulphate, phosphate) and total metals (Pb, Cu, Cd, Cr, Ni, Zn). The automatically collected storm event samples for Pond F/G were analysed in duplicate for suspended solids, conductivity, nutrients and total metals. Nutrient concentrations were determined by ion chromatography after making the appropriate dilution where necessary. Water samples for metal analysis were acidified with nitric acid, taken to dryness and then redissolved in 1% nitric acid. Graphite Furnace Atomic Absorption Spectroscopy was used for the determination of Pb, Cu, Cd, Cr and Ni, whereas Zn was measured by Inductively Coupled Plasma Atomic Emission Spectroscopy. The authenticities of the metal analytical procedures were verified by applying them to the determination of appropriate standard reference materials.

263

3. Results and discussion 3.1. Routine-monitoring results The trends in the inlet and outlet concentrations to both ponds during nonstorm event sampling/monitoring are reported in this section. The measured pH values throughout the 33-month monitoring period showed no significant deviations from neutrality, ranging from 6.3 to 8.1, and indicating effective buffering of incident rainfall by the respective catchment areas. The absence of acidic conditions reduces the possibility of metal release from the sedimentary to aqueous phases within the treatment systems (Yousef et al., 1990). A general trend of decreasing dissolved oxygen levels has been observed over the monitoring period, particularly at the pond outlets where occasional anoxic conditions have occurred. This intermittent but high demand for oxygen due to plant uptake and biodegradation processes is of concern because of its possible impact on the receiving water environment. There is evidence for increased BOD levels coincident with the large decreases in dissolved oxygen between inlet and outlet positions. The presence of decomposing algal mats in the silt traps and settlement ponds (Pond F/G) will certainly contribute to the BOD levels, although these only rose above 20 mg/l on a few occasions. The same processes were also believed to contribute to higher outletsuspended solids concentrations with the consequence that there was little evidence of removal within the treatment systems. However, the outlet-suspended solids concentrations rarely exceeded 30 mg/l. Measured conductivity values were normally between 600 and 800 AS cm 1 at both pond monitoring locations, except during some winter visits when higher values resulted due to the wash-off of deicing salts applied to the road surface. The incidences of high conductivity were matched by the expected increases in chloride concentrations with no evidence for significant removal occurring in either pond. The elevation of conductivity values by chloride ions emphasises the importance of planting macrophytes with a high tolerance of salinity in ponds receiving highway runoff. P. australis is known to grow well in brackish waters and therefore is appropriate for use in Pond B and at the front end of Pond F/G. Following the establishment of the vegetated systems, the inlet

264

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

nitrate levels decreased from a high of over 60 mg/l, and there was evidence for removal of this pollutant with the surface system pond (Pond B) behaving more efficiently. Sulphate concentrations have remained within the range, 50– 200 mg/l, with no clear removal patterns being discernible from the full range of routine-monitoring results. Phosphate concentrations were routinely below the analytical detection limit at all monitored locations. The measured total aqueous metal concentrations were all lower than those that would be expected in the runoff from a road with the traffic density associated with the Newbury Bypass (>35,000 vehicles day 1). This is probably caused by the relative newness of the road surface, which still retains the ability to preferentially adsorb metal ions rather than allowing them to be transported in drainage waters. Similar low runoff concentrations of metals have been reported by other workers researching the performance of other balancing ponds associated with the Newbury Bypass (Moy et al., 2003; Pontier et al., 2001). The observed averages and ranges of the metal concentrations for both ponds are shown in Table 1. These data show the low metal levels as well as their extreme variability. The reasons for some of these variations with respect to the individual metals have been previously discussed with respect to their behaviour in the individual ponds (Shutes et al., 2001). Within each pond, there is a tendency for the removal of each metal to be more significant as the inlet metal concentration increases. However, based on consideration of the metal concentrations in grab samples collected at the inlet and outlet positions, there is no clear evidence of consistent removal efficiencies. Previous research has identified a higher constructed

