Methods

Study area and emission sources

The study area was focused on the zinc smelter at Avonmouth, in south-west England.  The town of Avonmouth stands on the edge of the Bristol Channel, with the industrial cities of Newport about 20km to the northwest, and the major city of Bristol (population 380,000) about 9km to the southeast.  Other major settlements within the surrounding region include Bath (population 169,000) and Gloucester (110,000 population) (Figure 1).  The topography is for the most part low-lying, with an altitude range within the surrounding 20 km of 308 metres.  The climate is temperate oceanic, with prevailing winds from the south-west – a wind flow which is to some extent strengthened by the Bristol Channel.  

Avonmouth itself has long been an important area for manufacturing and heavy industrial activity, but the Britannica Zinc Smelter was the only major source of Cd emission in the region – and until it closed in 2002, the largest single emission source in the UK (Filzek et al. 2004).  The site lies in the northern quarter of Avonmouth (Figure 1), and the site comprised up to 12 operating stacks set within an area of about 60 ha. 

 Figure 1.  Map of study area

Modelling

Emission sources and exposure pathways

Computation of intake fractions from an industrial source such as the Avonmouth zinc smelter needs to take account both of the different emission sources and dispersion pathways involved, and of the long timescales during which released Cd may remain in the environment before being absorbed by humans.  As with other industrial sources, the main exposure pathways are via the air, drinking water and food.  Direct releases into the environment occur largely through stack emissions into the atmosphere, primarily in aerosol form (as cadmium oxide); the smaller amounts released in gaseous form quickly become adsorbed onto particles.  In addition, considerable quantities may be released as a result of fugitive emissions from the smelter, including escape from pipes and valves, and in the form of dust from stored ores, during transport and during milling. Coarser particles, including especially dusts emitted from ground-level sources, are likely to travel only short distances before removal by dry or wet deposition.  On the other hand, finer particles, released in the heated plumes with high emission velocities from the main stacks, can be carried to considerable elevations in the atmosphere and laterally over distances of thousands of kilometres.  Most of the cadmium in the atmosphere has thus been shown to be concentrated in finer particles (PM₁ or less). As such, it lies within the respirable particle size fraction (Milacic and Dolinsek 1994, Williams and Harrison 1984).

Removal of Cd from the atmosphere occurs in a number of ways: by dry deposition, under the influence of gravity, by wet deposition and washout by rain, as a result of filtering by vegetation and other physical barriers (e.g. ventilation systems in buildings), and inhalation by animals (including humans).  Over land, deposited cadmium tends to accumulate in the soil.  Cd binds strongly to organic matter and, to a lesser extent, to clays and is thus somewhat immobile, especially in humus-rich or fine-grained soils.  Residence times in the soil are therefore long, and most of the Cd accumulates in the topsoil in non-bioavailable form.  Plant uptake is therefore slow, and inputs into the human food chain restricted.  Leaching is also inhibited (Eriksson 1988, Li and Shuman1996), with the consequence that opportunities for Cd input into drinking waters is limited, and measured Cd concentrations in treated waters are generally low (Ryan et al. 2000).  Loss by soil erosion, on the other hand, can be substantial, with the consequence that much of the Cd reaching streams is in insoluble form, attached to soil particles.  Stream sediment loads tend to increase with slope angle because of the increased erosion potential (Korentajer et al. 1993). 

Based on these pathways, the intake of cadmium from the Avonmouth zinc smelter may be summarised as in Figure 2. As this indicates, the intake pathway via drinking water is considered to be negligible, and is ignored here.  Intake fractions are thus estimated only for the air and food pathways.  The former may be assumed to occur mainly in the near vicinity of the smelter (though long distance transport of small amounts of Cd must also be acknowledged); the latter affects local populations through consumption of locally-grown foodstuffs but also spreads more widely via commercial food distribution systems.  It should also be noted that numerous secondary intake pathways also exist, for example from solid wastes and sewage, or re-entrainment of dust and sediments.  These are not allowed for in the assessment, but their implications are discussed below.

Figure 2.  Cadmium pathways

Emissions

The zinc smelter at Avonmouth was operational between 1920 and 2002, during which time production levels, technologies and emissions varied substantially.  Under the Integrated Pollution Prevention and Control legislation, monitoring of emissions has been obligatory since the early 1990s, and data prior to that were collected by the operators.  For this study, however, detailed data could only be obtained on stack emissions, and these only for the period since 1996 (subsequently, annual data for earlier years were located and made available).  Over the period 1996-2002, these averaged 1235 kg/year.  Although this almost certainly under-estimates past stack emissions, the pattern of dispersion is likely to have remained largely unchanged; intake as a proportion of emissions (i.e. the intake fraction) is also likely to have remained broadly constant, though some variation will have occurred because of population changes in the local area.  Detailed data on fugitive emissions were not available, but were typically at least the equivalent of those via the stack (Bristol City Council 2003).  For the analysis used here, therefore, both sets of emissions were set to 950 kg/year.     Releases to water are considered to have been negligible.  

