Appendix I describes the methodology used to develop the CCW HBNs and provides the HBNs used in the screening analysis.
Sections 3.4 and 3.5), were much more effective at reducing risk for all constituents; 90th (and 50th) percentile risks with composite liners for landfills were zero 7 for arsenic and metals and
4.4 Variability and Uncertainty
4.4.2 Model Uncertainty
Model uncertainty is associated with all models used in a risk assessment because models and their mathematical expressions are simplifications of reality that are used to approximate real-world conditions and processes and their relationships. Computer models are simplifications of reality, requiring exclusion of some variables that influence predictions but that cannot be included in models either because of their complexity or because data are lacking on a particular parameter. Models do not include all parameters or equations necessary to express reality
because of the inherent complexity of the natural environment and the lack of sufficient data to describe the natural environment. Because this was a probabilistic assessment that predicted
what may occur with the management of CCW under actual scenarios, it is possible to compare the results of these models to specific situations.
The risk assessor needs to consider the importance of excluded variables on a case-by- case basis, because a given variable may be important in some instances and not important in others. A similar problem can occur when a model that is applicable under one set of conditions is used for a different set of conditions. In addition, in some instances, choosing the correct model form is difficult when conflicting theories seem to explain a phenomenon equally well. In other instances, EPA does not have established model forms from which to choose to address certain phenomena, such as facilitated groundwater transport.
The models used in this analysis were selected based on science, policy, and professional judgment. These models were selected because they provide the information needed for this assessment and because they are generally considered to reflect the state of the science. Even though the models used in this analysis are used widely and have been accepted for numerous applications, they each retain significant sources of uncertainty. These limitations are well documented in the model development references cited in Section 3.
Although the sources of model uncertainty in this assessment could result in either an overestimation or an underestimation of risk, the models used in this assessment have been developed over many years to support regulatory applications. As a result, they have been designed to be protective of the impacted populations that they represent. In other words, where simplifying assumptions are necessary, the assumptions are made in a way that will not
underestimate risk.
Assumption of Clean Closure of Surface Impoundments. As described in Section 3.5.1, the surface impoundment model treats a surface impoundment as a temporary waste management unit with a set operational life. At the end of this life, clean closure is assumed; all wastes are removed and there is no further release of waste constituents to groundwater.
Although this simplifying assumption is not consistent with the practice to close CCW surface impoundments with wastes in place, and it limits the length of potential exposure, the peak annual leachate concentrations on which the CCW risk results were based are not likely to be affected. Releases to groundwater are much higher during surface impoundment operation because the higher hydraulic head in an operating impoundment drives wastewater into the underlying soil with greater force than infiltration through the impoundment cover after the impoundment is closed. This higher head results in a greater flux of contaminants to groundwater during the active life of the surface impoundment, especially in unlined units. Thus, even if the post-closure period were modeled, the corresponding results would not be as high as the peak annual leachate concentrations used in the analysis.
Arsenic and Selenium Speciation. Because the models used in this assessment do not speciate metals during soil or groundwater transport, arsenic and selenium speciation in the subsurface is a significant groundwater modeling uncertainty in this analysis. Arsenic can occur in either a +3 (arsenic III) or +5 (arsenic V) oxidation state in groundwater, with arsenic III being the more mobile form. Selenium can occur in either a +4 (selenium IV) or +6 (selenium VI) oxidation state in groundwater, with selenium VI being the more mobile form. Because the soil and groundwater models assume one form for each model run, the risk results presented for
arsenic and selenium were originally based on 100% arsenic III and selenium VI, which is a high-end assumption (i.e., arsenic III has higher risks than arsenic V and selenium VI has higher risks than selenium IV). Although arsenic is generally thought to occur in the +3 form in
leachate, there is evidence from damage cases at CCW disposal sites that suggests that arsenic III is converted to arsenic V during subsurface transport at some sites (see, for example, U.S. EPA, 2000, 2003e; Lang and Schlictmann, 2004; Zillmer and Fauble, 2004). To address the
uncertainty of running the model with 100% arsenic III and selenium VI, the models were also run assuming 100% arsenic V and selenium IV. The results from the two species should bracket the results expected given some mixing of oxidation states.
Bioavailability of Constituents to Ecological Receptors. For the purposes of this analysis, the model assumed that all forms of a constituent were equally bioavailable to ecological receptors, and therefore, the actual exposures that may occur in the field tend to be overestimated, thus making this a high-end assumption. Both the chemical form and the environmental conditions influence bioavailability and ultimately the expression of adverse effects. For example, as discussed above, the form of arsenic has been shown to profoundly influence mobility and toxicity.
Compaction of Landfilled Waste. The source model did not consider potential compaction of CCW waste over time. Such compaction could decrease the hydraulic conductivity and the associated water infiltration. However, no readily available data were identified to support an analysis of the influence of CCW compaction on infiltration rates. The current approach would tend to overestimate infiltration rates compared to a model that would adjust the hydraulic conductivity over time due to compaction. EPA believes this is an
appropriately conservative assumption given the lack of the information needed to accurately model the effects of waste compaction.
Landfills Assumed to be Above Water Table. The landfill source model and EPACMTP assume that the source is above the water table. However, some actual CCW disposal units do extend below the water table. Because waste intersecting the saturated zone may increase groundwater concentrations, the approach may underestimate risk in some cases. However, including this effect would strengthen a general conclusion of the analysis that potentially unacceptable risks exist in some cases with unlined and clay lined CCW landfills.
