Fate and Behaviour of Pollutants in a Vegetated Pond System for Road Runoff.
Georgios Roinasa, Alexandros Tsavdarisa, John B Williamsb, Catherine Mantc
aPhD Student, bPrincipal Lecturer, cResearch Fellow, School of Civil Engineering and Surveying, University of Portsmouth, Portland Building, Portsmouth, PO1 3AH, UK
Corresponding Author:
John Williams
School of Civil Engineering and Surveying
University of Portsmouth
Portland Building
Portsmouth
PO1 3AH
Email:
Tel: 00 44 23 92 842404
Fax: 00 22 23 92 842521
Notation
ADP – Antecedent Dry Period
B - Basin
BOD – Biochemical Oxygen Demand
COD – Chemical Oxygen Demand
DO – Dissolved Oxygen
DP – Daily Precipitation
DRO – Diesel Range Organics
EC – Electric Conductivity
ERO – Engine Oil Range Organics
GC-MS – Gas Chromatograph-Mass Spectrometer
HEM – Hexane Extractable Material
ICP-MS – Inductively Coupled Plasma - Mass Spectrometer
PAHs – Polycyclic Aromatic Hydrocarbons
PGM – Platinum Group Metals
Q – Discharge
ST – Sediment Trap
TSS – Total Suspended Solids
VSS – Volatile Suspended Solids
SuDS – Sustainable Drainage Systems
WR = River Wallington
Keywords
SuDs, Road Runoff, First-flush, PAHs, Metals
Abstract
1 Introduction
There is increasing concern about the impact that runoff from developed areas, especially from roads, can have on aquatic ecosystems [1]. The variety of pollutants (includingsilt, organic matter, heavy metals and arange ofhydrocarbons and polycyclic aromatic hydrocarbons (PAHs)[4]) and intermittent loadings mean that predicting the impact is very complex.Most contaminants are attached to particles in the µm range and are attributable to automobile activity [6]. The pollution risk to receiving waters has led to the development of a more sustainable approach to the management of urban runoff; the Sustainable Drainage System concept (SuDS). One of the common SuDS featuresfor road runoff is a wet pond; these have various layouts, vegetation cover and inflow/outflow control devices[7].Although hydrological attenuation is relatively easy to address in design, pollutant removal is more difficult and treatment can be highly dependent on site specific parameters
Wet balancing ponds are one of the most efficient systems for treating highway runoff[8], as the complex ecology exposes pollutants to a range of treatment mechanisms. This includes adsorption, volatilization, photolysis, biodegradation and sedimentation . Many studies have reported large reductions in organic loads in ponds, often in excess of 90% [9,10]. However removal of suspended solids has often been less, with typical reductions of 60-65%, possibly due to biogenic debris from plants [8, 10].
Studies on heavy metals removal have been more varied, with reported removal rates between 0-84% [8, 9, 11]. This may be due to the variety of chemical properties of heavy metals affectingtheir behavior in SuDS. For many metals sorption and subsequent sedimentation are the dominant removal mechanism, most metals are associated with 0.45-75 μmparticles[6]. Particles greater than 125 µm are readily trapped by vegetated systems, where in 6-32µmrange particles are often difficult to remove [12]. The varying behavior of these particles, and in-situsorption/desorption processes, have given rise to varied patterns of metal deposition. Some studies have found that sediments located at the inlet of a pond have the highest metal concentrationswhile others have shown found the opposite [13]. Longer term studies have generally found an increase in concentrations associated with sediment accumulations over time [13, 14].
Hydrocarbons are also of concern in road runoff, with wear of road surfaces, tyres and brake pads combined with combustion by-products and “drip loss”[15]. Loadings vary with road use and vehicle behaviour.These emissions can be classified in a variety of ways by either carbon number (e.g. diesel range organics (DRO) C10-22 and engine oil range organics (ERO) C22-C40) or for pollutants of most concern by class e.g. PAHs. These organics have an even wider range of behaviour than metals, with adsorption characteristics, solubility, volatility and biodegradation characteristics giving a variety of possible fates.PAHs are of particular concern due to their toxicity [16]. A range PAHs are found in urban runoff with higher concentrations usually found the particulate phase [18].However, due to analytical costs,organic pollutants are often not included in monitoring of urban runoff and further investigation has been identified as a research priority [19, 20, 21].
