Filter properties of seam material from paved urban soils

Depositions of all kinds of urban dirt and dust including anthropogenic organic substances like soot change the filter properties of the seam filling material of pervious pavements and lead to the formation of a new soil substrate called seam material. In this study, the impact of the particular urban form of organic matter (OM) on the seam materials CECpot, the spe- cific surface area ( As ), the surface charge density (SCD), the adsorption energies (Ea) and the adsorption of Cd and Pb were assessed. The Cd and Pb displacement through the pavement system has been simulated in order to assess the risk of soil and groundwater contamination from infiltration of rainwater in paved urban soils. As , Ea and SCD derived from water vapor adsorption isotherms, CECpot, Pb and Cd adsorption isotherms where analyzed from adsorption experiments. The seam mate- rial is characterized by a darker munsell-color and a higher Corg (12 to 48 g kg 1 ) compared to the original seam fill- ing. Although, the increased Corg leads to higher As (16 m 2 g 1 ) and higher CECpot(0.7 to 4.8 cmolc kg 1 ), with 78 cmolc kg 1 C its specific CEC pot is low compared to OM of non-urban soils. This can be explained by a low SCD of 1.2◊10 6 molc m 2 and a low fraction of high adsorption energy sites which is likely caused by the non-polar character of the accumulated urban OM in the seam material. The seam material shows stronger sorption of Pb and Cd compared to the original construction sand. The retarda- tion capacity of seam material for Pb is similar, for Cd it is much smaller compared to natural sandy soils with similar Corg concentrations. The simulated long term displacement scenarios for a street in Berlin do not indicate an acute con- tamination risk for Pb. For Cd the infiltration from puddles can lead to a breakthrough of Cd through the pavement sys- tem during only one decade. Although they contain contam- inations itself, the accumulated forms of urban OM lead to improved filter properties of the seam material and may re- tard contaminations more effectively than the originally used construction sand.


Introduction
In urban areas, sealing of soil surfaces leads to ecological problems caused by the increased and accelerated runoff and reduced evapotranspiration compared to nonsealed soils (Fig. 1).Cities are normally drier and hotter than the surrounding areas.At heavy rainfalls the exceeding water causes mixed sewage systems to run over.This is one of the main threats to water quality of urban water bodies (Heinzmann, 1998).Therefore, run off reduction by increasing infiltration is among the main ideas of ecological urban planning.This goal can be reached by increased use of pervious

EGU
pavements.In many European cities, especially in the older parts, they are regularly found.They keep the convenience for the city dwellers but allow infiltration rates of 58 cm d −1 for a mosaic tile pavement with 20 % of seams, until 155 cm d −1 for so called grass pavers having an amount of open seams of up to 41% (Wessolek and Facklam, 1997).
The seams allow infiltration and the pavers reduce evaporation.Therefore, groundwater recharge rates are 99 to 208 mm a −1 in sealed areas compared to only 80 mm a −1 for a pine-oak forest around Berlin (Wessolek and Renger, 1998).If rain water accumulates in ponds on the pavements, the groundwater recharge can be greater than 300 mm a −1 .Ponds with up to 60 mm depth are no exception in old quarters.We gauged a pond in front of our department (52 • 31 ′ 4.03 ′′ N; 13 • 19 ′ 25.79 ′′ E) with a volume of 56 L at a horizontal projection of only 2 m 2 .Rainwater runoff in urban areas is often contaminated, e.g. by heavy metals.Dannecker et al. (1990) and Boller (1997) found Pb concentrations of up to 0.3 mg L −1 in the street runoff, while Cd concentrations reached up to 0.0076 mg L −1 (Dierkes and Geiger, 1999).The high infiltration rates together with even low contaminant concentrations may result in severe contaminant fluxes (Dannecker et al., 1990).An assessment of the risk of soil and groundwater contamination requires hydraulic and sorption parameters for the according materials, which are the pavers and the construction material, mainly sand.Furthermore, after a while, a new horizon, the "seam material", develops.The term "seam material" describes the soil material in between single pavestones of pavements, which developed from technogenic sand.It shows a very dark colour, is mostly only 1 cm thick and contains all kinds of deposited urban dirt like foliage, hairs, oil, dog faeces, food residues, cigarette stubs, glas -in short: any kind of urban waste, small enough to fit into the pavement seams, at least after being milled by pedestrians or cars.Thereby, pedestrians and cars abrase their soles and tires and these abrasions also end up in the seams.Furthermore seams contain great amounts of urban dust and therefore soot, concrete abrasions, aerosols and so on.While the hydraulical characteristics of the seam material have already been investigated (Nehls et al., 2006)

