Toxic metals in urban runoff: physico-chemical mobility assessment using speciation schemes

Physico-chemical characterization of lead, zinc and cadmium has been carried out on eight samples from both separate and combined sewers. Dissolved and particulate total metal levels have been determined. A speciation scheme has been used ta further divide these phases in two dissolved fractions, bioavailable and stable, and five particulate fractions, ion-exchangeable, acid-soluble, reducible, oxidizable and residual. TotaJ dissolved lead concentrations were found ta be largely below European Community (EC) directives for drinking water production. Lead was represented by stable forms while zinc and cadmium were mainly bioavailable. Particulate concentrations were higher than naturallevels found in the Seine-Nolmandie basin. Zinc was characteristic for municipal effluents, and lead and cadmium were more concentrated in separate sewers. Correlations have been observed between particulate heavy metals and volatile matter on the ring highway, or with iron at the Savigny site. The particulate meral speciation scheme has demonstrated the important dependence of heavy metal distribution on the sampling site, as weil as the potentially more mobile nature of zinc and cadmium, compared to that of lead. The structural modifications undergone by solids during their transport along the sewer networks had a significant impact on particulate metal mobility. This has been clearly noted for lead, whose mobility decreased from upstream ta downstream in the studied system.


INTROD CTIO
Originally, only hydraulic impacts of urban runoff were considered and assessed for waste water management.During the bst decade, it has been demonstrated that roof J'unoff (Forster. 1990), as weil a.' highway runoff, cantains high pollutant concentrations (Hémain, 1984;Torno el al., 1985).ln fuct, runoff appears ta be the main problem in improving surface water quality in highly urbanised areas (Marsalek, 1991).Sewer suspended salids and sediments are especially interesting since a very large fraction of pollutants is fixed on them rChebba, 1992).In most cases, the particulate matter di.rectly reaches slilface l.FLORES-ROORiGUEZ et al. waters by combined or separate sewer overflows.There, through transport and settling, the y may pose a threat to the aquatic environment.[n particular, solids attract toxic metals, i.e. zinc, lead, copper and cadmium, which may be released inta the dissolved phase.Mobility and toxicity of these metals must be determined for assessing urban runoff impact on receiving waters and for developing suitable treatment procedures.
Different kinds of experimental approaches have been developed ta study metal transport and biogeochemical cycling processes, as weil as to predict their biological impact, with acute or long term effect studies (Allen et al., 1980;Nelson and Donkin, 1985).Field measurements of river suspended solids and sediments surrounding sewer overflows give direct indications of the total amount of heavy metaJ pollution introduced into the receiving system and its accumulation by sedimentation in river beds (Estèbe, 1993).Caged or free bioindicatars have also been utilized to determine sewer overflow toxicity (Shutes et al., 1990; Fraboulet et al., 1993).Laboratory assays, like sediment re-suspension experiments in a reactor, can aJso yield interesting data on the reJease capacities of poJluted particJes (Lebreton et al., 1993, Bussy, 1993).Finally, numerous physico-chemical speciation schemes have been developed to give indications of the heavy metal distribution in sediments and waters (Florence, 1982;Welté and Montiel, 1983).[t is weil known that exact species determination in aqueous and solid environmental sampJes is excessiveJy difficult because of the compJexity of natural systems, and that heavy metaJ physico-chemical speciation schemes cannot really extract weil defined fractions, because of sam pIe handling mistakes (Kersten and Forstner, 1987), non-specific reactions ta particu lar minerai or organic fractions (Megellati, 1982), or metal redistribution during the fractionation procedure (Ajayi and Vanloon, 1989).However, the data obtained from these schemes are considered as operationaJ (Martin et al., 1987) and provide useful information on the relative mobility and bioavailability of toxic metals (Morisson el al., 1989;Bourg et al., 1992).
This paper presents sorne of the results obtained during a three year study of storm sewer discharges.The heavy metal speciation scheme included both dissolved and palticuJate elements (Flores-Rodrfguez et al., 1990).The palticuJate part, slightly modified from Tessier et al. (1979), consisted of five fractions.The dissolved part was simpler, with a unique organic resin extraction (Figura and McDuffie, 1977) as close as possible to the biological membrane uptake.A wide range of situations was reviewed, with pure road runoff, as weil as combined sewer overflows, separate sewers, and wet and dry detention ponds.Eight sewer f1ows, folJowing storm evenls, were studied at ,seven different sites, m(l,st uf them near Paris.Table 1 shuws the principal site characteristics.The sturm uutlet of a poltiun uf the Paris ring highway was chosen as a road runoff sample.La Morée separate sewer cunies waters frum a part of Seine-Saint-Denis

