K.6.0 UNCERTAINTY IN HUMAN HEALTH RISK
Human health risk assessment results are conditional estimates that depend on the assumptions made to account for uncertainties in biological processes or a lack of information on source data, transport, or receptor behavior. It is important to recognize these uncertainties to place the risk estimates in proper perspective. The uncertainties associated with the TWRS EIS risk estimates include parameters involved in the models used and historical data on worker risks and accidents. Volume Three, Appendix D presents some parameter uncertainties associated with remediation risk (Section D.4.14), anticipated post-remediation risk (Section D.5.14), ecological risk (Section D.6.5), and intruder risk (Section D.7.5), which are briefly discussed as follows.
To estimate risk, information must be available on dose-response relationships, which define the biological response from exposure to a contaminant. Although human epidemiological data are used for developing radiological and nonradiological chemical dose-response models, this information also is developed in laboratory tests using animals exposed to relatively high doses. Therefore, uncertainty is inherent in dose-response relationships, including extrapolating from effects in animals at high doses to potential effects in humans that most often are exposed at much lower doses.
Another important component of risk assessment is estimating exposure concentrations. Uncertainties associated with this component include estimating releases of contaminants from emission sources to different environmental media such as groundwater, soil, air, and surface water, the transport and transformation of contaminants in these media, and the pathway, frequency, and duration by which humans contact the contaminants.
The risk associated with the release of radionuclides or chemicals to ambient environmental media during routine operations was estimated using models. The risk estimates determined by these models have a greater uncertainty than those based on historical data. However, it is reasonable to assume that releases would occur on a routine basis over the operational lifetime of the facility. The risk estimates for post-remediation and intruder scenarios are associated with more uncertainty than facility routine operation risk and involve uncertainties associated with the hypothetical land use and intrusion in addition to modeling. Finally, the MEI risk estimates generally involve a greater level of uncertainty than population risk estimates.
K.6.1 POST-REMEDIATION LAND-USER RISK
The uncertainty analyses for post-remediation risk assessment were based on the Hanford Site Risk Assessment Methodology (HSRAM) uncertainty analysis. The carcinogenic and noncarcinogenic risks presented in the post-remediation risk evaluation were estimates based on multiple assumptions about exposures, toxicity, and other variables. Therefore, discussion of uncertainty was provided for this risk assessment. The uncertainties are inherent (e.g., toxicity values, default exposure parameters) or specific (e.g., data evaluation, contaminant identification) in the risk assessment process. Specific considerations in evaluating uncertainty were Site-specific factors, exposure assessment factors, toxicity assessment factors, and risk characterization factors, which are discussed as follows.
K.6.1.1 Site-Specific Uncertainty Factors
Uncertainty related to the source inventory, Site contamination, availability of information on Site-specific environmental conditions (e.g., climate, geology, and hydrogeology), and uncertainties in model application to the Site were important in assessment of risk associated with the Site. These uncertainties are addressed in Appendices A, B, and F.
K.6.1.2 Exposure Assessment Uncertainty Factors
Exposure assessment requires multiple assumptions that can affect the outcome of a risk assessment. Key factors contributing to uncertainty in the exposure assessment included the following:
- Identification of land use;
- Likelihood of future land use actually occurring;
- Model assumptions that affect exposure point concentrations;
- Use of standard default parameters (e.g., upper 95th percentile values for intake/contact rates, exposure frequency, and exposure duration);
- Uncertainty related to biotransfer factors;
- Uncertainty related to production and distribution of food; and
- Uncertainty related to lifestyle and diet of specific or referenced individuals.
K.6.1.3 Toxicity Assessment Uncertainty Factors
A high degree of uncertainty was associated with data used to derive toxicity values and resulted in less confidence in assessment of risk associated with exposure to a substance. Sources of uncertainty associated with published toxicity values include:
- Use of dose-response information from effects observed at high doses to predict effects at the low levels expected in the environment;
- Use of data from short-term exposure studies to extrapolate to long-term exposure or vice-versa;
- Use of data from animal studies to predict human effects; and
- Use of data from homogenous animal populations or healthy human populations to predict effects in the general population.
