The Science of the Total Environment 319 (2004) 215–224

Progressive changes in water and sediment quality in a wetland system for control of highway runoff H. Pontiera, J.B. Williamsa,*, E. Mayb a

Department of Civil Engineering, University of Portsmouth, Lion Gate Building, Lion Terrace, Portsmouth, Hants, PO1 3HF, UK b School of Biological Sciences, University of Portsmouth, Lion Gate Building, Lion Terrace, Portsmouth, Hants, PO1 3HF, UK Received 5 December 2002; received in revised form 18 June 2003; accepted 27 June 2003

Abstract Innovative wetland based systems were designed and installed on the Newbury Bypass, Berkshire, England to provide flow balancing and pollution control for road runoff. The systems were monitored over 18 months to evaluate performance, pollutant removal processes and offer improved design and operation codes for this new application of wetlands. Water quality, sediment accumulation rates, and metal concentrations in size-fractionated, settling solids and deposited sediments were determined in parts of the system to provide information on spatial and temporal variability. The results presented here show that over the long term, there were progressive changes in parts of the system for BOD and COD and for metal concentrations in the sediment fractions, which occurred with linear (or semi loglinear) rates, despite variability in flow rates, retention times and in pollutant loading to the system. Future work will continue monitoring to increase the data set, examine possible processes contributing to the regression constants, and test the potential use of the regressions in system modelling. Attempts at modelling road runoff treatment using wetlands must allow for progressions, since the systems can only be effective if they retain removed metals in the sediment sink. 䊚 2003 Elsevier B.V. All rights reserved. Keywords: Constructed wetland; Highway or road runoff; Metals; River pollution control; Sediments

1. Introduction The recognition that the discharge of runoff from rural roads carrying in excess of 30 000 vehicles per day can damage rivers has led to increasing interest in mitigation methods (Luker and Montague, 1994). Flood risk and scouring of river-beds due to elevated peak discharges from road drainage has traditionally been controlled by *Corresponding author. E-mail address: [email protected] (J.B. Williams).

flow balancing basins. These accommodate the runoff volume to allow discharge to the river at predetermined rates, similar to those of the preroad catchment. The quality of the runoff can also have impacts. Pollutant loads include sediments, with associated metals or hydrocarbons, and ionic concentrations from road de-icing salts, which can lead to loss of species diversity and amenity value (Luker and Montague, 1994). Constructed wetlands have been developed to treat a variety of wastewaters, including sewage,

0048-9697/04/$ - see front matter 䊚 2003 Elsevier B.V. All rights reserved. doi:10.1016/S0048-9697(03)00410-8

216

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

agricultural, industrial and mine wastewaters (Nuttall et al., 1997), it has been suggested that similar systems could be successfully used to control road runoff. However, road runoff is a highly variable and intermittent feedstock, which contrasts to more uniform or predictable flows and loadings of the other wastewaters treated using constructed wetlands. There have been several promising studies on the use of wetlands for road runoff (Lee et al., 1997a,b; Nix et al., 1988; Shutes et al., 1997; Munger et al., 1995), but as yet no established design criteria have emerged. The controversial Newbury Bypass, in Berkshire, England, crossed several high quality rivers and runoff control was a high priority (Pontier, 2002). Therefore, the bypass was equipped with innovative road runoff control systems, based on best available management practices. The systems offered a series of components to give progressive treatment or pollutant removal as well as flow balancing. Each of the 9 control systems differed in exact specification, offering generally unidirectional flow through a series of components. In downstream order, these comprised of an oil separator, sediment trap with grass slope leading to a constructed wetland, which was nested within a 1y25 or 1y100 year storm-balancing basin. The grass slope was planted with a mixture of species, including Alopecurcus sp. and Holcus sp. The design of the runoff control systems attempted to combine storm flow balancing and pollutant removal functions. The opportunity was taken to study the purpose-built systems and how they developed and matured after the road opened. This paper concerns a study centred on a monitoring programme to evaluate system performance, identify pollutant removal processes and provides improved design and operation codes. The latter aim included an investigation of potential modelling tools. 2. Method The 9 systems were designated letters for identification (A, B, C, D, E, ‘FG’, H, J, K), but security concerns means that their exact co-ordinates cannot be revealed. Treatment system C was the subject of an 18-month monitoring programme,