EGU
filter properties are still unknown.They can not be extrapolated from other soils because the organic carbon (C org ) of this material differs from non-urban, natural soils in origin, quality and function.For instance, the amount of black carbon, a "combustionproduced black particulate carbon, having a graphitic microstructure" (Novakov, 1984) is increased compared to non-urban soils (Nehls et al., 2007 1 ).
It is hypothesized, that because of its increased C org contents, seam material acts as a valuable filter and influences transport processes through the pavement seams.Therefore, the motivation of this study was to assess the seam materials retardation potential for heavy metals under consideration of its special forms of organic matter, its source or sink function and its thickness.Therefore, (i) general surface and filter properties of the seam material were investigated in relation to C org by ion exchange and water vapor adsorption measurements.Cation exchange capacity, surface area, surface charge density, and adsorption energy were studied.Thereby, the organic matter itself could be characterized.Furthermore (ii) the specific adsorption of Pb and Cd to the seam material and the according contamination status were studied.Thus, two relevant toxic heavy metals of different mobility and affinity to organic matter, were examplarily studied.Finally, (iii) the heavy metal solute transport through the pavement system considering the seam material have been simulated using the investigated parameters.

Sites and sampling
Samples were taken in Berlin and Warsaw, two cities with similar climatic, geologic and geomorphologic conditions, but different industrial activity, traffic intensity and contamination history.The sampling sites were located close to roads or directly on roads.

EGU
Samples consisted of the material that fills spaces between single stones of pavements.The dark seam material at 0 to 1 cm depth was always easily distinguishable from a much brighter 1 to 5 cm layer (Fig. 2).The upper layer is influenced by all external factors and deposits, while the deeper one was taken as representative of the original seem filling material, which is not severely altered by external deposits.The seam material was scraped up using one-way polyethylene teaspoons under suspicious observation of pedestrians.

General chemical properties
The pH(CaCl 2 ) and electric conductivity (EC) have been measured using standard methods (DIN-ISO 10390, DIN-ISO 11265).Munsells's color was determined for fresh samples.The potential cation exchange capacity (CEC pot ) was determined at pH 8.2 using a batch method with cation exchange and re-exchange (Mehlich, 1984) because of high EC of the samples.The total carbon content, C tot and C org (C org after HCltreatment to remove carbonates) were determined conductometrically after combustion (Carlo Erba, C/N analyzer).

Surface properties
Valuable substrate specific surface parameters can be obtained from very simple measurements, not requiring expensive equipment.The specific surface area (A s ) and the adsorption energies (E a ) were calculated from water vapor adsorption isotherms using the vacuum chamber method at the temperature T=294±0.1 K. Prior to the adsorption measurement, soil sample aliquots of 3 g were dried in a hermetic chamber with concentrated sulphuric acid until constancy of weights.Then, they were exposed to 20 stepwise rising water vapor pressures (p/p s ) ranging from 0.005 to 0.98 which were controlled by sulfuric acid solutions.The samples were equilibrated with the corresponding water vapour until constancy of weights.The higher p/p s is, the longer the time gets for equilibration, at the highest step more then one week was necessary.
The amount of adsorbed water, a, was the difference between the weight of the humid sample and the dry sample (dried for 24 h at 378 K).Results of triple weightings had a coefficient of variation (CV ) of 3.7%, averaged values were used for further calculations.
Aranovich's theory of polymolecular adsorption has been used to obtain the capacity of a mono-molecular layer from the experimental water vapour adsorption isotherms (Fig. 3).In contrast to the standard Brunauer-Emmett-Teller (BET) model (Brunauer et al., 1938), Aranovich's isotherm allows vacancies in the adsorbed layer and describes the polymolecular adsorption over a broader and more realistic range of p/p s (≈0.05<p/p s <0.8) than the BET does (≈0.05<p/ps <0.3) (Aranovich, 1992): where p/p s [hPa hPa Having a m , we calculated A s [m 2 ] of the sample using the following equation: Here, L is the Avogadro number (6.02×10 23 molecules per mole), M [g mol −1 ] is the molecular mass of the adsorbate and ω is the area occupied by a single adsorbate molecule (1.08×10 −19 m 2 for water).Knowing A s , the surface charge density SCD [cmol(+) m −2 ] can be calculated: The adsorption energy distribution functions showing fractions f of surface sites of different adsorption energies, (f (E i )), were calculated from adsorption isotherms data, using the theory of adsorption on heterogeneous surfaces (Jaroniec and Brauer, 1986).Aranovich's isotherm (Aranovich, 1992) was applied to describe local adsorption effects.
Here, C i is the value of the constant C for sites of kind i .Solving Eq. (4) for f (E i ) is an ill-conditioned problem.Small variation in experimental data caused by heterogeneity of the samples and detections capabilities of the balance leads to large variation in estimation of site fractions.Therefore, a condensation approximation CA was inserted Harris (1968).This method is based on the replacement of the local isotherm by a step-function.The final formula for the calculation of site fractions then becomes: (5) Introduction