K.6.1.4 Risk Characterization Uncertainty Factors
The summation of cancer risk across pathways or for multiple pathways would result in more conservative risks. This is because the slope factor for each chemical carcinogen is an upper 95th percentile estimate and such probability distributions are not strictly additive. The risk values calculated for the post-remediation scenario in the TWRS EIS were a conservative bounding estimate. The uncertainty in the risk values for certain receptors would increase as the time in the future increased. Less uncertainty would be associated with the risk values at 300 years than the risk estimates at 500, 2,500, 5,000, and 10,000 years.
The best approach to more fully characterize the uncertainty would be to conduct a probabilistic risk assessment from the start of the evaluation. A probabilistic assessment uses the range of variation in contaminant information, exposure parameters, and toxicity data to provide a risk distribution curve. This appendix examines the effect of variations on these parameters on the risk estimates to provide a better understanding of the uncertainties.
K.6.2 POST-REMEDIATION INTRUDER RISK
The greatest uncertainty in calculating the intruder risk was associated with the source data. Source terms were based on the estimated inventory and an average tank within the eight aggregated tank farms of the 200 Areas. Additional information regarding the source term would decrease the uncertainty in the risk estimate.
The relative uncertainty associated with the dose conversion factor was not as important as the source data, source terms, and exposure pathway parameters. The GENII computer code was used for the intruder dose calculation. GENII used the dosimetry model recommended by the International Commission on Radiation Protection (ICRP), in ICRP Publication 26 (ICRP 1977) and ICRP Publication 30 (ICRP 1979-1982), with updates from ICRP Publication 48 (ICRP 1986). The dose conversion factors used were equivalent to those currently recommended by the (DOE 1988). External dose factors were equivalent to Kocher (Kocher 1981; ORNL 1981). The overall uncertainty associated with risk in respect to GENII is discussed in Volume Three, Section D.4.14.
K.6.3 REMEDIATION ROUTINE RISK
By far the greatest uncertainty in the routine remediation risk was associated with the source data, which were based on the estimated inventory and source terms (i.e., the amount of chemicals and radionuclides released into the environment). The uncertainties associated with the source and source terms are discussed in Volume Two, Appendices A and B. Other contributors to the routine risk uncertainty were the airborne transport of the released chemicals and radionuclides, accumulation of contaminants in food products, production and distribution of food products, and lifestyle and diet of specific individuals, food consumption rates, and dose conversion factors which are discussed in this section.
Routine chemical emissions from the tank farm during remediation were based on existing tank farm emissions data (Jacobs 1996). Operational emissions from the tank farm, such as would occur while retrieving waste from tanks and gravel-filling the tanks, were appropriately scaled for potential increased emission rates during remediation.
The hazard index (HI) approach conservatively assumed that the noncarcinogenic health effects were additive for all chemicals (i.e., all chemicals would have the same mechanism of action and affect the same target organ). The HI is the sum of the hazard quotients (estimated intake/reference dose) for all chemicals. A HI greater than or equal to 1.0 indicates potential adverse health effects in the population of concern. Conversely, a HI less than 1.0 suggests that adverse health effects would be unlikely.
Carcinogenic risks were assumed to be additive. Consequently, the total ILCR is the sum of individual chemical cancer risks from each emission source for each alternative analyzed. Regulatory agencies have defined an acceptable level of risk to be between 1 in 10,000 (1.0E-04) and 1 in 1,000,000 (1.0E-06), with 1.0E-06 being the point of departure and referred to as de minimis (below which there is minimal concern) risk. For the purpose of this EIS, a risk below 1.0E-06 was considered low, and a risk greater than 1.0E-04 was considered high.