Table 1 Means (with standard deviations) and ranges of metal concentrations detected in Ponds B and F/G during the 33-month monitoring period Average concentration(Ag/l) Range of concentrations with standard deviation (Ag/l) Copper 6.74 F 5.20 Cadmium 0.66 F 0.90 Chromium 3.65 F 4.30 Nickel 5.75 F 3.78 Lead 1.41 F 1.99 Zinc 21.61 F 14.82

nd – 25.6 nd – 2.3 0.1 – 27.6 nd – 14.3 nd – 6.1 4.0 – 79.5

wetland retention of metals during summer and autumn compared to winter months (Goulet et al., 2002), but this trend is not confirmed by this study. Zinc was found to be the metal with the routinely highest concentrations, and the measured levels for all collected samples are shown in Fig. 1 for both Ponds B and F/G. The lack of a consistent trend between the inlet and outlet positions for this metal is clearly evident. For Pond F/G, there are 15 instances when zinc concentrations decreased between inlet and outlet positions, 7 instances when an increase in concentration occurred and 2 instances when identical concentrations were recorded. The comparable numbers for Pond B were 11, 4 and 1. Thus, the use of routinemonitoring data indicates the existence of a random removal of zinc within both treatment systems. The existence of similar trends for all pollutants suggests that routine monitoring, in the form of analysed grab samples from the inlet and outlet positions, cannot be reliably used to determine treatment performance. This is not completely unexpected as the consecutive collection of samples does not take into account the residence time of pollutants within the system. The unreliability will be further compounded by the collection of grab samples during wet weather conditions when incoming pollutants will be strongly influenced by rainfall conditions, whereas outlet samples may relate to runoff which entered the pond under relatively dry conditions. Therefore, the wet weather results obtained by grab sampling have been excluded by only considering data collected when there was negligible rainfall during the 48 h prior to sampling. The reduced data sets of nine for Pond F/G and seven for Pond B have been analysed statistically. The trends in the derived removal efficiencies are represented by box plots produced in MINITABk (Fig. 2). The continued variability of the dry weatheronly data is clearly demonstrated by consideration of the interquartile ranges (rectangular boxes) and the positions of the median values (horizontal lines within boxes) and mean values (solid dots). The extending vertical lines indicate upper and lower limits beyond the quartiles as F 1.5 times the interquartile ranges. The greatest variability occurs for Cr in Pond B, and the emphasis on negative removal can be explained by two high outlet Cr concentrations early in the monitoring period. In contrast, Pond F/G exhibited a consistent positive removal for Cr with

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

265

Fig. 1. Comparison of inlet and outlet concentrations of total aqueous zinc concentrations at Ponds F/G and B.

little variability in values for the different sampling dates. The behaviour of Ni in Pond F/G is also characterised by consistently positive removal efficiencies with mean and median values between 60% and 75%. The other parameters showing predominantly positive removals are Cd (in Pond F/G), Ni, SO4 and Pb (in Pond B) and NO3 and Zn (in both ponds). However, the variability in each of these cases does not generate confidence regarding the validity of using periodic dry weather grab sampling to assess the performance of treatment ponds receiving highway runoff. The mean removal efficiencies of

both suspended solids and BOD were indicated to be negative under dry weather conditions, although the median value was positive for BOD in Pond F/G. The absence of suspended solids removal during low flow conditions is unexpected and is probably a function of the generally low inlet concentrations which were monitored. Application of the Anderson – Darling normality test to the dry weather removal efficiency data sets provides evidence that all parameters are nonnormally distributed for both ponds. Analysis of the data sets using the Wilcoxon Signed Rank Test shows that there

266

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

Fig. 2. Statistical summary of removal efficiencies for Ponds B and F/G during dry weather conditions.

