- Email: [email protected]
The analysis of human health risk with a detailed procedure operating in a GIS environment
Environment International 32 (2006) 444 – 454 www.elsevier.com/locate/envint
The analysis of human health risk with a detailed procedure operating in a GIS environment P. Morra a, S. Bagli b, G. Spadoni a,* a
Dipartimento di Ingegneria Chimica, Mineraria e delle Tecnologie Ambientali, Alma Mater Studiorum, Universita` di Bologna, viale Risorgimento n.2, 40136 Bologna, Italy b Gecosistema srl-piazza Malatesta 21, 47900 Rimini, Italy Received 10 June 2005; accepted 7 October 2005 Available online 13 December 2005
Abstract An approach for quantifying the human health risk caused by industrial sources, which, daily or accidentally, emit dangerous pollutants able to impact on different environmental media, is introduced. The approach is performed by the HHRA-GIS tool which employs an integrated, multimedia, multi-exposure pathways and multi-receptors risk assessment model able to manage all the steps of the analysis in a georeferenced structure. Upper-bound excess lifetime cancer risk and noncarcinogenic hazards are the risk measures, the spatial distribution of which is calculated and mapped on the involved territory, once all the pathways and receptors of the study area are identified. A sensitivity analysis completes the calculations allowing to understand how risk estimates are dependent on variability in the factors contributing to risk. The last part of the paper makes use of a case study concerning a working industrial site to put in evidence in which way the designed tool can help local authorities and policy makers in managing risks and planning remedial and reduction actions. The considered geographical area is a hypothetical territory characterized by residential, agricultural and industrial zones. The presence of two sources of contamination, a municipal waste incinerator (MWI) and a contaminated site, are evaluated by the tool application. Various typologies of receptors have been taken into account, each of them characterized by different anatomical and dietary properties. The achieved results are analyzed, compared with acceptable and background values and alternatives of minor environmental impact calculated. D 2005 Elsevier Ltd. All rights reserved. Keywords: Quantitative human health risk; Sensitivity analysis; Fate and transport models; Municipal waste incinerator; Geographical information systems
1. Introduction Due to industrial and traffic emissions, agricultural activities and waste disposal facilities, many hazardous materials are released into the environment thus generating a potential hazard to human health and environment. In order to perform environmental risk evaluation, a great attempt is generally assigned to determine risk on human health owing to exposure at carcinogenic or toxic substances, which may be present in air, soil, water, foods, as consequence of anthropic activities. Procedures for performing this type of assessments are regularly delineated by government agencies, such as U.S. EPA (United States Environmental Protection Agency), EEA (European Environment Agency) and by European Directives and national regulations; most of them are based on the * Corresponding author. Tel.: +39 051 2093146; fax: +39 051 581200. E-mail address: [email protected] (G. Spadoni). 0160-4120/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.envint.2005.10.003
National Academy of Sciences model introduced in 1983 (NAS, 1983). A complete Environmental and Human Health Risk Assessment (EHHRA) requires the integration of information on environmental and chemicals database, inter and intra media dispersion models output, site description maps and demographic database. As a result, the designing and conducting analysis of the effects of contaminants on human health is a complex and time expensive procedure, particularly when the aim is to estimate risk in a geographical area characterized by the presence of several sources, exposure pathways and receptor typologies. Therefore in the last years, the need for defining and using procedures and tools, able to quantify risk for human beings, has increased especially for policy makers, local planners and, in general, for various stakeholders involved (Kuchuk et al., 1998). These ones are interested to manage and plan remedial actions in order to reduce environmental pollution by choosing options which minimize health risk.
P. Morra et al. / Environment International 32 (2006) 444 – 454
In a previous publication, a new Gis approach for quantifying the human health risk through a system able to manage all the steps in a georeferenced structure was introduced (Bagli et al., 2004). This paper presents the complete innovative methodology and explains the modified and improved tool, which is the result of a long timeconsuming work. The designed support system has been developed to help an expert user interested in assessing and managing risk for human beings living in a territory where industrial sources are present, which, daily or accidentally, emit dangerous pollutants able to impact on different environmental media. In order to estimate potential human health risk, the proposed tool, HHRA-GIS, employs an integrated, multimedia, multi-exposure pathways and multi-receptors risk model. Preserving the previous configuration, the system has been improved by adding new models and a sensitivity analysis module. This section allows to understand how risk estimates are dependent on variability in the factors contributing to risk. By assigning a probability distribution function to input variables and performing a Monte Carlo simulation, the sensitivity analysis provides a ranking of the model inputs based on their relative contributions. In the last part of the paper a case study concerning an hypothetical working industrial site is examined, to put in evidence in which way the designed tool can help local authorities, policy makers and stakeholders in managing risks and planning remedial and reduction actions. Since incineration is lately matter of great argument and debate, the presence of a municipal waste incinerator (MWI) as a source of contamination is evaluated by the tool application. In addition in the area of interest it has been assumed the presence of a contaminated site, delineated by soil sampling data. The presence of various typologies of receptors has been taken into account, each of them characterized by different anatomical and dietary properties. The achieved results are finally analyzed and compared with acceptability and background values. 2. Human health risk assessment The aim of human health risk analysis is the calculation of the upper-bound excess lifetime cancer risk and noncarcinogenic hazards. The procedure usually adopted in performing this evaluation, as assumed in the developed Gis tool, is based on the following paradigm: 1. Identification of Chemicals of Concerns (COC) 2. COC Fate and Transport Assessment 3. Exposure Scenario Definition and Exposure Intake Assessment 4. Dose– Response Assessment 5. Risk Characterization. The first step is the identification of substances which cause adverse effects and involves consultation of any toxicological and epidemiological data. In the second phase the distribution of concentrations of contaminants in various environmental media is analyzed. The Exposure assessment of the third step
445
requires the determination of the emissions, pathways and rates of movement of a substance and its transformation and degradation with the aim to estimate the ‘‘concentration/doses to which human populations (i.e. workers, consumers and men exposed indirectly via the environment) or environmental compartments (aquatic environment, terrestrial environment and air) are or may be exposed’’ (CEC, 1993). Dose –response assessment is obtained from epidemiological and toxicological data. Usually this is a tiered process that progresses from the use of short-term (acute) tests and conservative assumptions to longer-term (chronic) tests paired with more realistic assumptions. The risk characterization, which concludes the procedure, is essentially a summary of the data compiled in the risk assessment process including the uncertainties associated with each stage and the presentation of a risk estimate. Risk estimation consists in the computation of the upperbound excess lifetime cancer risk (risk) and noncarcinogenic hazards (hazard) for each of the pathways and receptors identified in the area of interest. Risk is defined as the probability that a receptor will develop cancer in his lifetime, assuming a unique set of exposure, model, and toxicity properties. In contrast, hazard is quantified as the potential for developing noncarcinogenic health effects as a result of exposure to COCs, averaged over an exposure period. It is worth noting that hazard is not a probability but, more exactly, a measure of the magnitude of a receptor’s potential exposure relative to a standard exposure level. The individual cancer risk of a receptor j set by exposure to multiple carcinogenic chemicals i, is calculated through the following equation: X Individual CancerRiskj ¼ LADDij CSFi i
where: LADDij Lifetime Average Daily Dose for a lifetime exposure of 70 years (mg/kg day) through multiple exposure pathways CSFi Cancer slope factor for COC i (mg/kg day) 1. Comparing an exposure estimate to a Reference Dose (RfD), the potential for noncarcinogenic health effects resulting from exposure to a chemical is evaluated. A RfD is defined as a daily intake rate that is estimated to cause no appreciable risk of adverse health effects, even to sensitive populations, over a specific exposure duration (USEPA, 1989). Generally, the more the Hazard Quotient value exceeds 1, the greater is the level of concern. In spite of this, because RfDs do not have equal accuracy or precision, and are not based on the same severity of effect, the level of concern does not increase linearly as the quotient approaches and exceeds 1. Based on similar COCs toxicological characteristics and additive health effects, the Hazard Quotient (HQ) for receptor j exposed to multiple chemicals i, is calculated as:
HQj ¼
X ADDij i
RfDi
446
P. Morra et al. / Environment International 32 (2006) 444 – 454
where: ADDij Average Daily Dose averaged for the exposure duration relative to the toxic i for the receptor j (mg/kg day) through multiple exposure pathways RfDi COC i Reference Dose (mg/kg day). Lifetime Average Daily Dose and Average Daily Dose represent the amount a person takes in as a result of exposure to a chemical in contaminated air, water, soil or food. The combinations of the potential pathways by which individuals can be exposed represent the exposure scenario for an individual. Human receptors may be exposed to COCs via multiple primary exposure routes, either directly or indirectly. The estimation of the exposure magnitude consists in the quantification of the potential dose that is of the quantity of agent that enters in contact with the human organism through the points of exchange with the environment (lungs, skin, etc.) in a specific interval time. The calculation of COC-specific exposure rates for each exposure pathway k to be considered requires the estimation of chemicals media concentrations, the consumption rate, the receptor body weight, the frequency and duration of exposure. The generic equation used to calculate chemical intake (LADDij for carcinogenic chemicals, ADDij for toxic chemicals) is: X Ci;em CRj;k EFj;k EDj;k Ii;j ð x; y; zÞ ¼ BWj ATj;k k where: I i,j (x, y, z) Intake for COC i and class of receptors j (mg/kg day) C i,em COC i concentration in exposure media (e.g. mg/kg for soil) CRj,k Consumption rate, the amount of contaminated medium consumed through the k exposure pathway per unit of time or event (e.g. kg/day for soil) EFj,k Exposure frequency (days/year) EDj,k Exposure duration (years), total time period over which contacts occur between receptor and COC BWj Average body weight of the receptor j over the exposure period (kg) ATj,k Averaging time, the period over which exposure is averaged (days); for carcinogens, the averaging time is 25,550 days, based on a lifetime exposure of 70 years; for toxics the averaging time is exposure duration expressed in days. In order to calculate chemicals concentrations in various exposure media, procedures extracted from USEPA (1998) are usually adopted, which allow the estimation of the concentration of a chemical in an exposure medium in terms of the chemical concentration in the environmental compartment, by means of an inter-media transfer factor: C ð x; y; zÞexposure ¼ f TF; C ð x; y; zÞenviromedia where: C(x, y, z)exposure Concentration in exposure media C(x, y, z)enviromedia Concentration in environmental media
TF
Inter-media transfer factor from an environmental compartment to an exposure medium.
