Lability of Pb in soil: effects of soil properties and contaminant source. Environmental Chemistry, (6).

the combined use of isotopic show how of Pb and affect Soil pH and soil Pb content are the dominant controls on Pb lability; the lability of recent petrol-derived Pb is similar to that of other sources in urban soils but greater than geogenic Pb in rural ABSTRACT Lability of lead (Pb) in soils is influenced by both soil properties and source(s) of contamination. We investigated factors controlling Pb lability in soils from (i) land adjacent to a major rural road, (ii) a sewage processing farm and (iii) an archive of the geochemical survey of London. We measured isotopically exchangeable Pb (E-values; Pb E ), phase fractionation of Pb by a sequential extraction procedure (SEP) and inferred source apportionment from measured Pb isotopic ratios. Isotopic ratios ( 206 Pb/ 207 Pb and 208 Pb/ 207 Pb) The following Tables provide additional details for the Rural Roadside, Sewage Farm and London archive soils, including; selected soil properties, Pb content and lability, and the isotopic ratios 206 Pb/ 207 Pb and 208 Pb/ 207 Pb in both the whole soil and the labile pool of soil Pb.


Introduction
Industrial pollution, combustion of leaded petrol and mining activities have increased environmental Pb concentrations since the Industrial Revolution. The Agency for Toxic Substances & Disease Registry (ATSDR, 2012) lists Pb as the second most dangerous substance that threatens human health, not only because of its high toxicity but also due to its wide dispersion around the globe (Nriagu, 1996;Watmough and Hutchinson, 2004;Bacon et al., 2006). Human exposure to Pb, via inhalation of particles and ingestion of contaminated soil, can result in physiological damage, particularly to the nervous system. Due to these health concerns, the commercial use of Pb has been considerably reduced over the last three decades and unleaded petrol introduced to reduce aerial Pb deposition Flament et al., 1996;Johnson et al., 1995). However, soils retain a large reservoir of anthropogenic Pb, derived partly from atmospheric deposition (Emmanuel and Erel, 2002;Sterckeman et al., 2000). Young children are at greatest risk from ingestion of soil contaminated with Pb because of their compulsive hand-to-mouth response (Wixson and Davies, 1994) and their greater sensitivity to Pb toxicity (Kayhanian, 2012). However, soil Pb may also enter the biogeochemical cycle through plant uptake, surface runoff and leaching into groundwater systems, depending upon the reactivity and solubility of the soil-borne Pb burden. Therefore, to fully quantify the risks associated with Pb it is also useful to assay the labile pool of Pb in soil, which is more closely linked to bioavailability and solubility than the total concentration in soil (Meers et al., 2007;Sauvé et al., 2000;Tongtavee et al., 2005).
The lability of cationic trace metals is generally lowest in alkaline soils with large mineral oxide and carbonate contents (Buekers et al., 2007;Tye et al., 2003). Trace metal binding to soil organic matter (SOM) and Fe/Mn oxides is also strongest at high pH (Tack, 2010); in contrast it appears that the effect of humus on lability is not consistent for all trace metals (Degryse et al., 2009). For Pb, lability may also be limited by formation of discrete mineral phases such as anglesite (PbSO 4 ) or Pb jarosite in acidic environments (Ruby et al., 1996) and chloropyromorphite (Pb 5 (PO 4 ) 3 Cl) which has been widely reported (Lang and Kaupenjohann, 2003). Association of Pb with discrete soil fractions has been investigated using sequential extraction procedures (SEP) by several workers (Bacon et al., 2006;Emmanuel and Erel, 2002;Imperato et al., 2003;Liu et al., 2003;Teutsch et al., 2001;Thornton et al., 2008). Attempts to identify the source of Pb in individual SEP extraction steps, using Pb isotopic ratios, have found that geogenic Pb was primarily associated with the residual fraction whilst anthropogenic Pb was distributed amongst carbonate, humus and Fe-oxide phases (Emmaunuel and Erel, 2002;Teutsch et al., 2001). However, the application of SEP assays in risk assessment may be limited because the labile Pb pool that controls solubility is unlikely to correspond to any single SEP fraction (Atkinson et al., 2011).