Vegetated detention ponds for road runoff are highly dynamic systems, with varyingflow, pollutant loads, plant growth and other seasonal factors, such as road saltingand temperature. Despite numerous studies [11, 22, 23]there are no established design criteria for optimising treatment processes. The toxicity and transport of metals and hydrocarbons depends upon their bioavailability which is influenced by variations in speciation, pH, redox potential, particle size distribution, organic matter and temperature [24]. Therefore an ongoing concern is that under unfavourable conditions accumulated pollutants could be released giving shock loads to receiving waters or exposing organisms such as invertebrates, fish, birds, amphibians and mammals attracted by the habitat potential of the ponds [22].
This paper aims to investigatethe fate of pollutants in a vegetated detention pond and contribute to further understanding of the treatment mechanismswhich will inform design and operational codes.
2 Materials and Methods
2.1 Study Site
Thestudy site at Waterlooville, Hampshire, UK (Latitude=50.881315, Longitude= -1.037575) is a greenfield Major Development Area (MDA) for 2,500 new homes. The impermeable clay soil means the site will be served by storage SUDs. This study considers a detention pond which was constructed to receive runoff from the access roads prior to house construction. The pond receives runoff from an urban commuter road (B2150) and roundabout with peak hour flows of approximately 3,100 cars and 100 lorries, which equates to a daily traffic flow of 40,000 (unpublished Traffic Survey 2009, Mayer Brown Ltd.). While mainly free flowing, peak time traffic is characterised by stop startcongestion associated with nearbytraffic lights.
The vegetated pond system receives runoff after aswale which receives piped inflow, as well as direct precipitation along its length. Figure 1 shows a schematic plan of the system with sampling points labelled by letters. The plan area is 51x26 m², the two flow-balancing basins are connected by a berm with an invert level of 1.1 m relative to the pond bed. The berm is designed to reduce short circuiting and increase the overall retention times. The basins have fixed sediment traps (ST)to collect settling solids. Basin 1` (B1) has 2 sediment traps (D and E) and Basin 2 (B2) has 1 trap (F). The storage capacity is 304 m³, the permanent water level is 1 m rising to 1.6 m at the overflow. A “hydro-brake” regulates the outflow to the River Wallington (WR). The design inflow for the 1:30 and 1:100 year events were 70 l/s and 100 l/s respectively. The system was planted with Phragmites australis and Typha latifoliain Spring 2009. By 2010 all the pond area was dominated by vegetation,differing in density with respect to depth of flow. Figure 2 is a photograph from the pond inlet taken in June 2009, approximately 3months after planting.
Figure 1: Schematic Plan of the Vegetated Pond System. Letters Indicate Sampling Points: A = Swale Inlet; B = Swale Mid-Point; C = Pond Inlet; D, E, F = Sediment Traps; G = Pond Outlet.
The site was equipped with a rain gauge and flumes/stage loggers on the inlet and outlet of the ponds. Unfortunately, this equipment was not operational during the monitoring so storms were characterised by total daily precipitation (DP) and antenacent dry period (ADP) since the last DP greater than 2.5 mm/d. These were obtained from closest rain gauge to the site (a private gauge approximately 1.5 km away: Station IHAMPSHI9 -
Figure 2: Photograph of the System in June 2009 looking from the Pond Inlet. The Sampling Points areLabelled as per Figure 1.
2.2 Experimental Methods
Studies into the fate of metal and organic pollutants have been the focus of two separate, but linked, studies, so have slightly different sampling strategies and sampling occasions.