is no significant overall difference between the removal efficiencies for Ponds B and F/G. The median removal efficiency for Pond F/G of 8.6% is significantly greater ( p < 0.05) than zero removal, which was not the case for Pond B, where the median removal efficiency was 5.6%. Therefore, both ponds exhibit low but positive removal efficiencies under dry weather conditions based on the analysis of grab samples collected simultaneously at the inlets and outlets. However, the large variabilities in the data raise questions concerning the validity of using instantaneous pollutant concentrations to determine re-

moval efficiencies despite the cost-effectiveness of this approach. This is particularly relevant when the monitored incoming water does not directly relate to that which is leaving the system. A more realistic approach would be to determine input and output pollutant loadings and to match these in terms of the retention time of the treatment system. This procedure has been attempted for Pond F/G during storm events when the pollutant removal capability of runoff is at its most critical due to the presence of high inlet loadings and the increased possibility of resuspension within the treatment ponds.

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

3.2. Storm event results Five discrete storm events have been monitored for Pond F/G by automatically collecting inlet samples during each rainfall event, and also automatically collecting the corresponding outlet samples in the poststorm period. Temporal plots of loading rates enable inlet loadings during the storm to be calculated and to be compared with the poststorm outlet loadings to determine a realistic removal efficiency. The relevant plots for Cu for a storm event occurring on 26 February 2001 are shown in Fig. 3. The outlet loadings were calculated from the area under the outlet

267

chemograph between 24 and 54 h after the commencement of the storm event, as the outflow during this period has been shown by tracer studies to correspond to the storm runoff entering the constructed wetland. The removal efficiencies calculated from the relevant inlet and outlet pollutant loadings have been analysed statistically for the five monitored storm events (Fig. 4). The box plot representations have the same meaning as those shown in Fig. 2. BOD was not monitored for the storm events, because it was not possible to refrigerate the automatically collected samples during the short storage period between sampling and analysis.

Fig. 3. Inlet and outlet chemographs for the storm even occurring on 26 February 2001 at Pond F/G.

268

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

Fig. 4. Statistical summary of the wet weather removal efficiencies for Pond F/G.

The pollutants demonstrating consistent positive removal efficiencies for all storms and also limited variability about the mean and median values were cadmium, chromium, nitrate and suspended solids. Nickel, zinc and sulphate exhibited overall positive mean and median values, although the variability indicated some inconsistency between storm events for these parameters. In the case of Zn, the removal efficiency distribution was strongly influenced by the storm event occurring on 22 April for which a value of 94.2% was recorded based on inlet and outlet loadings of 0.84 and 1.63 g, respectively. A more exaggerated disparity was observed for Pb for this storm with 96.8 mg entering the treatment system, and 372.6 mg leaving it to give an increasing contribution by Pond F/G of 285%. There was also a contribution during this storm event for Cu ( 40.3%). The reason for this unusual behaviour by certain metals during this April storm event is believed to be associated with the elevated chloride concentrations (up to 2910 mg/l) and conductivity values (up to 7200 AS cm 1), which were monitored at the outlet. The tenfold increase in both of these parameters between inlet and outlet strongly suggests that accumulated salts due to winter deicing activities are being removed by the increased flows (up to 17 l/s) associated with this storm event. The ability of chloride ions to complex with certain metals and increase their solubility has been shown

by metal speciation studies (Bewers and Yeats, 1989) and appears to be an important mechanism for Pb, Zn and Cu in this instance. The reason for the lack of impact by the increased conductivity/ chloride levels on Cr, Ni and Cd behaviour within Pond F/G is unclear, although in the case of Cd, it may be related to the low input loading (68.7 mg) in comparison to the other metals. Mitchell and Karathanasis (1995) have demonstrated the variability in metal removal from a sodium chloride-enriched wastewater passing through surface flow- and subsurface flow-simulated wetlands. The behaviour of Cu in Pond F/G is unusual and unexpected in that for three out of the five monitored storms, negative removal efficiencies were recorded resulting in the overall negative mean and median values shown in Fig. 4. Copper is known to have a strong affinity for organic materials (Bryan and Langston, 1992), and therefore, the presence of increasing amounts at the outlet to a constructed wetland is difficult to explain. Although the macrophytes demonstrated some progressive plant tissue uptake, there was no evidence of Cu accumulation on the substrate. Release of Cu due to decay of macrophyte tissue and filamentous algae in open water areas would be expected during the winter months, and although one winter storm exhibited increased outlet Cu loadings, this process also occurred for a summer and a spring storm.