commencing with the opening of the new road in November, 1998. System C received runoff from 1.6 Ha porous asphalt road catchment, with a peak discharge, continuation flow and time to concentration of 190 lys, 20 lys, and 12 min, respectively. The wetland was planted predominantly with Glyceria maxima, on benches of different depths, as shown in Fig. 1. Glyceria was chosen as it is more common in the general area than Phragmites or Typha, which are the predominant species used in constructed wetlands (Nuttall et al., 1997). Management plans for the wetland are being developed in response to ongoing monitoring. Balancing was provided by the storage in the basin configuration and a circular discharge orifice of 100 mm ID, to retain permanent water so the pool was 1 m deep, rising to 2 m deep at the 1y 25-year storm storage level. Water was collected as grab samples (single collection event) from two positions in each part of the system, the sediment trap, wetland inlet, pool and outlet, as shown in Fig. 1. Settling solids were sampled using catch traps, which were plastic cups of 95 mm ID, fixed to retrieval rods placed at fixed positions near the water sample points. The material accumulated in the traps was collected on each sample visit. Deposited, or bed sediments, were sampled using a plastic scoop to collect surface deposits from an area 30=80 mm and 30 mm deep (Pontier, 2002). The bed locations were based on a stratified random sampling. The wetland was stratified into three zones (in, pool, out) and random locations selected within these zones. Samples were stored in a cool box (approx. 4 8C) and dark and analysed or pretreated within 24 h. Various sensitive parameters were recorded using hand-held probes and meters at time of sampling, this included pH (Palin 900 meter and HI1332 probe), redox potential (Whatman stick meter) and dissolved oxygen (50B meter and 50 BYSI Ohio 45387 probe). BOD and COD were measured using standard methods (Greenberg et al., 1992) and the Hach娃 micro kit, respectively. At time of sampling, 250 ml aliquots of the water samples were filtered in series through 1.2 GF then 0.45 mm nucleopore Whatman filters using a hand pump, to separate size fractions of )1.2, 1.2 to 0.45 and -0.45 mm

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

217

Fig. 1. Design features and sampling locations of treatment system C.

corresponding to particulate matter, ‘macrocolloids’ and dissolved fractions, respectively. The residues were stored on the filter papers and sealed in plastic bags at approximately 4 8C. The final filtrate was preserved with a few drops of concentrated HNO3 in plastic containers. All the separated fractions were retained for metal analysis. The sediment samples were wet sieved using sample point water to separate grain size fractions ) and -63 mm. These were termed coarse grains and fines, respectively. Excess water was decanted after settlement, and the fractions were dried in the dark at 80 8C to calculate the dry weight, proportion of fines and coarse grains and the sediment accumulation rate (ACR, the mass accumulated per unit area over the time interval between sample retrieval). Solid samples for metal analysis were digested in a boiling water bath with concentrated HNO3 (Greenberg et al., 1992). Metal analysis of the

filtrate and solids was performed using acetylene flame AAS (Pye Unicam) according to standard methods (Greenberg et al., 1992). Blanks, fresh standards and a ‘standard sediment’ made from powdered -63 mm roofing slate were run concurrently. All glassware were acid leached and triple rinsed with de-ionised water. The size weighted or total metal concentration of whole sediments was calculated from the sum of each product for fines and coarse grains, of the proportion of grain size and metal concentration of the grain size. MINITAB娃 software was used for statistical analysis and graphical presentation. The data sets were grouped according to parts of the system and scanned for Pearson correlations, followed by graphical assessment and least squares regression. Correlation between BOD, COD or sediment metal concentration (y, the response variable) and time (as days elapsed since sampling began x, the predictor variable) indicated time trends, which

