A street deposit sampling method for metal and hydrocarbon contamination assessment

Urban surface contamination, by atmospheric deposits as well as human activities, is a major concern for urban pollution management. Besides coarse street deposits which are clearly perceived and easily removed, suspended Ž . solid SS surface loads and contamination by heavy metals and hydrocarbons are rarely assessed although they could Ž . be of major importance with regards to combined or separate server overflow CSO and SSO impacts. Both dry and wet vacuum sampling procedures have been first compared, in the laboratory, using dry and sieved clay or street deposits. Then the wet vacuum sampling procedure has been refined, coupling the injection of water and the hand-brushing of the surface prior to its vacuum cleaning, and evaluated on a car parking area close to the University. Finally this procedure has been assessed in Bearn Street within the ‘Le Marais’ district in Paris centre, ́ and 34 samples have been analysed for metal and eight for aromatic hydrocarbon contamination. Heavy metal Ž y1 y1 y1 . concentrations 0.1]1.7 g kg dry wt. Cu, 0.9]6.1 g kg dry wt. Pb and 1.5]4.6 g kg dry wt. Zn within street deposit samples collected in Paris centre, indicate a high contamination, especially for copper and zinc, as compared Ž . y1 to reported data. Total polyaromatic hydrocarbons PAHs are in the 3]11 mg kg dry wt. range, thus approximately 10 times less contaminated than dry atmospheric deposits. This paper presents data obtained and discusses the difficulties encountered when sampling street deposits in busy areas of a city like Paris. The water jet street cleaning procedure used by Paris city workers was tested for its efficiency, by comparison of surface loads before and Ž . after the cleaning procedure. Although solids cleaning efficiency is highly variable 20]65% and somewhat higher U Corresponding author. Tel.: q33-1-45-17-16 25; fax: q33-1-45-17-16-27. Ž . E-mail address: thevenot@univ-paris12.fr D.R. Thevenot ́ 0048-9697r99r$ see front matter Q 1999 Elsevier Science B.V. All rights reserved. Ž . PII: S 0 0 4 8 9 6 9 7 9 9 0 0 1 9 2 8 ( ) F.-J. Bris et al. r The Science of the Total En ̈ironment 235 1999 211]220 212 Ž . for particles larger than 100 mm, particulate metal cleaning efficiency is even more variable 0]75% and particulate PAHs appear not to be significantly removed. Q 1999 Elsevier Science B.V. All rights reserved.

bstract Urban surface contamination, by atmospheric deposits as well as human activities, is a major concern for urban pollution management. Besides coarse street deposits which are clearly perceived and easily removed, suspended Ž. solid SS surface loads and contamination by heavy metals and hydrocarbons are rarely assessed although they could Ž. be of major importance with regards to combined or separate server overflow CSO and SSO impacts. Both dry and wet vacuum sampling procedures have been first compared, in the laboratory, using dry and sieved clay or street deposits. Then the wet vacuum sampling procedure has been refined, coupling the injection of water and the hand-brushing of the surface prior to its vacuum cleaning, and evaluated on a car parking area close to the University. Finally this procedure has been assessed in Bearn Street within the 'Le Marais' district in Paris centre, and 34 samples have been analysed for metal and eight for aromatic hydrocarbon contamination. Heavy metal Ž y1 y1 y1 . concentrations 0.1᎐1.7 g kg dry wt. Cu, 0.9᎐6.1 g kg dry wt. Pb and 1.5᎐4.6 g kg dry wt. Zn within street deposit samples collected in Paris centre, indicate a high contamination, especially for copper and zinc, as compared Ž.
y 1 to reported data. Total polyaromatic hydrocarbons PAHs are in the 3᎐11 mg kg dry wt. range, thus approximately 10 times less contaminated than dry atmospheric deposits. This paper presents data obtained and discusses the difficulties encountered when sampling street deposits in busy areas of a city like Paris. The water jet street cleaning procedure used by Paris city workers was tested for its efficiency, by comparison of surface loads before and Ž. after the cleaning procedure. Although solids cleaning efficiency is highly variable 20᎐65% and somewhat higher

