Risk based characterisation of contaminated industrial site using multivariate and geostatistical tools
Introduction
Risk assessment has been internationally recognised as the most cost-effective and scientific tool for tackling the overwhelming problem of the contaminated sites management (US-EPA, 1989, CARACAS & NICOLE, 1997, Ferguson et al., 1999). In terms of human health, risk assessment involves identifying the potential for adverse health effects to be caused by chemicals of concern from a site, and thereby determine the need for remedial action or the development of target levels where remedial action is required.
Risk assessment procedures are generally based on the source–pathway–receptor model (Golden Software Inc, 1996, ASTM, 1995, CANCAWE, 1997), and encompass the examination of the site characteristics, the environmental behaviour and toxicity of the contaminants, the potential route of entry of the contaminants into the receptors (humans), the exposure of the receptors to the contaminants and their response to the dose.
Thus, site characterisation is the basis for risk assessment. Although much scientific literature is developing on risk assessment issues (Ferguson, 1996), comparatively little attention is paid to the characterisation.
A proper risk-oriented characterisation should provide a conceptual model of the site, a quali/quantitative representation of the contaminant sources and as much of the data necessary for modelling contaminant fate and transport (Ferguson et al., 1998). The identification of both primary and secondary sources is generally recommended (ASTM, 1995). The primary source is the cause of the actual contamination, and concerns the nature and the place of the discharge, the mechanisms of transport and the environmental processes occurred, whereas the secondary source is the impacted environmental media to which the receptor is exposed.
Since risk assessment is frequently a tiered or phased approach, moving from conservative assumptions to more site-specific and accurate characterisations, the characterisation is also a tiered approach, based on preliminary through to detailed investigations. The selection of the appropriate level of detail necessary for risk assessment depends on the complexity and the particular circumstances of the site, as well as cost and other project constraints (US-EPA, 1990).
Both the characterisation and the overall risk assessment, have to deal with a large number of uncertainties (Dakins et al., 1994). Due to the heterogeneity of the soil and the often accidental nature of contaminating processes, concentrations of pollutants may vary remarkably over very short distances. This often makes it difficult to obtain a meaningful picture of the contamination and to develop a conceptual model of the site. Moreover, it is common and often unavoidable to have to deal with small data sets and not exhaustive samplings in the horizontal and vertical dimension. The elimination of uncertainty is not feasible, cost-effective or necessary for estimating the risk and selecting appropriate remedies. The need to quantify and reduce the uncertainties and minimise the investigation costs, strongly encourages the use of geostatistical and multivariate statistical methods (ferguson, 1998, Ferguson et al., 1998). Kriging and principal component analysis (PCA) are two common examples of geostatistical and multivariate statistical methods, respectively.
Kriging is a linear-weighted gridding method, that has been already successfully used to produce illustrative contour plots of contaminant distribution on the basis of scattered observed concentration data (Leonte & Schofield, 1996, Juang & Lee, 1988, Nathanail et al., 1998). It generally allows one to increase the information retrieval from analytical data and thereby reduces investigation costs.
PCA has been extensively applied in many disciplines, but not yet in the risk assessment-oriented characterisation of contaminated sites. PCA enables the pollutant composition in different samples to be compared and also provides fingerprints for identifying the origin of the pollution (Burns et al., 1997).
The objective of this work is to extend, by the use of Kriging, PCA and additional data, the results of the risk-based characterisation of an industrial contaminated site presented by Carlon et al. (2000). This site, located near Fidenza (Parma, Italy), was selected as a case study in a project co-ordinated by the National Environmental Protection Agency of Italy (ANPA) for determining risk assessment guideline values. The characterisation of the site included analytical survey of both soil and groundwater. On the basis of the former industrial use of the site and the results of previous surveys, the sampling strategy focused on defining the extent of both total and tetraethyl Pb contamination in soil. However, the soil analyses also showed significant polycyclic aromatic hydrocarbon (PAH) contamination. A preliminary risk assessment referred to the possible commercial use of the site outlined a significant risk for human health derived from Pb and tetraethyl lead (Et4Pb) through dust ingestion and vapour inhalation, and only from benzo(a)pyrene, benzo(a)anthracene and naphthalene through leaching to groundwater.
Previous results suggested the need for further investigation of PAH distribution and the relation of the PAH contamination in the study area with that originated from a factory producing PAHs located on the East site.
The use of Kriging and PCA as explorative methods to describe the distribution and identify the primary and secondary sources of the site contamination is reported below.