Inter-media transfer factor is usually expressed by an appropriate chemical-specific partition coefficient, which describes the physicochemical attraction of the contaminant for the environment and the exposure medium. Concentration distributions in environmental media result from the application of contaminants fate and transport models or from measured data. 3. HHRA-GIS: a GIS approach to human health risk assessment The HHRA-GIS tool is an integrated approach for assessing the risks to human health; it is able to manage all the steps of the above described methodology in a georeferenced framework operating in a GIS environment, in order to display maps of iso-dose and iso-risk contours. In the numerical code the modules of the procedure are integrated in a system with a graphical user interface through which specific territorial pattern are represented in great detail. The tool, whose initial version has been improved and enriched with new modules, evaluates cancer risks and hazard quotients in a geographical area characterized by multimedia contamination, potential multi-exposure pathways and several receptor typologies. The system takes advantage of two typologies of database: the Gis Spatial Database, and the Relational Database in Microsoft Access linked to the ArcView Gis project using an Open Database Connectivity. Fig. 1 shows a block diagram that illustrates the architecture of the methodology to assess human health risk using HHRA-GIS. The digital site and facility description is an input of Gis Spatial DB; it consists in identifying and mapping contaminant sources, collecting facility and site information (e.g. the localization of rivers or lakes, soil properties, groundwater level, terrain elevation), estimating emission rates, and identifying Chemicals of Concern. For each chemical of concern identified in the contaminated area, estimates of the concentrations must be defined for each environmental medium of interest (e.g. soil, ground water, surface water, sediment, air, vegetation). User may derive concentrations from two broad methods: direct monitoring (e.g. sampling and chemical analysis of media at the site coupled with summary statistics) or environmental modelling (e.g. mathematical modelling to predict contaminant concentrations in various media). In both cases the output of the analysis is a series of maps describing the distribution of the toxic and carcinogenic chemicals concentrations in the environmental contaminated media. Chemicals environmental fate and transport models are implemented in the Gis tool so that the outputs of the models are imported into Gis and displayed as maps of distribution of concentration. The system graphical user interface allows the interaction with the different stages of the simulation, that are the data input (pre-processing), the run of the model and the
P. Morra et al. / Environment International 32 (2006) 444 – 454
GIS SPATIAL DB
ACCESS RELATIONAL DB
Contamination monitored data
Site characterization
Definition of emission sources
447
Fate and transport models
Definition of receptor characteristics
Distribution of COCs concentrations
Receptor typologies localization
Cancer Risk and Hazard Quotient calculation
Definition of calculation grid
Selection of exposure pathways
Chemical, phisical and toxicological database
Distribution of Cancer Risk and Hazard Quotient
Probability distribution of Cancer Risk and Hazard Quotient
Monte Carlo Analysis
Probability distribution functions
MONTE CARLO SOFTWARE Fig. 1. Block diagram of the methodology in HHRA-GIS. Yellow block: input data; light blue block: calculation module; orange block: output data.
output display (post-processing), taking advantage of Gis spatial analysis tools. The architecture of the designed system allows the integration of updated modelling components, or their replacement with models that match the specific features of the assessment. In the updated version of the system the following models are implemented: ISC3-GIS and AERMODGIS for the atmospheric dispersion of pollutants emitted by industrial sources; CALINE-GIS for the atmospheric dispersion of pollutants emitted by vehicles on roadways; HSSM-GIS and GMS for the transport of pollutants in soil and in groundwater; MULTIMEDIA models to estimate the partitioning of COCs in various environmental compartments; FOODCHAIN models to evaluate bioaccumulation in vegetables (belowground and above-ground), in food of animal origin (meat, eggs, milk), in aquatic organisms and in breast milk. Human exposure and risk assessment modules of the system are developed in the relational database; the input data are directly loaded by the user through interactive masks. The exposure model computes, for each specific human receptor, the intake of a Chemical of Concern due to multiple exposure media and to multiple environmental media. In order to run exposure models it is necessary to input the characteristics of individuals. The data required to characterize the human receptors include anatomical and dietary properties, food consumption rates, activity patterns and exposure times, other human factors as soil ingestion and breast milk intake,
and parameters related with local food producing. Receptors parameters may also be obtained from a default database containing information extrapolated from Exposure Factors Handbook (USEPA, 1997). In the identification phase of receptors and exposure scenarios for assessment, consideration is given to ensure that the most sensitive potentially exposed individual, based on physical characteristics (such as body weight and health status), and behaviours (such as dietary habits and time spent on-site), is included in the assessment. To such purpose localization of schools, hospitals or residential areas with children is recommended. For each identified exposure scenarios, it is possible to select one up to 21 specific exposure pathways, including ingestion of different types of food, soil and water, inhalation of gas and particle, dermal contact with water and soil. The calculation of the upper-bound excess lifetime cancer risk and noncarcinogenic hazards for each of the pathways and receptors identified in the previous steps is performed by the module of risk assessment (see Fig. 1). The tool evaluates the individual potential risk with the assumption that an individual lives and works with a certain exposure frequency defined by the user, in a point of the grid in which the risk is calculated. If data to describe variability and perform a sensitivity analysis are available, the tool finally calculates the probability
448
P. Morra et al. / Environment International 32 (2006) 444 – 454
distribution of risk by using a Monte Carlo analysis and the Gis displays risk maps through specific user-defined percentiles contours. It is worth noting that the power of the tool is especially its aptitude to examine several sources of contamination, COCs, pathways and receptors in a single system and its capacity to illustrate such in detail a specific territorial pattern in this way giving a complete picture of risks. The HHRA-GIS results, displayed on georeferenced maps, can improve risks communication to all stakeholders (public administrations, industry operators, regulator agencies) and risk managers, in order to select options to reduce health risk and plan remediation actions. 4. The case study 4.1. Site description and contamination The case study concerns a working industrial site. In European countries, industrial emissions are restricted and controlled by regulators and concepts such as Integrated Pollution Prevention and Control, Best Practicable Environmental Option, and Best Available Technology (Not Entailing Excessive Costs) are applied, in order to reduce environmental impact of industries. Essentially companies are not required to assess the risks posed by their emissions, but to assure that no emission limit values are exceeded. The authorization procedure of industrial emissions through the standard based principles does not account the real site conditions and the characteristics of sensitive receptors. As a consequence, the standard based approach could sometimes not guarantee health for all the potential receptors or in the opposite case define limits to much restrictive in respect to the territory vulnerability. Therefore the case study explores the ability of risk-based approach to produce a quantitative evaluation of the risk posed by industrial emissions taking into account the site and the receptor features. In this paper an hypothetical territory characterized by residential, agricultural and industrial zones is presented. To preserve the location identity, the region is imaginary but the utilized data are realistic, in order to obtain outputs comparable to those of a real circumstance. Recently in areas with high population density and increasing waste generation, local government implements wastes management plans to resolve waste problems. Since incineration is lately matter of great argument and debate, the case study analyzes the presence of a municipal waste incinerator (MWI) as a source of contamination. Waste combustion has been considered the most common source of environmental polychlorinated dibenzo-p-dioxins (Stach et al., 2000). In addition to polychlorinated-dibenzo-pdioxins (PCDDs) and polychlorinated dibenzo-furans (PCDFs) also other environmental pollutants are emitted from the incinerators including heavy metals, polyaromatic hydrocarbons, polychlorinated biphenyls, sulfur oxides and nitrogen oxides. Because PCDDs and PCDFs have generated the most public concern, as they are perceived as the most hazardous