2.2.1 Monitoring Strategy
Monitoring was undertaken of conditions in the pond system (monthly) and of individual storm events and for 2 years (03/2011-03/2013). The monthly monitoring aimed to assess the baseline water quality in the pond and characteristics of bed and settling sediments. The storm event monitoring aimed to characterise the water quality of runoff entering the pond and the transport of pollutants.
Monthly Monitoring: There were two sampling strategies in the monthly monitoring. The metal study has focussed on sedimentation in the pond and quality of deposited sediment at the inlet/outlet bank (C, G) , while the organic study focussed on soil in the swale (A, B, C). Grab samples of water were collected from the system and river (WR) via a hand pump to avoid aeration. Material accumulated in the sediment traps was removed. Soil cores and sediments were also collected from the swale and pond bed.
Storm Events: Potential storm events were identified from weather forecasts. The criteria for a storm event were (i) daily precipitation (DP) had to exceed 2.5 mm (ii) storm duration had to be greater than 3 hours and (iii) there had to be inflow to the pond for more than 3 hours. As the storm monitoring involved intensive multivariate sampling and testing, there were also logistical constraints on which storms could be monitored. Table 1 shows the characteristics of the 10 storms monitored: inflow was measured directly at the inlet flume (C) by the velocity area method using Valeport 801 electromagnetic flow meter. However only 4 of these events (3, 5. 6. 8) generated outflow due tostorage deficit in the Pond (low rainfall in 2011).
Table 1: Characteristics of the monitored storm events
Storm Event / Date / Antenacent Dry Period, ADP), d / Daily Precipitation (DP), mm(mm) / Qmax, m³/s
1 / 31/3/11 / 0 / 10.4 / mv
2 / 01/12/11 / 0 / 7.1 / 0.008
3 / 12/12/11 / 0 / 14.5 / 0.047
4 / 24/1/12 / 20 / 7.1 / 0.007
5 / 04/03/12 / 14 / 12.4 / 0.051
6 / 23/04/12 / 0 / 16.3 / 0.064
7 / 25/4/12 / 0 / 7.1 / mv
8 / 8/06/12 / 8 / 16.8 / 0.034
9 / 14/12/12 / 8 / 12.7 / mv
10 / 12/1/13 / 3 / 18.5 / mv
mv= missing value
Sampling logistics meant that a sub-set of storms were monitored for general water quality and metals (7 events - 1,2,3,4,5,6,8), water samples were taken from the inlet (C) and outlet of the ponds (G) at specific time intervals. The study of hydrocarbons examined another sub-set (5 events - 1, 4, 7, 9, 10). Water samples for hydrocarbon extraction were taken directly from the swale inlet pipe (A).
All sample types were stored in a cool box (4 OC) and analysed or pre-treated within 24 h.
2.2.2Water Quality
Biochemical oxygen demand (BOD) and total suspended solids (TSS) were measured using standard methods [26] and chemical oxygen demand (COD) by the Hach™ micro kit. VSS was measured via the loss on ignition method [26]. Ammoniacal nitrogen (AmmN) was measured via the Palintest™ kit. Other variables were measured in-situ with probes, e.g.EC (Palintest Micrcomputer 900), pH (Hanna HI1925) and DO (YSI 50B).
2.2.3Metals in Water and Sediments
Water: During on-site sampling 50 ml aliquots were filtered through 0.45 μm Whatman cellulose nitrate filters using a hand pump to separate particulate matter and dissolved fractions for metal analysis, these fractions were preserved with HNO3[13, 6]. Metal content was analysed using an Agilent 7500ce Inductively Coupled Plasma - Mass Spectrometer (ICP-MS) with octopole reaction cell using the semi-quantitative method in He mode. Samples were introduced using an integrated auto-sampler and calibration was by a tuning solution of 10 ppb of 6 elements across the mass range.
Sediments: The settling solids and bed sediments were wet sieved in-situ using pond water to two size fractions; namely the 2mm to >63 μmand <63 μm fractions[12, 13]. These were termed coarse grains and fines respectively. The fractions were then dried in the dark at 80°C and digested for metal analysis with HNO3[13] and analysed using ICP-MS.