Introduction
Urban surfaces receive deposits issued from Ž more or less remote sources car traffic, indus-. tries, waste incineration, domestic heating , through atmospheric transport as well as from local human activities. The sampling of dry and wet atmospheric deposits can be easily achieved either directly or through a roof outlet. The sampling of street and footpath deposits might be even simpler but it is actually much more difficult to achieve if the sample representativeness is taken into account. Soils are generally much more polluted in the immediate vicinity of major roads ŽHarrison and Johnston, 1985;Hewitt and Rashed, . 1991 , which suggests that a significant part of pollution due to traffic has a very short residence time in the atmosphere, and cannot be measured by atmospheric deposition sampling techniques. However, both remote and locally produced deposits contribute to urban run-off pollution. Their quantity, their composition, but also their physical Ž. characteristics at least their grain size should be characterised in order to determine their transport under wet or dry conditions, related both to urban wash-off as well as to the efficiency of usual street cleaning procedures.
Street deposits have been mostly sampled by Ž sweeping Duggan, 1984;Gulson et al., 1994;Madany et al., 1994;Ogunsola et al., 1994;Al-.Ž Rajhi et al., 1996 or brushing Archer andBarratt, 1976;Fergusson and Simmonds, 1983;Davies . et al., 1987;Culbard et al., 1988 . Since these techniques may discriminate between finer and coarser particles, vacuum sampling has also been Ž. suggested Butler et al., 1992;Ball et al., 1996 . Besides dry sampling procedures, alternative methods for collecting street deposits have used a water jet connected to the fire hydrant or a rain simulator, they showed collection efficiencies Ž larger than sweeping or brushing Sartor et al., . 1974;Ellis, 1986;Ellis et al., 1986 but they require heavy equipment. It was decided to develop a lighter wet method based on a domestic 'floor and carpet' wet vacuum cleaner and to compare its collection efficiency with that of a standard vacuum cleaner. After validation of these sampling procedures both in the laboratory and in a car-park area, street deposit surface loads and their contamination were assessed within the experimental urban catchment of central Paris in Ž the 'Le Marais' district Gromaire-Mertz et al., . 1998a . As this experimental catchment contained Ž. atmospheric deposits Garnaud et al., 1998 as well as roof, street and yard run-off water sampling devices, it was possible to compare the contamination of several types of particles inside this densely urbanised catchment. Finally, we attempted to assess the efficiency of one of the street cleaning procedures commonly used within Paris City, i.e. the water jet on road and footpath. . with a water jet 1 l 1 min at 1 bar whose Ž outlet is 1 cm apart from the vacuum inlet 22 = . 0.4 cm wide . Both devices have been modified by introduction of a 34-l high-density polyethylene container. In the case of dry vacuum cleaner, this container was initially filled with 1 l of water. Containers were designed to act as hydrocyclons and enable fine particle sample collection. Fi- nally, for field experiments, a 220-V AC generator Ž. 2500 W was used for power supply. As shown in Fig. 1 the sampling area was delimited by a 80 = 80-cm 2 plastic frame.

Sampling sites
Three sites have been used, two for validation and one for urban deposit characterisation.
The floor of the Cereve laboratory at Creteiĺ was used for the primary experiments. It is made of 5 = 5-cm 2 tiles connected with 3-mm wide and approximately 1-mm deep cement junctions.
An asphalt covered car park area close to the University was used for assessing the efficiency of hand-brushing combined with wet vacuum sampling on 25 April 1997, which followed an exceptional period of 50 days without rain, and on 31 May 1997 after 7 days of dry weather, and finally on 7 June 1997 immediately after a heavy rain.
Within the urban experimental catchment set-Ž up at Paris centre in the 'Le Marais' district 42 .Ž ha 78% impervious, 12 500 inhabitants Gro-. maire-Mertz et al., 1998a , Bearn Street was selected for field experiments, even though it is not situated upstream of any street run-off samplers Ž. Fig. 2 . Reasons for this choice are the relatively Ž. large roadway width 8 m with gutters as well as Ž. the width of its footpath 2.5 m , with low traffic Ž. approx. 100 vehicles per day and commercial Ž activity one school, two restaurants and a police .Ž station and homogeneous surfaces paved foot-. path and asphalt road . Two sampling campaigns have been organised in Bearn Street The first campaign was performed on 12 June 1997, 24 h Ž y1 after a small storm 3 mm total, 22 mm h maximum ᎏ evaluated on 0.2-mm rainfall depth y1 . steps ᎏ and 2.7 mm h mean intensity . The second campaign was performed on 19 June, 5 Ž days after a stronger storm 11 mm total, 76 mm  . h maximum and 4.5 mm h mean intensity . On 19 June 1997 a careful cleaning of Bearń Street was organised with a water jet vehicle, commonly used by the Paris city street management department, in order to assess the street cleaning efficiency with such equipment.