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

3.3. Comparison of dry and wet weather removal efficiencies The median removal efficiencies for Pond F/G during dry weather monitoring and based on storm loadings are shown in Table 2. Chromium and nickel appear to be removed equally well during both types of weather conditions, with Pb showing similar but poor removal during dry and wet conditions. In contrast, Cd and nitrate are removed more efficiently during storm events with examination of the data by the Mann – Whitney test showing that this difference is significant ( p < 0.05). There is a similar emphasis on more favourable removal under wet weather conditions for Zn, suspended solids and sulphate; although in each of these cases, the comparison with dry weather conditions is not significantly different. Only Cu is predicted to have a higher removal during dry weather conditions, and this is a consequence of the unexpected behaviour previously described for Cu during storm event monitoring. The considerations described above assume that the analysis of grab samples obtained simultaneously from inlet and outlet positions during dry weather conditions can be compared directly to storm event monitoring. Ideally, a series of time-based inlet samples should have been collected and compared with similarly obtained outlet samples taking into account the residence time of Pond F/G under dry conditions. This would have provided a direct comparison between the performances during the two types of extreme weather conditions. In the absence of such a comparison, an Table 2 Comparison of dry and wet weather removal efficiencies for Pond F/G Parameter Cd* Cr Cu Ni Pb Zn SS NO*3 SO4

Median dry weather removal efficiency 0.0 47.2 4.0 72.6 0.0 5.3 9.7 5.3 5.4

Median wet weather removal efficiency 84.7 42.8 40.3 77.5 9.1 66.2 57.7 65.5 44.1

* Indicates that the wet removal is significantly better than the dry removal (Mann – Whitney test).

269

explanation of the results is not straightforward. Thus, the indicated preferable removal of the two monitored nutrients (nitrate and sulphate) during wet weather would not have been expected as more time for plant uptake would be available during dry conditions, and a previous study of the performance of a constructed wetland treating urban runoff has suggested that nitrate removal occurred primarily between, rather than during, storm events (Carleton et al., 2000). Similarly, the settling out of suspended solids should be more efficient under quiescent conditions, whereas a higher removal during storm events is predicted by the results. However, this phenomenon is partly a function of the inlet-suspended solids concentrations which did not exceed 20 mg/l for routine monitoring but regularly approached 100 mg/l during storm runoff conditions. Lead is commonly found to be strongly associated with particulate material (Revitt and Morrison, 1987), but the absence of a marked inlet concentration difference between dry (maximum 4.5 Ag/l) and wet (maximum 10.1 Ag/l) weather conditions results in a median removal efficiency value (9.1%) for the latter conditions, which is only slightly higher than the dry weather value (0.0%). Cadmium is the most effectively removed metal during storm events and is the most significantly different from the dry weather results ( p < 0.05; Mann – Whitney test). This finding is again unexpected given the predicted high solubility of Cd in highway and urban runoff (Revitt and Morrison, 1987). Metal removal by a constructed wetland receiving highway runoff can generally be seen to be efficient during carefully designed storm event-monitoring conditions (Table 2), with only Cu showing an aberrant behaviour and Pb demonstrating a small positive removal.

