Health-related environmental indices and environmental ... - CiteSeerX

Environment and Planning A 2004, volume 36, pages 803 ^ 822

DOI:10.1068/a3691

Health-related environmental indices and environmental equity in England and Wales Benedict W Wheeler ô

Department of Social Medicine, University of Bristol, Canynge Hall, Whiteladies Road, Bristol BS8 2PR, England Received 28 March 2003; in revised form 11 August 2003

Abstract. This study developed small-area health-related environmental indices for England and Wales in the context of an investigation into socioeconomic inequity in the distribution of environmental risk. Selection of environmental hazards, relevant datasets, and their attribution to standard smallarea geography using a geographic information system are described. Four indices for 1991 Census wards are proposed, relating to ambient air quality, atmospheric chemical releases from large-scale industrial processes, landfills, and sites registered under Control of Major Accident Hazard regulations. Ecological measures of association between these indices and the Carstairs material deprivation index and its components are presented in the context of variations by urban ^ rural status. Based on these analyses, the study generally supports previous findings of environmental inequity in England and Wales, but highlights that associations are dependent on the environmental and deprivation measures under consideration, and urban ^ rural context. It is proposed that environmental indices such as those described here should be included in considerations of area deprivation, could assist with equitable environmental decisionmaking and planning, and that measures of environmental inequity could be considered as indicators of progress towards sustainable development.

Introduction Calls for environmental justice developed in the context of the US civil rights movement in the early 1980s (Bullard, 1994). The environmental justice movement contended that one of the many injustices faced by minority US communities was disproportionate exposure to environmental risk, particularly with regard to the disposal of hazardous wastes. The terms ènvironmental equity' and ènvironmental justice' have since been used interchangeably to describe the fair and proportionate distribution of environmental risk across society. However, different meanings have also been attributed to the terms, with èquity' implying equal distribution of risk, and `justice' involving action to correct prior injustices, including overall reduction of environmental exposures (Brainard et al, 2002). According to the US Environmental Protection Agency, `Ènvironmental justice is achieved when everyone, regardless of race, culture, or income, enjoys the same degree of protection from environmental and health hazards and equal access to the decision-making process to have a healthy environment in which to live, learn, and work'' (USEPA, 2003). Much research over the past twenty years in the USA has contributed to the environmental equity debate, and studies have frequently found evidence of inequity (Graham et al, 1999; Neumann et al, 1998; Perlin et al, 1999; UCC Commission for Racial Justice, 1987). The US government has responded to this and, in 1994, President Clinton issued an Executive Order stating that `èach Federal agency shall make achieving environmental justice part of its mission'' (Clinton, 1994, paragraph 1-101). The environmental equity agenda has arisen more recently in the United Kingdom (see below), in Canada (Jerrett et al, 2001), and in New Zealand (Salmond et al, 1999). ô Current address: Health Intelligence Applications Laboratory, School of Earth Sciences, Victoria University of Wellington, PO Box 600, Wellington, New Zealand. e-mail: [email protected].