Collection efficiency assessment on the laboratory floor
A cleaned section of the laboratory floor was given a hand-spread surface load of calcinated and sieved particles consisting of 0.5᎐10 g m y2 of Ž. dried and sieved solids: either clay -50 m Ž. kaolin or 200᎐2000 m gutter sediment collected at the 'Le Marais' experimental catchment site in central Paris. In the case of clay, a suspen-Ž y2 . sion in water approx. 50 ml m was spread on the experimental floor and air-dried. Calcination of 'Le Marais' particles avoids further aggregation of particles and allows a significant evaluation of grain sizes. Both types of particles were sampled twice with both types of vacuum cleaners. SS were collected from the 34-l container, which was further rinsed with 0.5 l water, and the collected suspension was centrifuged at 5000 Ž. rev.rmin during 20 min Jouan GR 412 . Between 0.3 and 6 g of solids were collected from the centrifugation tube, dried and weighed.

Collection efficiency assessment on the Uni¨ersity car park
After selection of an apparently homogeneous area of the University car park, the wet sampling procedure was repeated four times on the same area either with or without a hand-brushing step. In each case, 4 l of water were injected and the vacuum inlet was motioned at approximately 5 cm s y1 , in parallel stripes, along four possible directions inside the area delimited by the 80 = 80-cm 2 frame. For the hand-brushing step, approximately 250 ml of water was first introduced over onequarter of the delimited area, particles were hand brushed in the direction of the centre of the frame and this SS slurry was sucked without further water injection. Such injection-brushing-suction procedure was repeated four times before a 'regular' injection-suction step was achieved in one of the four possible directions. Whereas the regular wet vacuum sampling procedure needed approximately 20 min per 80 = 80-cm 2 unit area, the wet and brush vacuum sampling needed approximately 30 min, taking into account the 34-l container emptying and rinsing time. In both cases SS samples were sieved, centrifuged, air-dried and weighed.

Street deposit sampling
Road, gutter and footpath surfaces were sampled using water jet and hand-brushing, the vacuum inlet being moved only in two opposite direc-Ž y1 . tions at a lower velocity approx. 2.5 cm s in Ž. order to use the same amount of water 4 l . Both operations were separated by a hand-brushing step, as described above. Gutter areas were not suitable for the frame set-up: instead a 40-cm wide and 1-m long area was sampled. In order to avoid heterogeneity of collected samples, coarse solids ) 4 mm was discarded on all sites.
In order to avoid a too long sampling procedure which should have required car and pedestrian traffic interruption, the operation was conducted with four people: two persons for the vacuum equipment handling, one person for street brushing and one person for sample collection and vessel cleaning. It was necessary to carry the vacuum cleaner, the power supply and the necessary containers in a standard commercial vehicle. Ž There was 25 bottles used for sample collection 4 .Ž l initially filled with reverse osmosis water after . nitric acid cleaning procedure , three vessels for Ž. hydrocyclon 34 l , i.e. for road, gutter and footpath sampling, respectively, and a 20-l water tank for their cleaning between each sample collection.