4. Conclusions The results presented in this paper highlight the limitations of utilising analysed grab samples as the basis for estimating pollutant removal efficiencies between the inlet and outlet of a water treatment system. This is particularly true in wet weather conditions, and there is only a marginal improvement when dry weather conditions prevail both before and during sampling. Comparison of the performances of a constructed wetland and a vegetated balancing pond

270

D.M. Revitt et al. / Science of the Total Environment 334 – 335 (2004) 261–270

receiving highway runoff, with respect to the removal of pollutants during dry weather, shows the presence of similar but pronounced variabilities with only the constructed wetland showing a statistically significant removal efficiency greater than zero. Carefully planned storm event sampling can provide reliable removal efficiencies calculated from inlet and outlet loadings. Data obtained for the constructed wetland show evidence of increased or equivalent removal of several pollutants during storm events in comparison to dry weather conditions. Despite the variability in the pollutant removal efficiencies, particularly during dry conditions, the generally low inlet concentrations result in pond discharges, which do not threaten the environmental quality of the receiving waters.

References Bewers JM, Yeats PA. Transport of river derived trace metals through the coastal zone. Neth J Sea Res 1989;23:359 – 68. Bryan GW, Langston WJ. Bioavailability, accumulation and effects of heavy metals in sediments with special reference to UK estuaries—a review. Environ Pollut 1992;76:89 – 131. Carleton JN, Grizzard TJ, Godrej AN, Post HE, Lampe L, Kenel PP. Performance of a constructed wetland in treating urban stormwater runoff. Water Environ Res 2000;72:295 – 304. Farm C. Evaluation of the accumulation of sediment and heavy metals in a storm-water detention pond. Water Sci Technol 2002;45:105 – 12. Goulet RR, Pick FR, Droste RL. Test of the first-order removal model for metal retention in a young constructed wetland. Ecol Eng 2002;17:357 – 71. Hares RJ, Ward NI. Comparison of heavy metal content of motorway stormwater following discharge into wet biofiltration and dry detention ponds along the London Orbital (M25) motorway. Sci Total Environ 1999;235:169 – 78.

Kadlec RH, Knight RL. Treatment wetlands. Boca Raton (USA): Lewis, 1996. Lee PK, Touray JC, Bailif P, Ildefonse JP. Heavy metal contamination of settling particles along the A-71 motorway in Sologne France. Sci Total Environ 1997;201:1 – 15. Mitchell LK, Karathanasis AD. Treatment of metal-chlorideenriched waste-water by simulated wetlands. Environ Geochem Health 1995;17:119 – 26. Moy F, Crabtree RW, Simms T. The long term monitoring of pollution from highway runoff: Final report. Environment Agency R&D Technical Report P2-038/TR1. Swindon, UK: WRC Publications, 2003. Pontier H, May E, Williams JB. Constructed wetlands for the treatment of runoff from the Newbury Bypass. J Inst Water Environ Mang 2001;15:125 – 9. Revitt DM, Ellis JB. Drainage, runoff and groundwater. Guidelines for the environmental management of highways. London: Institution of Highways and Transportation; 2001. p. 67 – 102. Revitt DM, Morrison GM. Metal speciation variations within separate stormwater systems. Environ Technol Lett 1987;8: 373 – 80. Shutes RBE, Revitt DM, Lagerberg IM, Barraud VCE. The design of vegetative constructed wetlands for the treatment of highway runoff. Sci Total Environ 1999;235:189 – 97. Shutes RBE, Revitt DM, Scholes LNL, Forshaw M, Winter B. An experimental constructed wetland system for the treatment of highway runoff in the UK. Water Sci Technol 2001;11/12: 571 – 8. Stahre P, Urbonas B. Stormwater detention for drainage, water quality and CSO management. New Jersey: Prentice Hall, 1990. Strecker EW, Priscoll ED, Shelley PE, Gaboury DR. Use of wetlands for stormwater pollution control. US Environmental Protection Agency, 1992. Yousef Y, Hvitved-Jacobsen T, Harper H, Lin L. Heavy metal accumulation and transport through detention ponds receiving highway runoff. Sci Total Environ 1990;93:433 – 40. Yousef YA, Baker DM, Hvitved-Jacobsen T. Modeling and impact of metal accumulation in bottom sediments of wet ponds. Sci Total Environ 1996;189/190:349 – 54.