Isotopic dilution techniques can be used to determine the labile trace metal content of soils (Degryse et al., 2009). These methods can define the fraction of metal both in solid and solution phases that is 'isotopically exchangeable,' known as the 'E-value' (Smolders et al., 1999;Young et al., 2000). A small number of studies have used isotopic exchange with enriched stable isotopes to investigate lability of Pb in soils. Although variation in soil Pb solubility and lability is commonly ascribed to soil characteristics that affect adsorption strength (Gustafsson et al., 2011) it is likely that the original source of Pb is also a determining factor (Atkinson et al., 2011). Degryse et al. (2007) measured Pb E in soils from 3 historically industrialized sites (pH 6.6-7.5) and found %Pb E ranged from 45% to 78% (mean 58%). They suggested that high values of %Pb E in soils contaminated with smelter fallout or battery production waste may arise because the Pb from these sources is more likely to enter the soil in a relatively soluble form. Similar %Pb E values have been measured in acidic woodland and grassland soils (Marzouk et al., 2013a) whereas, in the same study, in alkaline minespoil soils with extremely high Pb concentrations (> 20,000 mg kg -1 ) %Pb E values were very low because of poorly soluble secondary carbonate or primary sulphide mineralogy; overall the range in %Pb E was 7-99% when all soils were considered. In a second study of Pb lability in a contaminated catchment, Marzouk et al. (2013b) observed an inconsistent trend in %Pb E value with soil metal:. They suggested that whereas an increase in soil Pb loading would reduce soil Pb adsorption strength it also increased the likelihood of poorly soluble minespoil minerals being present. Overall the range of %Pb E values reported was <10% to >90% (n= 246).
The current study was an extension of work by Atkinson et al. (2011) who investigated Pb fractionation in four soils with distinctive metal sources and soil properties. The objective of this study was to broaden the investigation of the effect of soil properties and sources of contamination on Pb lability in soils, as determined by isotopic dilution. Soils (108 samples in total), covering a wide range of soil properties with both natural and anthropogenic sources of Pb, were obtained from three different localities (designated as 'Rural Roadside', 'Sewage Farm' and urban 'London' soils). Soil samples were analysed for chemical properties likely to affect Pb solubility, total Pb isotopic signature, fractionation by sequential extraction and isotopically exchangeable Pb (Pb E ); multivariate regression was used to examine the influence of soil properties within each set of soils. The sources of Pb in both labile and nonlabile pools in the Rural Roadside soils were estimated by measuring the isotopic ratios 206 Pb/ 207 Pb and 208 Pb/ 207 Pb in those fractions. An attempt was made to quantify the relative contribution from two likely sources (geogenic Pb and leaded petrol) to the isotopically exchangeable soil Pb fraction (%Pb E ) to determine whether Pb from petrol remains more labile than geogenic soil Pb despite the withdrawal of leaded petrol for several decades in the UK. Initially we regard 'Broken Hill Type' lead (BHT-Pb) as being synonymous with 'petrol-derived lead' but the limitations to this assumption are discussed in Section 3.5.

Soil sampling
Three sets of soils, with different sources of Pb contamination, were used in this study.
Twenty one topsoils (0-20 cm) and twenty one subsoils at a range of depth intervals (between 20-60 cm) were collected from four sites along a major road (the A6) leading to the M1 motorway in a rural area of Nottinghamshire, UK (52°48'N, 1°16'W), where soils were likely to have received petrol-derived Pb from heavy traffic (designated 'Rural Roadside').
Sixteen topsoils were sampled from five fields at a sewage processing farm in Nottinghamshire, UK (52°57'N, 1°02'W) (designated 'Sewage Farm' soils). A further 50 topsoils (5-20 cm) were sub-sampled from the Geochemical Baseline Survey of the Environment (G-BASE) "London Earth" sample archive of the British Geological Survey (BGS) (Johnson et al., 2005). The 'London' soils selected were a subset from the systematic survey of the Greater London Authority area, chosen to cover 13 different land uses and a range of soil properties including pH, organic matter content and total Pb concentration.