Conclusions References
Tables Figures Printer-friendly Version Interactive Discussion

EGU
From f (E i ) values, the average water vapor adsorption energy, Ēa , can be calculated by the following equation as described by Jozefaciuk and Shin (1996): The calculation of adsorption energy distribution functions (f (E i ) against E i dependencies) was performed using Eq. ( 5).Energy values were expressed in the units of thermal energy (R T), as scaled energies ( A scaled energy equal to 0 represents an adsorption energy equal to the condensation energy of the vapor.The maximum adsorption energy in the condensation approximation is related to the minimum value of p/p s which was around 0.004 in our case.This corresponds to an energy adsorption of around -5.5.Thus, it was reasonable to assume -6 as the maximum energy for the samples.However, this value can be considered only as a first estimate of the maximum energy because of the lack of experimental data at small relative pressures.We arbitrarily set the maximum energy to -8 assuming that if there were no sites with higher adsorption energies than -6, the corresponding values of f (E i ) will be close or equal to zero.In order to reduce the noise of the measurements, only pairs of values, which differ for at least 2 units were used.

Adsorption isotherms for Cd and Pb
Adsorption isotherms were obtained for C org affected, immobile Pb and mobile Cd according to OECD guideline 106 (OECD, 2000).One gram of soil has been equilibrated with 45 ml 0.01 M CaCl 2 overnight.Then, heavy metals were added as a typical cocktail (Pb, Cd, Ni, Cu, Zn) in 5 ml 0.01 M CaCl 2 solution for 48 h of equilibration.Thus, the resulting isotherms include realistic adsorption concurrence.For all samples five stepwise rising solute concentrations including destilled water were added.Therefore, the resulting adsorption isothermes include one desorption step and span over a wide range of concentrations (liquid concentrations range from 0.05 to 10 mg L −1 for Pb and from 0.001 to 5 mg L −1 for Cd).After two hours of shaking, the original pH of Introduction