owing to possible carcinogenic effects, the present study focuses on these substances. Most dioxins are known to be resistant to environmental and biological degradation, chemically stable and poorly metabolised; they can concentrate in body fat and build up under conditions of long-term exposure, both in animals and humans, and accumulate as they move through the food chain. These properties suggest that even low levels of dioxins in the environment may eventually pose risks to animals and humans (USEPA, 2000). PCDD/Fs are formed through certain heterogeneous catalytic reactions in the temperature range 200 – 500 -C. Several studies have been conducted to elucidate the formation mechanisms of dioxins and to find out a correlation between dioxin emissions and waste composition, the flue gas composition and the operating conditions of incinerators (McKay, 2002; Gan et al., 2003). The wide variability of material incinerated and the combustion technologies with variable temperatures, oxygen requirements and residence times make this problem extremely complex. All data are transformed to 2,3,7,8-TCDD ‘‘toxic equivalents’’ by the application of weighting factors (I-TEFs or International Toxic Equivalent Factors) developed by EPA. This allows conversions of a concentration of any PCDD or PCDF congener into the toxicologically equivalent concentration of 2,3,7,8-TCDD. The use of TEFs to estimate the potency of dioxin-like compounds is a procedure in accordance with the definition of dioxins and furans specified in actual European directives, applying the concept of additivity for the PCDDs, PCDFs and dioxin-like PCBs to provide total concentrations estimates (EC Directive, 2000). In addition in the area of interest it has been assumed the presence of a contaminated site, delineated by soil sampling data. The considered Chemical of Concern is Hexachlorobenzene: a carcinogenic pesticide widely utilized in agriculture in the past as a selective fungicide for seed treatment of wheat against bunt. It adsorbs strongly to soil and is thought not to be mobile in soil. Hexachlorobenzene is a very persistent environmental chemical due to its chemical stability and resistance to biodegradation; this contaminant may persist in soil over 20 years. The Italian national regulation regarding contamination of soil has set upper concentration limits as 0.05 mg/kg for residential and public sites, 5 mg/kg for commercial and industrial sites (DM 471, 1999). 4.2. Maps of contaminants distribution Spatial interpolation of sampled point data of concentration was performed to map the distribution of hexachlorobenzene contamination of surface soil. To estimate the atmospheric dispersion of dioxins and their dry deposition on soil, due to the emission of the incinerator, the EPA model AERMOD (USEPA, 2002), integrated in the system, has been applied. A municipal waste incinerator stack of 60 m height and an emission factor of 0.075 ng I-TEQ/Nm3, were considered in the simulation, taking into account that all installations in Europe will achieve to comply with the 0.1 ng ITEQ/Nm3 emission limit until 2005 (EC Directive, 2000).
P. Morra et al. / Environment International 32 (2006) 444 – 454
Predicted ambient concentration and deposition of PCDD/Fs estimated by the atmospheric dispersion model were imported in HHRA-GIS project. 4.3. Characterization of receptors To take account of various typologies of receptors, the presence of two principal classes of receptors has been considered: the less vulnerable (healthy adult) and the most vulnerable (child). In addition, because ingestion of contaminated human milk may result in newborn infants in higher exposure to PCDDs, the infant receptor class exposed to breastfeeding is analyzed too. The exposure pathway considered in this assessment include inhalation, dermal absorption from soil, ingestion of soil, vegetation, meat, egg, water and milk. In the simulation the receptor population is exposed to emissions for a 70-year lifetime and is divided into three age groups: young children (up to 6 years old, 15 kg average body weight), adults (age 15 – 70 years old, 70 kg) and infants (up to 1 year, 5 kg). Inhalation exposure to emissions is calculated by assuming that individuals have different inhalation rate when exposed outdoor or indoor to contaminated air. As far as ingestion exposure is concerned, HHRA-GIS requires information about the fraction of food produced locally, since home-grown vegetables cultivated in the vicinity of the incinerator and the contaminated site are assumed to be consumed by local people in varying quantities. The assumption is that the individuals taken into account would consume products sold commercially, which contain a certain percentage of contaminated foods. Among ingestion receptors properties, soil intake is included: human ingest small amounts of soil indirectly from food and other sources. In addition children may ingest soil directly as a normal behaviour during childhood or as a result of pica.
449
4.4. Intake and risk calculation: results and discussion Risk assessment was performed combining chemical, toxicological data and model calculations. The modelling was performed on a grid domain of 17 16 km2 with a spatial resolution of 300 m. Multimedia models are applied to calculate total intake: atmospheric substances are partitioned to soil and vegetation and bioaccumulate in livestock (cows milk and meat). Concentrations of dioxins in vegetables, eggs, milk and meat were calculated using bioaccumulation food-chain models. Chemical accumulation in vegetation is estimated by adding dioxins deposition, atmospheric concentration uptake from the leaves and root uptake from soil. In addition farm animals can be exposed to contaminants via inhalation, ingestion of contaminated food and ingestion of contaminated soil. As a result, these compounds bioaccumulate in animals. In regard to milk, PCDDs and PCDFs bioconcentrate in animals due to their long half-life in adipose tissue. As a first example of risk distribution maps generated by the tool, Fig. 2 presents the map of cancer risk for adults due to hexachlorobenzene contamination. If a decontamination of the site after 2 years is hypothesized, with the aim of analyze the effects of remedial action on receptor health, the modified map should show a cancer risk level 20 times lower, because receptor is exposed for a duration period of 2 years instead of 40 years and risk depends linearly from exposure duration. Exploring HHRA-GIS total intake distribution maps, there is evidence that children are subjected to higher exposure then adults, as a consequence of differing eating habits and lower body weight. With respect to input data concerning fraction of local products consumed by children receptors, two extreme situations are explored (Figs. 3 and 4): the first one assumes that no local products are consumed by children, while the second one supposes that the fraction of local products is 1.
Fig. 2. Map of hexachlorobenzene cancer risk for adults receptors.
450
P. Morra et al. / Environment International 32 (2006) 444 – 454
Fig. 3. Map of dioxin cancer risk for children receptors—fraction of local products = 0.
Dioxin cancer risk for infant class of receptors is calculated, considering a diet based almost exclusively on breast milk intake, and showed maximum cancer risk of about 10 10. Examining results in the light of generally acceptable lifetime cancer risk range, 10 4 – 10 6, the maximum human health risk estimated caused by dioxins for receptors living in the study area (10 9 for adults and children, 10 10 for infants) has to be considered absolutely acceptable. The hexachlorobenzene contaminated site poses a cancer risk that exceeds the upper bound acceptable risk (10 4) and a hazard quotient that exceeds 1, but involves a rather limited area (0.6 km2); for this reason the analysis of total cancer risk posed by both contaminants has been considered not remarkable. Decontam-
ination of the site after 2 years leads to values approaching the acceptable range. Another useful comparison may be made with human exposure standards, generally referred to as tolerable daily or weekly intakes: World Health Organization has established a Tolerable Daily Intake (TDI) for dioxins as 1 –4 pg TCDD or TEQ/kg body weight day. The calculated total intakes of the case study (10 5 pg/kg day for adults and children, 10 6 pg/ kg day for infants) are significantly lower than the recommended TDI. A recent Italian study (Caserini et al., 2004) refers the measured values of background contamination and determines baseline contamination of polychlorinated dibenzo-p-dioxins
Fig. 4. Map of dioxin cancer risk for children receptors—fraction of local products = 1.