Section snippets
The study area
The study area was described in detail by Carlon et al. (2000) and only the key features and additional details are present here. The study area is located in the industrial district North of Fidenza (Parma, Italy), and covers some 3 km2 out of which approximately one-third is occupied by disused factory buildings. Until 1976 this area was occupied by a factory (called CIP in this paper) producing Et4Pb, prior to being abandoned. To the east the study area adjoins a factory (called CARB in this
Analytical methods
The results of 46 soil samples collected in the CIP area (18 sampling points labelled with 1–5 and E3–14) and analysed using the mobile chemical laboratory of the Joint European Research Centre (JRC) in 1996 (Carlon et al., 2000), were combined with those of 25 soil samples collected in the east CIP area and CARB area (six sampling points labelled Car5, Car7, Car9, Car10, Car14 and Car16) and analysed by the CARB laboratory in 1995 (Ambert et al., 1995). The sampling methods used in the two
Results
The descriptive statistics of the PAHs concentrations in CIP and CARB soil are presented in Table 1. The PAH concentration distribution shows extremely high skewness and kurtosis values. After the data log-transformation, the consistent reduction of skewness and kurtosis indicates a better agreement with a normal distribution (the probability plot before and after the log-transformation are not presented but reported as supplemental material and available on request).
Discussion
PAH distribution in CIP soil is remarkably different from those of Pb and Et4Pb. Whereas Pb and Et4Pb contamination affects only the west CIP area and is mainly confined to the surface soil, PAH contamination also affects the east CIP area and the deeper soil layers. This is mainly due to different pollution sources. As with Pb and Et4Pb, the PAHs have been directly discharged into the settling tanks and on CIP soil surface, as inferred from the presence of hot spots in the unsaturated zone
Conclusions
In this case study, additional information from the data set, which is relatively small and affected by sampling gaps, was extracted by Kriging and PCA. It provided fundamental results to risk assessment, such as the determination of the primary and secondary pollution sources, and allowed the development of a reliable conceptual model of the site. The study of the spatial correlation of the concentration data allows preliminary information on the general pattern of contaminative processes and
References (30)
- et al.
Identification of petroleum sources in an area impacted by the Amoco Cadiz oil spill
Marine Pollution Bulletin
(1988) - et al.
Hydrocarbon source identification and weathering characterization of intertidal and subtidal sediments along the saudi Arabian coast after the Gulf War oil spill
Marine Pollution Bulletin
(1993) - et al.
Risk Assessment Study of the Carbo Chemicals Site in Fidenza for Addressing Remedial Actions (Final report)
(1995) - ANPA, CCR, ENEA, UNICHIM, 1998. Contaminated Soil and Groundwater — The CIP Abandoned Industrial Site —...
- ASTM, 1995. Standard Guide for Risk-based Corrective Action Applied at Petroleum Release Sites-RBCA, E 1739-95. ASTM,...
- et al.
Structure-biodegradability relationships of polycyclic aromatic hydrocarbons in soil
Bulletin of Environmental Contaminant Toxicology
(1986) - et al.
A principal component and least squares method for allocating polycyclic aromatic hydrocarbons in sediment to multiple sources
Environmental Toxicology and Chemistry
(1997) - CARACAS and NICOLE, 1997. Joint Statement. Available at: http://www.caracas.at/joint_statement.rtf. EU,...
- et al.
Risk analysis: chemical characterisation of a contaminated industrial site
Annali di Chimica
(2000) The scree test for the number of factors
Multivariate Behavioral Research
(1966)
European Oil Industry Guideline for Risk-based Assessment of Contaminated Sites (Final Report by the Water Quality Management Group)
Risk-based environmental remediation: decision framework and role of uncertainty
Environmental Toxicology and Chemistry
Environmental stability of selected petroleum hydrocarbon source and weathering ratios
Environmental Science and Technology
Chemometrics in Environmental Analysis
Assessing human health risks from exposure to contaminated land: a review of recent research
Land Contamination & Reclamation
Cited by (138)
Understanding the driving mechanisms of site contamination in China through a data-driven approach
2024, Environmental PollutionAssessing the ecological risk of croplands in loess drylands by combining environmental disturbance with ecosystem vulnerability
2023, Journal of Environmental ManagementThree-dimensional distribution characteristics of multiple pollutants in the soil at a steelworks mega-site based on multi-source information
2023, Journal of Hazardous MaterialsHow to describe organic contamination in soils: A model comparison for chlorinated solvent concentrations at industrial site scale
2018, Science of the Total EnvironmentCitation Excerpt :Geostatistics allows modelling the spatial distribution of the contamination, estimating concentrations for mapping and calculating the associated uncertainty. In geostatistics, the concentrations are interpolated to assess the vertical and horizontal distribution of pollutants and to identify the sources of contamination (Carlon et al., 2001). However, the data set of soil pollution is never exhaustive and the estimation of high contaminated areas can be associated with a large uncertainty.
Chemical accident hazard assessment by spatial analysis of chemical factories and accident records in South Korea
2018, International Journal of Disaster Risk Reduction