Brownfields and health

Housing is often built on soils which have residual contamination from previous uses. Certainly more and more Canadian cities are starting to use old downtown industrial sites, or “brownfields,” for new housing developments. The former Expo Lands in Vancouver is just one example.

Before they are reoccupied, these lands are cleaned up to standards set by provincial authorities. Most provinces use generic soil criteria for such clean-ups but some are now also permitting site specific risk assessments (SSRA) as an alternative. Canada Mortgage and Housing Corporation (CMHC), the federal housing department, initiated research to look at the variability of site specific risk assessments on contaminated sites which might be intended for residential land use.

Recently, provincial authorities have been allowing site specific risk assessments to justify only partial clean-up of a site’s soils if it can be shown that the inhabitants of the housing are at an acceptable level of risk from exposure to the contaminants. These assessments integrate various scientific disciplines (e.g. toxicology, hydrogeology, physics, chemistry, transport modelling) and use them along with the risk assessor’s professional judgment to ascertain if the level of contamination presents unacceptable health risks to people or animals. During the process, the assessor must provide an assessment of the health risk for risk management purposes and yet has to contend with uncertainty in the available data. The uncertainty may arise from natural variation, or imperfect information on variables such as the level, fate and transport mode of the contaminant, and human receptor behaviour (Figure 1).

These circumstances led CMHC to query how far screening risk assessments differ among risk assessors. Golder Associates’ Burnaby office was retained to compare the results of risk assessments performed by an array of consultants in a “round-robin” study, and to comment on the differences found between them. The results of the study suggest that screening level risk assessments should be conducted cautiously and by experienced practitioners if they are to be used in important business or remediation decisions.

Golder started by surveying the expertise of Canadian firms performing environmental risk assessments. Responses came from 25 firms, representing a range of company size and expertise. We used a scoring system based on the type and number of risk assessments conducted annually and the reported depth of knowledge and experience of the staff.

We selected 10 firms to represent both geographical diversity and a range of expertise. The Golder-assigned consultant scores ranged from 5-39 points, 39 being the highest possible. Of the nine consultants who eventually performed the risk assessments, four had more than 30 points, three had between 20 and 30 points, and two had scores between 10 and 20. Firms with very low scores, reflecting minimal experience and capability, were not chosen.

To help standardize the level of effort, funds were provided to each participant for approximately eight person-days to complete the task. Participants were instructed to use the standards and procedures relevant to their provinces in their analyses. Golder provided no strategic guidance to aid the participants.

The hypothetical case study was a vacant industrial site in a mid-sized city. A mock phase two site assessment identified several metals, benzene, and vinyl chloride in the soil, based on 30 soil core samples and 15 groundwater wells. The developer was going to build 60 single detached dwellings, typical three-bedroom houses of about 1,800 s.f. (167 m2). The houses would have full basements with forced air heating systems. The municipality would supply drinking water.

Given the analyses from the participants, Golder evaluated the calculated health risks. We used criteria which reflect the policies of many regulatory agencies to establish if the health risks would be considered acceptable or unacceptable: an “acceptable health risk” was one where the hazard quotient (dose rate divided by reference dose) was less than one, or the incremental lifetime cancer risk was smaller than one in a million.

Disparity

The risk assessments submitted by the nine firms varied widely. The cancer risks associated with vinyl chloride exposure, for example, as summarized in Figure 2, show a variability spanning some seven orders of magnitude depending on the receptor chosen. Similarly, cancer risks associated with benzene exposure and also non-carcinogenic contaminants exhibited a wide range of responses.

Not surprisingly, when we examined whether the widely varying estimated health risks would be considered acceptable to regulators and society, we found some agreement, but little consensus. This situation can be seen for non-carcinogenic compounds in Figure 3. The study is predicated on “acceptable” health risks being below unity and “unacceptable” health risk being above unity. Note that it was Golder who applied these common criteria respecting “acceptability” of risk, and not the participants themselves.

All respondents clearly found lead concentrations to be a concern. The risks from the other chemicals were sometimes seen as acceptable and sometimes as unacceptable, depending upon the analysis of the consultant.

Similarly, there was no consensus between the nine consultants on whether contaminants on this site pose an acceptable or unacceptable carcinogenic risk. Furthermore, there were differences in the type, number and input parameters for the exposure pathways used in the analyses, reflecting the variability among consultants in formulating the problem at the screening level of environmental risk assessment.