P. Morra et al. / Environment International 32 (2006) 444 – 454 90
451
sensitivity analysis applied in many different models in science, engineering, and economics, otherwise known as the elasticity equation. The ratio is equal to the percentage change in risk divided by the percentage change in input for a specific input variable. Risk estimates are considered more sensitive to input variables that yield the highest absolute value for SR. Sensitivity ratios can generally be grouped into two categories: local SR and range SR. For the local sensitivity ratio method, an input variable is varied by a small amount, usually T 5% of the nominal point estimate, and the corresponding change in the model output is observed. For the range sensitivity ratio method, an input variable is varied across the entire range (plausible minimum and maximum values). So, taking into account dioxin cancer risk for adults receptors, the following sensitivity ratios are calculated (Fig. 6). It is worth noting that equal values of local and range sensitivity ratios of a specific variable suggest that risk depends linearly from the variable. In particular local and range sensitivity ratio values relative to exposure duration and exposure frequency variables are equal to 1; body weight and inhalation rate are the succeeding variables in decreasing order of sensitivity according to both local and range SR. A variation of the sensitivity ratio approach may provide more information, but requires that additional information is available for the input variables: as a matter of fact the Sensitivity Score is the sensitivity ratio weighted by a normalized measure of the variability of the input variable. A probability distribution function of each input variable must be assigned: for normal distributions, arithmetic mean and standard deviation have been defined, for triangular distribution minimum, mode and maximum are the defined parameters. By normalizing the measure of variability (i.e. dividing by the mean), this method effectively weights the ratios in a manner that is independent of the units of the input variable, and provides a more robust method of ranking contributions to the
80 70 60
%
50 40 30 20 10 0 INHALATION INGESTION DERMAL CHILDREN ADULTS
Fig. 5. Percentage contribution of exposure pathways to total dioxin cancer risk.
and polychlorinated dibenzofurans; the typical concentration range is 0.02– 0.05 pg I-TEQ/m3 for rural sites, and 0.1– 0.4 pg I-TEQ/m3 for urban/industrial sites. As a result the contribution of incinerator emissions appears to be not influent. 4.5. Sensitivity analysis The case study has been completed with a sensitivity analysis, the aim being to understand how risk estimates are dependent on variability in the factors contributing to risk. In particular dioxin cancer risk is analyzed. According to the usual first step approach to sensitivity analysis (USEPA, 2001), the percentage contribution of exposure pathways to total risk is observed. Fig. 5 shows the important contribution to the risk of inhalation exposure pathway. The first step of sensitivity analysis proceeds with the calculation of the Sensitivity Ratio (SR). This is a method of
Sensitivity Ratio -1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
ED EF Body Weight Inhalation Rate Indoor Inhalation Rate Aboveground Intake Beef Intake Pork Intake Milk Intake Chicken Intake Egg Intake
Range SR
Local SR
Fig. 6. Local and range sensitivity ratio.
0.6
0.8
1
452
P. Morra et al. / Environment International 32 (2006) 444 – 454 Sensitivity Score -0.16
-0.12
-0.08
-0.04
0
0.04
0.08
0.12
0.16
IR indoor IR ED BW EF Aboveground Intake Beef Intake Pork Intake Milk Intake Egg Intake Chicken Intake Local SS
Range SS
Fig. 7. Local and range sensitivity score.
risk estimates than the sensitivity ratio alone. Fig. 7 illustrates the sensitivity scores values obtained in the case study. According to these results, variables which provide major contribution to risk value are inhalation rates, followed by exposure duration, body weight and exposure frequency variables, in decreasing order of sensitivity. A comparison with sensitivity ratios shows a different order of importance. In the second step approach of sensitivity analysis a commercial statistical software (Decisioneering, Inc., 2004) was utilized in order to obtain results of Monte Carlo
simulations, which allow multiple input variables to vary simultaneously. As a matter of fact, a probability distribution function of the risk has been estimated using the probability distributions functions of the input variables. A 1-D Monte Carlo Analysis for variability with 30,000 trials has been performed, which provided a new ranking of model inputs. Fig. 8 represents in the top panel the bar graph showing Spearman rank correlation coefficient, a metric for the dependence of cancer risk on exposure factors, and in the bottom panel a bar graph, sometimes referred to as ‘‘tornado
}
Fig. 8. Sensitivity chart obtained from Monte Carlo code.
P. Morra et al. / Environment International 32 (2006) 444 – 454
453
Fig. 9. Percentiles representation of dioxin cancer risk for adults.
plot’’, showing contributions to variance. The contribution to the variance is calculated by squaring the rank correlation coefficients and normalizing them to 100%. This graph is effective in showing both the relative magnitude and direction of influence (positive or negative) for each variable. The greater is the absolute value of the correlation coefficient, the stronger is the relationship. The sensitivity scores of Fig. 7 and correlation coefficients of Fig. 8 yield similar results, and suggest that, if the variation coefficients of the input variables can be estimated, the first step analysis can suffice to give help in focusing efforts on the variables that mostly contribute to the variability in risk. On the contrary, sensitivity ratio approach suggested that exposure frequency is the most significant variable, while the Monte Carlo analysis evaluates for it an importance less than 15% to the variability in the risk. These results suggest that sensitivity ratios are best applied to identify dominant exposure pathways rather than dominant exposure variables in the risk equation. But probability distribution functions of the input variables have a second and significant use: the drawing of contours lines at different percentiles. An example is given in Fig. 9 where the probability distribution of dioxin cancer risk for adults is displayed through the drawing of 10 9 contours at 20th, 50th and 90th percentiles. The magnitude of involved areas is a right measure of how variabilities in parameters influence risk calculations. 5. Final considerations The estimates of risk for human being caused by antropic activities is profitably done if the complete procedure of risk calculation can be adopted. The paper shows that HHRA-GIS code employs a risk calculation detailed procedure because it
may consider an integrated, multimedia, multi-exposure pathways and multi-receptor model. The case study discussed puts in evidence benefits, in terms of problem knowledge, quantitative risk evaluation of actual contamination state and of alternative scenarios, that the code can give to an assessor interested in assuming decisions for safeguarding citizen health. Acknowledgments Financial support from CNR in the form of a grant to one of the authors (P. Morra) is gratefully acknowledged. References Bagli S, Morra P, Spadoni G. The EHHRA tool: a decision support system for assessing and managing human health risk from industrial activities. Proceedings of 11th international symposium loss prevention, Praha; 2004. Caserini S, Cernuschi S, Giugliano M, Grosso M, Lonati G, Mattaini P. Air and soil dioxin levels at three sites in Italy in proximity to MSW incineration plants. Chemosphere 2004;54:1279 – 87. CEC (Commission of the European Communities). Commission Directive 93/67/EEC, OJ L227/10, 8.9.1993. Brussels; 1993. Decisioneering, Inc., Crystal Ball \ User Manual; 2004. D.M. n.471 of the 25/10/1999, published on Gazz. Uff. Suppl. Ordin. n.293 of the 15/12/1999. EC Directive 2000/76/EC. European Parliament and Council of the 4 December 2000 on the incineration of waste. Official Journal L 332/91; 2000. Gan S, Goh YR, Clarkson PJ, Parracho A, Nasserzadeh V, Swithenbank J. Postcombustion formation of dioxins/furans in waste incinerator plants. Journal of the Institute of Energy 2003;76:11 – 21. Kuchuk AA, Krzyzanowski M, Huysmans K. The application of WHO’s Health and Environment Geographic Information System (HEGIS) in mapping environmental health risks for the European region. Journal of Hazardous Materials 1998;61:287 – 90. McKay G. Dioxin characterisation, formation and minimisation during municipal solid waste (MSW) incineration: review. Chemical Engineering Journal 2002;86:343 – 68.
454
P. Morra et al. / Environment International 32 (2006) 444 – 454
NAS (National Academy of Sciences). Risk assessment in the federal government: managing the process. Washington, D.C.’ National Academy Press; 1983. Stach J, Peka´rek V, Grabic R, Lojka´sek M, Paca´kova´ V. Dechlorination of polychlorinated biphenyls, dibenzo-p-dioxins and dibenzofurans on fly ash. Chemosphere 2000;41:1881 – 7. USEPA (United States Environmental Protection Agency). Risk assessment guidance for superfund: volume I Human health evaluation manual (Part A). Washington, D.C.’ OERR; 1989. USEPA (United States Environmental Protection Agency). Exposure factors handbook Office of research and developmentR NCEA; 1997. USEPA (United States Environmental Protection Agency). Human health risk assessment protocol for hazardous waste combustion facilities, vol. 1; 1998.
USEPA (United States Environmental Protection Agency). Exposure and human health reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and related: Part IR Estimating exposure to Dioxin-like compounds. Washington Office; 2000. USEPA (United States Environmental Protection Agency) V. Risk assessment guidance for superfund: volume III—part AR Process for conducting probabilistic risk assessment. Washington’ OERR; 2001. USEPA (United States Environmental Protection Agency). User’s guide for the Ams/Epa regulatory model aermod Office of air quality planning and standards emissions, monitoring, and analysis. North Carolina’ DRTP; 2002.