Soil characterization
Soil samples were air-dried and sieved to < 2 mm. Soil pH was measured in 0.01 M CaCl 2 (1 : 2.5 soil : solution ratio) after shaking for 30 min. Soil organic matter content (SOM) was estimated from loss on ignition (%LOI) at 550°C for 7 hr. Available phosphate was determined using the Olsen method and a colorimetric assay (Rowell, 1994). Sub-samples of Rural Roadside and Sewage Farm soils were agate ball-milled and total Pb concentration was measured by ICP-MS (Thermo-Fisher Scientific, X-Series II ), following HF/HClO 4 /HNO 3 mixed acid digestion, at the University of Nottingham (UoN). The Pb concentration in the London soil samples was determined on powder pellets by wavelength dispersive x-ray fluorescence spectrometry using PANalytical MagiX-PRO PW2440 and Axios-Pro spectrometers in BGS, each fitted with 4kW Rh-anode x-ray tubes. The manufacturer's SuperQ software was used to account for matrix effects and correct for spectral line overlap interference. The XRF lower detection limit was 1.3 mg kg -1 , the calibration extended to 1,000 mg kg -1 and data were reported up to 10,000 mg kg -1 . Three of the London soils, in which total Pb concentrations exceeded 10,000 mg kg -1 were assayed by ICP-MS following acid digestion at the UoN.

Measurement of Pb isotopic abundances
The isotopic abundances of Pb isotopes (IA) in the HF/HClO 4 /HNO 3 acid digestates of Rural  (Rohl, 1996). For Pb in UK coal published average IR values for England and Wales were used (Farmer et al., 1999): 206 Pb/ 207 Pb = 1.184 ± 0.005 and 208 Pb/ 207 Pb = 2.461 ± 0.012. Regional Pb ore and coal are also used here as a proxy for soil parent material (i.e. geogenic Pb). The ratio of 206 Pb/ 207 Pb is generally adopted as the most sensitive to change in environmental studies (Bacon et al., 1996) because the similar isotopic abundance of the isotopes minimises instrumental bias and maximises precision.

Lead lability measured by isotopic dilution
A method adapted from Gäbler et al. (1999) where M Pbsoil and M Pbspk are the average atomic masses of Pb in soil and spike respectively, W is the weight of soil (kg), C spk is the gravimetric concentration (mg L -1 ) of Pb in the spike solution, V spk is the volume of spike added (L), IA is the isotopic abundance (on a mole basis) of 204 Pb or 208 Pb in the spike or soil (0.01 M Ca(NO 3 ) 2 extracts), and R ss is the IA ratio of 204 Pb to 208 Pb in the separated solution phase of the spiked soil suspension.
The isotopic compositions of the labile and non-labile Pb fractions were determined to investigate the source of Pb in each case. The isotopic abundance in the non-labile pool was The isotopic ratios ( 206 Pb/ 207 Pb and 208 Pb/ 207 Pb) of petrol Pb and coal Pb were used to estimate the proportion of each source of Pb in the labile and non-labile pools. The assumption that Pb came only from these two sources is a simplification but it indicates whether Pb from petrol is more labile than Pb from local parent material which has an isotopic signature similar to that of UK coal or Pb ore minerals. The concentration of petrolderived Pb which is labile was estimated by linear interpolation of isotopic ratios ( Pb total is total soil Pb concentration (mg kg -1 ) measured by acid digestion. The proportion of the petrol-derived Pb which is labile can then be calculated by dividing Pb E(petrol) by Pb total(petrol) . Equations 3a and 3b are extensions of a widely used simple binary mixture model, based on interpolation between the isotopic ratios of prescribed end-members, discussed by Komárek et al. (2008). It should be noted that, although interpolation between the IR values of end members is commonly used, this approach incurs a small error. By assuming an IA value of 0.014 for 204 Pb in both end members and comparing results from IR interpolation with calculations of isotopic abundance (IA) we estimate that there is a maximum error of about 1.3% in source apportionment; this occurs at the midpoint across the full range of possible source compositions (i.e. 0 -100% petrol-derived Pb).

