Document wgRk6ZzLK5yy6NQgVyE496QO6

AR226-2349 U.S. Environmental Protection Agency Draft October 27,2003 EPA Comments on DuPont's Proposed Air Modeling Approach to Verify the Modeled Results, and Use the Results to Extermine Where to Sample Ground Water (as presented to EPA on October 27,2003). EPA has reviewed the DuPont presentation on air modeling verification and has prepared comments-based on our understanding of the proposed approach, as presented on October 22, 2003. DuPont presented their approach to verifiying that their intended use o f ISCST3 will alone adequately serve as a direct pointer for where there would be the most efficient underground water monitoring. A t the October 22,2003 meeting, EPA experts in the Office o f Pollution Prevention and Toxics, the Office o f Research and Development,the Office o f Air Qualily Planning and Standards from Headquarters, Research Triangle Park, North Carolina, Ada, Oklahoma, and Athens, Georgia all stated that they believe soil sampling, especially in th first few inches above the area the ISCST3 suggests highest air levels, is a more reliable means o f verifying that the air model is accurately predicting where to sample ground water than air sampling alone. DuPont stated it intended to use the verification, i.e. limited air sampling, at one place (Washington Works) to suffice for verifying the model's applicability elsewhere without further site specific verification via additional sampling. EPA's experts disagreed with this approach, stating that the model would need to he calibrated to the other sites. Our comments are grouped into three sections in this document: 1) Comments on DuPont's proposal to evaluate the ISCST3 model to determine maximum air; 2) Comments on DuPont's suggested proposal for using the modeled air concentrations to determine where ground water monitoring will be performed; and 3) Comments on DuPont's modeling/measurement approach. The Agency's comments appear below. ' 1. DuPont's proposal for evaluation o f ISCST3 model to determine maximum air concentrations (emphasis on particles) Use 1SCST3 to develop concentration gradients (based on annual data) . M onitor at 5 locations within fenceline (cardinal ordinates plus two locations with highest modeled concentration) . . Collocated air sampler at highest location to provide estimate o f sampler precision One field control sample for a total of 7 samples each period to estimate error associated with blank m aterial M onitor for 7 different periods over 6 weeks Sampling inlet approximate to breathing zone . Sample collection protocol following NAMS/SLAMS every sixth day timeframe - Analyze samples and provide data to EPA within approximately 18 weeks Potential issues Sampling regime needs to be revisited to ensure that sample collection occurs during periods where maximum concentrations are anticipated at the collocated site. Since this ASH027X65 EID780445 E ID 780445 U.S, Environmental Protection Agency Draft October 27,2003 effort is to evaluate the usefulness o fthe model in lieu o f status and trends, sampling back-to-back days might be more appropriate in lieu of every 6* day sampling. Sampling on a fixed every sixth day regime may result in sampling during rain events or other conditions that might result in increased numbers of-sampling non-detects Sampling on a fixed calendar basis in lieu o f a meteorological/industrial process basis might result in sampling occurring when the plant emissions are lowest or when meteorological conditions are not favorable related to the fixed sampler layout The proposed activity is intended to provide model evaluation and not status and trends data. The sampling design and schedule needs to be reconsidered to ensure it focuses on the primary objective. Sampling and analyzing a couple o f back-to-back maximum day events early would provide early input for the usefulness o f this activity, and opportunities for adjustments, in lieu of sampling 7 events and then trying to interpret the completed results Sampling within the fenceline is a good first approach, but it does not provide the corresponding data to evaluate the usefulness o f employing this model for assessing exposures outside the fenceline The possibility o f long range transport o f particulates is not considered with this approach. Field and lab spikes should be included to demonstrate analyte recoveries and potential interference from handling, storage, and shipping. . If the ambient air model results presented are an accurate representation o fthe expected air concentration field, there is a good probability that the size and nature o f the air monitoring data set proposed will be inadequate to verify the model predictions. Sites 2 and 3 are duplicates, and sites 4 ,5 and 6 would be expected to have little or PFOA in the air. Consequently, the ability to estimate magnitudes o f concentration are based primarily on repeated measures in time at two sites, not 6. A total o f 35 samples is characterized