218

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

were assessed by linear regression. MINITAB reports the equation of the fitted line (ysmxqc), plus the r-value indicating the goodness of fit along with a P value which represents the probability of making a Type 1 error, the rejection of the null hypothesis was made at P values of less than 0.05 (MINITAB, 1997y1998b). 3. Results Fe and Zn were detected in the water column but Cu was below the detection limit in all phases (i.e. -1 mgyl dissolved, -40 mgyl macrocolloid and -100 mgyl particulate matter fractions). Fe and Zn were detected in the dissolved (-0.45 mm) fraction, with means of 304(294) mg Feyl and 10(9) mg Znyl in the sediment trap and wetland collectively (n, 110). However, the majority of these metals were found in the solid phase ()45 mm), with means of 946(743) mg Feyl and 157(164) mg Znyl. In the system the runoff was generally well oxygenated with neutral to basic pH, with mean (and S.D.) in the sediment trap and wetland of 75.6 (S.D. 22.5) and 73.5 (S.D. 19.5)% saturation and 8.4 (0.8) and 7.4 (0.7) pH units, respectively. COD was an order of magnitude greater than BOD (Fig. 2), this ratio is typical for road runoff but the overall levels of organic matter were lower than reported in other studies (Ellis and Revitt, 1991). The BOD and COD in the wetland were highly variable, however, there was a general

Table 1 Regression coefficients for log BOD and log COD against time in days c (log mgyl)

m (log mgyl d)

R

Log BOD Sediment trap Wetland inlet Wetland pool Wetland outlet

0.370 0.228 y0.156 0.037

0.001 0.001 0.002 0.001

0.56 0.72 0.88 0.62

Log COD Sediment trap Wetland inlet Wetland pool

1.31 1.30 0.941

0.001 0.001 0.002

0.50 0.47 0.74

P-0.0.1.

increase with time of both parameters as shown in Fig. 2. When these increases were considered for each part of the system, they followed a semi-log linear relationship, with P values of -0.001 in almost all occasions. The results of the regression analysis of Log BOD and Log COD against time (in days) are shown in Table 1, with. BOD and COD generally increased over time at similar rates, as shown by the similar regression coefficient (m) for the log BOD and log COD vs. time elapsed. The doubling time was estimated as 300 days for both COD and BOD in the sediment trap and inlet, but as 150 days in the pool. BOD correlated with COD in the inlet, and pool (rs0.877, P-0.001 and rs0.645, P-0.001, respectively) while COD correlated with dissolved oxygen in the sediment

Fig. 2. Time trends of log BOD and log COD in parts of the System.

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

219

Table 2 Mean settling solid accumulation rates (ACR) in parts of the system Grain size

Fines ACR, gym2 day

(S.D.)

N

Coarse ACR, gym2 day

(S.D.)

N

Sed trap Inlet Pool Outlet

21.01 16.18 19.23 1.30

25.31 20.74 15.52 1.04

28 28 14 28

8.38 5.59 3.85 1.24

9.14 7.78 1.25 1.07

28 28 16 28

Fig. 3. Time trends of Cu in sediments in different parts of the System.

Fig. 4. Time trends of Cu in the pool and Fe in the outlet.

trap and inlet (rsy0.502, P-0.006 and rsy 0.740, P-0.001, respectively). The means of the sediment accumulation rates (ACR) of fines and coarse settling solids are given in Table 2. The greatest ACR and variability occurred in the sediment trap for both fractions with general attenuation downstream, with the lowest ACR in the wetland outlet. Variability in sediment ACR and attenuation downstream is

described elsewhere (Pontier et al., 2001, 2002). The ACRs did not show correlation with time, water quality descriptors, metals in water or sediments. However, the metal concentrations of the sediments showed many correlations with time and water quality in parts of the system. The metal concentration in sediment size fractions and whole sediments (Figs. 3–5), were comparable to those reported in other studies (Lee

220

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

coarse and consequently total deposits, as shown in Fig. 5. In the inlet, Zn in fine deposits correlated with settling solid counterparts (rsy0.533, P0.003) and with the proportion of fines in deposits (rs0.574, P-0.001). 4. Discussion

Fig. 5. Zn concentrations in inlet deposits vs. time.