804

B W Wheeler

Some research disputes findings of inequity (Anderton et al, 1994), and there is debate on the validity of analytical methodologies (Anderton, 1996; Bowen, 2002; Bullard, 1996; Mohai, 1996; Yandle and Burton, 1996). Key criticisms include the effects of areal unit selection on quantitative results, lack of consideration of actual impacts on exposed populations, poor environmental data quality, and the need for assessment of causal direction. In a recent review of environmental justice research, Bowen (2002) concluded that ``the empirical foundations of environmental justice are so underdeveloped that little can be said with scientific authority regarding the existence of geographical patterns of disproportionate distributions'' (page 3). The focus of environmental inequity concern in the United Kingdom is largely deprivation or socioeconomic status, rather than racial injustices as in the USA (Mitchell and Dorling, 2003). Results of UK research have been mixed, but generally supportive of hypotheses that people living in deprived communities are more likely to be subjected to environmental exposures that may be detrimental to health (Brainard et al, 2002; Friends of the Earth, 2001; McLaren et al, 1999; McLeod et al, 1998; Mitchell and Dorling, 2003; Pye et al, 2001). There is also recognition from government and regulators that the issue has relevance in the United Kingdom. For example, in the ministerial foreword to the 1999 review of the National Air Quality Standards, John Prescott stated `Àir pollution ... has shaped our cities. Even when our cities were being formed, those who could afford to paid to be upwind of the stink and smoke, while the poor were left to suffer the penalties of living downwind'' (Prescott, 1999). The Environment Agency held its 2000 AGM debate on `Àchieving Environmental Equality''. In the context of that debate, Sir John Harman, Chairman of the Agency, said ``surely one factor in social exclusion is a poor environment. It is clear that the poorest people in our society live with some of the most pressing environmental problems. Deprived inner-city communities are particularly likely to suffer poor air quality or land contamination. Just as with health inequalities, more needs to be done to remove these disbenefits'' (Harman, 2000). A recent report from the government's Sustainable Development Commission (SDC, 2002) suggests that recognition of the association between environment and poverty is a key element of the sustainable regeneration of communities affected by ``social, economic and environmental problems of deprivation'' (page 1). Despite the uncertainties of environmental equity research, ascertainment of the spatial and social distribution of potential environmental health risk is a useful approach toward empirical assessment of environmental inequity. The use of census and other sociodemographic data to create indices giving an indication of socioeconomic circumstances within small geographic areas is well established (Carstairs and Morris, 1989; DETR, 2000a; Townsend et al, 1988). Similarly, environmental indices or indicators can be used to simplify complex data to produce summary measures with pertinence to a particular issue (Ott, 1978). Briggs (2000) suggests that environmental health indicators may be designed ``to detect temporal trends or spatial patterns, as simple or composite indicators, at the local, national or international scale [and/or] for the purpose of policy/management, epidemiological research or awareness raising'' (page 58). A number of environmental indicators have been proposed, such as UK government sustainable development indices (DEFRA, 2002a), composite air pollution indices for public information in the United Kingdom (DEFRA, 2002b) and USA (USEPA, 2000), and more general `state of the environment' indicators (Hope et al, 1992; Odemerho and Chokor, 1991; Young, 1997). However, these are not designed to assess spatial variation in environmental health risk. Only two studies that explicitly

Health-related environmental indices

805

set out to construct small-area, health-related indices were discovered, both in the Netherlands. Sol et al (1995) describe the construction of an environmental index intended for land-use zoning purposes. Environmental hazards were identified from the Dutch National Environmental Policy Plan and classified according to annoyance, toxicity, or mortality. Index values were constructed using expert judgment and relating substance concentrations to their NOAELs (no observable adverse effect levels). A single composite index value was calculated, in order to indicate the most acceptable locations for the development of different land uses. An alternative approach by Pruppers et al (1998) constructed maps for high-resolution grids of the Netherlands for a variety of environmental exposures, using a comparable scale of `risk'. The authors considered carcinogens, accidents, and radiation exposure in terms of mortality risk, and noise in terms of probability of annoyance. They then created maps using the same colours to represent the same ranges of risk, allowing comparability between the different exposure sources. The attribution of environmental data to common areal units, such as census areas, is frequently used in environmental equity and epidemiological studies. For example, in a recent paper examining the equity of the distribution of air pollution across Great Britain, Mitchell and Dorling (2003) allocated nitrogen dioxide (NO2) concentrations to wards by relating ward centroids to proximal NO2 concentration estimates. Similarly, a Friends of the Earth study allocated factory locations to postcode sectors (McLaren et al, 1999). However, it is the intention of this study to create an ènvironmental indices' dataset for small areas, and to construct the indices with particular regard to human health impacts and environmental equity assessment. Environmental index construction The intent of the proposed environmental indices is to provide an indication of relative levels of potential environmental health risk for small areas across England and Wales, and the four key stages of index development are outlined below. The reasons for this focus are related to the ways in which various datasets are collected and made available. For example, some environmental data used for this study are collected and produced by the Environment Agency for England and Wales. Equivalent data for the rest of the United Kingdom are collected and produced in different forms by the relevant agencies for Scotland and Northern Ireland. Much of the pertinent legislation and regulations are also constructed separately for Scotland and Northern Ireland. Selection of pertinent environmental hazards