SAMPLING SITES
Toxic mews in urban runoff county, a n011hem Paris suburb, as weU as the Sausset, which also has an outfall in the wet detention pond of Savigny.La Molette dry detention pond is located in the same county but receives combined waters.Two storms of different intensities were studied at this site.South of Paris, Créteil Lake is used as a wet detention pond for a separate sewer draining the town and, particularly, an important shopping centre.The most important overf1ow of the combined sewer draining Paris and sorne of the surrounding towns is located at Clichy.Béquigneaux dry detention pond is the only site situated in the Bordeaux area, in the South of France, having a different soil type, and with sorne industrial inputs into its separate sewer system.

LABORATORY ANALYSES
Laboratory samples were split into dissolved and particulate phases using a 30 min 6000 g centrifugation (Jouan, model Plasma R (000), followed by a tangential membrane filtration (Millipore, model Minitan S) using 0.45 /lm Millipore Durapore membranes.Total dissolved lead, zinc and cadmium were determined directly by fumace atomic absorption spectrophotometry (AAS) with a HOA 700 Perkin Elmer spectrophotometer and an AS 70 autosampler.The bioavaiJable fraction was extracted by a chelating resin 24 hour contact.The chosen resin.Chelex 100, had previously been transformed into its calcium form by means of al M CaCJ 2 solution as indicated by Figura and McDuftïe (1977).It was estimated to be one of the closest resins to biological membrane uptake, except for S-compounds (Monison et al., 1989).Stable dissolved metals were measured by AAS in the resin-extracted sub-sample supematants.Bioavailable metals were determined by subtracting stable fractions from total dissolved levels (Flores-Rodrfguez et al., 1990).
Total particulate metals were determined by fla me or fumace AAS, after an acid digestion by a mixture of concentrated HN0 3 and HCI0 4 .The sediment samples also underwent a chemical speciation.using a five successive extraction speciation scheme adapted from Tessier et al. (1979) by Flores-Rodrfguez (1992).
Following the extraction order, these fractions were: ion exchangeable, obtained after a 1 M CH j COONH 4 extraction at pH 7 and room temperature, during 2 hours, acid-soluble, obtained after a M CH 3 COOH/CH 3 COONa extraction, at pH 5 and room temperature, during 2 hours, reducible, obtained after a 0.04 M NH 2 0H,HCI in 25% CHjCOOH extraction, at pH 2 and 8YC during 5 hours, oxidizable, obtained after a 3 hour 30% H 2 0 2 in 0.02 M HNO j extraction at 96°C, in a water bath, residual, obtained after the addition of a HNO j /HCI0 4 concentrated mixture, evaporation to dryness on a hot plate, and a second extraction with 15% HNO j under a slight heating.