Atmospheric dispersion

Atmospheric dispersion of the emitted Cd was modelled using the ADMS-Urban model.  This is a semi-Gaussian (non-Gaussian in the vertical profile in convective conditions) dispersion model.  It shows a number of improvements on previous generations of model including more advanced methods for modelling the near-surface layer and more powerful techniques for parameterising meteorological conditions (CERC 1999).  It has been widely validated under laboratory and field conditions, and shown to give reliable estimates of concentrations in well-parameterised conditions (Carruthers et al. 1999, 2000).  Uncertainties for mean annual concentration have been reported to be 20% or less (Colvile et al. 2002).  The model includes a terrain function to take account of topographic effects on dispersion and an intelligent interpolator for generating high resolution maps of pollutant surfaces.

The ADMS-Urban model was run for the year 2001 using hourly meteorological data obtained from Filton national weather station, 8km to the east of Avonmouth.  The emissions arising from the 12 operating stacks were modelled in point source mode whereas the fugitive emissions were modelled in area source mode.  Table 1 shows the model inputs and assumptions.  Cadmium concentrations were modelled for a 40 x 40km grid around the smelter with a 400m resolution.  Model performance was investigated by comparing predicted concentrations with measured concentrations at six sites for which monitoring had been done during 2000 and 2001. 

Intake from the air

Human intake of atmospheric Cd occurs through inhalation of contaminated air in a range of different environments and in different physiological circumstances.  Intake rates thus depend both on where people spend their time and on their activities in those micro-environments.  Detailed data on time activity were not available for the study population, so estimates were made for relevant parameters from the literature. 

In this context an important distinction needs to be made between indoor and outdoor exposures.  Most people spend a large proportion of their time indoors.  Data from the UK Time Use Survey (National Statistics 2002), for example, show that about 50% of the day, on average, is spent at home indoors, and 30% at work or school; a further 10% of time is spent in other indoor premises, such as the houses of friends or shops.  Indoor exposures of Cd released from the smelter thus depend fundamentally on its ability to penetrate buildings.  Few measurements of indoor and outdoor concentrations of Cd in the UK have been undertaken, but since most Cd is in aerosol form ratios may be imputed from studies of fine particulates.  In homes without major indoor sources (e.g. smoking, solid-fuel fires), these suggest that indoor concentrations are typically 50-80% of ambient concentrations, though with marked seasonal variations depending on ventilation practices (Dimitroulopoulou et al. 2001, Kingham et al. 2000, Milner et al. 2004, Wallace 1996).  In buildings with air conditioning, such as offices, indoor concentrations tend to be lower, generally in the order of 30-50% of outdoor concentrations (Janssen et al. 200, Wallace 1996).  Here, a mid-point ratio of 0.65 was used for homes and 0.40 for workplace environments.  These two micro-environmental factors were then time-weighted, by 0.5 and 0.4 respectively, to take account of the proportion of the day spent in each micro-environment.  Ambient concentrations were weighted by 0.1 to reflect time spent out of doors (Table 1).

Table 1.  Parameters used for modelling Cd intake from the atmosphere

Physical activity levels in these different micro-environments vary.  Again, detailed data are not available for the study population, but the UK Time Use Survey (2000) suggests that people spend less than 10% of their time in moderate to intense physical activities, such as heavy work or exercise-taking.  Inhalation rates likewise vary, both in response to activity levels and because of differences in physiological function (e.g. associated with age, gender, body mass and physical health).  Estimates of inhalation rates reported in the literature show a considerable range (e.g. US Environmental Protection Agency 1996, Layton 1993, Marshall et al. 2005), but in the UK an average value of 0.85 m³/hour is widely used for dose estimates for adults, while Layton (1993) has suggested rates of 0.2 m³/hour and 0.36 m³/hour for children aged <1 year and 1-10 years, respectively.  Here we assume inhalation rates for of 0.5 m³/hr and 1.5 m³/hr for indoors and outdoors respectively, with values of half these for children less than one year of age (Table 1). 