Relating the labile fraction of Pb to soil properties
Two approaches were used to investigate the relationship between Pb lability (%Pb E ) and soil properties; a logistic model and multiple regression. The parameters in the logistic model (pH 50 and k S Eq. 4) were optimized using the Solver function in Microsoft Excel, on all three sets of soils to examine the relationship between %Pb E and pH : where pH 50 is the pH at which %Pb E is 50% and k S is a spreading factor that determines the shape of the curve. Equation 4 provides limits of 0 and 100% on the value of %Pb E and assumes that pH value is the primary determinant of %E-value in a manner analogous to the adsorption behaviour of trace metals in soils.
The relationships between the labile fraction of Pb (Pb E ) and soil characteristics including pH, %LOI, available phosphate (P Olsen , mg kg -1 ) and total Pb concentration (Pb total , mg kg -1 ) in soils were also investigated using multiple regression (Eq. 5); the constants (k 1 − k 4 ) were optimized using the stepwise regression function in Minitab 16.
log 10 Pb E = k 1 log 10 Pb total + k 2 pH + k 3 %LOI + k 4 P Olsen (5) Correlation coefficients between measured and modelled values of log 10 Pb E and residual standard deviations (RSD) for the model fit were used to assess the contribution of each soil variable to the prediction of labile Pb.

Sequential extraction of Pb
A sequential extraction procedure (SEP, adapted from Li and Thornton, 2001) was applied to 32 soil samples selected to cover a range of values of the variables in Eq. 5 (Pb total , pH, %LOI and P Olsen ). Soils included 10 Rural Roadside (both topsoils and subsoils), 12 Sewage Farm and 10 London soils with a range of land uses. Full details of the SEP are summarized in Table 1. The concentration of Pb in the residual fraction (F5) was determined by difference between Pb total , measured independently, and the summation of F1 to F4 of the SEP.

Soil characteristics
Soil characteristics are shown in Table 2. Rural Roadside sites had a wide range of pH values (3.9 -7.6) with slightly lower values of %LOI at depth in the subsoils. Available phosphate concentration (P Olsen ) and total Pb concentration were relatively low compared to the other two sets of soils. The average Pb concentration in Rural Roadside topsoils was approximately double that in subsoils suggesting anthropogenic inputs. Sewage Farm soils had a narrower pH range (5.9 -6.7), greater %LOI (mean=16%) and very large available phosphate content (53-380 mg kg -1 ). Lead concentrations varied from 55-712 mg kg -1 with lower concentrations in fields where pH was slightly higher and %LOI and available phosphate were lower suggesting lower historical sewage sludge inputs. For comparison, the BGS G-BASE dataset for soils in Nottinghamshire (BGS; 636 samples) gives a background Pb concentration range from 13.8 to 976 (mg kg -1 ) with an average of 145 mg kg -1 and median of 100 mg kg -1 . The London soils covered a wide pH range (3.3-7.3) and had similar values of %LOI (mean = 9.7%) to those of the Rural Roadside soils. Available phosphate was greater than for the Rural Roadside soils but much lower than for soils from the Sewage Farm. Lead concentrations were very high in the London soils (median= 940 mg kg -1 ) and showed greater variation (99-22600 mg kg -1 ) between different land uses, than soils sampled elsewhere.
Many urban sites in the UK are heavily contaminated with metals; the G-BASE dataset for urban soils (BGS; 13583 samples) gives a median value of 128 mg kg -1 and a range of 2.1 to 10,000 mg kg -1 . The latter figure is certainly exceeded and simply represents the maximum quantification limit for XRF.
The isotopic characteristics of the 108 soils are shown in Fig For London soils however, the correlation coefficient between isotopic ratio ( 206 Pb/ 207 Pb) and total Pb content was not significant, suggesting a more complex mix of source materials. Farmer et al. (2011) studied the relationship between total Pb concentration and the isotopic ratio 206 Pb/ 207 Pb in 27 urban soils in the city of Glasgow, Scotland and also suggested that the isotopic ratio of total soil Pb represented a complex mixture of Pb deposited from a variety of contaminant sources since the Industrial Revolution.