as a very large number o f samples for the analysis but this number is only large when taken as a whole: However, they will only be sampling from 35 locations at selected company boundary locations according to a compass points (N-S-E-W). Therefore, they will only have seven data points per location and no data beyond the facility boundary. It would appear that for their statisticians to validly reject the null hypothesis (i.e., accept that model results "match" measured samples) would require several additional sampling locations at greater distances than those proposed and would require seasonal effects be considered as well (e.g., wind speeds, directions, precipitation, barometric pressure changes, etc.). This would greatly delay final study results because data collection could be extensive, could greatly delay obtaining final study results because data collection would essentially take a year to complete. However, they are asking EPA to accept their conclusions to what appears to be an inadequate study. Additional information on the statistical analysis that will be utilized in this study would be helpful in understanding the power o fthe analysis. The emissions estimate will be based on production volume capacity data. This is an important parameter. Additional information is needed to understand how well the capacity data predict the actual emissions, and to determine whether this is an acceptable ASH027X66 EID780446 E ID 780446 U.S. Environmental Protection Agency method for this purpose. Draft October 27,2003 2. DuPont's Suggested Proposal for Using the Modeled A ir Concentrations to Determine Where Ground W ater Monitoring W ill Be Performed Using the W ashington Works data, DuPont observed that there was an apparent correlation between the modeled (sampled) air concentration data and the measured groundwater results from the extensive network o f wells at the Washington Works facility DuPont suggests, once the ISGST3 model is evaluated for the Washington Works, that this model could be used at the other facilities to estimate ambient air concentrations without the need for confirmatory air measurements . DuPont suggests that the modeled air concentrations (based in part on measured data) could be used as a first assessment for selecting areas for conducting ground water m onitoring Use this methodology to determine if and when to sample groundwater at all 4 fluoropolymer manufacturing facilities. Potential Issues The initial assessment o f the data provided by DuPont does not allow EPA to directly assess the correlations/relationships among the Washington Works air, soil, and groundwater measurements (modeled values). This may result in EPA not fidly . . understanding die tim ing and sequencing o f the air, soil, and groundwater sample results, Le., when were the air samples collected in time relationship with the water samples; what data is available to demonstrate the phased or total wash-out of the PFOA from the soil to the groundwater. . Groundwater samples containing some of the higher concentrations o f PFOA were collected in locations where,air concentrations were predicted to be relatively low. Likewise, some groundwater samples containing some of the lower concentrations of PFOA were collected in locations where air concentrations were predicted to be relatively high. Factors associated with each industrial plant/location (meteorological conditions, industrial processes, sampling site plan, soil type, etc.) may adversely influence the key modeling parameters if model sensitivity testing and/or some confirmatory monitoring is not performed to evaluate the robustness o f the model for each specific site. Sampling the ambient concentrations without parallel results from other key environmental parameters does not allow the direct assessment o f the usefulness o f using the air concentration data to directly determine locations for ground water sampling The ambient air modeling, as presented at the 10/22/2003 meeting, is insufficient to justify or support a monitoring program whose goal is to establish the source o f contaminated ground water in the vicinity o f the Washington Works manufacturing facility. Without ASH027167 EID780447 EID 780447 U.S. Environmental Protection Agency Draft October 27,2003 - surface soil sampling to establish that PFOA is derived from an aerial source, the actual PFOA source o f contaminated ground water would remain an uncertainty. Using some variation o f the air dispersion model to locate monitoring wells may be useful, but not in the manner which has been proposed. No scientific justification was presented at the 10/22/2003 meeting as to why air concentration is, in their opinion, a good predictor o f ground-water concentration for PFOA. They noted a coincidental co-occurrence of predicted contaminated ambient air and contaminated ground water, but has not established the source o f PFOA in ground water. There was no technical information or discussion of a route o f