Oissolved Metals
Table 2 displays mean lead, zinc and cadmium concentrations in seven dissolved samples coUected dUling, or juS! after, stolm events.Levels stayed usually below or neaf surface water EC directives for drinking water production (EC Oil'.80/77'1.).The total dissolved metals appeared to be much sm aller than values previousiy measured in Paris urban area (STU, 1995).However, it may be important to notice that these latter data were obtained with bulk waters first1y aciditied and then filtered, allowing possible metal release trnm suspended snlids into the final dissolved samples.
The extraction canied out with the chelating resin revealed that cadmium and zinc were mostly found in binavailable fOlm in separate and combined water samples, whereas the majority of lead was often stable (Fig. I).The different pattem observed for Créteil lake may be due ta ils wet detention pond statu s, as weil as its large surface (40 ha versus 4 ha for Savigny pond) which tends to favour the influence of the groundwater.By comparing sewer meta] concentrations to pond levels at Savigny site (samples 2 s and 2 p ), a purification capacity of this pond for dissolved metals was observed, especially for zinc and lead.For the same storm event as the studied one on Savigny site (but for a larger catchment area), La Morée sewer (sample 4) displayed much higher zinc levels.If totallead and cadmium concentrations were slightly lower, their relative bioavailable fractions were increased.The storm intensity did not modify the total levels of lead and cadmium at La Molette site.However, their bioavailable fractions were more important for a small rain (sam pIe 7) th an for a big storm (sam pIe 6).Among the three metals, zinc was the most affected by rainfall intensity, with an observed doubled bioavailable fraction.Finally.Béquigeaux dry detention pond (sample 5) displayed similar patterns to sites in Paris, except for the cadmium level of its dissolved samples, which was strongly increased by industrial inputs.1)2).It'ihowed higher particuJate metal levels than our analysis, as weil a significant tendency to leach out, which could be attributed to its lack of water contact.ParticuJate lead and cadmium are also carried down ta Créteil Lake SSO (sample 3) From a catchment area which principally consists of an important shopping centre and its extensive CCLr park areas, the latter exerting an important influence on sewer effluent quaJity.A battery production plant was identified in the Béquigneaux detention pond catch ment area (sam pIe 5) which explains the high levels of cadmium and zinc found in this sam pIe.Savigny and La Morée separa te sewers (samples 2 s and 4) showed lower heavy metal concentrations than combined sewer samples Laken from the same area.
The combined sewer sam pie levels did not vary significantJy with catchment area (Clichy -sam pIe 8), with a concentration very similar to La Molette, but for a much larger drained surface.ParticuJate lead and cadmium concentrations were simiJar for separate and combined samples.By contrast.zinc levels were notably more impoltant on combined se weI' sites, as weU as less variable among the storm events.A study of French combined sewage metal composition (Colin, 1986) has shown that zinc is the heavy metal which is the most important in flow coming into a sewage plant, by at least a factor 4. Thus, when a notable storm event occurred, runoff solids, drained by La Molette combined sewer, produced a dilution of the particulate zinc concentration normally found in the municipal sewage effluent.(see text for sample identification) Finally, most of our samples reach the river Seine.Considering the naturallevels of Seine basin sediments, which have been statistically calculated (Pereira-Ramos, 1989 -Table 3), our results demonstrated that urban runoff and CSO outfaJls may have a very negative influence on receiving water bodies.Thus, it should be useful to have suspended soliu separation treatment before the sewage reaches the river Seine.A recent study of the Seine sediment tïrst layer, on a site located downstream From Paris and its main combined seweI' outfalls supported this, shmving great .similality to our sam pies and a significant increase of natural heavy metal levels (Estèbe et (J/" 1993).They imply a significant sewage impact on river sediment quality.Measured metal concentrations, panicularly lead.often exceed alloweu limils for agricultural use of .solidwaste (AFNOR U44-04l/1975).If such solid.swere removed, using for instance dry detention ponds, the resulting sludge could not always be recycled, However, Savigny wet detention ponu showed real purifying capacity, if eflluent levels are compared with levels found in ilS connected sewer (samples 2 p and 2 ), Local s water authorities have detïned a pollution factllr for sediments.by uividing measured metal concentrations by calculated natllral levels (Pereira-Ramos, 1l)~9).This faclClr aJiowed us to more easily evaluate .seuimemheavy metal pollution, since natural soil concentration may vary significantly, uepending on the site (Robbe. 19~2).The right pan (If Table 3 displays pollution factors for lead, zinc, caumium and iron for ail our "Palisian" samples.One principal characteristic of separate sewer waters should be noted; the significant variability of their quality, which has been the argument for not treating these waters.Pollution factors calculated for combined sewers are more homogenous since they only vary from 12 to 47, while the separate samples vary from 4 to 106 (except Savigny pond, which can be considered as unpolluted).Another remarkable fact is that iron levels were roughly constant among the samples, with a maximum pollution factor of 2. We can conclude that natura!levels of this meta!were found in our samples.This major soil characteristic can be considered as typical in the Seine basin, independent from the generaJ heavy metal pollution level.