Pb lability in soils affected by soil properties and sources of contamination
In a study of Pb bioaccessibility, the source of Pb was found to be less important than the concentration and physico-chemical influences of soil properties (Farmer et al., 2011). In the current investigation, the median Pb lability values (%Pb E ) measured for the Rural Roadside topsoil and subsoil, Sewage Farm and London soils were quite similar at 31%, 30%, 24% and 22% respectively. However, Fig. 2 also shows a wider range of Pb lability in the Rural Roadside subsoils than in both the Sewage Farm and London soils.

Rural Roadside soils
Alkaline subsoils from the Rural Roadside sites, had the lowest values of %Pb E , ranging from 0.6-12.7%, where Pb from parent material is likely to be fixed within soil particles or may be in the form of discrete Pb minerals such as cerrusite (PbCO 3 ) or pyromorphites (Li and Thornton, 2001). Lead-containing particles from petrol would be expected to dissolve in soils and adsorb to geocolloidal phases. Teutsch et al. (2001) found that petrol Pb was predominately bound to carbonate and Fe-oxides but presumably this will depend on soil composition.

Sewage Farm soils
Although the mix of sources of Pb contamination in the Sewage Farm soils may have been similar to those from the Rural Roadside sites, 75% of %Pb E values fell in the lower range of 20-27% (Fig. 2). This was probably because of the relatively limited range of pH values in the Sewage Farm soils and their extraordinarily high phosphate concentration (Tye et al., 2003). Contaminant Pb in the Sewage Farm soils is almost exclusively from sewage sludge but the history of sludge application also determines the concentrations of soil humus and available phosphate. There was therefore a strong co-variance between the soil Pb content and the factors likely to affect its lability (Brazauskiene et al., 2008): soil Pb content was highly correlated with both %LOI (r= 0.97) and P Olsen (r= 0.75). This, coupled with an on-site liming policy to maintain a limited range of soil pH values, around pH 6.5 (Severn Trent Ltd, pers. comm.), probably explains the low and very restricted range of %Pb E values observed so that Pb total explained 96% of the variation in Pb E . A flat, non-significant, relationship was observed between %Pb E and P Olsen (not shown) which would appear to suggest that phosphate had no effect on Pb fixation in these soils. This seems intuitively unlikely as a soil subject to increasing phosphate inputs might be expected to show a negative trend in %Pb E against soil P content. However, the absence of such a trend may arise as a consequence of the coaddition of Pb and phosphate to the soil, in a 'pre-reacted' form, in sewage sludge.
Furthermore, there was a strong negative correlation between P Olsen and soil pH (r = -0.73), probably arising from oxidation of sludge components (organic carbon, ammonium), so that the lower pH values at high phosphate (and sludge) loadings would elevate %Pb E against the expected trend arising from phosphate fixation of Pb.

London soils
In the fifty London soils, although land use, soil properties and possible sources of Pb all covered a wider range than at the other two sites, the range of Pb lability was relatively small with 75% of the %Pb E values falling in the range 16−26% (Fig. 2). This was a larger range than found in the Sewage Farm soils but much smaller than seen for the Rural Roadside soils and so soil properties appeared to have only a minor effect on Pb E with Pb total accounting for was from a park planted with deciduous woodland with an exceptionally low pH value (for this dataset) of 5.28.

Predicting Pb E from soil properties
Predicted values of %Pb E as a function of soil pH using a logistic model (Eq. 4), is shown in Fig. 3a, with model constants, RSD and R 2 given in Table 3. Using the logistic model in preference to linear regression ensures an asymptote in %Pb E of 100% at low pH and zero at high pH. Plotting data from all three sources together suggests only a very broad trend in Pb lability (%Pb E ) with soil pH. Whilst pH was a reasonable predictor of %Pb E for the Rural Roadside soils this was not the case for the London and Sewage Farm soils which had a more restricted range of pH and Pb E values. It is difficult to incorporate further soil variables (eg P Olsen , %LOI) into the structure of Eq. 4 in a meaningful form and so it is not possible to say whether the scatter around the model line (Fig 3a) arises from variation in soil properties or contaminant characteristics. Therefore to test the importance of other soil properties we also used a regression approach (Eq. 5) to test the dependency of Pb E on total soil Pb concentration (Pb total ; mg kg -1 , log 10 scale), pH, %LOI and P Olsen. Both %LOI and P Olsen (k 3 and k 4 in Eq. 5) were found to be non-significant in improving the prediction of log 10 Pb E when all three sets of soils were treated as a single dataset. The inclusion of Pb total and pH produced an almost identical model performance to the logistic model (  Figure 4 shows the mean proportion of Pb in each SEP fraction and the proportion of Pb that was labile (isotopically exchangeable) or non-labile. The SEP results indicate that the 'inert' residual pool (F5) was the largest fraction for all three sets of soils (37.2% for Rural Roadside soils, 77.6% for Sewage Farm soils, and 41.8% for London soils), which corresponds qualitatively with the results from other studies of Pb in contaminated soils ; Thornton et al. 2008;Liu et al. 2003