2.2.4DRO, ERO and PAH
Water sampleswere taken in amber bottles, 50 or 100 mlwas used for hexane extraction using Solid Phase Extraction Empore C18 Discs (3M) as per EPA method 1664 revision A. The extract was passed through a 1g anhydrous Na2SO4 cartridge (Bond Elut) to remove residual water. 50 µl of nonane was added andsamples concentrated down to 1ml at 40oC in a stream of N2prior injection on the Gas Chromatograph-Mass Spectrometer (GC-MS).
The extraction of the PAHs used the same procedure usingdichloromethane as the solvent, based on EPA Method 550.1[27] using application note 54 from SUPERLCO (Sigma Aldrich) for C18 discs [28].
Soils: mild steel tubes 60mm long and 50.8mm diameter were hammered into the ground, cores were extracted and transported to the laboratory covered in foil inside sealed bags. Accelerated solvent extraction (ASE200 Dionex) was used to extract the hydrocarbons in soils following manufacturer’s application Note 324 [29] for DRO and ERO and Note 313 for PAHs, both these methods are based on EPA method 3545.In the extraction of DROs and EROs a weighed sample of was mixed with equal parts of drying agent HYDROMATRIX and packed between washed sand and cellulose filters in the metal cells and placed in the ASE, a 50:50 solvent mixture of hexane:acetone at a pressure of 1500 psi was applied using N2 gas at an oven a temperature of 200oC. The solvent extract was passed through a Bond Elut anhydrous Na2SO4 1GM cartridges to remove any remaining water. The extract was filtered (0.45 µmChromacol), 50 µl of nonane was added. Heat and N2 were used to blow down the samples to 1ml prior to injection in to the GC-MS.
PAHs extraction was similar but a 50:50 mixture of acetone:dichloromethane was used [30]at an oven temperature of 100oC. The extract was dried using anhydrous Na2SO4Bond Elut tubes. After concentration down to 1ml there was a further clean up stage using Bond Elut silica gel cartridges 500 MG to extract the PAHs from the solvent (EPA Method 3630C) [31], this involved conditioning the cartridge with hexane then adding the sample and then 5ml of hexane:dichloromethane (60:40) to elute the PAH’s, 50µl of nonane was added and samples blown down to 1ml again before GCMS analysis.
2.2.5 GC-MS
A Varian 430GC with a VF-5 Column and a Varian 210- MS detector were used. The operational specifications of the GC-MS were:
DRO/ERO: Initial temperature 50˚C for 1.5 min. Increase at rate of 15˚C/min to 300˚C in splitless mode. Trap: 220˚C. Manifold: 80˚C. Transfer-line: 300˚C. Injection volume of sample: 1 µl. Injector temperature: 280˚C.
PAH’s: Injector temperature 250oC. Temperature ramp: initial temperature of 60˚C hold for 1 minute; then increase to 150oC at a rate of 30oC/min; and then increase to 186 at a rate of 6oC/min; and finally increase to 280 at a rate of 4oC/min and hold for 20 minutes.
The DRO and ERO concentrations were calculated by baseline to baseline integration over the carbon ranges and individual PAHs peaks were integrated and calibrated against standards.
3 Results and Discussion
Statistical analysis and graphical presentation of results was performed using Minitab 16. The variables were tested for normality and where appropriate, if it showed a better approximation to normality, log transformed data was used for statistical analysis.
3.1General Water Quality
3.1.2 Water Quality in the Pond
Table 1 shows the median, minimum and maximum values of the water quality descriptors in the basins (B1. B2) and river (WR) measured during monthly monitoring and provides a baseline of conditions in the system. Upstream, the River Wallington passes through a built-up area and so receives other sources of urban runoff, it is therefore not pristine with a BOD of up to 20 mg/l. AmmN concentrations in the pond basins are lower than the river; while BOD and EC are of approximately similar values. This suggests that the ponds will not have a significant impact on the oxygen balance of the receiving water. However the ponds do have higher COD and turbidity than the river. There is a notable increase in COD between B1 and B2 (154mg/l compared to about 118mg/l) and also smaller increases in TSS and BOD. The majority of solids suspended in the water column were composed of volatile matter (B1 53% and B2 61%). The overall COD:BOD ratio increases from about 9:1 in B1 to 19:1 in B2, suggesting that much of the accumulated organic material not very biodegradable. COD:BOD ratios of 30:1 have been reported in other road runoff studies, so this is not unusual [13]. This transformation between basins suggests that the nature of solids changes within the system, which could be due to preferential transport or accumulations of plant derived debris.