Street run-off water sampling
Sampling procedures for urban run-off within the 'Le Marais' experimental urban catchment Ž have been previously described Gromaire-Mertz . et al., 1998a,b : during individual rain events, bulk Ž samples were collected within three gullies street . run-off and one combined sewer, at the catchment outlet. These bulk samples were immediately filtered for dissolved and particulate pollu-tant determination. Whereas hydrocarbons were separately determined on both of these fractions, particulate metal concentration was calculated by subtracting dissolved metal levels from bulk con-Ž. centrations Garnaud et al., 1998 . 2.3.5. Street deposit analysis SS samples were sieved at 100 m on a Nylon Ž. mesh Bioblock , and further filtrated on a 0.7-m Ž GFrF Whatman membrane pre-calcinated at .Ž . 525ЊC during 2 h for total suspended solid TSS determination after drying at 60ЊC and volatile Ž. suspended solid VSS determination after 2-h Ž. calcination at 525ЊC Thermolyne, 48000 .
Duplicate sieved samples were digested using 18 q 2 ml concentrated nitric᎐perchloric acid Ž. mixture Merck Suprapure added to 1 g dry wt.
Ž large particle i.e. ) 100 m, collected above the .Ž Nylon mesh or 20-ml fine particle i.e. -100 . m suspension. The 50-ml filtrate was acidified at pH 1 with nitric acid. Metal concentrations in digested sub-samples were determined by flame or electrothermic furnace atomic absorption spec-Ž. trophotometry AAS . Both particulate and dissolved sample digestion and quantification were validated using certified samples: NIST 1648 Ž. urban particulate matter , BCR 146 R and 144 R Ž. industrial and domestic sludge , NIST 1643 c Ž.
Ž fresh water , SPS WW1 and SW2 waste and .Ž . surface water Promochem, Molshelm, France . Mean relative deviations from certified values Ž. were generally lower than 10% n s 6᎐10 . .Ž . pyrene-D , Euriso-top Estebe et al., 1997 10 added into the Soxlet apparatus before extraction. PAH extraction and GC-MS quantification procedures were validated using a NIST 1941a marine Ž sediment certified sample NIST, Gaythersburg, . USA relative deviations as well as dispersion for triplicates were usually below 20%

Validation of the sampling technique
Model sampling experiments performed with calcinated and sieved clay or gutter sediments deposited on the cleaned laboratory floor surface enable a clear discrimination between dry and Ž. wet vacuum cleaning procedures Fig. 3 . Although SS collection efficiencies reach 95% with dry sampling for 200᎐2000-m particles and sur-Ž. Ž . Fig. 3. Collection efficiencies of model experiments using dry left diagram or wet vacuum sampling right one and various surface Ž. Ž . Ž concentrations of calcinated and sieved-clay -50 m or gutter sediments 200᎐2000 m deposited on the laboratory floor see . Fig. 1 : mean " dispersion of duplicates. face loads larger than 5 g m y2 , wet vacuum collection efficiencies are always above 95% whatever the particle size and surface loads.
As the laboratory floor, although presenting approximately 3-mm wide and 1-mm deep cement junctions, is much smoother than the asphalt road surface, the wet vacuum sampling procedure was applied to the University car park area. The wet vacuum sampling allowed a collection of 12.2 g dry wt. m y2 for the first assay and 1.5, 1.1 and 1.0 g dry wt. m y2 for three successive assays on the same area: the sum of these four samples represent a surface load of 15.9 g dry wt. m y2 , which means that the first assay yielded only 80% of this sum. A similar wet assay was duplicated in another area of the same car park. The surface load obtained was 10.0 instead of 12.2 g dry wt. m y2 for the first step. A week later, the same car park was sampled for solid deposits, combining wet vacuum sampling and wet brushing. The surface loads obtained successively on the same area were 32.5, 9.3, 3.5 and 2.7 g dry wt. m y2 , respectively, the sum reaching 48 g dry wt. m y2 . This demonstrates the efficiency of the wet brushing both for Ž. the first assay 2.7 times larger surface loads and Ž for the sum of the four successive ones three . times larger , although the dry weather period before the experiment with brushing was very Ž. short -1 vs. 7 days . When comparing the fine particles fraction, i.e. -100 m, 25.1 and 5.4 g dry wt. m y2 were obtained after four repetitions, either with or without hand brushing, while large particle collection was only doubled. This shows that the main effect of hand brushing is the mobilisation of finer particles. As such particles are usually described as more contaminated, it was decided to further use wet vacuum with wet brushing procedures. One should note that a 4-l injection, even coupled with hand brushing, is insufficient for a total urban deposit sampling on an asphalt surface. Successive amounts of dust collected after each step do not constitute a geometric sequence, which would mean that a constant fraction of a total dust residual on the surface is sampled at any step. The ratio between Ž. successively collected fractions R rR inmq 1 m creases with increasing step number. This result is understood as the occurrence of the easy collec-Ž tion of a limited amount of dust mainly during . the first step followed by the more difficult collection of a large bulk of more strongly bound dust. The finer fraction appears more strongly bound to the asphalt surface than the coarser fraction, as long as the no-brushing technique is used, but the difference between both fractions becomes negligible as soon as the brushing technique is used. Brushing would therefore introduce less discrimination in sampling. Although the first sample represented only 70% of the surface load determined after four repetitions, it was decided to limit sampling to one set of 4-l water injection and brushing in order to obtain a larger number of samples per campaign for field work on a street of Paris centre.