Indirect Ecological Effects. Indirect ecological effects (e.g., depletion of food resources) were not considered in the analysis. For any given facility, the spatial scale of potential
contamination would affect a very small proportion of the home range for typical species; determining impacts on food supply and habitat quality with regard to the landscape and overall health of the animals is not currently possible in a national-level assessment (and difficult to understand or estimate in the majority of site-specific assessments). In addition, many species are opportunistic feeders and will seek other areas if food sources decline, regardless of the source of the stress to the food supply. For these reasons, EPA does not believe that it is possible to consider indirect ecological effects in a national risk assessment like CCW.
Aquifer pH. As explained in Section 3.4, aquifer pH was used to select the metal sorption coefficients that were in turn used to calculate retardation coefficients for groundwater transport of the CCW constituents. To estimate pH in an aquifer impacted by CCW leachate, the
CCW risk analysis assumed that, after entering the aquifer, the leachate plume thoroughly mixes with the ambient, uncontaminated groundwater. However, because this mixing zone is largely at the periphery of the groundwater plume, thorough mixing may or may not occur at actual sites. The full mixing assumption results in higher receptor point concentrations for most metals, because metal sorption and precipitation tend to increase (i.e., Kd goes up) with higher pH and full mixing tends to reduce the pH of CCW leachate, which is normally alkaline (i.e., assuming full mixing results in a lower groundwater pH and lower sorption).
To assess the effect of this simplifying assumption on the risk results, we compared two landfill Monte Carlo simulations for coal ash waste containing As(III) and coal ash waste containing As(V): (1) the fully mixed aquifer assumption and (2) an assumption that no mixing occurs in the aquifer and the leachate pH is the governing pH for Kd selection. These two metal species were selected because their sorption isotherm behavior with pH change differs; Kds derived from As(III) isotherms tend to decrease as pH increases (which is typical of most metal species examined in the risk assessment), while Kds derived from As(V) isotherms tend to increase with increasing pH.
Percentiles of peak receptor well concentration from the As (III) and As (V) simulations were selected and compared by calculating the percent change with mixing assumption as follows: 100 C C C Mix Full Mix Full Mix No − × = Change % where
CNo Mix = Simulated peak receptor well concentration for a select percentile based on a
no mixing assumption (mg/L)
CFull Mix = Simulated peak receptor well concentration for a select percentile based on a
fully mixed assumption (mg/L)
Table 4-29 compares the percent change in peak receptor well As (III) and As (V) concentrations between the well mixed and no mixing scenarios over a range of peak well percentiles. The results indicate that As(V) has a sensitivity to pH that leads to increased
receptor well concentrations under the no mixing assumption (i.e., when the leachate pH is used to determine Kd in the saturated zone) relative to the well-mixed assumption used in the risk assessment. These results suggest that a change in the complete leachate mixing assumption could raise the receptor well concentrations (and therefore risks) for metal constituents whose Kd values decrease with increasing pH.
Table 4-29. Change in Peak Receptor Well Concentrations for Ash Disposed in Landfills Assuming Leachate Does Not Mix in Aquifer
Percentile of Peak Concentration
Percent Change in Peak Concentration
As(III) As(V) 10 0.00% 0.00% 20 0.00% 0.00% 30 0.91% 0.00% 40 0.25% 0.00% 50 0.31% 2.28% 60 0.00% 15.57% 70 0.23% 57.97% 80 0.00% 18.31% 90 0.00% 11.75%
Goethite Versus Hydrous Ferric Oxide Sorbent. The choice of iron sorbent is
important because goethite is a much poorer adsorbent than hydrous ferric oxide and will result in larger leachate contaminant concentration. With respect to the use of goethite versus the use of hydrous ferric oxide, EPA had discussions with Dr. David Dzombak and Dr. Samir Mathur (developer of the goethite database). In these discussions, the group discussed the sorbent question extensively, and EPA chose to use goethite rather than hydrous ferric oxide as a best estimate that would not underestimate risk. However, because actual CCW disposal sites could have hydrous ferric oxide present in their soils, the risks for arsenic could be overestimated.
Multiple Constituent Exposures. The individual human risk from each CCW constituent was considered separately in this analysis. However, the CCW waste constituent database and recent field studies such as U.S. EPA (2006c) and U.S. EPA (2008c) suggest that exposure to multiple constituents is highly likely. Because multiple constituent exposure may be synergistic depending on the constituents, certain constituent combinations may cause adverse health impacts that a single-constituent approach may underestimate. However, the quantitative human health benchmarks used by EPA are based on the toxicity of individual chemicals. With only one carcinogen present in CCW (arsenic), it was not necessary to add carcinogenic risks. Noncarcinogenic risks can be added only for chemicals with toxic effects on the same target organs, and this could have been done for fish and drinking water ingestion risks by accounting for transit time and adding HQs for contaminants with noncancer effects on the same target organs that arrive at the same time to the receptor point.
However additivity across chemicals was not considered in this risk assessment; neither was synergism or antagonism. Noncancer hazard may, therefore, be under- or overestimated. Nevertheless, risks were high enough from human exposure to single chemicals (notably arsenic, the single carcinogen) that this would not have changed the basic conclusion of the risk
assessment: that there are potentially significant risks to human health from CCW disposal in landfills and surface impoundments.