et al., 1997a,b). Note that the total metal (in whole sediments) is related to the proportion and metal concentration of each grain size. The proportions of fines tended to increase downstream towards the outlet (Table 2). In the sediment trap, Cu was the only metal in sediments showing time trends (with P-0.001), increasing in the coarse and total settling solids and deposits, illustrated by Fig. 3 and by the correlation coefficients of Cu concentrations against time in Table 3. In the sediment trap the Cu in coarse settling solids correlated with deposited coarse solids. Fig. 3 and Table 3 also show that Cu increased over time in the coarse deposits of the inlet and outlet. In the inlet, Cu concentrations in coarse deposits correlated with the BOD, COD and DO (%) of the water phase, with r values of 0.734, 0.672, and y0.680, respectively, and P-0.001 for all associations. Fig. 4 gives the time trends for Cu in the pool, where the rate of increase of Cu concentration in fine and total deposits was greater than for the other sediments in other parts of the system (Table 3). Fig. 4 also shows that Fe increased in fine deposits and total deposits of the outlet, and here Fe in fine deposits correlated with the settling counterpart, and in coarse deposits (Table 3). Concentrations of Fe and Cu in the coarse deposits also had a significant positive association (rs 0.615, P-0.001, ns30). Zn showed contrasting behaviour to Cu in the inlet. Concentrations fluctuated over time in fine,

The neutral to basic pH reflects the chalkderived soils and weathering of fresh concrete of the catchment and drainage system. The pH range and oxidising conditions probably favoured partitioning of metals to the solid phase, by precipitation and sorption to solids such as organic matter, Fe oxyhydroxides and clay minerals (Langmuir, 1997; Lee et al., 1997a,b; Donahoe and Liu, 1998). The ‘dissolved fraction’ could have included colloidal material, which was indistinguishable from the thermodynamically defined dissolved metals by the separation technique (Lee et al., 1997a,b). Removal of solids would thus promote removal of metals, but more acid catchments may not show similar metal behaviour (Langmuir, 1997; Luker and Montague, 1994). The settling behaviour of discrete particles obeys Stokes law, and depends on the size and density of the particles as well as the density and Table 3 Regression coefficients for sediment metal concentrations against time in days c (mgyg)

m (mgyg d)

R

14

0.03

0.66

14

0.05

0.74

20

0.03

0.62

8

0.04

0.70

19

0.03

0.59

9 12

0.07 0.06

0.59 0.59

Iron

c (mgyg)

m (mgyg.d)

Outlet: fine deposits Outlet: total deposits

19.1 17.0

0.04 0.04

Location: grain size Copper Sediment trap: coarse deposits Sediment trap: coarse settling Sediment trap: total deposits Inlet coarse: Deposits Outlet: coarse Deposits Pool: fine deposits Pool: total deposits

0.52 0.52

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

viscosity of the water column medium (Nix et al., 1988; O’Melia, 1985; Van Capellen and Wang, 1995). Road sediments consist predominantly of fines, which contain of an array of materials including large proportions of organic material and clay that are associated with the greatest metal loads. Other studies have demonstrated correlations between organic or clay fractions of fines and metal concentration (Roger et al., 1998). Sediment transport from the road depends on antecedent dry periods for accumulation, particle size and type, and the hydrological regime of the road and drainage system (Luker and Montague, 1994). Within the wetland, autochthonous organic material and local runoff from the slopes of the storm storage basin also contributes to sediments. The former is likely to have a strong seasonal pattern due to the growth cycles of vegetation in temperate climates. The sediment ACR is best referred to as the apparent ACR, due to the difficulty in distinguishing sediment sources and between settlement and re-suspension (Lee et al., 1997b). The samples were collected under different conditions of storm flow balancing or between storm storage. Variable flow rates in the system during the rise and fall of the storm hydrograph and quiescent conditions between storms influenced the sediment loading to the system, but the combined effects of preferential transport of more buoyant solids and settlement of less buoyant solids influence sediment loadings downstream. Furthermore, variable flow rates, and visible development of preferred paths resulted in scouring and erosion, with re-suspension and transport of deposited sediments. This transport, and additional autochthonous material and solids from localised basin runoff contributed to spatial and temporal variability in sediment ACR. The physical presence of the plants also enhances sediment removal by filtration and localised effects on flow. Some evidence suggests that particle removals over distance downstream in wetlands may be modelled by first order decay equations (Wong et al., 2000), but road runoff flows and particles may be too variable for application of similar models (Nix et al., 1988). As solids settle from the water column, the accumulation and subsequent release of the organic

