summary report. Therefore, readers of the NEHC draft summary report cannot understand the modeling results, their interpretation, and their significance. More details should be included. The discussion of dispersion modeling on p. 13 of the Pioneer (2000) draft report should include at least a summary of the assumptions, data sources, and methods.
Page 4-7 of the Radian report (2000a) states that results of a small number of studies indicate that the maximal discrepancies between the values predicted by the Industrial Source Complex-Short Term (ISCST3) model and measured concentrations are generally less than 30% for well-characterized sources. References should be provided for that statement. One validation study using that model gave a correlation coefficient of 0.97 between observed and predicted concentrations of particles at an industry fence line, but that was for long-term estimates (Heron et al. 1984). Correlations for short-term estimates can be much worse than for long-term estimates.
On p. 4-8, the Radian (2000a) report notes that atmospheric stability class was not directly measured and lists it as a source of uncertainty. However, Appendix I states that atmospheric stability class was determined on the basis of solar-radiation and temperature-gradient measurements made on site (Radian 2000d; p.1-3). That method is one of the best for determining stability class—better than the most commonly used approach based on wind speed and cloud cover. Therefore, very little uncertainty would result from that determination.
On p. 4-8 (Radian 2000a), another listed source of uncertainty is the assumption that each of the three incinerator stacks always had equal emission rates. The approach for estimating emission rates on p. 4-6, however, allows calculation of the emission rates separately, as well as lumped together. It might not have been possible to distinguish the impact of one stack from another with the method described, but no rationale for using the lumped approach is given.
APPENDIX C
Uncertainty
There is inadequate discussion of uncertainty in the NEHC draft summary report (2000). The draft summary report mentions uncertainty only in the context of saying that it is minimized. It fails to disclose the types or magnitudes of any source of uncertainty and to discuss the impact of uncertainty in the context of the risk-assessment results. For example, p. 2 states:
Throughout the monitoring program and development of the health risk assessment, NEHC consulted with USEPA experts in various scientific fields. They provided the latest scientific information to address health concerns, as well as input on sampling and analytical methodologies, to ensure that the health risk assessment would be developed with the minimum degree of uncertainty.
The purpose and meaning of “minimum degree of uncertainty” is not clear. Combining this statement with the precise and unqualified estimates of risk in the draft summary report, readers might infer that the risk-assessment results are certain. That is not consistent with the limitations of risk assessment in general.
On p. 11, the NEHC draft summary report (2000) states:
Because this study was designed to support risk management decisions regarding the health risks at NAF Atsugi, it was critical to reduce, as much as possible, the uncertainties regarding sampling and analytical procedures. To reduce these uncertainties in the sampling methodology, sampling methods were selected based on their ability to collect samples with sufficiently low detection
limits to perform health-based risk analysis. The number and frequency of samples were determined using a statistical approach to control the possibility of over- or underestimating the health risks. This statistical method determined that air samples would be collected approximately every six days. The sampling frequency for more than one year guaranteed that each day of the week, as well each week of the year, would be represented. Thus any variability due to the day of the week, season, or other temporal effects could be assessed.
Nevertheless, the subcommittee recommends that NEHC characterize the magnitude of uncertainties before and after their minimization and determine their impacts on the results of the risk assessment.
A discussion of uncertainty appears on pp. 45-46, but the discussion is not adequate. The text fails to convey the uncertainty surrounding the risks and could result in confusion about risk-assessment methods. For example, it discusses issues related to extrapolation of animal-toxicity data to humans, discusses generic uncertainty issues, and states that “in calculating toxicity values for each chemical, safety factors of 10 to 1,000 are applied to the toxicity values to account for these extrapolations.” The sentence apparently is directed at interpreting the application of uncertainty and modifying factors in estimating reference doses and reference concentrations for noncancer health effects, but the complete bases for the safety factors of 10-1,000 and when they are used should be clarified.
The final paragraph in the section on uncertainties, on p. 46, implies that it is more controversial to evaluate uncertainties than it is not to, and that doing so requires more judgment than is required for a screening risk assessment. The subcommittee disagrees and recommends that the uncertainty in the point estimates of risks presented (such as a statement that actual risks are likely to be somewhere between zero and the upper-bound estimates provided) be more fully characterized and that NEHC reconsider the discussion of the benefits of characterizing uncertainty.