The analysis of human health risk with a detailed procedure operating in a GIS environment P. Morra a, S. Bagli b, G. Spadoni a,* a
Dipartimento di Ingegneria Chimica, Mineraria e delle Tecnologie Ambientali, Alma Mater Studiorum, Universita` di Bologna, viale Risorgimento n.2, 40136 Bologna, Italy b Gecosistema srl-piazza Malatesta 21, 47900 Rimini, Italy Received 10 June 2005; accepted 7 October 2005 Available online 13 December 2005
Abstract An approach for quantifying the human health risk caused by industrial sources, which, daily or accidentally, emit dangerous pollutants able to impact on different environmental media, is introduced. The approach is performed by the HHRA-GIS tool which employs an integrated, multimedia, multi-exposure pathways and multi-receptors risk assessment model able to manage all the steps of the analysis in a georeferenced structure. Upper-bound excess lifetime cancer risk and noncarcinogenic hazards are the risk measures, the spatial distribution of which is calculated and mapped on the involved territory, once all the pathways and receptors of the study area are identified. A sensitivity analysis completes the calculations allowing to understand how risk estimates are dependent on variability in the factors contributing to risk. The last part of the paper makes use of a case study concerning a working industrial site to put in evidence in which way the designed tool can help local authorities and policy makers in managing risks and planning remedial and reduction actions. The considered geographical area is a hypothetical territory characterized by residential, agricultural and industrial zones. The presence of two sources of contamination, a municipal waste incinerator (MWI) and a contaminated site, are evaluated by the tool application. Various typologies of receptors have been taken into account, each of them characterized by different anatomical and dietary properties. The achieved results are analyzed, compared with acceptable and background values and alternatives of minor environmental impact calculated. D 2005 Elsevier Ltd. All rights reserved. Keywords: Quantitative human health risk; Sensitivity analysis; Fate and transport models; Municipal waste incinerator; Geographical information systems
1. Introduction Due to industrial and traffic emissions, agricultural activities and waste disposal facilities, many hazardous materials are released into the environment thus generating a potential hazard to human health and environment. In order to perform environmental risk evaluation, a great attempt is generally assigned to determine risk on human health owing to exposure at carcinogenic or toxic substances, which may be present in air, soil, water, foods, as consequence of anthropic activities. Procedures for performing this type of assessments are regularly delineated by government agencies, such as U.S. EPA (United States Environmental Protection Agency), EEA (European Environment Agency) and by European Directives and national regulations; most of them are based on the * Corresponding author. Tel.: +39 051 2093146; fax: +39 051 581200. E-mail address: [email protected] (G. Spadoni). 0160-4120/$ - see front matter D 2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.envint.2005.10.003
National Academy of Sciences model introduced in 1983 (NAS, 1983). A complete Environmental and Human Health Risk Assessment (EHHRA) requires the integration of information on environmental and chemicals database, inter and intra media dispersion models output, site description maps and demographic database. As a result, the designing and conducting analysis of the effects of contaminants on human health is a complex and time expensive procedure, particularly when the aim is to estimate risk in a geographical area characterized by the presence of several sources, exposure pathways and receptor typologies. Therefore in the last years, the need for defining and using procedures and tools, able to quantify risk for human beings, has increased especially for policy makers, local planners and, in general, for various stakeholders involved (Kuchuk et al., 1998). These ones are interested to manage and plan remedial actions in order to reduce environmental pollution by choosing options which minimize health risk.
P. Morra et al. / Environment International 32 (2006) 444 – 454
In a previous publication, a new Gis approach for quantifying the human health risk through a system able to manage all the steps in a georeferenced structure was introduced (Bagli et al., 2004). This paper presents the complete innovative methodology and explains the modified and improved tool, which is the result of a long timeconsuming work. The designed support system has been developed to help an expert user interested in assessing and managing risk for human beings living in a territory where industrial sources are present, which, daily or accidentally, emit dangerous pollutants able to impact on different environmental media. In order to estimate potential human health risk, the proposed tool, HHRA-GIS, employs an integrated, multimedia, multi-exposure pathways and multi-receptors risk model. Preserving the previous configuration, the system has been improved by adding new models and a sensitivity analysis module. This section allows to understand how risk estimates are dependent on variability in the factors contributing to risk. By assigning a probability distribution function to input variables and performing a Monte Carlo simulation, the sensitivity analysis provides a ranking of the model inputs based on their relative contributions. In the last part of the paper a case study concerning an hypothetical working industrial site is examined, to put in evidence in which way the designed tool can help local authorities, policy makers and stakeholders in managing risks and planning remedial and reduction actions. Since incineration is lately matter of great argument and debate, the presence of a municipal waste incinerator (MWI) as a source of contamination is evaluated by the tool application. In addition in the area of interest it has been assumed the presence of a contaminated site, delineated by soil sampling data. The presence of various typologies of receptors has been taken into account, each of them characterized by different anatomical and dietary properties. The achieved results are finally analyzed and compared with acceptability and background values. 2. Human health risk assessment The aim of human health risk analysis is the calculation of the upper-bound excess lifetime cancer risk and noncarcinogenic hazards. The procedure usually adopted in performing this evaluation, as assumed in the developed Gis tool, is based on the following paradigm: 1. Identification of Chemicals of Concerns (COC) 2. COC Fate and Transport Assessment 3. Exposure Scenario Definition and Exposure Intake Assessment 4. Dose– Response Assessment 5. Risk Characterization. The first step is the identification of substances which cause adverse effects and involves consultation of any toxicological and epidemiological data. In the second phase the distribution of concentrations of contaminants in various environmental media is analyzed. The Exposure assessment of the third step
445
requires the determination of the emissions, pathways and rates of movement of a substance and its transformation and degradation with the aim to estimate the ‘‘concentration/doses to which human populations (i.e. workers, consumers and men exposed indirectly via the environment) or environmental compartments (aquatic environment, terrestrial environment and air) are or may be exposed’’ (CEC, 1993). Dose –response assessment is obtained from epidemiological and toxicological data. Usually this is a tiered process that progresses from the use of short-term (acute) tests and conservative assumptions to longer-term (chronic) tests paired with more realistic assumptions. The risk characterization, which concludes the procedure, is essentially a summary of the data compiled in the risk assessment process including the uncertainties associated with each stage and the presentation of a risk estimate. Risk estimation consists in the computation of the upperbound excess lifetime cancer risk (risk) and noncarcinogenic hazards (hazard) for each of the pathways and receptors identified in the area of interest. Risk is defined as the probability that a receptor will develop cancer in his lifetime, assuming a unique set of exposure, model, and toxicity properties. In contrast, hazard is quantified as the potential for developing noncarcinogenic health effects as a result of exposure to COCs, averaged over an exposure period. It is worth noting that hazard is not a probability but, more exactly, a measure of the magnitude of a receptor’s potential exposure relative to a standard exposure level. The individual cancer risk of a receptor j set by exposure to multiple carcinogenic chemicals i, is calculated through the following equation: X Individual CancerRiskj ¼ LADDij CSFi i
where: LADDij Lifetime Average Daily Dose for a lifetime exposure of 70 years (mg/kg day) through multiple exposure pathways CSFi Cancer slope factor for COC i (mg/kg day) 1. Comparing an exposure estimate to a Reference Dose (RfD), the potential for noncarcinogenic health effects resulting from exposure to a chemical is evaluated. A RfD is defined as a daily intake rate that is estimated to cause no appreciable risk of adverse health effects, even to sensitive populations, over a specific exposure duration (USEPA, 1989). Generally, the more the Hazard Quotient value exceeds 1, the greater is the level of concern. In spite of this, because RfDs do not have equal accuracy or precision, and are not based on the same severity of effect, the level of concern does not increase linearly as the quotient approaches and exceeds 1. Based on similar COCs toxicological characteristics and additive health effects, the Hazard Quotient (HQ) for receptor j exposed to multiple chemicals i, is calculated as:
HQj ¼
X ADDij i
RfDi
446
P. Morra et al. / Environment International 32 (2006) 444 – 454
where: ADDij Average Daily Dose averaged for the exposure duration relative to the toxic i for the receptor j (mg/kg day) through multiple exposure pathways RfDi COC i Reference Dose (mg/kg day). Lifetime Average Daily Dose and Average Daily Dose represent the amount a person takes in as a result of exposure to a chemical in contaminated air, water, soil or food. The combinations of the potential pathways by which individuals can be exposed represent the exposure scenario for an individual. Human receptors may be exposed to COCs via multiple primary exposure routes, either directly or indirectly. The estimation of the exposure magnitude consists in the quantification of the potential dose that is of the quantity of agent that enters in contact with the human organism through the points of exchange with the environment (lungs, skin, etc.) in a specific interval time. The calculation of COC-specific exposure rates for each exposure pathway k to be considered requires the estimation of chemicals media concentrations, the consumption rate, the receptor body weight, the frequency and duration of exposure. The generic equation used to calculate chemical intake (LADDij for carcinogenic chemicals, ADDij for toxic chemicals) is: X Ci;em CRj;k EFj;k EDj;k Ii;j ð x; y; zÞ ¼ BWj ATj;k k where: I i,j (x, y, z) Intake for COC i and class of receptors j (mg/kg day) C i,em COC i concentration in exposure media (e.g. mg/kg for soil) CRj,k Consumption rate, the amount of contaminated medium consumed through the k exposure pathway per unit of time or event (e.g. kg/day for soil) EFj,k Exposure frequency (days/year) EDj,k Exposure duration (years), total time period over which contacts occur between receptor and COC BWj Average body weight of the receptor j over the exposure period (kg) ATj,k Averaging time, the period over which exposure is averaged (days); for carcinogens, the averaging time is 25,550 days, based on a lifetime exposure of 70 years; for toxics the averaging time is exposure duration expressed in days. In order to calculate chemicals concentrations in various exposure media, procedures extracted from USEPA (1998) are usually adopted, which allow the estimation of the concentration of a chemical in an exposure medium in terms of the chemical concentration in the environmental compartment, by means of an inter-media transfer factor: C ð x; y; zÞexposure ¼ f TF; C ð x; y; zÞenviromedia where: C(x, y, z)exposure Concentration in exposure media C(x, y, z)enviromedia Concentration in environmental media
TF
Inter-media transfer factor from an environmental compartment to an exposure medium.