K.6.4 REMEDIATION ACCIDENT RISK
The objective of this section is to summarize the results of the Monte Carlo uncertainty and sensitivity analyses of the LCF predictions associated with the potential accidental release of contaminants from each TWRS EIS remedial alternative.
A detailed description of the general methodology used in the Monte Carlo approach is presented in Section K.5.0. The methodology used to estimate the uncertainty in the LCF predictions was similar to that used to predict the uncertainty in the human health exposure factors (see Section K.5.0). In this approach, the variables used to predict the LCF were separated into variables which can be described as PDFs and those having constant or fixed point estimates. A computer simulation was then run in order to produce a PDF for LCF. The results of the computer simulation were then compared to the results of the fixed point estimate. The equation used to predict LCF from an accidental release is:
Where:
LCF = latent cancer fatality
Chi/Q = atmospheric dispersion coefficient (second/m3)
V = release volume (rem/liter)
IR = inhalation rate (m3/second)
ULD = unit liter dose (committed effective dose equivalent/liter)
C = conversion factor (LCFs/rem)
The Monte Carlo analysis described each of the variables in the above equation (i.e., Chi/Q, V, IR, ULD and C) as PDFs and not as a single value.
K.6.4.1 Accident Release Scenarios
This uncertainty analysis evaluated the consequences to four receptors as a result of the spray release accident scenario presented in Volume Four, Appendix E:
- MEI noninvolved worker;
- Noninvolved worker population;
- MEI general population; and
- General public population.
The radiological dose to a receptor would depend on the receptors location relative to the point of release of the radioactive material. Doses for a MEI and population dose were computed for each receptor (noninvolved worker and general public). Noninvolved workers would be onsite workers not involved in the proposed action. The general public would be people located off the Hanford Site. The MEI for each of these receptor categories would be a single individual assumed to receive the highest exposure in the category. Volume Four, Appendix E of this report contains a more detailed description of the receptors associated with each accidental release scenario.
K.6.4.2 Monte Carlo Uncertainty Analysis
The PDFs for the variables in the equation used to calculate LCF were assumed to be triangular distributions (Finley et al. 1994). Triangular distributions can be viewed as conservative characterizations of truncated normal or lognormal distributions. The triangular distribution was conservative in that it resulted in more frequent selection of values in the extremes of the factor's distribution.
The inhalation rate triangular PDF was assumed to have a minimum value of 6.9E-05 m3/s (6 m3/day), mean value of 2.1E-04 m3/s (18.9 m3/day), and maximum value of 3.7E-04 m3/s (32 m3/day) based on worker ventilation rates under light activity levels (EPA 1985).
The remaining variables in the equation for calculating LCF were also assumed to have a triangular PDF. The values were chosen to correspond to conditions associated with a nominal accidental release value as well as an upper bounding value for accidental release. A more detailed description of the accidental release scenarios and the rationale for the selection of the nominal and bounding values is presented in Section K.6.5.
A Monte Carlo analysis was conducted based on the above algorithm for calculating the LCF. The sensitivity analysis indicated that the parameters which contributed the most to the uncertainty in the LCF were as measured by rank correlation: the unit limit dose, the atmospheric dispersion coefficient, the release volume, the inhalation rate, and the conversion factor. The detailed results of the Monte Carlo analysis are presented in Attachment 1 and are summarized in Table K.6.4.1.
Table K.6.4.1 contrasts the mean and percentile estimates of the LCF distributions for the four accidental release scenarios with the fixed point estimate derived using the upper-bound values. The results show that the LCF derived using the upper-bound values is in all cases greater than the 100th percentile of the LCF PDF. The mean of the LCF PDF was approximately one order of magnitude less than the upper-bound fixed point estimate. These results demonstrate that the predicted LCF estimates would be upper bound estimates of cancer probability and/or fatality rates. The true probability of contracting cancer or fatalities as a result of cancer could actually be much less than the predicted value.
