Peer Review of the Surface Impoundment Study (SIS) Technical Plan for Human Health and Ecological Risk Assessment
General Comments and Recommendations by Alan Eschenroeder, Ph.D.
This report is one of the best EPA efforts at developing and documenting a risk assessment methodology that this reviewer has seen. It reflects a good degree of realism on the part of the authors in recognition of the obstacles in performing such a complex and large-scale task as that mandated by the statute and the consent decision. Alternatives and backup plans reflect anticipation of the difficulties in carrying out such an extensive risk assessment project. The authors are to be especially commended for trying to use existing models wherever possible, and for carrying the ecological risk assessment into levels of detail seldom seen in such risk assessments. At the end of the day, one wonders, however, whether there will ever be enough resources to really perform all that is promised. Perhaps additional fallback alternatives could be presented.
The intended readership of this report is not made clear, but it may likely involve judges and lawyers that need to be convinced. That being the case, it is always helpful to engage the services of a well educated,but non-technical editorial reviewer. Get that reviewer's comments and take them seriously in revising the document. This is primarily a linguistic exercise, which goes beyond the issues that we as peer reviewers will discuss. A master flow chart with each box keyed to a subsection number would provide a wonderful visual outline of the report.
This peer review addresses a series of areas where improvements are possible in the documentation or the study design. In most instances, good judgment has been exercised; however, there are other cases where some improvements are indicated. In most cases, the improvements focus on presentation issues rather than substantive technical problems. Before presenting the comments and questions, it is helpful to outline the organization of the review.
This review moves from the general to the particular for each of the areas specified in the CHARGE TO PEER REVIEWERS. Comments in each area begin with general suggestions for improvements in both the writing style and the technical content. The comments then move to specific points where clarification, revisions or expansions might improve the document.
Derivation of Human Health Screening Factors
The methodology for deriving human health screening factors seems valid by assuming that the media themselves are inhaled, ingested or contacted by the receptors. Exceptions to this direct principle are recognized in the report by providing that chemicals that bioaccumulate automatically get kicked into a Phase II analysis. The methodology for identifying these chemicals is only presented in a sketchy way much later in the report. This concern should be addressed up front with a rational approach to screening for these substances. I would suggest that the authors consider a fugacity approach employing a Mackay compartment model (at least to level II) with some thresholds for identifying exceptions. Look at the original Mackay work (e.g., Multimedia Environmental Models-The Fugacity Approach; Lewis Publishers, Boca Raton, Florida, 1991), and also consider recent work like Gouin, T., D. Mackay, E. Webster and F. Wania, ES&T 34 (5) 881 - 884, 2000 to formulate this approach. The report does not make it clear whether or not the spreadsheet tool (U.S. EPA, 1998e in the report) does this already. The authors seem unaware that the word criteria is the plural form, and the word criterion is the singular form. The text uses the plural at all times. This is but one example of how a good editor can improve the document.
Several sources are cited for supplying data missing from traditional databases like IRIS. This was well thought out as a backup. The authors are cautioned in using State of California dose-response data because these are often biased to be more conservative than EPA data. A study commission appointed by the State of California recognized this a few years ago and suggested that they conform with EPA data. Although California has adopted the Sec 112 list of Hazardous Air Pollutants, it still has its own list of slope factors and reference doses. Thus, the commission's recommendations have not yet been fully implemented to the best of my knowledge. One alternative that should be considered is cross assignment based on other health criteria such as MCLs. In the case of lead, the MCL can be cross-assigned to get an RfD. This should be done despite the fact that they will implement the IEUBK model. It should be noted that in this specific case, the authors do not describe how the IEUBK model will be implemented. At low exposure media concentrations this model does not provide output because of code limitations. Straightforward scaling is not possible because of non linearities. Alternative sources for supplying missing data are the Risk Based Concentrations derived by Region III (available by searching in the Regional Office part of www.epa.gov), and the Minimal Risk Levels tabulation compiled by the Agency for Toxic Substances and Disease Registry (available at www.atsdr.cdc.gov/mrls.html). Perhaps these were mentioned in the report, and I missed them.