The selection of indicators should be determined primarily by relevance to health outcomes and policy priorities, but is also influenced by data availability. Because selection is bound to be subjective, a number of information sources were consulted to establish what types of indicator data might be included. Three key sources were consulted in order to provide a list of policy-relevant environmental public health issues. These were the National Environmental Health Action Plan (DETR, 1998), the 1997 annual report of the Chief Medical Officer (CMO, 1998), and the UK government's indicators of sustainable development (DEFRA, 2002c; 2002d). In order to be of relevance, it is proposed that the environmental indicator should be widespread across England and Wales, amenable to policy-level intervention, and of relevance to public health policy. Additionally, it is suggested that the environmental factors of interest are those that are anthropogenic. Although naturally occurring environmental phenomena can have considerable health effects, the equity implications of human-sourced environmental exposures are quite different. The equity question here is whether the external costs of human activities are borne in an equitable manner

806

B W Wheeler

across society. The major factors that fit these criteria, based on specific mention in the sources above, are: 1. outdoor ambient air quality (largely related to road traffic emissions); 2. indoor air quality; 3. routine industrial emissions, for example, from chemical plant and incinerators; 4. waste landfill sites and contaminated land; 5. major chemical or nuclear accident hazards; 6. noise; 7. water supply quality; 8. food (contamination, not nutritional value); 9. electromagnetic fields (EMF); 10. road traffic collisions. The public health impact of some of these factors is contentious, but whether perceived or actual, the potential risks indicate that they are still of relevance to public health and environmental policy. Two prominent environmental risks that feature in policy on environment and health, but are excluded here, are radon and ultraviolet radiation exposure. These are naturally occurring phenomena, and they therefore lie outside the intended purpose of these indices. There are, of course, numerous other environmental factors beyond this simplified list that could potentially have an impact on public health, from green spaces for urban populations to access to sufficient food and warm homes. These wider issues have been raised as having important environmental equity implications (Stephens et al, 2001), extending the research question beyond the usual consideration of air quality and industrial facilities. However, the intent of this study is to create indices with direct relevance to current environmental health policy, and to initiate a set of indices that could perhaps be extended to encompass some of these broader environmental issues in the future. Selection of hazards and data for indices

It is proposed that each indicator to be included should: have relevance to environmental health policy; be supported by epidemiological or toxicological evidence for human health effects; have readily available data for England and Wales with geographic referencing at small-area level, without significant cost; and be based on data that will be updated. Table 1 describes the national datasets available for index construction, including comments with regard to these criteria. A number of the potential indicators have no appropriate national database available. These include contaminated land, noise, and EMF exposures. Additionally, some of the environmental health hazards do not lend themselves to consideration of simple spatial variation: indoor air quality (dependent on internal and external sources along with building characteristics), drinking water quality (dependent on supply location and treatment or consumption characteristics), and food contamination (dependent on sources, purchasing, and consumption behaviour). One environmental risk of substantial public health importance is that of injuries and mortality due to road traffic collisions (RTCs). These outcomes are inherently caused by an environmental risk (exposure to road traffic), and studies have demonstrated socioeconomic inequalities in RTC risk and related health outcomes (Laflamme and Diderichsen, 2000; Plasencia and Borrell, 2001). The link between RTCs and RTC-related injuries and deaths is explicit, and proximity to a road at the time of the accident is fundamental to the risk. The best indicator for RTC health risk is the rate or number of RTC-related injuries or deaths occurring in an area or on a given stretch of road. This is an ènvironment-related health indicator', rather than a `health-related environmental indicator' (Wills and Briggs, 1995). RTCs are therefore


807

Table 1. Environmental datasets and assessment against selection criteria for environmental health indices (policy relevance; evidence of health effects; data availability; data will be updated). Data (source)

Relevant hazard

Dataset description

Criteria comments

National Air Quality Archive (AEA Technology on behalf of Department for the Environment, Food and Rural Affairs)

Ambient air quality

1 km grid cell annual ambient pollutant concentration estimates, based on monitoring and emissions modelling

Fully meets all four criteria.