Section 6 of the draft human health risk assessment by Pioneer Technologies Corporation (Pioneer 2000) provides the direction and a qualitative assessment of the magnitude of uncertainty related to each of the following:
-
Identification of chemicals of concern (COCs) present in the soil at NAF Atsugi.
-
Identification of COCs present in ambient air at NAF Atsugi.
-
Identification of COCs present in indoor air at NAF Atsugi.
-
Identification of COCs present in indoor dust at NAF Atsugi.
-
Quality of analytic data.
-
Assumption that no attenuation or enrichment of constituent, concentrations in soil or indoor dust occurs over time
-
Exposure assumptions.
-
Experimental dermal absorption rates.
-
Failure to include all constituents because of lack of EPA-approved toxicity values.
-
Extrapolation to humans of toxicity observed in animal studies.
-
Lack of constituent-specific dermal toxicity values.
-
Use of dose-response information from homogeneous animal populations or healthy human populations to predict effects in the general population, including susceptible subpopulations.
-
Assumption that health effects of constituents are additive.
Of those 13 sources of uncertainty, the NEHC draft summary report (2000, pp. 45-46) mentions only the last four; the subcommittee believes that other sources of uncertainty also warrant mention. The criteria used to assign qualitative magnitudes should be provided. For example, the report should state why extrapolation from animal studies to humans is classified as a 3 and the use of dose-response information from homoge-
neous animal populations or healthy human populations to predict effects in the general population, including susceptible subpopulations, is classified as a 1. In addition, the magnitude rating of 2 for uncertainty associated with soil, indoor dust, and dermal absorption seems to be high, considering that little of the risk was attributable to those media. Therefore, absorption might not have a “medium” effect on risk estimates, as stated in the supporting document (Pioneer 2000; p. 88). The subcommittee recommends that NEHC state the meanings of “negligible”, “small”, “medium”, and “large” effects on risk calculations.
On p. 88 of the Pioneer (2000) draft report, the column labeled “action or result” in the table is confusing, and it is not clear whether the magnitude classification applies before or after the listed actions have been taken. It is not clear whether those magnitude classifications are related to the uncertainties in the characteristics themselves or to their impact on the overall results.
Discussion of the exposure assumptions and scenarios should be expanded to convey the variability and uncertainty in exposure estimates. Some uncertainties mentioned elsewhere in the text do not appear in the section on uncertainty. For example, the partitioning of the substances into particulate and gas phases is stated to be uncertain in some places but is not mentioned in the section on uncertainty.
Measurements of particular compounds should be discussed as possible sources of uncertainty. For example, uncertainty might be associated with dioxin measurements (on the basis of the following statement p. 29, NEHC 2000):
Maximum detections for dioxins were always found in the samples taken near the fenceline north of the incinerator and at the golf course between the third and fourth holes, east of the incinerator. Elevated dioxin concentrations found at the golf course between the third and fourth holes, east of the incinerator appear to be an anomaly perhaps associated with blowing ash.
Because of the interpretation of the hydrogen chloride data, reported concentrations might have been higher than the actual concentrations at NAF Atsugi. FTIR did not detect hydrogen chloride (Radian 2000a; p. 3-3), although, according to conventional measurements, it was expected that FTIR would detect hydrogen chloride. It was assumed that FTIR failed to detect hydrogen chloride because it was in particulate or aerosol form, which would not be detected by FTIR. That assumption is illustrated in the following paragraph (Radian 2000a; p. 3-4) :
On seven occasions between January and April 1999, the ambient air monitoring station at the GEMB site reported 24-hour hydrogen acid concentrations near or above the FTIR system's minimum detection limit of 7 µg/m3. The FTIR system did not detect hydrochloric acid on any of these occasions, and these results were carefully checked. As mentioned in Section 3.1 above, the FTIR only detects chemicals in the form of a gas. Hydrochloric acid in the form of aerosols or particles is not detected. It must be assumed that the hydrochloric acid measured by conventional means on the seven occasions consisted largely of aerosols and particles since the chemical was not detected in the form of a gas by the FTIR system.
However, the following is stated about possible interference by other chloride compounds, on the basis of denuder measurements (Radian 2000a; p. 2-14):
Therefore, particles captured on the filter should only contain negligible amounts of HCl and HF, but could contain an interferent such as sodium chloride NaCl (metallic chloride salts). Chloride salts on the filter would cause a high bias in the estimate of HCl. In fact, for most of the highest HCl values reported, the major contribution was found on the filter, with much smaller amounts on the denuder sections.