Urban sampling in Paris centre: street dust surface loads
As particle surface loads present large spatial variability, it was not possible to differentiate between samples collected after one and five days dry weather periods. Footpath loads were much Ž y2 . lower 1.6᎐3.8 g dry wt. m , n s 6 than road Ž y2 . 8.5᎐17 g dry wt. m , n s 6 and especially gut-Ž y2 . ter loads 17᎐144 g dry wt. m , n s 6 . Organic matter content of particles, as assessed by their Ž. volatile suspended solid ratio VSS contents, i.e. their weight loss at 550ЊC, were not significantly different for footpath and road particles, i.e. 16᎐22 and 17᎐23%, respectively, with gutter sedi-Ž ments being somewhat less organic 7᎐16% VSS, . n s 6 . This difference between footpath and road dust on the one hand and gutter sediments, on the other, is also evidenced by the difference in relative loads of particles larger than 100 m; they reach 30᎐51, 22᎐37 and 41᎐76%, respectively for footpath, road and gutter particles. This result is probably related to the street cleaning procedures of this district of historical Paris: street hydrants are opened daily, for 0.5᎐1 h, and the finest sediments in gutters are transported to gullies.
Taking into account the respective widths of Ž. Ž . footpath 2.5 m , road 3.6 m for half width and Ž. gutter 0.4 m in Bearn Street, the average esti-

Urban sampling in Paris centre: metal and PAH concentration in street dust
Particles sampled during the 12 June and 19 June 1997 campaigns have been analysed for their Ž. metal and aromatic hydrocarbon PAH content.
Particulate metal concentrations present a large Ž. spatial variability Fig. 4 , the copper values being approximately one order of magnitude lower than lead and zinc. Taking into account the median, lower and upper deciles, it appears that footpath dust is much more concentrated in terms of copper than gutter and road samples; corrosion of copper-containing urban furniture and building Ž may be a significant source of contamination Ta-. ble 1 . As shown in Fig. 4, these street deposits appear to be as contaminated as particles sampled at any of the gullies within the 'Le Marais' Ž. experimental catchment Fig. 2: 15 samples . Their contamination is also similar to those of dry fallout particles sampled on the same site in 1996᎐1997, i.e. 0.8᎐2.9, 2.5᎐14.4 and 3.3᎐24.8 g kg y1 dry wt. for copper, lead and zinc, respec-Ž. Ž tively lower-upper decile range, n s 28 Garnaud . et al., 1998 . Unlike the studies of Sansalone and Ž. Buchberger 1996 , we observed similar or higher concentrations of copper and lead in the coarser Ž. fraction ) 100 m compared to the finer fraction, the concentrations were higher in the finer fraction only for zinc.
Based on average values of concentrations in solids collected at Paris, rue du Bearn, and at other urban sites where data have been collected, the centre of Paris appears to be severely polluted. Copper and zinc contamination are the highest among all reported urban deposit data. Taking into account the dust surface loads and their metal contamination, it appears that copper, lead and zinc surface loads are significantly higher at the gutter than on the road surface, with the latter values being higher than the footpath, ex-Ž. cept for zinc Table 1, Fig. 7 . These metal surface loads at footpath, gutter and road sites, enabled the assessment of the total Bearn Street load as being 2, 18 and 22 mg m y2 copper, lead Ž. and zinc, respectively Table 1 .
Total PAH concentrations on particles deposited on streets appear to be significantly lower than those collected from dry atmospheric de-Ž. posits Fig. 5 but similar to those collected at the Ž. outlet of the 'Le Marais' urban catchment Fig. 2 both during storm events and dry weather. This result shows that atmospheric deposits present a significant source of particulate PAHs in this urban catchment, apparently much higher than sources related to leakage from vehicles.