Conclusions References
Tables Figures Printer-friendly Version

Interactive Discussion
EGU the samples were re-adjusted with KOH.Liquid phase HM concentrations were measured in the supernatant after 20 min of centrifugation at 1000× g.The according solid phase concentration was analyzed like following.In order to consider only the HM fraction which is involved in adsorption processes, the 0.025 M (NH 4 ) 2 EDTA (pH 4.6) extractable portion has been determined (Welp and Brummer, 1999) in advance for the individual samples.The difference of HNO 3 and EDTA extract is the residual amount which has not been considered as adsorbed phase.The samples were extracted with concentrated HNO 3 (soil/acid ratio = 1/20) for 360 min at 453 K under pressure.The concentrations of Pb and Cd were measured with a maximum CV of 2% and 5%, respectively.
For the fitting of an adsorption isotherm to the experimental data with complex solid phases like soils, the use of the Freundlich model is widely recommended (Stumm and Morgan, 1996) before Langmuir, because it does not expect similar energy sites but can be seen as a sum function of Langmuir isotherms for the individual energy sites (Sposito, 1980).So it is adequate to describe the seam material which is a mixture of sorbents with varying surface characteristics: where is the Freundlich distribution coefficient, describing the soil material's affinity to the solute.The exponent m is a measure for the linearity of the adsorption.
The adsorption characteristics are studied to describe the mobility of the heavy metals.A good method to compare different adsorption characteristics with different degrees of linearity is to calculate the retardation factor R, which derives from the convection-dispersion equation and describes the retardation of substances compared to a conservative, non sorbing tracer:  −3 ] is the water content of the soil.For non-linear adsorption of solutes R becomes: Simulation of the Cd and Pb displacement The filter effect of the seam material has been assessed by a numerical HM displacement simulation using the software HYDRUS 1D v2.02 (U.S. Salinity Laboratory USDA-ARS, 1991, Riverside, CA, USA).It solves Richard's equation numerically and applies the convection dispersion equation for the solute transport.It can work with Freundlich's adsorption model as well as with other concepts (Simunek et al., 1999).
The employed input data are shown in Table 1.
It was the goal to assess the impact of the chemical properties (special form of organic matter), the importance (thickness) and the source/sink function (contamination status) of the seam material on heavy metal displacement.The travel times were simulated through a 20 cm long column of (i) construction sand, (ii) a 1 cm thick seam material and 19 cm of construction sand and (iii) 20 cm of seam material.The soil hydraulical parameters of the site B9 have been applied while the chemical features of the sites B1, B2, B4 and B9 have been used for the simulation.Travel time in that context indicates the period of time which is required to recover 95 % of the upper boundary concentration at the lower boundary of the column.The measured background concentrations in the soil were (a) ignored (no contamination) and (b) considered as initial conditions.So, six scenarios ia, ib, iia, iib, iiia, iiib have been simulated.Scenarios ia and iiia compare the maximum retardation capacity of the original construction material with that of the seam material, while in iia the maximum retardation of the realistic soil material mixture is assessed.The scenarios a and b are then used to assess the actual source/sink potential of the construction sand and the seam material.
Groundwater recharge rates of 120 mm a −1 (Wessolek and Renger, 1998)  EGU seam area of 27 %, measured at site B9, was included to get the daily groundwater recharge rates of 0.12 and 0.48 cm d −1 for the open seam area as the net infiltration rate at the upper boundary.Thus, annual evapotranspiration has been indirectly considered, but seasonal variations have been disregarded.The water flow was not simulated but realized by accordingly chosen boundary conditions and derived effective parameters for the site B9: K * s of 0.12 and 0.48 cm d −1 , respectively and pressure heads of 0 cm for the upper and lower boundary conditions.The according artificial saturated water contents, Θ * s corresponded to K * s and was derived from K (Ψ) and Θ(Ψ) functions fitted to measured porosity (Nehls et al., 2006) and K s data (Table 1).For these measurements, 100 cm 3 cylinders were filled with seam material up to bulk densities which have been determined using small cylinders which fitted in the intersection of seams on a cobble stone paved street (Fig. 4) (Nehls et al., 2006).For the original construction sand, undisturbed cylinders were taken from the layer beneath the pavestones.
For simulating the contaminated material, measured liquid equilibrium concentrations were applied (see Table 1) which were in the range of the measured adsorption isotherms.
The rain water runoff concentrations at the upper boundary for Pb and Cd were chosen to be 0.3 and 0.01 mg L −1 , respectively.The solute transport balance for the simulations always had an relative error of less than 5%.To reach this level, sometimes two different time and spatial discretizations must have been used for the two solutes.

Basic and surface properties
The seam material shows a neutral soil reaction with an average pH of 7.1 (SD = 0.2, N = 17).That is typical for urban soils.It is most likely caused by concrete abrasions from buildings or salts (Burghardt, 1994).For the seam material, additionally the adja-Introduction