K.6.5 ANALYSIS OF NOMINAL VERSUS BOUNDING RISK ESTIMATES
The bounding risk estimates in the TWRS EIS used a series of conservative assumptions about source and release terms, environmental transport parameters, and the effects of a given exposure on cancer risk and noncancer health effects to account for the uncertainties involved in the alternatives. This section analyses the effect of using less conservative values for several of the source term, release term, and environmental transport assumptions on the risk estimates. No change was made in the SIFs used to estimate the risk from each exposure.
Based upon available data, the assumption for the distribution coefficient (Kd) for Np-237, a major contributor to the groundwater risks, was changed from zero, which implies that Np would move at the same rate as water, to 1.0, which implies that interaction with the soil would slow its movement to and through the aquifer. For the ex situ alternatives, assumptions about tank residuals were changed, as described in the following sections.
Table K.6.4.1 Comparison of Monte Carlo-Based and Fixed Point Estimates
K.6.5.1 Tank Residuals Nominal Case
A nominal case retrieval release and residual tank inventory was developed to assess the impacts that would result from nominal assumptions for tank releases during retrieval and the residual waste left in the tanks following retrieval. Details are presented in Volume Two, Appendix B. The nominal release inventory was developed by assuming that the waste would be diluted by one-third by adding liquids for sluicing during retrieval. Possible dilution ratios that would be used during waste retrieval ranged from 3:1 to 10:1. Thus, the dilution factor of one-third assumed for the nominal case was a conservative assumption. These dilution ratios represent the amount of liquid required to mobilize the waste solids and would be made of existing tank liquids and water additions. The nominal case retrieval release volume was assumed to be 15,000 L (4,000 gal) from each SST, and the contaminant concentrations were assumed to be two-thirds of the bounding case. The average volume of waste released from each SST during retrieval was not reduced for the nominal case, because insufficient information was available to support a lower average release volume. The volume released would depend on the ability to detect a leak and take corrective action.
The nominal tank residual inventory was developed by modifying the bounding tank residual inventory to reduce the mobile constituents of concern based on solubility. The mobile constituents of concern were evaluated because of their contribution to post-remediation risk. The isotopes C-14, Tc-99, and I-129 were reduced to the nominal case residual inventory to 10 percent of the bounding residual inventory. This was based on the assumption that 90 percent of the residual inventory of these isotopes would be soluble in the retrieval liquids and would be retrieved from the tanks for ex situ treatment. Typical sludge wash factors, representing the water solubility of these isotopes, were as high as 99 percent. The nominal case residual was limited to 90 percent to account for conditions in which the scale and hardened sludges would not see the sluicing liquid during retrieval. Table K.6.5.1 shows the nominal and bounding residual inventories for select mobile constituents.
Table K.6.5.1 Tank Residual Inventory, Curies
K.6.5.2 Nominal vs. Bounding Risk Results
The ILCR and HI results for the nominal and bounding cases are presented in Volume One, Table 5.11.7. The overall effect of the changes in the nominal case was to reduce the estimated risks. The size of the effect would vary with the exposure scenario, the alternative, and the future time examined (Volume One, Table 5.11.7). For some scenarios, for some points in time, the nominal case risks were higher than those for the bounding case. For example, the No Action alternative showed higher risks for the nominal than the bounding case for all scenarios at 2,500 years. This occurred because one of the key assumptions, decreasing the mobility of Np-237, caused the exposure to Np in groundwater to be delayed, but did not change its ultimate impact on the risk. In the bounding case, the risk from Np occurred early, because the Np was assumed to move quickly, and then decreased as the Np was removed by attenuation and ultimate loss to the Columbia River. Thus, relaxing a conservative assumption about contaminant mobility could have more effect on the timing than on the degree of risk. Nonetheless, Volume One, Table 5.11.8 demonstrates that the total cancer incidence over the 10,000-year period of interest is decreased in the nominal case.
|
NEWSLETTER
|
| Join the GlobalSecurity.org mailing list |
|
|
|