Intake from food

As noted, Cd enters foodstuffs largely as a result of uptake by plants growing in Cd-contaminated soils.  Soil contamination from the smelter is likely to occur mainly as a result of wet and dry deposition from the atmosphere, and might thus be expected broadly to reflect atmospheric concentrations in the surrounding area.  Measurements of soil Cd were, in fact, made during the 1980s and 1990s as part of a study by Long Ashton Research Station but data on Cd uptake by crops or animals in the area are not available.  Data have been collected on consumption of locally grown vegetables, and these along with locally produced meat and dairy products may be a significant exposure source (Hellström et al. 2007).  Nevertheless, it cannot be assumed that intake via food is restricted to the local population; instead, agricultural produce is likely to find its way to a large and widely dispersed population, via commercial food distribution. 

Cd intake via food was therefore estimated at the general population level by modelling transfers from soils into foodstuffs.  The first step in this sequence involves uptake by growing crops.  As many studies have indicated, the processes involved are complex and Cd concentrations vary substantially between different crops, management regimes and soils – so much so that Olsson et al. (2005) conclude that “There is no known means at present of predicting Cd uptake by crops in a specific field with any degree of accuracy”.  Higher rates of uptake, for example, tend to be seen for vegetables and salad crops (e.g. spinach, carrots, lettuce, radishes), with lower rates for cereals and grasses (Davis 1984, Olsson et al. 2005, (Mueller and Anke 1994, van Driel and Smilde 1990, van Lune and Zwart 1997).  Cd mobility in the soil is also strongly influenced by cation exchange capacity, so uptake rates tend to be reduced with higher pH, organic matter content and clay content (Eriksson 1988, Gray et al. 1999a, 1999b, Guttormsen et al. 1995, He and Singh 1994).  Liming may likewise reduce Cd uptake (Eriksson 1988, Hatch et al. 1988, McBride 2002, Li and Shuman1996).  In addition, factors such as fertiliser practice, harvesting regime and crop rotations may affect uptake rates (Davis 1984). 

Although a general relationship is commonly observed between Cd concentrations in the soil and in crops, therefore, the slope, strength and shape of the association seems to vary to a considerable degree.   While some studies have suggested a linear relationship (Jackson and Alloway 1991, Guttormsen et al. 1995, He and Singh 1994, Hooda et al. 1997), others have reported non-linear relationships, characterised by logarithmic or power models (Davis 1984, Dudka et al. 1998, Garrett et al. 1998, Gerritse et al. 1983, Hooda et al. 1997, Krauss et al. 2002).   These differences may reflect a number of factors, including methodological variables such as the sampling and extraction medium, as well as soil or management conditions.   Some of the linear associations seen in some studies may be a reflection of the limited range of soil Cd concentration studied; inspection of some of the reported data also suggests that non-linear models might fit the data at least as well.  Overall, a convex association seems most plausible, on the basis that plant uptake of Cd will be inhibited to some degree at higher concentrations.  Krauss et al. (2002) thus suggest that Freundlich-type equations, of the form Y = a Ln(X) – ln(b), provide a general characterisation of the associations. 

For the analysis here, data on agricultural land use in the study area were obtained from the June Agricultural Returns, which report crop areas and livestock numbers for each farm in the UK (aggregated to ward level).  Crops were classified into seven broad groups: potatoes, other root crops, green vegetables, wheat, barley, other cereals and grass and mapped at ward level.  Average annual production rates (tonnes/ha) were estimated for each group based on national statistics (DEFRA 2004, 2006) and these used to determine the total annual biomass production for each crop group within each ward; for grass yields recent data were not available, so estimates were made by the authors based on a range of study data.  Uptake functions for each of the crop groups were then derived from the literature (Table 2).  All functions used relate to total soil Cd. 

The soil Cd and pH concentrations were then interpolated using inverse distance weighting, in ArcGIS, and averages calculated for each ward.  Measured data on clay content (required for the assessment of uptake by wheat, root vegetables and green vegetables) were not available, but inspection of soil survey records for the study area suggest that most of the soils are loams or clay-loams, in which the clay content is likely to range between about 15 and 30%.  An average of 20% was imputed here.

Total Cd uptake by each crop in each ward was then assessed as follows:

 where: Cd_crop is the total Cd uptake (kg in edible dry matter), is the average Cd concentration in the crop (ug/g edible dry biomass) calculated from the relevant equation in Table 2, A is the planted area of the crop (ha), Y is the average crop yield (tonnes/ha), and i and j define the crop group and ward, respectively.

Table 2.  Crop uptake functions for major crop groups

Notes: a – losses in livestock feeding; b – field losses during grazing; c – function for barley reported by Dudka et al. applied to other cereals