Pollution Inventory (Environment Agency)

Chemical emissions from large industrial processes

Location and mass of specified chemicals released

Fully meets all four criteria (potential health impact depends on the nature of emissions and exposure pathways).

Landfill sites (Environment Agency registers)

Landfill sites

Location and type of landfill site

Meets all criteria, although limited evidence for public health impact (Vrijheid, 2000).

Sites registered under Control of Major Accident Hazards regulations (Health and Safety Executive)

Major accident hazards

Location and type of registered site

Meets all criteria except evidence for health effectsÐ with the exception of acute exposure due to accidents, there is little evidence for long-term physical effects of residence near to an accident hazard, although possibility of psychosocial effects (Luginaah et al, 2002; Moffatt et al, 1995).

not included here because they are very different to the other environmental risks under consideration. Two datasets identified in table 1 include information on specific substances: ambient air quality (AAQ), and emissions from industrial processes licensed by the Environment Agency (under Part A of the Integrated Pollution Control regulations, Environmental Protection Act 1990). Emissions from `Part B' processes (smaller industrial processes licensed by local authorities) were also considered, but excluded, because resources were not available to collect and collate data held by local authorities. Specific substances featuring in the AAQ and Pollution Inventory (PI) data were selected according to the World Health Organisation's (WHO's) air quality guidelines for Europe (WHO, 1996). These guidelines were published for the substances listed in table 2 (see over), which were selected by WHO on the basis of potential widespread exposure and public health impact. A UK-specific list of priority substances with respect to environmental health has not been defined. The WHO Europe list was therefore used as a rational basis for selection. Emissions to air only were selected from PI processes, because these would be likely to have impacts on the immediate locality of the facility location. Emissions to water and land (including off-site disposal) also feature in the PI data; these were excluded because they are not necessarily indicative of local environmental quality.

808

B W Wheeler

Table 2. Substances considered in the WHO Air Quality Guidelines for Europe and respective data sources used in this study. Substance

Data sourcea

Substance

Data sourcea

Carbon monoxide (CO) Ozone Nitrogen dioxide (NO2) Sulphur dioxide (SO2) Particulate matter (PM10) Benzene 1, 3-Butadiene Dichloromethane Formaldehyde Polyaromatic hydrocarbons (PAH) Polychlorinated biphenyls (PCBs) Polychlorinated dibenzo-dioxins (PCDDs) Polychlorinated dibenzo-furans (PCDFs)

AAQ na AAQ AAQ AAQ AAQ na PI PI PI na na na

Styrene Tetrachloroethylene Toluene Trichloroethylene Arsenic Cadmium Chromium (Cr VI) Fluoride Lead Manganese Mercury Nickel Platinum

PI PI PI PI PI PI PI na PI PI PI PI na

a PI:

Pollution Inventory; AAQ: ambient air quality; na: not included in studyÐsee appendix A.

Those substances available in the AAQ data are largely related to road traffic, and are all considered in the National Air Quality Strategy (DETR, 2000b). Other substances are largely associated with industrial emissions, and were considered using data from the PI. As table 2 indicates, several substances that feature in the WHO Guidelines have been excluded from consideration; reasons for exclusion are given in appendix A. Landfill and Control of Major Accident Hazard (COMAH) site data do not detail specific substances or environmental releases. The licensing data simply contain information on location and regulatory details such as classification of activities at landfill sites. These potential environmental hazards are therefore considered in terms of presence or absence in the locality, as described below. Attribution to small-area geography

Once the appropriate indicators had been selected, they were attributed to small-area boundaries. Selection of the boundary system to be used is significant, because the modifiable areal unit problem may influence results of subsequent analyses, and this has been highlighted as a pertinent issue in environmental equity analyses (Anderton, 1996; Cutter et al, 1996). Because selection is bound to be subjective, the following describes the rationale by which a boundary system was chosen. The indices are intended to be appropriate for analyses with common socioeconomic and health datasets, and also pertinent for planners and other decisionmakers in assessing needs for environmental remediation or protection. They therefore need to be constructed for common, small geographic units allowing analyses with other datasets and contiguity with administrative boundaries. 1991 Census wards, of which there are 9527 (excluding shipping wards) in England and Wales, have a mean area of around 16 km2 (range 0.02 to 448 km2 ) and mean 1991 resident population of 5240 (77 to 31 609). Wards were selected because, first, they are small enough to be sensitive to small-area variation in environmental conditions ö larger areas (for example, local authority districts) may have high internal heterogeneity in terms of both socioeconomic and environmental characteristics. Second, they are appropriately sized for the resolution of the environmental dataösmaller areas such as enumeration districts (EDs) would be inappropriately precise. Third, many health and socioeconomic datasets are available for wards, such as census, mortality, and hospital-episode data. Fourth, wards have commonly been