Street cleaning efficiency for SS and micropollutants
In order to assess the efficiency of the usual street cleaning procedure, a procedure performed two to three times per week in the 'Le Marais' district of Paris centre, SS surface loads and their associated metal contamination were compared before and after the street water jet cleaning procedure commonly undertaken by workers from Paris city.
Although the spatial variability of SS surface concentration is large, the road values appeared significantly lower after the street cleaning proce-Ž. dure Fig. 6 . The removal efficiency was better for the road than for footpath and gutter and somewhat better for larger particles. Water entering gullies during street washing of a 29.5-m 2 area within Bearn Street was also collected and it waś found to contain the equivalent of 3.5" 1gm y 2 Ž. surface load Dulac, 1997 . Compared to our assessment of the total initial surface load of Bearń Street on 19 June 1997, i.e. 14.2 g m y2 and to its value after the cleaning, i.e. 8.9 g m y2 , it appears that a reasonable mass balance was obtained and that the water jet procedure used by Paris city workers removes approximately 25% of street deposits. As street run-off particles collected at gullies of this urban catchment during individual storm events represented a surface load in the y2 Ž. 0.5᎐1.5 g m range Dulac, 1997 , rain events appear to transport only approximately 10% of the available street deposits and are clearly less efficient than water jet street cleaning. Such low cleaning efficiency values are probably even lower when the street cleaning is performed by Paris city street workers, who spend much less time and Fig. 6. Assessment of street cleaning efficiency on particle Ž. surface loads at Bearn Street 19 June 1997 : comparison of bulk particles surface loads prior to and after street cleaning. care on such operations. Fig. 7 presents the metal surface load evolution during the water jet street cleaning procedure of Bearn Street Although a large spatial variabilitý was noted, metal removal was observed on footpath, gutter and road surfaces. The lower metal removals obtained for gutter sites are probably related to the high roughness of the gutter surface and to its very high SS surface loads. Furthermore, metal removal rates appears to be somewhat larger for coarse particles, i.e. larger than 100 m.

Conclusions
Based on model experiments performed with dry and sieved fine and coarse sediments, as well as on sampling in car park areas, it appears that the sampling method employed for urban dust and sediment plays an important role on the assessment of SS surface concentrations. A simple vacuum cleaning is less efficient, especially for fine particles, than suction of a suspension obtained with a local water jet; such a wet vacuum cleaning may be achieved with a commercial 'floor and carpet' vacuum cleaner. Wetting the finer particles improves their recovery inside the 'hydrocyclon' system; on the contrary, in the dry collector system, a fraction of the finer particles can easily cross the water-containing vessel and is not recovered. Furthermore, introduction of a brushing step improves, by a factor of approximately 3, the amount of solids collected on a given asphalt surface. Such brushing and suction procedure is clearly not complete since its replication on the same asphalt surface shows that only 70᎐80% is collected on the first step as compared to four replicates. This last result should be used with caution since it is clearly related to urban surface structure and roughness. Furthermore, a repetitive water jet and brushing procedure may contribute to the partial degradation of the asphalt surface and introduces a bias in the assessment of road dust surface concentration. Taking this into account, it was admitted that a good compromise for urban deposit sampling was the association of wet brushing and suction steps performed only once on a given area. Such a procedure allows a good discrimination between footpath, gutter and road surface loads. It can be performed, using an electric generator, on any street, although such sampling obviously necessitates deviation of the car and pedestrian traffic.
This sampling procedure allowed the assessment of street surface concentrations both in Ž. solids, metals Table 1 or PAHs and the demonstration of a high mean contamination in copper and zinc, as compared to other urban sites, reaching a mean value of 2.2 and 18 mg m y2 , respectively. Within the studied street of this historical part of Paris centre, both of these metals may originate from the corrosion of urban furniture Ž and local construction materials Gromaire-Mertz . et al., 1998b . When used prior and after the street cleaning procedure, the sampling of street deposits allows a direct assessment of the cleaning efficiency. For example the water jet cleaning procedure was found to remove 20᎐65% SS but 0᎐75% metals and no significant PAHs: roughness of the street surface and granulometric distribution of micropollutants are clearly important factors for this removal. If the removal ratio of particles below 4 mm appears to be limited, it should be noted that, averaged on the whole street surface, the water jet removes a much larger amount than rain run-off, i.e. 25 and 10% of the total solids recovered by the wet vacuum sampling technique, respectively.