If the authors really believe the formula in words given on page 2-4, then the "risk criterion" factor cancels out when you substitute directly from the formulas on Table 2-1. It is doubtful that the authors intended this to be the case. If my guess is correct, the problem is fixed by omitting the words "risk criterion" (criteria was meant to be in the singular form, probably) from the formula on page 2-4 because the equations for the risk screening factors already contain the risk criterion.
Derivation of Ecological Screening Factors
This part of the document is extremely well-presented and offers a better level of detail than most ecological risk assessments. Fish seem to come in and out of the picture with scant notice given to the reader. For example, Table 2-7 lists fish, but omits piscivorous birds. Then on the next page both fish and their predators are missing from the diagram. Are fish really in or are they out? They definitely should be in because of their potential for biomagnification of some pollutant concentrations. The table on page 2-48 lists most of the sources we have ever used and then some. Since people are very interested in the welfare of warm furry animals, it might be advisable to add criteria for mammalian and avian species such as the water quality criteria from Sample et, al., 1996 (already referenced in the report, but missing from Table 2-8). They treat 85 chemicals and their impacts on 9 representative mammalian wildlife species and 11 avian wildlife species. Fish and piscivorous birds suddenly reappear at the bottom of page 2-50; therefore, it seems hopeful that they will be counted in the final analysis. The discussions of screening factor formulation are very thorough.
Level of Protectiveness
The three-tier structure involving impoundment, facility and constituent seems like a somewhat redundant overkill in the effort to establish an acceptable level of protectiveness. A single tier system should be tested run in parallel with the proposed three-tier system. The one-tier scheme works like this: Do everything at the impoundment level; establish quantity-distance tables based on engineering and geological judgement to group impoundments where they might act in concert; run the analysis keeping constituents in tabulated detail to keep track of what's going on. Run the proposed three-tier scheme in parallel with this one-level scheme for a trial batch to see whether or not you get any difference in risk distribution. I doubt that there will be a difference. The idea is to take out some of the mumbo-jumbo that could confuse the non-technical audience referred to above.
Another bit of obscuration might be removed by doing away with the margin of protection. In the interest of clarity, the same objective is met by simply redefining the bright line risk that is acceptable. After all, these bright lines are not really carved in stone. This recommendation is in the spirit of clarifying what's being done to the non-technical members of the audience.
Another issue of protectiveness arises in the chemicals designated as exceptions. The procedure for identifying exceptions needs to be described at least briefly early on in the process because these could be the drivers. The multimedia compartment approach suggested above offers one approach to systematizing this decision. I am sure that there are others. Most of the chemicals in this category are the really scary ones in the eyes of the activist community. Along these lines it is advisable to speciate the fate and toxicity analyses where there are multiple forms or multiple congeners. Polychlorinated dibenzo-p-dioxins and -furans, polycyclic aromatic hydrocarbons and coplanar polychlorinated biphenyls are examples of such families. In the cases of the chlorinated organic compounds PCDDs, PCDFs and PCBs, the fate properties are vastly different as are the toxicities. Toxic equivalency factors are available for each of these families of compounds.
Approaches For Dealing With Lack Of Information
It is difficult for me to find an explicit statement of the alternatives spelled out in the Charge to Reviewers. Section 3.4.2 comes as close as I could come up with to a plan for data acquisition, but it falls short of telling the whole story. It promises that "Specific issues associated with this effort are discussed in the following section", but the following section is the reference list. Maybe the report is incomplete, and there really is another section. As mentioned above, the document generally makes adequate provisions for anticipating contingencies. Of the alternatives offered in the charge for supplying missing data, I would vote for using scale factors on data that is available from impoundments representing the same SIC as the unknown case. This is suggested in the spirit of emission factors as defined in AP42 where the factor is derived from actual measurements and then is scaled by some metric of activity or production rate.