Conclusions References
Tables Figures Printer-friendly Version

Interactive Discussion
EGU cent pavers can be direct sources of constant carbonate intake as pedestrians not only abrase the soles of their shoes but also the surface of concrete or limestone pavers.The difference of C tot and C org , most probably carbonates, is 5 g kg −1 (SD = 2 g kg −1 , N = 18).The slightly higher pH of the Warsaw samples may be due to salinity effects.At least the EC is significantly higher in Warsaw compared to Berlin (p<0.05,N = 20) (Table 2).The reason for the higher ECs is a more frequently use of deicing salts on streets of Warsaw compared to Berlin.There was no difference between the seam material and the original construction sand detectable.However, from the point of view of the high pH a pronounced HM displacement must not be suspected.
The seam material is dark.The Munsell colors objectively prove and quantify this observation (Table 2).Like in "natural" soils this coloration can be explained by an enrichment of organic matter, which in this case could be black soot (Nehls et al.,  2007  1 ).Measured C org varies from 12 to 48 g kg −1 in the 0 to 1 cm layer and is higher than in the 1 to 5 cm layer (less than 3 g kg −1 ) (Table 2).A displacement of C org into the 1 to 5 cm layer cannot be excluded, but must be expected (Fig. 2).However, the low C org contents in the deeper layer allows to regard the 1 to 5 cm layer as original and not altered.Besides that, it cannot be excluded, that the construction sand already contained different amounts and forms of organic matter.
The CEC pot of the 0 to 1 cm layer in Berlin is not significantly different from that in Warsaw.With 1.2 to 4.8 cmol(+) kg −1 the CEC pot of the seam material is low compared to values of 12 to 20 cmol(+) kg −1 for non-urban German sandy soils (Renger, 1965) or compared to values of 8 to 36 cmol(+) kg −1 for Berlin sandy forest soils with similar C org contents (Wilczynski et al., 1993).
With 0.5 cmol(+) kg −1 soil, the CEC pot of the 1 to 5 cm layer is significantly smaller than in the first layer.Compared to the original seam filling represented by the 1 to 5 cm layer, the depositions of urban dirt lead to an approximately four times higher CEC pot in the seam material.Like for other sandy soils (Renger, 1965) the CEC of the seam material depends mainly on organic material.The CEC pot of seam materials is propor-Introduction

Conclusions References
Tables Figures Printer-friendly Version

Interactive Discussion
EGU tional to A s which depends on C org (Figs. 5 and 6) and both regression lines intersect the y-axis close to (0;0), which indicates that only C org compounds contribute substantially to A s and after all to CEC pot .Due to it's special quality, the contribution of the seam material's C org to CEC pot is only 74 cmol(+) kg −1 C, (Fig. 5).However, the CEC of common, more natural soil organic matter has been estimated to be around 300 cmol(+) kg −1 C by Parfitt et al. (1995) for New Zealand soils and by Krogh et al. (2000) for Danish soils.Higher values of up to 680 cmol(+) kg −1 C were observed in acidic sandy forest soils (Wilczynski et al., 1993).
As mentioned above, the organic matter consists of different constituents, not everything can be referred to as humus.Combustion residues (BC) for instance account for up to one third of the C org fraction (Nehls et al., 2007  1 ).Although they contribute to the CEC of soils, its contribution is lower than that of humus (Tryon, 1948).Consequently, the CEC pot of the OM rises, if not C org but (C org -BC) is applied for the correlation (83 cmol(+) kg −1 C (r 2 =0.74) for C org and 93 cmol(+) kg −1 C (r 2 =0.74) if C=C org -BC is applied).The low contribution of C org to CEC pot can be explained by its low specific surface area (Fig. 6) and its low surface charge density.The first of the two reasons can be ascribed to the spheriodal particular character of the C org (Nehls et al., 2006) resulting in low surface areas per weight (presuming similar substance densities) compared to humus.The samples from Berlin and Warsaw show a very similar behavior resulting in a good correlation between C org and specific surface area (r 2 = 0.8).As it can be seen from the slope of the linear regression line, the specific surface area of the seam material's C org is about 516 m 2 g −1 .Wilczynski et al. (1993) investigated a Berlin forest soil using the same methodology for measuring specific surface areas.They found the contribution of C org to A s to be upto 1062 m 2 g −1 .The SCD of a soil sample is a measure for the quantity of polar functional groups.The SCD of the seam material is seven times lower than that of sandy forest soils (Wilczynski et al., 1993;Hajnos et al., 2003)

EGU
CEC at similar C org levels (Fig. 5. Hajnos et al. (2003) report on SCD of 2.9 to 10.6×10 −6 mol(+) m −2 for Berlin sandy forest soils, while the SCD of seam material is not greater than 1.2×10 −6 mol(+) m −2 .The low SCD of the seam material's C org is caused by unpolar organic substances.
This statement is supported by the investigated distribution of adsorption energies (Fig. 7).Medium energy sites dominate in the original construction sand.In the studied seam materials the water binding forces are lower.Most of the seam materials show a high number of low energy sites and only a low number of higher energy sites.This indicates a higher fraction of polar surfaces in the lower layer.The shift between the construction sand and the seam material is caused by the accumulation of nonpolar substances, mainly organic material.Additionally to the accumulation, organic substances may have coated the polar mineral surfaces in the upper layer.Higher polarity usually results in higher SCD.However, there is no dependence between E a and SCD for the seam material.Possibly, it has been shadowed by the salt accumulation effects, because E a was found to be related to the EC of the studied seam materials.The E a decreases with an increase of EC which is connected logarithmicaly with low hydration energy of the salt cations.Similar relations of adsorption energy and parameters related to soil salinity are observed in natural saline environments (Toth and Jozefaciuk, 2002).
The investigation of SCD and E a revealed characteristics of the seam materials OM, which should be validated by analysis of the quality and quantity of functional groups, e.g. using solid phase 13 C nuclear magnetic resonance spectroscopy (Preston, 1996;Baldock et al., 1992) or FT-IR spectroscopy (Ellerbrock et al., 1999).