221

fraction of accumulated material may contribute to the progressive increase in BOD and COD in the water column, shown in Fig. 2. The linear regression fit of Log BOD and COD against time, although not perfect, is remarkable considering the many sources of variation found in runoff. The general constant rate of increase in the oxygen demand occurred despite periodic runoff flushing through the system, and variability in quantity and quality of sediment loadings, flow rates and other conditions such as seasonal variation in temperature or availability of organic material. Other studies have demonstrated natural colloids or dissolved organic matter (DOM) fluxes between the water column and sediments over time, and the sorption of these colloids to solids (Valsaraj et al., 1993). Organic matter provides an important energy source to sediment bacteria, and the decomposition affects metal behaviour in deposited sediments. Indirect effects are due to the consumption of oxygen with decay of organic matter. Slow rates of oxygen diffusion through the sediment depth profile may promote the development of pH and redox gradients and anaerobic degradation processes (Faulkner and Richardson, 1989). This influences reductive dissolution and oxidative precipitation of inorganic FeyMn oxyhydroxides, which are ubiquitous in aquatic environments and excellent metal scavengers (Langmuir, 1997; Lee et al., 1997a,b; Donahoe and Liu, 1998). These redox reactions, coupled with diffusion of mobile, reduced metals, govern the metal distribution in the sediment depth profile. FeyMn redox reactions may also be microbially mediated (Donahoe and Liu, 1998; Van Capellen and Wang, 1995). More direct effects are due to the consumption of the organic matrix, which may lead to an increase in the metal concentration associated with non-detrital fraction as reported in detention ponds (Lee et al., 1997a). The change in sediment or pore water constituents over time, was put simply by Van Capellen and Wang (1995) as Eq. (1): dC s DJq8Sq8R dt 8

(1)

222

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

Where dC is the change in concentration of the constituent per unit volume of sediment, dt is the time interval, DJ is the divergence of the local transport flux affecting the constituent, S is the source or sink of the constituent from non-local transport, R is the rate of biogeochemical transformation In the wetland sediments, DJ could be due to exchange with the water phase or adjacent sediments due to processes such as advection, porewater diffusion, local mixing; S the background sediment pollutant level and also the pollutants present in solids washed off the road; and R represents processes such as chemical or redox reaction, microbial degradation, which can release or transform pollutants. Therefore, dCydt at any location is the net effect of a number of processes causing increases and decreases in concentration. These generalised terms give a structure that can assist in interpreting the patterns of pollutants in the wetland sediments (Pontier, 2002). In system C, the time trends shown by Cu and Fe suggest that the dCydt term is constant, but different for each metal in different grain sizes and parts of the system despite the potential variability in the other terms. The time trends for Cu (shown in Table 3) and the correlation with BOD, COD and DO suggests that this metal was most strongly influenced by degradation of organic matter, or the term for R, in different sediments in different parts of the system. However, the S term may also be important. The settling material may add Cu to deposits (Sigg, 1985). Estimation of the Cu concentration of accumulated settling solids divided by the time taken shows that the accumulation of settling solids could increase the Cu concentration of deposited sediments in the inlet and pool by 0.05 and 0.1 mgyg sediment per day, respectively. This is greater than the rate determined from the time trends for the coarse and total deposits, and suggests that here may be considerable transport of material, and copper to other parts of the treatment system. Fig. 5 shows that the patterns of Zn concentrations in the inlet contrast to those shown by Cu. The behaviour of Zn is more difficult to interpret. Zn may be associated with solids and such as organic matter or clay particles, which may be