transport from the contaminant source (alleged to be Washington Works) to the ground water (in the Little Hocking well field) which would circumvent transport through the soil, or unsaturated zone consistent w ith their assertion that ambient air concentrations are better predictor o f the location o fcontaminated ground water than soil concentrations. Lack of detections in soil samples cited in support o f their position that soils should not be sampled, can be ^ logically attributed to field methods used to sample soils which did not target surface soil samples. . The map presented at the meeting represented the estimated average annual ambient concentrations for only a single unspecified year (2002?). They indicated verbally that other years showed veiy different patterns, perhaps due to mitigation?? They provided no information on how model inputs reflected that mitigation or if wind patterns change substantially from year to year (very possible). Note also that no information was presented on the stack heights and historical discharges, the discharge rate, or the pattern o f particulates deposited in relation to total (gaseous plus particulate) PFOA concentrations. It is impossible to know the validity of the air modeling with virtually no information on the inputs presented. Year-to-year as well as or day-to-day variability ia ambient air concentrations was hinted at but not demonstrated. One can speculate that high annual mean air concentrations around the plant would be associated with stationary air, or temperature inversions, rather than with atmospheric conditions likely to move the chemical offsite. ^ The variability in deposition pattern of PFOA associated with stock emissions would be expected to be a critical parameter in determining the source of PFOA contaminated ground water (presuming the Washington Works is the sole PFOA source). Neither the proposed air sampling nor the annual concentrations presented would be adequate to Hetermine the source with adequate certainty. Th air dispersion simulations should be altered to consider variations in emission rates with varying meteorolgical conditions and then dispersion model should be used to determine where soil sampling should take place. The soil sampling could then be used for well location. Using the yearly average meteorological conditions may not actually provide a location o f maximum deposition. The location o fmaximum deposition will depend on a combination o f simultaneously occurring processes,'namely the emission rate and the prevailing meteorological conditions that occur during that particular emission. If emissions are constant throughout the year, then perhaps the average yearly meteorological conditions may be appropriate. However, if there are substantial day-to-day or seasonal variations in the ASH027168 EID780448 EXD780448 U.S. Environmental Protection Agency , Draft October 27,2003 emission rate throughout the year, then using the average meteorological condition would be inappropriate. It such a case, it would be preferable to create daily deposition contour maps each using the daily emissions rate for individual emission sourcs and the corresponding daily meteorological. It may also be worthwhile to check evening vs. day meteorological conditions and day vs. evening emission rates. Integrating these maps would create an overall contour map, such that a more appropriate location for maximum deposition could be obtained. The location determined by these integrated maps should be used to give an indication o f where soil sampling should take place. Soil sampling should occur across a large enough transect that the maximum location o f deposition could be determined (assuming that there is a single rea o f maximum deposition). Wells could then be located in this area. Note that soil sampling across a wide swath should be much more economical than the expense o f drilling a well in an inappropriate area. ASH027169 EID780449 EXD780449 U.S. Environmental Protection Agency 3. Comments on the modeling/measurement approach. Draft October 27,2003 DuPont suggests a 0.1 microgram/cubic meter detection rate based on a 1 liter/min flow rate. This would result in ~150 ng o f analyte if fully extracted: 0.1 microgram/m3 * 1 liter/m in *1 m3/1000 liters * 1440 mins in a day = 0.144 micrograms 0. 144 micrograms * 1000 nanogratn/1 microgram = 144 nanograms O fcourse, there are dilution and matrix factors that influence the actual extract concentration, 1. e., the final volume the extraction was reduced down to foT the purposes o f analyses. Potential Issues . This would suggest ~150 nanograms are available for analysis EPA experiences picogram levels of detection with LC/MS technology, additional orders o f magnitude in sensitivity. However this difference may result from differences in EPA understanding in the instrumentation and estimates o f precision and accuracy associated with both the sampling and analytical instrumentation. ASH027170 EID780450 EID 780450