Heavy metal adsorption
The variations for the Freundlich parameters as well as for the background contaminations are high, especially regarding comparable pH values.The Freundlich K f values for Pb and Cd are higher in the 0 to 1 cm layer than in the 1 to 5 cm layer.

EGU
for m are similar for both depths.However, a comparison of K f without considering m is not meaningful.Therefore, the retardation factors were calculated, which consider both parameters.They vary from 2056 to around 95 000 for Pb and from 26 to 608 for Cd for the described case (Table 3).Thereby, the comparison between the seam material and the original seam filling reveals a clear positive effect of the depositions in the seam material.

Pb adsorption
For samples from Berlin, the higher adsorption potential of the seam material compared to the construction sand corresponds to higher C org contents, which is expressed by the correlation of the retardation factor, R and C org [g g −1 ] (R = 281986 C org + 3600; r 2 = 0.82).Thereby, two samples have been excluded from the dataset: B1 and B2.For both samples, the Freundlich parameters were fitted with a specified range for m: 0≤m≤1 to meet the conditions for Freundlich's model stated by Sposito (1980).However, in all investigated parameters they behave not different from the other samples, except for the total concentrations of Pb, which are comparably low.
There is no correlation of C org and R for the Warsaw samples (R = 10 6 C org + 9920, r 2 = 0.15 ).If the sample W1 is excluded from the dataset for the same reasons like B1 and B2, the corellation is more significant but the resulting regression function makes no sense (R = 2×10 6 C org -4243, r 2 = 0.64 ).However, the samples from Warsaw have higher R (p<0.05,N = 9) at similar C org , caused by higher K f (p<0.05,N = 9) and similar values for m.Following Sposito (1980)  EGU the Warsaw samples.In Berlin, Pb tot correlates very well with C org , indicating same sources, e.g.combustion processes and/or strong adsorption of Pb on organic matter.However, the lacking correlation for the Warsaw samples could also be an effect of (iii) the small number of samples compared to the high spatial variability in urban soils.

Cd adsorption
The adsorption of Cd to the soil substrate is higher in the seam material compared to the original construction material.All retardation factors are higher for the seam material compared to the construction sand (Table 3).Compared to Pb, for Cd the improved sorption capacity is not a result of the deposition of organic material, as there is no corellation between R and C org , nor Cd tot and C org .So the deposition of other adsorbers, which have not been studied here, must be relevant.However, there is also no difference in the adsorption behaviour of samples from Warsaw or Berlin towards Cd.
Compared to natural sandy soils, the seam material shows a similar retardation towards Pb but a very low retardation of Cd.Kocher (2005) investigated sandy soils along federal highways in Germany for their HM retention while Adhikari and Singh (2003) investigated sandy soils from India and England (Table 4).The soils A1, A2, and A3 had higher C org (31, 50, and 50 g kg −1 ) and slightly lower pH values (6.5,6.6,and6.1) but were comparable in terms of clay content.For a comparison, the retardation factors were computed for the same case like in table 3.For Pb, the values are are slightly, but not significantly higher.This is due to similar K f , but very low values for m (0.24, 0.18, and 0.08) indicating a highly non-linear adsorption favouring adsorption at low concentrations.
For Cd Kocher (2005) reports of very similar m values, while the values for K f are significantly higher in road side soils compared to seam materials.So, the Cd retardation in roadside soils is clearly higher compared to seam materials.