easily transported (Roger et al., 1998). Zn bearing particles may be delivered to or exported from the inlet deposits under some flow conditions by preferential re-suspension and downstream transport. Alternation between flooding and air exposure of the shallow inlet sediments may promote micro-scale pH changes in the pore waters, favouring mobilisation of Zn especially in the shallow inlet zone (Tessier et al., 1989). The third possibility is that accumulation of other materials, such as increasing Cu concentration, effectively diluted Zn concentrations over time. All of these processes may act to different degrees at different times, so that Zn concentrations appear to show fluctuations rather than time trends. This suggests variability in all terms of the equation at different times, depending on the sensitivity of Zn to the conditions. The time trends shown by Fe in the outlet (Table 3) are probably related to sorption or scavenging of Fe from the water column. Only 1.5 kg sediment ym2 and 39.9 g Feym2 were delivered to the outlet, giving a potential increase in deposited Fe concentration of 0.04 g Feykg sediment per day. This is remarkably similar to the rate of increasing Fe concentration shown by the fine and total deposits in Table 3. The accumulation in the outlet contrasts to 17.6 kg sedimentym2 and 189 g Feym2 delivered to the sediment trap, or a potential increase in Fe concentration of 0.01 g Feykg sediment per day over the monitoring period. Although it appears that the S term is strongly related to the increasing concentration of deposited Fe, in the outlet, the terms for DJ and R, probably have greatest influence on the Fe concentrations of surface sediments throughout the wetland. The sediment processes include sorption or ion exchange reactions and advectionydiffusion of cations and solutes through the pore waters. It is the R term, the effects of bacterial activities, which governs the Fe distribution in sediments, both in the vertical gradient and in different ecological zones (Donahoe and Liu, 1998). Fe has a particularly complex set of possible biochemical and physico-chemical interactions and the heterogeneous wetland environment means that many of these are probably operating simultaneously (Donahoe

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

and Liu, 1998; Lee et al., 1997a,b). The relative rates of these reactions (i.e. the sum of DJ and R) determines whether the wetland sediments act as a source or a sink for metals in the water column (Donahoe and Liu, 1998). The rates of bacterial activity depend on many factors, such as temperature, substrate availability, and presence of activity inhibitors, this means that this could also change over time. 5. Conclusions The time trends suggest that simple models of metal removal by the system cannot be based on black box studies of input–output, but must include a component to account for the long retention times of solids and progressive changes as organic matter accumulate and decays. Ideally, effective treatment would promote retention of sediments and associated metals by provision of appropriate conditions to promote settlement, while preventing re-suspension and promoting conditions which favour partitioning of metals to the solid phase. However, the essential storm flow balancing function of road runoff control systems means that the system is periodically flushed with waters of variable quality. It is especially prone to perturbations during salt flushes of winter de-icing salts, changes in redox and pH with fluctuation in water levels and seasonal variation in availability of organic leaf litter. These fluctuations in conditions affect metal partitioning, both in settling solids and the post-depositional environment. Furthermore, the variable flow rates during storms and quiescent conditions between storm storage affect sediment transport or settlement behaviour, making it difficult to identify original road runoff solids and estimate their residence time in the system. The progressive changes in metal concentrations of sediments appear to occur at a linear rate and may indicate the influence of processes involving biological decay of organic fractions andyor residence times as well as loadings of the solids in parts of the system. The spatial variability in sediment accumulation may favour metal enrichment processes in sediments near the loading zone, where fluctuations in water levels and storm flow balancing may pro-

223

mote the enrichment, and provide a source of enriched material for re-suspension and downstream transport during subsequent storms. Wetland-based systems for control and treatment of road runoff can only be effective provided that removed metals are retained by the sediment sink. Given longer-term study, the time trends showing accumulation of metals in sediments could offer a way to predicting metal concentrations of settling solids and deposited sediments in the long term, but some metals appear to be less predictable than others. Thus, a better understanding of the progressive changes, especially enrichment patterns of settling and post depositional solids, could contribute to improved design and operation. Acknowledgments The research was supported by Mott MacDonald, the Highways Agency, the Environment Agency, the Worshipful Company of Paviors and the University of Portsmouth with help from Mr P. Wilson (HA), Mr C. Walker (MM), Dr C. Mant (UoP) and Mr M. Woodhatch. References Donahoe RJ, Liu C. Pore water geochemistry near the sediment-water interface of a zoned, freshwater wetland in the southeastern United States. Environ Geol 1998; 33(2y3):143 –153. Ellis JB, Revitt DM. Drainage from roads control and treatment of highway runoff, Report NRA 43804-MID.012, National Rivers Authority, London, 1991. Faulkner SP, Richardson CJ. Physical and chemical characteristics of freshwater wetland soils. In: Hammer DA, editor. Constructed Wetlands for Waste Water Treatment, Municipal, Industrial, Agricultural—Proceedings from 1st International Conference on Constructed Wetlands for Waste Water Treatment, Chattanooga, Tennessee, 1988. Chelsea, Michigan: Lewis Publishers, 1989. p. 41 –71. Greenberg AE, Clesceri LS, Eaton AD. (editors) Standard Methods for the Examination of Water and Waste Water 18th Edition, American Public Health Association (APHA) 1992. Langmuir D. Aqueous Environmental Geochemistry. New Jersey: Prentice Hall Inc, 1997. Lee P-K, Baillif P, Touray J-C. Geochemical behaviour and relative mobility of metals (Mn, Cd, Zn and Pb) in recent sediments of a retention pond along the A-71 motorway in Sologne, France. Environ Geol 1997a;32(2):142 –153.