809

used as the spatial basis for health inequalities and other socioeconomic research, and this work maintains consistency. Digital ward boundary files were obtained (see acknowledgements), and ArcView and Arc/Info GIS software packages (ESRI Inc., Redlands, CA) were used to clean and analyse the boundary and environmental datasets. AAQ datasets consisting of 1996 annual mean concentration estimates for the relevant substances for 1 km grid cells were obtained from the National Environmental Technology Centre (NETCEN), who produce the data on behalf of the government. Each grid dataset was overlaid with the boundary polygons in the geographic information system, and for each ward polygon, a simple mean of the intersecting cells' pollutant concentration values was calculated. A public-access version of the PI database (1 November 2000 version) was obtained from the Environment Agency. Data were extracted for the year 1999. Data for 1991 ^ 98 also appear in the database. However, from 1991 to 1997, the data relate to the Chemical Release Inventory, the precursor of the PI, and the two inventory systems are not compatible. As 1998 was the first year of collection for the PI, the Environment Agency also suggest that data for subsequent years may be more reliable (Environment Agency, 2002). The data include grid coordinates for each site location, and these were used to map all facilities running licensed processes in 1999. Eight of the 1279 sites had missing grid references; four were located using their postcode and a lookup table based on the Central Postcode Directory. No locational information was available for the other four sites, and they could not be georeferenced, although subsequent analyses suggested that any error introduced here would be negligible. The omission of these four sites is in addition to those emissions that have been excluded from public versions of the data, such as those where commercial confidentiality is claimed by the operator. The data are therefore not perfect, but do include the majority of relevant chemical releases. In an ideal situation, plume-dispersal modelling would have been undertaken to estimate the likely spatial extent of releases of each particular substance from each site. However, this modelling process requires a large quantity of site-specific and substance-specific information, such as stack heights and temperatures, prevailing weather conditions, chemical release rates, and so on. The substantial resources required to carry out this complex modelling for many substances for all sites across England and Wales were not available. An alternative method would have been to use a simple `point-in-polygon' approach, attributing PI release data to the ward polygon containing the point representing the location of the industrial facility. However, this process has the potential to introduce artificial boundary effects. First, the point locations of the sites are accurate only to approximately 100 m. This means that a site lying very near to a ward boundary may be calculated to fall within the adjacent ward to that in which it truly lies. Second, the procedure assumes that any potential effects of the site on the local population are distributed only within the ward in which the point location is found. This is unlikely as, for example, if a site lies near the edge of one ward, its presence should be attributed both to the ward in which it lies and to the nearby contiguous ward. Third, the representation of these sites as points is somewhat spurious, as some cover very large areas. To negate boundary effects buffer polygons (circles) of 1 km radius were constructed around each site and sites attributed to intersecting wards by proportional area. For example, if a ward intersected 50% of the area of a PI site buffer, it would be attributed with 0.5 of a PI site. The choice of buffer distance is necessarily arbitrary, but 1 km was considered to be reasonable to decrease artificial boundary effects without overestimating the sphere of influence.