Inter-media transfer factor is usually expressed by an appropriate chemical-specific partition coefficient, which describes the physicochemical attraction of the contaminant for the environment and the exposure medium. Concentration distributions in environmental media result from the application of contaminants fate and transport models or from measured data. 3. HHRA-GIS: a GIS approach to human health risk assessment The HHRA-GIS tool is an integrated approach for assessing the risks to human health; it is able to manage all the steps of the above described methodology in a georeferenced framework operating in a GIS environment, in order to display maps of iso-dose and iso-risk contours. In the numerical code the modules of the procedure are integrated in a system with a graphical user interface through which specific territorial pattern are represented in great detail. The tool, whose initial version has been improved and enriched with new modules, evaluates cancer risks and hazard quotients in a geographical area characterized by multimedia contamination, potential multi-exposure pathways and several receptor typologies. The system takes advantage of two typologies of database: the Gis Spatial Database, and the Relational Database in Microsoft Access linked to the ArcView Gis project using an Open Database Connectivity. Fig. 1 shows a block diagram that illustrates the architecture of the methodology to assess human health risk using HHRA-GIS. The digital site and facility description is an input of Gis Spatial DB; it consists in identifying and mapping contaminant sources, collecting facility and site information (e.g. the localization of rivers or lakes, soil properties, groundwater level, terrain elevation), estimating emission rates, and identifying Chemicals of Concern. For each chemical of concern identified in the contaminated area, estimates of the concentrations must be defined for each environmental medium of interest (e.g. soil, ground water, surface water, sediment, air, vegetation). User may derive concentrations from two broad methods: direct monitoring (e.g. sampling and chemical analysis of media at the site coupled with summary statistics) or environmental modelling (e.g. mathematical modelling to predict contaminant concentrations in various media). In both cases the output of the analysis is a series of maps describing the distribution of the toxic and carcinogenic chemicals concentrations in the environmental contaminated media. Chemicals environmental fate and transport models are implemented in the Gis tool so that the outputs of the models are imported into Gis and displayed as maps of distribution of concentration. The system graphical user interface allows the interaction with the different stages of the simulation, that are the data input (pre-processing), the run of the model and the
P. Morra et al. / Environment International 32 (2006) 444 – 454
GIS SPATIAL DB
ACCESS RELATIONAL DB
Contamination monitored data
Site characterization
Definition of emission sources
447
Fate and transport models
Definition of receptor characteristics
Distribution of COCs concentrations
Receptor typologies localization
Cancer Risk and Hazard Quotient calculation
Definition of calculation grid
Selection of exposure pathways
Chemical, phisical and toxicological database
Distribution of Cancer Risk and Hazard Quotient
Probability distribution of Cancer Risk and Hazard Quotient
Monte Carlo Analysis
Probability distribution functions
MONTE CARLO SOFTWARE Fig. 1. Block diagram of the methodology in HHRA-GIS. Yellow block: input data; light blue block: calculation module; orange block: output data.
output display (post-processing), taking advantage of Gis spatial analysis tools. The architecture of the designed system allows the integration of updated modelling components, or their replacement with models that match the specific features of the assessment. In the updated version of the system the following models are implemented: ISC3-GIS and AERMODGIS for the atmospheric dispersion of pollutants emitted by industrial sources; CALINE-GIS for the atmospheric dispersion of pollutants emitted by vehicles on roadways; HSSM-GIS and GMS for the transport of pollutants in soil and in groundwater; MULTIMEDIA models to estimate the partitioning of COCs in various environmental compartments; FOODCHAIN models to evaluate bioaccumulation in vegetables (belowground and above-ground), in food of animal origin (meat, eggs, milk), in aquatic organisms and in breast milk. Human exposure and risk assessment modules of the system are developed in the relational database; the input data are directly loaded by the user through interactive masks. The exposure model computes, for each specific human receptor, the intake of a Chemical of Concern due to multiple exposure media and to multiple environmental media. In order to run exposure models it is necessary to input the characteristics of individuals. The data required to characterize the human receptors include anatomical and dietary properties, food consumption rates, activity patterns and exposure times, other human factors as soil ingestion and breast milk intake,
and parameters related with local food producing. Receptors parameters may also be obtained from a default database containing information extrapolated from Exposure Factors Handbook (USEPA, 1997). In the identification phase of receptors and exposure scenarios for assessment, consideration is given to ensure that the most sensitive potentially exposed individual, based on physical characteristics (such as body weight and health status), and behaviours (such as dietary habits and time spent on-site), is included in the assessment. To such purpose localization of schools, hospitals or residential areas with children is recommended. For each identified exposure scenarios, it is possible to select one up to 21 specific exposure pathways, including ingestion of different types of food, soil and water, inhalation of gas and particle, dermal contact with water and soil. The calculation of the upper-bound excess lifetime cancer risk and noncarcinogenic hazards for each of the pathways and receptors identified in the previous steps is performed by the module of risk assessment (see Fig. 1). The tool evaluates the individual potential risk with the assumption that an individual lives and works with a certain exposure frequency defined by the user, in a point of the grid in which the risk is calculated. If data to describe variability and perform a sensitivity analysis are available, the tool finally calculates the probability
448
P. Morra et al. / Environment International 32 (2006) 444 – 454
distribution of risk by using a Monte Carlo analysis and the Gis displays risk maps through specific user-defined percentiles contours. It is worth noting that the power of the tool is especially its aptitude to examine several sources of contamination, COCs, pathways and receptors in a single system and its capacity to illustrate such in detail a specific territorial pattern in this way giving a complete picture of risks. The HHRA-GIS results, displayed on georeferenced maps, can improve risks communication to all stakeholders (public administrations, industry operators, regulator agencies) and risk managers, in order to select options to reduce health risk and plan remediation actions. 4. The case study 4.1. Site description and contamination The case study concerns a working industrial site. In European countries, industrial emissions are restricted and controlled by regulators and concepts such as Integrated Pollution Prevention and Control, Best Practicable Environmental Option, and Best Available Technology (Not Entailing Excessive Costs) are applied, in order to reduce environmental impact of industries. Essentially companies are not required to assess the risks posed by their emissions, but to assure that no emission limit values are exceeded. The authorization procedure of industrial emissions through the standard based principles does not account the real site conditions and the characteristics of sensitive receptors. As a consequence, the standard based approach could sometimes not guarantee health for all the potential receptors or in the opposite case define limits to much restrictive in respect to the territory vulnerability. Therefore the case study explores the ability of risk-based approach to produce a quantitative evaluation of the risk posed by industrial emissions taking into account the site and the receptor features. In this paper an hypothetical territory characterized by residential, agricultural and industrial zones is presented. To preserve the location identity, the region is imaginary but the utilized data are realistic, in order to obtain outputs comparable to those of a real circumstance. Recently in areas with high population density and increasing waste generation, local government implements wastes management plans to resolve waste problems. Since incineration is lately matter of great argument and debate, the case study analyzes the presence of a municipal waste incinerator (MWI) as a source of contamination. Waste combustion has been considered the most common source of environmental polychlorinated dibenzo-p-dioxins (Stach et al., 2000). In addition to polychlorinated-dibenzo-pdioxins (PCDDs) and polychlorinated dibenzo-furans (PCDFs) also other environmental pollutants are emitted from the incinerators including heavy metals, polyaromatic hydrocarbons, polychlorinated biphenyls, sulfur oxides and nitrogen oxides. Because PCDDs and PCDFs have generated the most public concern, as they are perceived as the most hazardous