224

H. Pontier et al. / The Science of the Total Environment 319 (2004) 215–224

Lee P-K, Touray J, Baillif P, Ildefonse J. Heavy metal contamination of settling particles in a retention pond along the A71 motorway in Sologne, France. Sci Total Environ 1997b;201:1 –15. Luker M, Montague K. Control of Pollution from Highway Discharges CIRIA Report 142. London: CIRIA, 1994. MINITAB Inc. (1997y1998 b) Users Guide 2: Data Analysis and Quality Tools. Munger AS, Shutes RBE, Revitt DM, House MA. An assessment of metal removal from highway runoff by a natural wetland. Water Sci Technol 1995;32(3):169 –175. Nix ST, Heaney JP, Huber WC. Suspended solids removal in detention basins. J Environ Eng 1988;114(6):1331 –1342. Nuttall PM, Boon AG, Rowell MR. Review of the Design and Management of Constructed Wetlands CIRIA Report 180. London: CIRIA, 1997. O’Melia CE. The influence of coagulation and sedimentation on the fate of particles associated pollutants and nutrients in lakes. In: Stumm W, editor. Chemical Processes in Lakes. Wiley 1985. Pontier H. An Evaluation of Combined Conventional and Wetland Systems fpr the Control and Management of Road Runoff, PhD Thesis (University of Portsmouth) 2002. Pontier H, Williams JB, May E. Metals in combined conventional and vegetated road runoff control systems. Water Sci Technol 2001;44(11–12):607 –614. Pontier H, Williams JB, May E. Behaviour of sediments associated metals in a wetland based road runoff control system on the Newbury Bypass, 8th International Confer-

ence on Wetland Systems for Water Pollution Control, (IWA). Tanzania 2002;1:688 –697. Roger S, Montrejaud-Vignoles M, Andral MC. Mineral, physical and chemical analysis of the solid matter carried by motorway runoff water. Water Res 1998;32(4):1119 –1125. Shutes RBE, Revitt DM, Munger AS, Scholes LN. The design of wetland systems for the control of urban runoff. Water Sci Technol 1997;35(5):19 –25. Sigg L. Metals transfer mechanisms in lakes, the role of settling particles. In: Stumm W, editor. Chemical Processes in Lakes. Wiley 1985. Tessier A, Carignan R, Dubreuil B, Rapin F. Partitioning of zinc between the water column and the oxic sediments in lakes. Geochimica et Cosmochimica Acta 1989;53:1511 – 1522. Valsaraj KT, Thoma GJ, Porter DD, Reible DD, Thibodeaux LJ. Transport of dissolved organic carbon-derived natural colloids, from bed sediments to overlying water: laboratory simulations. Water Sci Technol 1993;28(8–9):139 –148. Van Capellen P, Wang Y. Metal cycling in surface sediments modelling the interplay of transport and reaction. In: Allen E, editor. Metal Contaminated Aquatic Sediments. MI USA: Ann Arbor Press Inc, 1995. Wong THF, Breen PF, Lloyd S, Walker T, Dahnke B, Wooton R. Suspended solids removal in storm water wetlands: Quantifying the role of aquatic macrophytes, 7th International Conference on Wetland Systems for Water Pollution Control, (IWA), November 11–16, 2000, Lake Buena Vista, Florida, Vol. 3, pp, 1545–1552.