We examined the sources of variability behind the risk estimates to explain the diverse results. Part of the variability might be attributed to the fact that the project involved the “screening level” risk assessment in which coarse assumptions are made to enable rapid analysis. With the extra scrutiny and effort involved in a “detailed” risk assessment, there may have been more congruency.

There are two general areas where variability was found: in the toxicity reference values (the safe level of exposure defined by regulators) and in the dose rates (the estimated level of exposure that would occur at the hypothetical housing development).

The toxicity reference values should have been relatively uniform for the contaminants because these are typically dictated by provincial and federal regulators, and their contribution to the variability was indeed less than that found in dose rates. However, some exceptions were still noted. For example, while reference values were consistent for gaseous contaminants such as benzene and vinyl chloride, those for lead and, to a lesser extent, for copper and zinc varied up to three orders of magnitude. These differences are largely attributable to differing sensitivities between children and adults. But the information source (e.g. Health Canada, provincial documents, journal articles, and the U.S. EPA IRIS database) also introduced variability. Cancer slope factors, which characterize the toxic potency of carcinogens such as vinyl chloride and benzene, had far more consistency than toxicity reference values for non-carcinogens.

Dose rate variability was more complex. In the calculated dose rates of metals through soil ingestion and absorption through the skin, the characteristics of the human receptors accounted for the majority of the variation. Ingestion rate, exposed dermal area, and body weight were influential factors, as was the frequency of a person’s exposure. With inhaled contaminants, be they dust or gases, similar variations were found.

However, with inhaled contaminants the differences in predicted exposure concentrations were larger than the differences in the characteristics of the human receptors, indicating that additional and significant sources of variability were present. Variability in the predicted exposure concentration for airborne contaminants involves many factors. For example, the
participating firms had to select which of the several concentrations given by the assessment would be used: the mean, the maximum reported, or some percentile of those values.

The availability of several soil gas transport models introduced further variability. In addition, the house itself would affect the dilution rate of vapour phase contaminants as they exit the soil and infiltrate the house. Very few of the participants had a strong grasp of the housing factors that would affect this dilution ratio. It is likely that the more complex the exposure pathway model, the greater would be the deviation among the participants’ estimates of contaminant exposure and risk due to uncertainty in the numerous model variables. This variability appears to have been the case for soil-to-house infiltration modelling.

When the consultants calculated indoor airborne benzene concentrations, for example, their use of different assumptions resulted in benzene concentrations ranging from 0.0000027 mg/m3 to 9.64 mg/m3.

The study suggests that excessive and repetitious conservative assumptions (“blanket conservatism”) by some participants probably led to some of the highly deviant predictions. Such conservatism tends to inflate the final risk estimates far beyond those where conservatism was limited to one or two key factors. There was no apparent trend in this small sample of nine participants between their capability (as defined within the study) and the degree of conservatism they employed.

Finally, it is important to note that the project results are also affected by the contaminant concentrations provided to the participants. All of them, for example, produced soil ingestion risks for zinc that were acceptable by Golder’s criteria. However, if the zinc level introduced in the hypothetical case was roughly three orders of magnitude higher, zinc ingestion risks would have been split, with some above and some below a hazard quotient of one.

Summary

1. Nine consultants performed screening level risk assessments on a hypothetical house to be located on a brownfield. There was significant variability in the predicted level of health risk and no consensus on the acceptability of risk from various sources.

2. The disagreements on the risk were mostly due to variations on the calculated dose and, to a lesser extent, the toxicity reference values.

3. Much of the variation was based on differences in modelling of soil gas flows.

4. Some consultants applied a conservative bias to all aspects of the risk assessment, instead of for key parameters only, resulting in an unrealistic estimate.

5. The study suggests that screening level risk assessments should be conducted cautiously and by experienced practitioners if they are to be used in important business or remediation decisions. CCE

Don Fugler, P.Eng., is a senior researcher in the Research Division of Canada Mortgage and Housing Corporation, Ottawa, Ontario. Michael Rankin, M.Sc., is an associate and senior toxicologist/risk assessor at Golder Associates Ltd., Burnaby, B.C.

For more information tel. Don Fulger at (613) 748-2658. To obtain the report or summary: Evaluation of Site-Specific Risk Assessment for Contaminated Lands, 1997 contact the CMHC library at 1-800-668-2642/ (613) 748-2367.

This paper is adapted from one given at the 1998 Air and Waste Management Association conference in San Diego.

Many cities want to use old industrial downtown sites, or “brownfields,” for housing developments. But when researchers asked environmental assessment firms to gauge the health risks of a hypothetical contaminated site, they turned up some surprising results.