owing to possible carcinogenic effects, the present study focuses on these substances. Most dioxins are known to be resistant to environmental and biological degradation, chemically stable and poorly metabolised; they can concentrate in body fat and build up under conditions of long-term exposure, both in animals and humans, and accumulate as they move through the food chain. These properties suggest that even low levels of dioxins in the environment may eventually pose risks to animals and humans (USEPA, 2000). PCDD/Fs are formed through certain heterogeneous catalytic reactions in the temperature range 200 – 500 -C. Several studies have been conducted to elucidate the formation mechanisms of dioxins and to find out a correlation between dioxin emissions and waste composition, the flue gas composition and the operating conditions of incinerators (McKay, 2002; Gan et al., 2003). The wide variability of material incinerated and the combustion technologies with variable temperatures, oxygen requirements and residence times make this problem extremely complex. All data are transformed to 2,3,7,8-TCDD ‘‘toxic equivalents’’ by the application of weighting factors (I-TEFs or International Toxic Equivalent Factors) developed by EPA. This allows conversions of a concentration of any PCDD or PCDF congener into the toxicologically equivalent concentration of 2,3,7,8-TCDD. The use of TEFs to estimate the potency of dioxin-like compounds is a procedure in accordance with the definition of dioxins and furans specified in actual European directives, applying the concept of additivity for the PCDDs, PCDFs and dioxin-like PCBs to provide total concentrations estimates (EC Directive, 2000). In addition in the area of interest it has been assumed the presence of a contaminated site, delineated by soil sampling data. The considered Chemical of Concern is Hexachlorobenzene: a carcinogenic pesticide widely utilized in agriculture in the past as a selective fungicide for seed treatment of wheat against bunt. It adsorbs strongly to soil and is thought not to be mobile in soil. Hexachlorobenzene is a very persistent environmental chemical due to its chemical stability and resistance to biodegradation; this contaminant may persist in soil over 20 years. The Italian national regulation regarding contamination of soil has set upper concentration limits as 0.05 mg/kg for residential and public sites, 5 mg/kg for commercial and industrial sites (DM 471, 1999). 4.2. Maps of contaminants distribution Spatial interpolation of sampled point data of concentration was performed to map the distribution of hexachlorobenzene contamination of surface soil. To estimate the atmospheric dispersion of dioxins and their dry deposition on soil, due to the emission of the incinerator, the EPA model AERMOD (USEPA, 2002), integrated in the system, has been applied. A municipal waste incinerator stack of 60 m height and an emission factor of 0.075 ng I-TEQ/Nm3, were considered in the simulation, taking into account that all installations in Europe will achieve to comply with the 0.1 ng ITEQ/Nm3 emission limit until 2005 (EC Directive, 2000).
P. Morra et al. / Environment International 32 (2006) 444 – 454
Predicted ambient concentration and deposition of PCDD/Fs estimated by the atmospheric dispersion model were imported in HHRA-GIS project. 4.3. Characterization of receptors To take account of various typologies of receptors, the presence of two principal classes of receptors has been considered: the less vulnerable (healthy adult) and the most vulnerable (child). In addition, because ingestion of contaminated human milk may result in newborn infants in higher exposure to PCDDs, the infant receptor class exposed to breastfeeding is analyzed too. The exposure pathway considered in this assessment include inhalation, dermal absorption from soil, ingestion of soil, vegetation, meat, egg, water and milk. In the simulation the receptor population is exposed to emissions for a 70-year lifetime and is divided into three age groups: young children (up to 6 years old, 15 kg average body weight), adults (age 15 – 70 years old, 70 kg) and infants (up to 1 year, 5 kg). Inhalation exposure to emissions is calculated by assuming that individuals have different inhalation rate when exposed outdoor or indoor to contaminated air. As far as ingestion exposure is concerned, HHRA-GIS requires information about the fraction of food produced locally, since home-grown vegetables cultivated in the vicinity of the incinerator and the contaminated site are assumed to be consumed by local people in varying quantities. The assumption is that the individuals taken into account would consume products sold commercially, which contain a certain percentage of contaminated foods. Among ingestion receptors properties, soil intake is included: human ingest small amounts of soil indirectly from food and other sources. In addition children may ingest soil directly as a normal behaviour during childhood or as a result of pica.
449
4.4. Intake and risk calculation: results and discussion Risk assessment was performed combining chemical, toxicological data and model calculations. The modelling was performed on a grid domain of 17 16 km2 with a spatial resolution of 300 m. Multimedia models are applied to calculate total intake: atmospheric substances are partitioned to soil and vegetation and bioaccumulate in livestock (cows milk and meat). Concentrations of dioxins in vegetables, eggs, milk and meat were calculated using bioaccumulation food-chain models. Chemical accumulation in vegetation is estimated by adding dioxins deposition, atmospheric concentration uptake from the leaves and root uptake from soil. In addition farm animals can be exposed to contaminants via inhalation, ingestion of contaminated food and ingestion of contaminated soil. As a result, these compounds bioaccumulate in animals. In regard to milk, PCDDs and PCDFs bioconcentrate in animals due to their long half-life in adipose tissue. As a first example of risk distribution maps generated by the tool, Fig. 2 presents the map of cancer risk for adults due to hexachlorobenzene contamination. If a decontamination of the site after 2 years is hypothesized, with the aim of analyze the effects of remedial action on receptor health, the modified map should show a cancer risk level 20 times lower, because receptor is exposed for a duration period of 2 years instead of 40 years and risk depends linearly from exposure duration. Exploring HHRA-GIS total intake distribution maps, there is evidence that children are subjected to higher exposure then adults, as a consequence of differing eating habits and lower body weight. With respect to input data concerning fraction of local products consumed by children receptors, two extreme situations are explored (Figs. 3 and 4): the first one assumes that no local products are consumed by children, while the second one supposes that the fraction of local products is 1.
Fig. 2. Map of hexachlorobenzene cancer risk for adults receptors.
450
P. Morra et al. / Environment International 32 (2006) 444 – 454
Fig. 3. Map of dioxin cancer risk for children receptors—fraction of local products = 0.
Dioxin cancer risk for infant class of receptors is calculated, considering a diet based almost exclusively on breast milk intake, and showed maximum cancer risk of about 10 10. Examining results in the light of generally acceptable lifetime cancer risk range, 10 4 – 10 6, the maximum human health risk estimated caused by dioxins for receptors living in the study area (10 9 for adults and children, 10 10 for infants) has to be considered absolutely acceptable. The hexachlorobenzene contaminated site poses a cancer risk that exceeds the upper bound acceptable risk (10 4) and a hazard quotient that exceeds 1, but involves a rather limited area (0.6 km2); for this reason the analysis of total cancer risk posed by both contaminants has been considered not remarkable. Decontam-
ination of the site after 2 years leads to values approaching the acceptable range. Another useful comparison may be made with human exposure standards, generally referred to as tolerable daily or weekly intakes: World Health Organization has established a Tolerable Daily Intake (TDI) for dioxins as 1 –4 pg TCDD or TEQ/kg body weight day. The calculated total intakes of the case study (10 5 pg/kg day for adults and children, 10 6 pg/ kg day for infants) are significantly lower than the recommended TDI. A recent Italian study (Caserini et al., 2004) refers the measured values of background contamination and determines baseline contamination of polychlorinated dibenzo-p-dioxins
Fig. 4. Map of dioxin cancer risk for children receptors—fraction of local products = 1.