Simulated HM displacement
For the groundwater recharge rate of 120 mm a −1 , the simulated displacement times vary from 315 to 2350 a for Pb and from 37 to 253 a for Cd (Table 5).At the site B2, the measured Pb concentration for the liquid phase in the soil is already higher than the concentration in the rainwater runoff.For Cd, however all sites show higher soil water than rainwater concentrations.In these scenarios, every rainfall is a net HM displacement!Apart from these cases, the positive filter effects of urban depositions, which were already discussed by means of retardation factors, could be approved by the simulations for all cases but one.In the simulations, besides adsorption behaviour, also the soil hydraulic properties and densities were considered.The effect of the smaller dry bulk density of the seam material can compensate the effect of stronger adsorption.The displacement of Cd at the site B4 is the one example.If the scenarios ia and iiia for the site B4 are calculated with the original seam filling's dry bulk densities, the travel time through the seam material would be 105 a and therefore higher than that of the original seam filling.So except for the displacement of Cd at the site B4 we conclude that in general, the seam material has a positive impact on the retardation of Cd and Pb.Although the impact is detectable, it is not very high and the formation of only 1 cm thick layer of seam material leads not to substantial longer travel times.However, the potential of the urban dirt as a Pb filter is clearly to be seen if the scenarios iiib are compared to scenarios ia: even contaminated seam material is a much better filter than any clean construction sand can be.For Cd, the background contamination is that high, that both seam material and original construction sand act rather as a source than as a sink.For Pb however, the calculated traveltimes do not imply an acute risk of groundwater contamination.
Because of the abundance of ponds at streets and sidewalks in urban areas, we simulated the infiltration of accumulated rainwater runoff from ponds.Realizing a higher infiltration rate of 0.48 instead of 0.12 cm d −1 influences the soil water content and therefore the pore water velocities.The average ratio between the traveltimes for Introduction

Conclusions References
Tables Figures Full Screen / Esc Printer-friendly Version

Interactive Discussion
EGU 120 mm a −1 and 480 mm a −1 is 1/4.The calculated traveltimes range from 10 to 100 a for Cd.As the pavements at the investigated sites were constructed from the early 1900s on, a substantial amount of heavy metals may already have left the upper soil layers.The displacement risk for Pb even from ponds is rather low.However, the simulated infiltration from ponds shows the high importance of non-uniform infiltration in urban areas.It therefore underlines the need of knowledge on soil surface properties for realistic risk assessments, e.g. for more mobile substances such as glyphosate (Kempenaar et al., 2007).

Conclusions
Depositions of all kinds of urban dirt form the seam material, which has different properties than the original material.Compared to the original construction sand, the depositions lead to increased specific surface area and cation exchange capacity.Compared to natural organic matter this anthropogenic form of organic material has a rather small surface area and surface charge density which results in comparable low cation exchange capacity.
In terms of heavy metal mobilisation and retention the seam material can act as a filter and a source depending on the element.Thereby the source as well as the filtereffect are low because of the typical thin layer of only 1 cm.However, the simulated break through times for one dimensional matric flow suggest, that there is no general risk of a groundwater contamination from traffic released Cd and Pb.However, the infiltration of rainwater runoff from ponds can lead to fast displacement of Cd, which may leave the pavement system after only a decade.We conclude, that the seam material is a interesting model substrate to show the positive impacts of deposited dust on ecological soil functions.Fertilizer, 29, 218-225, 1996. 2632 Kempenaar, C., Lotz, L. A. P., van
, the chemical Introduction , which causes the comparable low Introduction The values Introduction that indicates a higher variance of adsorbers in Warsaw, providing adsorption sites with a broader range of Pb adsorption energies.This might be due to, (i) a different kind of organic matter in Warsaw, which provides more adsorption sites for Pb compared to Berlin.However, this is not supported by different water adsorption energies (E a ) nor by different surface areas or different Pb amounts whether total or EDTA extractable amounts.More likely, it shows (ii) the greater importance of other adsorbers than the studied organic substances for Introduction

Table 1 .
der Horst, C. L. M., Beltman, W. H. J., Leemnans, K. J. M., and Bannink, A. D.: Trade off between costs and environmental effects of weed control on pavements, Crop Protection, 26, 430-435, 2007.2642 Kocher, B.: Eintr äge und Verlagerung strassenverkehrsbedingter Schwermetalle in Sandb öden Introduction Input parameters for the numerical simulation of heavy metal displacement through a pavement soil column.K S * and Θ s * indicate derived effective model parameters for realizing the water flow.

Table 2 .
Surface properties and general characteristics of seam materials and original seam fillings from Berlin and Warsaw.Values in one column followed by the same letter are not significantly different at p<0.05; N=20.

Table 5 .
Simulated heavy metal displacement through a 20 cm paved soil column at different sites in Berlin.a The travel times indicate the period which is needed to recover 95 % of the applied concentration at the lower boundary of the soil column.