Table 2: Water quality in the pond basins (B1, B2) and River (WR) (n=17)
Variable / Statistics / LocationMedian / Min / Max
BOD (mg/l) / 11.6 / 2.2 / 17.9 / B1
13.5 / 1.7 / 28.0 / B2
8.1 / 1.1 / 20.8 / WR
SBOD (mg/l) / 8.2 / 0.8 / 13.1 / B1
8.2 / 1.3 / 12.8 / B2
4.3 / 0.2 / 12.7 / WR
COD (mg/l) / 118 / 30 / 541 / B1
154 / 17 / 832 / B2
53 / 2 / 641 / WR
SCOD (mg/l) / 67 / 17 / 279 / B1
75 / 7 / 148 / B2
32 / 0 / 128 / WR
TSS (mg/l) / 28.8 / 6.4 / 74.0 / B1
33.2 / 10.4 / 88.7 / B2
17.6 / 2.4 / 55.3 / WR
VSS (mg/l) / 15.3 / 5.0 / 35.1 / B1
20.1 / 6.0 / 78.0 / B2
8 / 1.2 / 49.3 / WR
Turbidity (NTU) / 14.5 / 2.0 / 60.6 / B1
10.0 / 2.5 / 68.0 / B2
4.5 / 2.0 / 61.0 / WR
EC (µS/cm) / 725 / 355 / 1853 / B1
594 / 337 / 1168 / B2
756 / 225 / 1054 / WR
Amm-N (mg/l) / 0.17 / 0.01 / 1.00 / B1
0.11 / 0.00 / 0.47 / B2
0.29 / 0.07 / 0.96 / WR
pH / 6.93 / 6.54 / 7.63 / B1
6.88 / 6.61 / 7.25 / B2
7.25 / 6.8 / 7.65 / WR
3.1.1. General Water Quality Storm Events
The storm monitoring covered a range of events, but the lack of automatic monitoring data meant that rather coarse descriptions have been used to characterise them(Table 1). Figure 3 shows the plots of COD, SS and flow rate at the inlet to the pond (C)over the first 3 hours of the 7 storm events studied for general water quality, There is a clear “first flush” phenomena as inlet water quality progressively improves in terms of COD and TSS overstorm events (Fig 3i and ii), this was also seen for BOD, VSS and turbidity (data not shown). The initial rate of decay of TSS and COD over the first 15 mins was faster than over the later stages and did not fit well to an exponential decay model, perhaps indicating that several process were interacting.
There were significant different pollutant loadings between the storm events (LogBOD, LogCOD, LogTSS, LogVSS, turbidity, AmmN: all ANOVA p>0.00). There was general trend for pollutant loads to increase over the study period (Fig 3i and ii), but there is also a trend for increasing flow rates and velocities of the storm runoff over time (Fig 3iii) which makes interpretation difficult. Increased influent pollutant loads have been seen in other systems as they become established [9], but this systemhad been operating for 2 years prior to the start of this studyso this may not be the case. There was significant association between TSS at the start of the storms and initial flow velocity which may suggest that increased pollutant transport at high flows was an important factor(Log TSS = 16.6 Q(m3/s) + 2.17; n=6; r=0.88; p=0.000) as soluble pollutants such as EC, did not have a significant association with flow rate. Although there were several significant differences between pollutant loadings and the ADP and DP, none of these showed a clear trend, perhaps due to the relatively small number of storm data points.