810

B W Wheeler

The same buffering and proportional area overlay method was used for the point locations representing landfill and COMAH sites. When the research was undertaken, the landfills data were publicly available only from locally held Environment Agency licence registers. This information was, however, collated to a national database by Landmark Information Group, who supplied the data for this research (these data are now freely available from the Environment Agency). Information was extracted from the September 2000 database on all operational licensed landfill sites. Data included grid references, used to map the landfill locations. The Health and Safety Executive maintains a database of sites registered under the COMAH regulations, and supplied a public-access database of sites registered at June 2000. Site postcodes were used to geocode the database as for the PI sites with missing grid references. A large number of sites in the COMAH database had postcodes that were missing or had no match in the geocoding database. In order to map these sites, a selection of Internet-based mapping and aerial photography sites were used to attribute a location to each site (http://www.multimap.com, http://www.royalmail.co.uk, and http://www.yell.co.uk). It is worth noting that there is some overlap between the COMAH-registered sites and those in the PI. Ideally, sites overlapping the two datasets would be identified, and analyses could consider this. However, the data are collected by separate government agencies, in different ways, and for different purposes. This makes identification of duplicates difficult, and detailed investigations of the data revealed that company names, addresses, and locations are often recorded in different ways in the two databases. The only way that this could have been rectified would have been manual trawling of the datasets to attempt to flag duplicates. Even with the possibility of using locational information to focus the comparison of the two datasets, it was decided that the gain from this process would not justify the resources needed. The impact of this overlap is that patterns of the PI and COMAH indices and their associations with other variables may be somewhat similar. Calculation of index values

By following the above steps, each ward polygon was attributed with values for each of the indicators: annual mean ambient concentrations of carbon monoxide (CO), nitrogen dioxide (NO2), sulphur dioxide (SO2 ), particulate matter (PM10), and benzene; and counts of PI, landfill, and COMAH sites. The landfill and COMAH values remain simply as counts of sites, as originally intended for these indices. Methods were then developed to create single-figure indices for the AAQ and PI data (detailed in appendix B). Environmental equity analysis Once constructed, the environmental indices were used to investigate the extent of environmental inequity in England and Wales. Simple measures of association were calculated between the environmental indices and the Carstairs index (Carstairs and Morris, 1989), a census-based index of material deprivation commonly used in health inequalities research. Further analyses were conducted using regression models to investigate the nature of associations between the environmental indices and component variables of the Carstairs index, in order to assess any differential associations that may be masked by the composite measure. Analyses were stratified using an indicator of urban ^ rural status in order to investigate possible differences in environmental equity associations, especially with regard to the AAQ index. This possibility arises for two reasons. First, the Carstairs index may not be an adequate representation of deprivation in rural areas, and the nature of deprivation may differ according to the degree of urbanisation (Haynes and Gale, 2000; Martin et al, 2000). Second, urbanisation is likely to be associated with


811

higher levels of the AAQ index, because it is largely based on road-traffic-related pollutants, and also with greater deprivation in terms of components of the Carstairs index, particularly low car ownership and overcrowded housing. An urban ^ rural indicator was derived from an Office for National Statistics indicator from the 1991 Census based on the proportion of EDs in a ward whose population was considered to be within an urban area. Wards were classified as wholly urban (100% of EDs in ward classified urban), predominantly urban (75 ^ 99%), mixed urban ^ rural (1 ED to 74%), or wholly rural (no urban EDs). Findings Environmental indices

The AAQ index is approximately normally distributed with a mean of 1.5 and maximum of just under 4. Of 1279 PI process sites in 1999, 676 were associated with an atmospheric release of at least one of the selected substances in table 2. With the 1 km buffers, 1550 wards had a nonzero PI index. Similarly, 1273 landfills resulted in 2265 wards with a nonzero landfills index, and 1139 COMAH site buffers intersected 2429 wards. These three site-based indices exhibit highly positively skewed distributions across wards, with the majority of wards scoring zero, and most index scores at the low end of the range. Table 3 details descriptive statistics for all four environmental indices. Table 3. Descriptive statistics for the environmental indices across 9527 wards of England and Wales. Environmental index

Total sites

Wards with index > 0 (1 km buffer)

Minimum

Maximum

Mean

Median Standard deviation

Ambient air quality a Pollution Inventory Landfills COMAH sites c

na 676 1273 1139

na 1550 2265 2429

0.68 0 0 0

3.96 78.48 5.48 15.88

1.50 0.48 0.13 0.12

1.40 0 0 0

0.53 ÿb ÿ ÿ

a na:

Ambient air quality index not based on site locations. deviation inappropriate because of highly skewed distributions. c COMAHÐControl of Major Accident Hazards. b Standard

As illustrated in figure 1 (see over), the AAQ index demonstrates a strong geographical pattern, as might be expected, with the highest values in urban areas and close to major roads, and low values in remote and rural areas. The geographical patterns of the site-based indices are less obvious, given that most wards score zero for these indices. The distribution of each index across the four urban ^ rural categories is illustrated in table 4 (over). The AAQ index is highest in urban categories and lowest in rural categories, in concordance with figure 1. Given the highly skewed nature of the site-based index values, binary indicator variables were used to indicate increased risk. These allow simple calculation of the proportion of wards within, for example, 1 km of a COMAH site, or within 1 km of a release of one of the selected PI substances. The figures in table 4 for the PI, landfill, and COMAH indices therefore indicate the percentage of wards with a positive score on the relevant index. Most PI and COMAH sites are located in the more urban wards, in terms both of numbers of sites and of proportion of wards attributed with a nonzero score. In absolute terms, most landfills are in wards categorised as wholly urban or wholly rural.

812

B W Wheeler

AAQ index quintiles 0.7 ± 1.1 1.2 ± 1.3 1.4 ± 1.5 1.6 ± 1.8 1.9 ± 4.0

0

25

50

100 km

Figure 1. Ambient air quality (AAQ) index, constructed from 1996 annual mean concentrations of NO2, SO2, PM10, and benzene for wards of England and Wales. Table 4. Distribution of the four environmental indices across urban ^ rural categories. Urban ± rural indicator

Ambient air quality index mean

Pollution Inventory index

Landfill sites index

COMAH sites index

sitesa % of wards with index > 0



1 (Urban) 1.71 (1.70, 1.73) 435 2 1.33 (1.32, 1.35) 110 3 1.23 (1.21, 1.24) 64 4 (Rural) 1.11 (1.10, 1.12) 67

19.1 (18.0, 20.1) 16.8(14.6, 19.2) 15.4 (13.1, 17.9) 8.4 (7.3, 9.8)

932 373 371 589

16.8 (15.8, 17.8) 1676 30.2 (29.0, 31.5) 34.1 (31.3, 37.0) 272 24.8 (22.3, 27.5) 39.6 (36.5, 42.9) 197 21.0 (18.5, 23.8) 30.2 (28.1, 32.2) 284 14.5 (13.0, 16.2)

Figures in parentheses are 95% confidence intervals (exact binomial for site-based indices). a Site counts are based on 1 km buffers. Pollution Inventory site count includes only sites releasing one of the fourteen selected substances.


813

However, the intermediate rural wards are proportionately most likely to score positively on the landfills index. This seems reasonable, given that landfills are likely to be located at the periphery of urban areas, where there is sufficient space and proximity to waste production. Environmental equity

Table 5 reports summary index values by quintiles of Carstairs index and urban ^ rural indicator. These results suggest that in urban areas the association between deprivation and AAQ is indicative of environmental inequity, in that more deprived wards are subject to poorer air quality. However, in the more rural wards, the association is reversed, with poorer air quality in the less deprived areas. This is probably explained by the distribution of wards in the more rural categories, which include `commuter belt' wards on the periphery of urban centres as well as more remote rural wards. The commuter belt wards tend to be less deprived and have higher levels of ambient air pollution with reference to the more remote wards, giving rise to the figures in table 5 for the site-based indices which are also indicative of substantial environmental inequity. For example, 7.9% (95% confidence interval 6.0, 10.2) of wholly urban wards in the least deprived fifth of all wards are attributed with a nonzero PI index, while the Table 5. Summary environmental index values across quintiles of the Carstairs index, by urban ^ rural indicator. Carstairs quintile Wholly urban (n 5512)

Predominantly Urban ± rural urban (n 1095) mix (n 924)

Ambient Air Quality index [mean (95% 1 (least deprived) 1.53 (1.50, 1.55) 2 1.57 (1.54, 1.60) 3 1.57 (1.54, 1.60) 4 1.64 (1.62, 1.67) 5 (most deprived) 1.97 (1.94, 2.00) p (trend)