A spatial statistical methodology to assess the contribution of land use to soil contamination and its influence on human health

Soil is a crucial component of rural and urban environments and soil quality in both these environments can be influenced significantly by land management. Heavy metals occur naturally in soils in small amounts. However, occasionally high heavy metal loads may originate from parent rocks through weathering, volcanic eruptions, and forest fires, or are attributed to human activities. Potentially contaminated soils may occur at old landfill sites, particularly those that accepted industrial wastes; fields that had past applications of waste water or municipal sludge; areas in or around mining waste piles and tailings; industrial areas where chemicals may have been dumped on the ground; or in areas downwind from industrial sites. Excess heavy metal accumulation in soils is toxic to humans and animals. Exposure to heavy metals is normally chronic, while acute poisoning from heavy metals is rare through ingestion or dermal contact, but is possible. Cadmium (Cd), and Lead (Pb) have been drawing a lot of attention from geochemists and environmental scientists due to the greater understanding of their toxicological affects on agriculture, ecosystems and human health. Chronic exposure to Cd is known to adversely affect kidney, liver, and the gastrointestinal function, while Pb is well known to affect the nervous system. One interesting question for governments, regulators and the community as a whole is to be able to attribute the sources of heavy metals to either natural sources or land management practices, which may span decades or even centuries considering heavy metals tend to accumulate. Redevelopment and reuse of these soils may pose a threat to human health through uptake of contaminants via ingestion, inhalation, and dermal contact. From the human health protection point of view it is important to assess whether the contamination present in soil reaches human receptors through intake/uptake since plausible source-pathway-receptor linkages exists. In that case, health effects may be related to the source and pathway of heavy metals. In addressing these topics, the objectives of this research were to develop spatial statistical methodologies that can be used: i) to correlate geochemical data with historic land uses and geological data in order to evaluate the influence of historic land use on soil contaminant levels ii) to correlate modelled contaminant levels with cancer incidence data in order to ascertain whether contaminated land may influence human health and assess the strength of any putative relationships. The correlation between heavy metals in soils and their origins (geological and anthropogenic sources) was investigated and quantitatively analysed through model regression, while geostatistical methods were used to analyse the spatial aspects of soil contamination autocorrelation. A probabilistic modelling method was developed to assess whether hot-spot areas within a study region can be better defined using geological and historical land use data. The methodology developed includes indicator kriging, logistic regression and the Bayes theorem as its main building blocks. In order to assess the health risk attributable to soil contamination, spatial autocorrelation and data clustering analyses were employed on cancer incidence data in order to identify whether living in a contaminated area can be one of the factors that attribute to developing cancer. Assessment of the correlation between contamination levels and cancer incidence on different geospatial levels was carried out. It was concluded that it is possible to identify and model the relative contribution of different land use based heavy metal sources to soil contamination. It was also shown that integration of spatial autocorrelation in the modelling has some advantages in terms of model fitting and prediction ability. Correlation between estimated soil contamination levels and cancer incidence data was not shown to be significant and no spatial clustering was found on census geospatial levels