Differences between the isotopic signature of labile and non-labile pools of Pb
Previous studies have suggested that Pb from petrol and coal combustion particulates is more soluble in dilute acid extractions than geogenic soil Pb which tends to be associated with the residual pool of SEPs (Erel et al., 1997;Li et al., 2011). Pb from petrol and geogenic sources in these soils has become distributed between labile and non-labile forms in the same proportions following prolonged contact with the soil. Of course, effects arising from the original form of Pb cannot be discounted. For example reaction between phosphate and Pb during sewage sludge production could reduce the lability of petrol-derived Pb prior to introduction to the soil. By contrast, the Rural Roadside topsoils provided some evidence of a link between isotopic signature and Pb E , with consistent and significant elevation above the R NL,L = 1.0 line. This suggests greater lability within petrol-derived Pb compared to the geogenic soil Pb pool. An isotopic distinction between labile and non-labile Pb is most likely where there is a small background Pb concentration into which the major pollutant source is petrol-derived Pb. With approximately 1/8 th the Pb content of the Sewage Farm soils and 1/40 th that of the London soils (Table 2) It is also conceivable that the displacement from an R NL,L value of 1.0 seen in Rural Roadside topsoils could arise from a small systematic error in ICP-MS measurement. We took all the measures we could to avoid this, including 'in sample run' mass bias correction with an isotopic standard (NIST 981), diluting all samples to a restricted Pb concentration range (typically 5 -50 µg L -1 ) to avoid detector mode changes, running inter-laboratory comparisons between UoN and BGS and correctly setting the detector dead time correction factor to avoid systematic shifts in IR measurements with Pb concentration. However, although the random errors for R NL,L values in Figure 5 were extremely small there is no way to be certain that the very small shifts in IR being measured are valid. Figure 6 shows the proportion of labile Pb (%Pb E ) in petrol-derived and geogenic fractions (from Eq. 3) for topsoils from the three sites as a box and whisker diagram. The Rural Roadside subsoils were not included in this figure, as it is difficult to give a confident estimation of the proportion of petrol-derived Pb in these soils. The values of IR total were very close to or beyond the isotopic signature of Pb coal (1.181) from England and Wales (Fig. 1).
In Rural Roadside topsoils, there was a small but significant difference in Pb lability between petrol-derived Pb (35%) and geogenic Pb (27%) (paired t-test). As found for the overall %Pb E values, there was a reasonably strong correlation between soil pH and the labile proportion of both geogenic Pb (r = -0.70；p<0.001) and petrol-derived Pb (r = -0.60; p<0.005). For Sewage Farm and London soils, the values of %Pb E for Pb originating from petrol and geogenic sources were not significantly different. For some of the Sewage Farm soils, as suggested by Atkinson et al. (2011), the co-existence of large phosphate concentrations may cause rapid fixation of petrol-derived Pb during sewage sludge processing which then contradicts the pattern seen in the Rural Roadside soils where petrolderived Pb was clearly more labile. For the London soils, despite having the widest range of possible Pb sources (Fig. 1), the proportions of petrol-and geogenic Pb that were isotopically exchangeable were very close (average of 23.1% and 22.1% respectively). This suggests a remarkably consistent level of assimilation into the soil and a geochemical "aging" to a similar reactivity for Pb from all sources and is consistent with the narrow range of Pb E seen across this dataset (Fig. 2).