This property has allowed us to use iron as a natura!minerai fraction tracer.At the same time, we chose the volatile matter ratio as the organic matter tracer, since it is generally utilized ta measure the organic matter content of sewage effluent~ (AFNOR T90-0291l970).By plotting correlations between these two parameters and the heavy meta!concentrations, we wanted to roughly distinguish the origin of metallic pollutants or, at least, their transport mode.These coo'elations are illustrated for zinc in Figure 2. The relations between sites are c1early displayed on the graphs.Similar conclusions were made for lead and cadmium.Créteil sediment (sam pie 3) characteristics were consistent with the ring highway (sample 1) results, probably because of the numerous parking areas incJuded in the catch ment area.La Morée (sample 4) was closer to Savigny (sam pie 2), which is located in the same county.The smail rainfall we studied at La Molette pond (sample 7) presented some analogies with Clichy outfall (sam pIe 8), but during an important storm event La Molette samples were more similar to the separate sewer samples collected in the same county at the Savigny and La Morée sites.The non-homogeneous repanition of solids in a dry detention pond was clearly notable within La Molette pond.The weak intensity of the sm ail rainfall increased the phenomenon of selective settling which induces sampling representativity problems.4. Significant cOlTelations were only found on two sites.Solids sam pied in the runoff detention cham ber of Paris ring highway (sam pie 1) indicated a strong link between heavy metal concentrations and volatile matter content By contrast, Savigny pond (sample 2) metal levels showed a strong correlation with the level of paniculate iron.The fil'st site, typically representative of automobile pollution, may concentrate pollutants from organic materials like tyres.oils, etc.These product$ have also been listed by M,HTison (19~5) as important heavy metal sources.On the second "ite, sediment s'Impies showed a very strong correlation between iron and the three studied elements (correlation cpeftïcients ranged l'rom ().~3 tu n.9S).This may be explained by the fact that location of the catchment area, outside Paris and not very urban, favoUl's the presence of soil particles rather than pure road runotl snlid".Tessier 1'1 (JI.(J ')~2) have also ()bserved important correlations between iron and heavy metals for rivel' sediments, as weil as Brilhante (1990), for different types of solids.The relationship to vobtile matter, although not as goud a.' for the ring highway.
was notable.This lasl revealed the impact of road runoff on the qua lity of these samples.However, the minerai nature of the source, or at least of the transport mode of heavy me tais on this site, is more important.Although Béquigneaux (sam pie 5) is a separate se weI' site Iike the two former, its sediments did not show good correlation with the two studied parameters.Particularly, lead did not show a relationship with iron, even though it is strongly bound to inorganic p3l1icles (Aualiitia and Pickering, 1987).The industrial outfall may create this difference by modifying deeply the sam pIe characteristics.La Molette large storm (sam pIe 6) showed some links between heavy me tais and iron, as weil as volatile matter, like it was observed for Savigny pond.The small rainfall (sam pie 7), which occurred on the same site, presented the lowest correlation coefficient for both parameters.Thus, the nature and transport of the metallic pollutants in combined sewers is SU'ongly dependent on storm event duration.A heavy rainfall gave results closer to separate sewer ones (like Savigny samples) th an did small events.This difference between the storms at La Molette was verified by calculating the regression coefficients for ail the samples which were collected on this site.We generally obtained lower values, except for lead (which links with iron), which was greater for both events.As these solids 31'e disch3l'ged directly to the rivers, it was interesting to evaluate the portion of the particulate heavy metals which could be released w pollute the natural environment.Total metal cannot provide this kind of indication.With these sequential chemical extractions, physico-chemical sediment speciation may offer more information.We used this process upon some paniculate samples.Figure 3 displays the distribution of heavy metals among the five fractions of the chosen scheme.Regardless of the sites or the metals, the most important fractions were the acid-soluble and reducible fractions, if the residual fraction was not taken into account.This latter introduced greater differences for lead and iron than for zinl.: and cadmium.Thus, these two latter heavy metals showed similar distributions, while the lead pattern appeared to be closer to iran.This may be linked w the good I.:orrelations usually found between the total particulate cpncentrations of these two elements.Moreover.the large particulate lead residual fraction impJied that ilS stable nature may render it essentially non-toxic for the environment.In 1'17'), Wilber fi ul.
also found this metal mostly in residual forms in runoff waters.The ring highway (sample 1) was the only site where samples containing important jon-exchangeabJe fractions for ail the metals were found.This agrees with previous work carried out on the same site, but with tunnel road aerosols (Lebreton and Thévenot. 1992).and with the study of Revitt el ul.(1 0H7).which found that particulate lead in mad runoff and gully pot outflow WolS mainly present in the ion-exchangeable fraction.ln addition to this fraction, we found heavy metals in acid-soluble and reducible fOnTIS in our road runoff sam pie.However, the correlations previously plotted for this site displayed notable links between heavy me taIs and volatile matter, which is related 10 organic matter, destroyed with the oxidizable fraction.
[Pb] (g/kg d.w.) [Cd] (mg/kg d.w.)This discrepancy mOlY be uue to the reagent useu for the extraction, baseu on the breaking down of humic acidslmetals bonus, while heavy metals were more probably adsorbeu on different organic matter with weaker chemical bonus.The Créteil Lake SSO (sam pie 3) was not as clo.se to the ring highway sampJe as the correlations inuicateu.Particulate heavy metals were extJacteu plimarily in more stable fractions.This could be explained by the sewer sample, which allowed an in-pipe evolution of sediments.The result differed l'rom those obtained with longer separate sewers (samples 2 and 4), probably because of the shon travel time.
A comparison of metal uistributi,1I1s in these samples, particulw'ly with lead, indicates their progressive stabilization hom upstream (roau runoff) !Cl uownstream (the sewer system).The signitïcant runoff proce.sses in the larger catchment areas implied a greater susp@nueu solius transp(lrt.These solius tenu tu be rieh in humic organic matter.which may fonn more stable chemical bonds than with runoff l'rom small catchment areas like Créteil.Finally, it is important ta remark that for substantial metal concentrations, the largest fractions were most easily released (i.e.Créteil and Béquigneaux), while less concentrated samples had a large residual fraction.