P. Morra et al. / Environment International 32 (2006) 444 – 454 90
451
sensitivity analysis applied in many different models in science, engineering, and economics, otherwise known as the elasticity equation. The ratio is equal to the percentage change in risk divided by the percentage change in input for a specific input variable. Risk estimates are considered more sensitive to input variables that yield the highest absolute value for SR. Sensitivity ratios can generally be grouped into two categories: local SR and range SR. For the local sensitivity ratio method, an input variable is varied by a small amount, usually T 5% of the nominal point estimate, and the corresponding change in the model output is observed. For the range sensitivity ratio method, an input variable is varied across the entire range (plausible minimum and maximum values). So, taking into account dioxin cancer risk for adults receptors, the following sensitivity ratios are calculated (Fig. 6). It is worth noting that equal values of local and range sensitivity ratios of a specific variable suggest that risk depends linearly from the variable. In particular local and range sensitivity ratio values relative to exposure duration and exposure frequency variables are equal to 1; body weight and inhalation rate are the succeeding variables in decreasing order of sensitivity according to both local and range SR. A variation of the sensitivity ratio approach may provide more information, but requires that additional information is available for the input variables: as a matter of fact the Sensitivity Score is the sensitivity ratio weighted by a normalized measure of the variability of the input variable. A probability distribution function of each input variable must be assigned: for normal distributions, arithmetic mean and standard deviation have been defined, for triangular distribution minimum, mode and maximum are the defined parameters. By normalizing the measure of variability (i.e. dividing by the mean), this method effectively weights the ratios in a manner that is independent of the units of the input variable, and provides a more robust method of ranking contributions to the
80 70 60
%
50 40 30 20 10 0 INHALATION INGESTION DERMAL CHILDREN ADULTS
Fig. 5. Percentage contribution of exposure pathways to total dioxin cancer risk.
and polychlorinated dibenzofurans; the typical concentration range is 0.02– 0.05 pg I-TEQ/m3 for rural sites, and 0.1– 0.4 pg I-TEQ/m3 for urban/industrial sites. As a result the contribution of incinerator emissions appears to be not influent. 4.5. Sensitivity analysis The case study has been completed with a sensitivity analysis, the aim being to understand how risk estimates are dependent on variability in the factors contributing to risk. In particular dioxin cancer risk is analyzed. According to the usual first step approach to sensitivity analysis (USEPA, 2001), the percentage contribution of exposure pathways to total risk is observed. Fig. 5 shows the important contribution to the risk of inhalation exposure pathway. The first step of sensitivity analysis proceeds with the calculation of the Sensitivity Ratio (SR). This is a method of
Sensitivity Ratio -1
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
ED EF Body Weight Inhalation Rate Indoor Inhalation Rate Aboveground Intake Beef Intake Pork Intake Milk Intake Chicken Intake Egg Intake
Range SR
Local SR
Fig. 6. Local and range sensitivity ratio.
0.6
0.8
1
452
P. Morra et al. / Environment International 32 (2006) 444 – 454 Sensitivity Score -0.16
-0.12
-0.08
-0.04
0
0.04
0.08
0.12
0.16
IR indoor IR ED BW EF Aboveground Intake Beef Intake Pork Intake Milk Intake Egg Intake Chicken Intake Local SS
Range SS
Fig. 7. Local and range sensitivity score.
risk estimates than the sensitivity ratio alone. Fig. 7 illustrates the sensitivity scores values obtained in the case study. According to these results, variables which provide major contribution to risk value are inhalation rates, followed by exposure duration, body weight and exposure frequency variables, in decreasing order of sensitivity. A comparison with sensitivity ratios shows a different order of importance. In the second step approach of sensitivity analysis a commercial statistical software (Decisioneering, Inc., 2004) was utilized in order to obtain results of Monte Carlo
simulations, which allow multiple input variables to vary simultaneously. As a matter of fact, a probability distribution function of the risk has been estimated using the probability distributions functions of the input variables. A 1-D Monte Carlo Analysis for variability with 30,000 trials has been performed, which provided a new ranking of model inputs. Fig. 8 represents in the top panel the bar graph showing Spearman rank correlation coefficient, a metric for the dependence of cancer risk on exposure factors, and in the bottom panel a bar graph, sometimes referred to as ‘‘tornado
}
Fig. 8. Sensitivity chart obtained from Monte Carlo code.
P. Morra et al. / Environment International 32 (2006) 444 – 454
453
Fig. 9. Percentiles representation of dioxin cancer risk for adults.
plot’’, showing contributions to variance. The contribution to the variance is calculated by squaring the rank correlation coefficients and normalizing them to 100%. This graph is effective in showing both the relative magnitude and direction of influence (positive or negative) for each variable. The greater is the absolute value of the correlation coefficient, the stronger is the relationship. The sensitivity scores of Fig. 7 and correlation coefficients of Fig. 8 yield similar results, and suggest that, if the variation coefficients of the input variables can be estimated, the first step analysis can suffice to give help in focusing efforts on the variables that mostly contribute to the variability in risk. On the contrary, sensitivity ratio approach suggested that exposure frequency is the most significant variable, while the Monte Carlo analysis evaluates for it an importance less than 15% to the variability in the risk. These results suggest that sensitivity ratios are best applied to identify dominant exposure pathways rather than dominant exposure variables in the risk equation. But probability distribution functions of the input variables have a second and significant use: the drawing of contours lines at different percentiles. An example is given in Fig. 9 where the probability distribution of dioxin cancer risk for adults is displayed through the drawing of 10 9 contours at 20th, 50th and 90th percentiles. The magnitude of involved areas is a right measure of how variabilities in parameters influence risk calculations. 5. Final considerations The estimates of risk for human being caused by antropic activities is profitably done if the complete procedure of risk calculation can be adopted. The paper shows that HHRA-GIS code employs a risk calculation detailed procedure because it
may consider an integrated, multimedia, multi-exposure pathways and multi-receptor model. The case study discussed puts in evidence benefits, in terms of problem knowledge, quantitative risk evaluation of actual contamination state and of alternative scenarios, that the code can give to an assessor interested in assuming decisions for safeguarding citizen health. Acknowledgments Financial support from CNR in the form of a grant to one of the authors (P. Morra) is gratefully acknowledged. References Bagli S, Morra P, Spadoni G. The EHHRA tool: a decision support system for assessing and managing human health risk from industrial activities. Proceedings of 11th international symposium loss prevention, Praha; 2004. Caserini S, Cernuschi S, Giugliano M, Grosso M, Lonati G, Mattaini P. Air and soil dioxin levels at three sites in Italy in proximity to MSW incineration plants. Chemosphere 2004;54:1279 – 87. CEC (Commission of the European Communities). Commission Directive 93/67/EEC, OJ L227/10, 8.9.1993. Brussels; 1993. Decisioneering, Inc., Crystal Ball \ User Manual; 2004. D.M. n.471 of the 25/10/1999, published on Gazz. Uff. Suppl. Ordin. n.293 of the 15/12/1999. EC Directive 2000/76/EC. European Parliament and Council of the 4 December 2000 on the incineration of waste. Official Journal L 332/91; 2000. Gan S, Goh YR, Clarkson PJ, Parracho A, Nasserzadeh V, Swithenbank J. Postcombustion formation of dioxins/furans in waste incinerator plants. Journal of the Institute of Energy 2003;76:11 – 21. Kuchuk AA, Krzyzanowski M, Huysmans K. The application of WHO’s Health and Environment Geographic Information System (HEGIS) in mapping environmental health risks for the European region. Journal of Hazardous Materials 1998;61:287 – 90. McKay G. Dioxin characterisation, formation and minimisation during municipal solid waste (MSW) incineration: review. Chemical Engineering Journal 2002;86:343 – 68.
454
P. Morra et al. / Environment International 32 (2006) 444 – 454
NAS (National Academy of Sciences). Risk assessment in the federal government: managing the process. Washington, D.C.’ National Academy Press; 1983. Stach J, Peka´rek V, Grabic R, Lojka´sek M, Paca´kova´ V. Dechlorination of polychlorinated biphenyls, dibenzo-p-dioxins and dibenzofurans on fly ash. Chemosphere 2000;41:1881 – 7. USEPA (United States Environmental Protection Agency). Risk assessment guidance for superfund: volume I Human health evaluation manual (Part A). Washington, D.C.’ OERR; 1989. USEPA (United States Environmental Protection Agency). Exposure factors handbook Office of research and developmentR NCEA; 1997. USEPA (United States Environmental Protection Agency). Human health risk assessment protocol for hazardous waste combustion facilities, vol. 1; 1998.
USEPA (United States Environmental Protection Agency). Exposure and human health reassessment of 2,3,7,8-Tetrachlorodibenzo-p-Dioxin (TCDD) and related: Part IR Estimating exposure to Dioxin-like compounds. Washington Office; 2000. USEPA (United States Environmental Protection Agency) V. Risk assessment guidance for superfund: volume III—part AR Process for conducting probabilistic risk assessment. Washington’ OERR; 2001. USEPA (United States Environmental Protection Agency). User’s guide for the Ams/Epa regulatory model aermod Office of air quality planning and standards emissions, monitoring, and analysis. North Carolina’ DRTP; 2002.