Limitations to the binary model of Pb source apportionment
There are substantial limitations to the assumption that Pb in soil originates from just two major sources. First, the assumption that a clear linear mixing line between two end members indicates the presence of just two sources can be challenged. The analytical constraints of quadrupole ICP-MS, which preclude accurate determination of small differences in 204 Pb isotopic abundance, have encouraged a reliance on plots such as 208/207 Pb against 206/207 Pb (Fig. 1) in source apportionment studies. However, as Ellam (2010) has shown, multiple geological sources of Pb are likely to cover a limited range of values on such plots. This factor combined with the low IR value for 206 Pb/ 207 Pb in the 'Broken Hill Type' lead (BHT-Pb) used in petrol, tends to produce an apparently linear plot, erroneously suggesting that there are mixtures of just two end members. The second, more obvious, problem arises from the simple assumption that BHT-Pb is synonymous with petrol-derived Pb. In fact, imported BHT-Pb was widely used in industrial applications in the UK in the 20 th century (Vane et al., 2011;Chenery et al., 2012). Thus Bacon et al. (1996) found that the historical trend in the 206/207 Pb ratio in archived Park Grass herbage samples declined ahead of the introduction of tetra-ethyl Pb in petrol, indicating industrial inputs from BHT-Pb into atmospheric aerosols in the UK from as early as 1900. For the three datasets presented in the current study, it may be reasonable to assume that Pb in the Rural Roadside soils is predominantly a combination of petrol-derived and geogenic Pb. The Sewage Farm soils, historically amended with sewage from Nottingham, must have a greater input of non-petrol BHT-Pb even though Pb from urban road runoff is likely to form a substantial part of the soil Pb burden. The London soils, although also substantially affected by petrol Pb, are likely to be the most affected by non-petrol BHT-Pb. An extreme example from the London dataset is Sample 654432 (Electronic Annexe 1) with a soil Pb content of almost 23,000 mg kg -1 and the most pronounced BHT signature of all the soils studied ( 206/207 Pb ratio = 1.081 (BGS, total Pb) and 1.082 (UoN, labile Pb)). Intuitively it is highly unlikely that most of the soil Pb originated from petrol; we estimate that this would have required the equivalent of Pb from more than 30 L petrol per kg soil. The sample site is located close to a dockland area and scrutiny of a local map from the early 20 th century, shows the historical presence of a 'White Lead Works' (2PbCO 3 . Pb(OH 2 ) for paint etc.) within 400 m of the site -possibly using imported Australian or Canadian lead.
The current challenge to source apportionment is perhaps to combine Pb isotopic ratio data, including consideration of 204 Pb, with other markers of industrial and road traffic sources, such as Sb and Cu from brake liners (Fujiwara et al., 2011;Huang, 1994;Weckwerth, 2001)

Conclusions
In summary, both soil properties and sources of contamination influence the isotopic exchangeability (lability) of Pb. Although soil pH was the only soil factor tested which There was no consistent agreement between Pb E and any single SEP fraction in all three sets of soils. However, the non-labile Pb in both Rural Roadside and London soils was likely to be a combination of Pb occlusion within Fe oxides (F3) and primary minerals (F5 of the SEP), whilst for Sewage Farm soils, the SEP residual fraction (78%) could be identified very closely with the (isotopically) non-labile pool of Pb (76%) -possibly a consequence of Pbphosphate interactions prior to biosolid application.
There was evidence in Rural Roadside topsoils that petrol-derived Pb remained more labile than Pb from geogenic ore or coal, and the proportion of petrol-derived Pb that was labile was strongly correlated with soil pH. There was also limited evidence, from comparison with  :   Table 1: Summary of the modified Li and Thornton sequential extraction procedure (SEP). Table 2: Land use and soil characteristics for the three site locations studied. Table 3: Constants, RSD and R 2 values for the prediction of Pb E using logistic (Eq.4) and linear regression (Eq. 5) models. The logistic model (Eq. 4) was parameterized against measured values of %Pb E ; modelled values of %Pb E were then multiplied by measured Pb total values to estimate log 10 (Pb E ). The regression model (Eq. 5; three variants) was used to predict log 10 (Pb E ) directly.