Methodology for Representing Cumulative Risks
Again, in an effort to improve clarity, I would seriously question the necessity of keeping the three-tier system proposed for facility, impoundment and constituent levels. Obviously there is a high degree of overlap, and the document's intent is to assure that no high-risk situation is erroneously screened out in Phase I. It was suggested above that a clustering algorithm be developed that aggregates impoundments that are in close proximity. Then each impoundment unit will show a breakdown of constituent risks. Effectively, this kills three birds with one stone and spares the reader the agony of trying to decipher the tangled web of scores that emerge from the three tier scheme. If the agency fears that this is too risky, all it need do is run both methods and test for sensitivity. My guess is that the one tier system will give essentially identical results and will reduce the haze factor to the lay reader.
The authors have made an admirable attempt to use off-the-shelf EPA models. This effects clear economies in dealing with such a large-scale problem, but still places large data demands on the user. Here again, there is an implication that the HWIR work has already developed much of the data needed. The document mentions the "HWIR model", the "3MRA model" and the "HWIR 3MRA model" in various places in the discussion of Phase II. I suspect that these three are one and the same model. It would help if they settled on a single name if this is the case.
Consideration should be given to adding a fugacity-based screening model somewhere in the Phase I decision tree. This would be especially useful in defining the set of exceptions that automatically pass to Phase II based on bioaccumulation or other similar indirect pathway mechanisms that might escape the direct ingestion/ inhalation/ contact approach to screening. As a matter of presentation style, the treatment of exceptions should appear earlier in the document than it does. This reviewer was troubled by the possibility of specific chemicals falling through the crack in the implementation of the basic screening procedure. It wasn't until the part way back in the text about exceptions that these concerns were allayed. Here is a comment aimed at the experienced reader rather than the lay reader.
Another hole that needs filling is the case of acute or catastrophic events. These cases get only three lines of mention back on page 3-66. This is probably like the joke about the drunk looking under the streetlight for the door key that was lost. Sudden failure risk analysis is mostly outside of the EPA repertoire; therefore, it is often overlooked; however, some significant advances have been made in the case of Risk Management Plans required under Section 112 of the Clean Air Act Amendments. One important example is liner failure. Membrane liners can fail because somebody drops a bulldozer blade a little too low during some maintenance operation. Considering the 5000-year time horizon, simple rotting of the plastic or diffusion of organic molecules through the polymer structure are very likely to occur. Seismic events crack clay liners. The methodology needed begins with an event tree analysis that assigns proximate causes and probabilities to each link in the chains leading to failure modes. Failures are quantified by severity (unlike those in health risk assessments, which are essentially quantal). Then each set of consequences is described by one of a family of physical models (several of which are those already chosen for the plan) to generate risk profiles. Unlike the term "profile" as used in the plan report, these profiles display probability versus consequences for each event. As in the case of nuclear waste repositories, the long time horizon brings in factors not usually considered in conventional health and ecological risk analyses, which generally envision exposure durations of no longer than a century.
At the risk of repetition, I offer brief comments on some areas outside the topics highlighted in the charge. The reference list is impressive, but it is worthwhile to devote a few additional words in the text to describing how each reference fits in; however, the brief overview is extremely well done in the model descriptions except for the modeling of soil levels caused by deposition. It was good judgment to omit the large draft compendium of indirect pathway risk assessment methodology in the set of volumes entitled Human Health Risk Assessment Protocol for Hazardous Waste Combustion Facilities because that document has problems that must be resolved in its own peer review. Appendix A could be enhanced by relying more on some references and less on others as detailed above. Also, consider cross-assignments in cases of missing values. Appendix B looks as if the in-house statistical expert was asked to write an essay on weighting techniques. Both in the text and in this Appendix additional wordsmithing should improve the linkage between this analysis and the main theme of the report. Appendices C and D are excellent mines of information about detailed sources of ecological benchmarks and all you want to know about 3MRA respectively. Finally, the site-specific data availability issue needs amplification. Markups on pages supplied with this review provide further comments and suggestions.&