CONCLUSION
Concentrations of dissolved lead.zinc and cadmium in separate and combined samples were generally below EEC water quality directives for sources of drinking water.ln some cases, zinc and cadmium exceeded these values.Moreover, they were mostly found in bioavailable fonn, whereas lead was more stable.Particulate heavy metal concentrations vaJied over a wide range.Ali were much higher th an natural levers calculated for the Seine-Normandie basin, and for measured Seine sediment levels, and higher than French regulations for residual metals in sludge used for agricultural recycling.Lead and cadmium were mainly found in runoff, but zinc was concentrated in municipal effluents diluted at the final outfal!.Few cOITelations cou Id be found between iron and volatile matter, except for very specifie sites.Dry detention ponds presented very different results because of a kind of selective settlement which occurred during and al'ter storms.It appeared that the structural modifications undergone by solids dUling their transport along the sewer networks had a significant effect on particulate heavy metal distributions (and their related mobilities).These were also very site-dependent.Runoff l'rom roads and car parks showed significant metal mobility.Particulate pollutant~ were transformed during travel through the sewage network to more stable fonns found in the detention ponds.This was particularly notable for lead, whose exchangeable forms disappeared, and whose residual fraction increased.
Fig. 2. Correlations between total particulate zinc and either total particulate iron, or volatile matter Sites: 0 : 1, x : 2, * : 3, " : 4, + : 5, 0 : 6, 18I : 7, X : li (see text for sample identification) Linear regressions were done on sample sites when possible.Correlation coefficients are listed in Table4.Significant cOlTelations were only found on two sites.Solids sam pied in the runoff detention cham ber of Paris ring highway (sam pie 1) indicated a strong link between heavy metal concentrations and volatile matter content By contrast, Savigny pond (sample 2) metal levels showed a strong correlation with the level of paniculate iron.The fil'st site, typically representative of automobile pollution, may concentrate pollutants from organic materials like tyres.oils, etc.These product$ have also been listed by M,HTison (19~5) as important heavy metal sources.On the second "ite, sediment s'Impies showed a very strong correlation between iron and the three studied elements (correlation cpeftïcients ranged l'rom ().~3 tu n.9S).This may be explained by the fact that location of the catchment area, outside Paris and not very urban, favoUl's the Fig. 3. Physico-chcmic<ll uistrihution of parliculate Ph, Cd and Zn.fraclions: 0 : ion-exchangc<lhlt:, EJ2J: aciu-soluhle, ~: rcuucihle, .: oxidizahlt:, ,,,.: residual (sce text l(Jr sample iuentilïcalion)

TABLE J .
Main Site Characteristics: Liguid and Particulate Samples

TABLE 2
Values of Total Dissolved Metals in Sewer Water Samples

TABLE 4
Correlation Coefficients (?) for Paniculate Heavy Metal Versus Volatile Matter and Iron Linear Regressions -Site by Site Study