The following editorial suggestions and clarifications were made:

"Page v, Tables 8 and 9: Change 'Form' to 'From.'"
"Page 5, line 14: Change 'a' to 'at.'"
"Page 5, lines 18-20: Recommend delete last sentence" because military training exercises at Vieques are scheduled to cease in May 2003.
"Page 15, paragraph 3, line 2: Recommend change 'marine' to 'Naval.'"
"Page 18, line 19: Change April 19 to April 20. Exercises did occur on April 19, 1999."
"Page 24, line 11: Change 'D.' to 'G.'"
"Page 44, Section V, Line No. 23. '...no exposures would occur that present a public health hazard' should read '...no exposures would occur that present a public health hazard.'"

Several minor editorial revisions were made in the PHA to address this comment.

"Recommend replace 'practice bombs' with 'non-explosive ordnance' and replace 'live bombs' with 'explosive ordnance.' Recommend universal change throughout document."

ATSDR recognizes that military personnel and civilians often use different terminology when describing military training exercises. While we try to be sensitive to differences in terminology, we also strive to prepare documents that communicate information in language that the public appreciates and understands. After working with community members on Vieques and reviewing local media accounts of the Navy's activities, ATSDR decided that the public–the intended audience of the PHA–is more familiar with the terms practice bombs and live bombs, rather than the terms non-explosive ordnance and explosive ordnance. Because of this, the terms practice bombs and live bombs are used consistently throughout the PHA. To highlight the different terms used to characterize the exercises, the PHA includes a text box titled "Terminology Used in this PHA to Characterize Military Training Exercises." Because we still believe that practice bombs and live bombs are the most appropriate terms to convey different military training exercises to the public, we have not made any changes in the PHA to respond to this comment.

"Page 63, line 21: Incorrect statement regarding use of chaff. The use of chaff is prohibited over Vieques as well as the adjoining Warning and Restricted Areas (W-428 and R-1704)."

The comment addresses a sentence in the public comment release PHA in which ATSDR recommended actions to characterize potential exposures to chaff during future military training exercises. Since releasing the public comment draft, however, the Navy has ceased its military training exercises at Vieques altogether and there is no longer a need to characterize potential future exposures to chaff at the island. As a result, ATSDR has deleted the sentence of concern from the final version of the PHA.

"This is a well written document that provides detailed information in regards to the evaluation of the air exposure pathway. The findings and recommendations from ATSDR are clearly stated and fully justified. Although all studies are not complete, the available environmental data is of sufficient quality for ATSDR to make a valid public health assessment for the air exposure pathway at this time."

No changes were made in the PHA to address this comment.

"The section describing Demographics on page 12 would be more complete in regards to the census data shown in Table 1, if the increasing elderly population, decreasing number of children, and decreasing number of women of child-bearing age were mentioned."

ATSDR revised the section on Demographics (Section III.B) to indicate the population trends mentioned in the comment. We also added statements comparing demographic data for Vieques to that for the entire Commonwealth of Puerto Rico.

"The information in Tables 4, 5, and 6 would be more descriptive if standard deviations or standard errors of the mean for the average concentrations as well as the number of samples were provided in each table, as appropriate."

ATSDR made several changes to the tables in question. First, we updated the tables based on the latest ambient air sampling data available from PREQB. The tables now reflect sampling data PREQB has reported to EPA through March 2003. Second, we included standard deviations in Table 6, which summarizes the particulate sampling that PREQB has conducted to date. We did not include standard deviations in Tables 4 and 5, as these tables present estimated exposure concentrations for a highly ideal scenario. Specifically, the scenario assumes that all airborne contaminants in the residential areas of Vieques originate from the bombing range. This scenario was used to illustrate the upper-bound impacts that the military training exercises using practice bombs might have had on air quality at the island. ATSDR does not believe that presenting standard deviations for upper-bound estimates adds any additional insights into this evaluation.

"The data used in describing the air contaminants are the ones collected by PREQB since 2000. In relation with the other reviews of air sampling studies, ATSDR believes that the quality is questionable."

Appendix C presents ATSDR's review of air sampling studies that have been conducted to date at Vieques. The comment implies that we had questions about the quality of PREQB's particulate sampling data collected since 2000. To the contrary, Appendix C.1 states: "For several reasons, ATSDR believes the data collected by PREQB are of a known and high quality." To respond to this comment, we have added several statements to the main body of the report that echo this finding.

For reference, Appendix C reviews two additional sampling studies conducted by PREQB during the 1970s (see Sections C.4 and C.5). We concluded that the quality of these studies cannot be assessed, because original documentation for these sampling studies could not be located.

"No data is available related to PM2.5. 'EPA proposed regulating ambient air concentrations of PM2.5 in 1997 based on evidence linking inhalation of fine particles to adverse health effects in children and other sensitive populations.'"

The comment correctly notes that EPA has recently placed a greater emphasis on measuring the levels of fine particulate matter (or PM2.5) in the air that people breathe, due to the findings of health studies linking exposure to these particles and adverse health effects. ATSDR does not necessarily view the lack of PM2.5 sampling data as a shortcoming in this evaluation, based on the results of our modeling analysis, which found that the military training exercises at Vieques using live bombs had only minimal air quality impacts in the residential areas of the island. Specifically, the modeling indicated that the exercises using live bombs caused annual average PM10 concentrations and maximum 24-hour average PM10 concentrations to increase by 0.04 µg/m³ and 10.2 µg/m³, respectively. Even if we make the worst-case assumption that all particulate matter emitted from the exercises was in the form of fine particulates, such incremental increases in PM2.5 are still considerably lower than EPA's current health-based standards (15 µg/m³ for annual average, and 65 µg/m³ for 24-hour average concentrations). Therefore, the modeling data indicate that air emissions of particulate matter, whether coarse or fine, from the military training exercises were not at levels of health concern for the residential areas of Vieques. We have added a few sentences to Section V.C of the PHA to explain why we think that ambient air concentrations of fine particulates were not of health concern.

"The air sampling does not specify the time intervals of the readings (5 min., 10 min.,?)."

Appendix C of the public comment release of the PHA presents a detailed summary of the ambient air sampling studies previously conducted at Vieques. The appendix indicates that PREQB has been collecting 24-hour average air samples for particulate matter at Vieques in its most recent sampling efforts. This averaging time is entirely appropriate for characterizing air quality impacts for particulate matter and is routinely used at monitoring stations throughout the United States. Further, this averaging time is the basis of EPA's health-based standards for particulate matter. We have added text to the main body of the report noting the averaging times for PREQB's air sampling.

"The model used by ATSDR to study the air quality does not specify the emissions punctually and the meteorological characteristics. This is due to lack of information and working with proxy. This makes the model weak due to lack of standards."

ATSDR considered many different approaches for how to model air emissions from the military training exercises. The comment correctly states that our modeling analysis did not account for the specific time of day and day of the year on which each detonation occurred at the bombing range. This fact does not make the modeling "weak," as the comment suggests. To the contrary, our modeling approach accounts for realistic worst case scenarios, as follows:

• We estimated maximum 24-hour average concentrations by reviewing range utilization statistics and meteorological data to identify the day with the most intense level of bombing activity and the day with the least favorable meteorological conditions (i.e., the conditions that would cause the highest levels of air pollution in the residential parts of the island). We assumed that the highest level of bombing activity occurred on the day with the least favorable meteorological conditions. Even when we added the air quality impacts from such an upper-bound scenario to the highest background concentrations recorded to date, the exposure levels predicted for the residential population were still considerably lower than levels of health concern. If we had modeled "the emissions punctually," as the comment suggests, we would have predicted lower air quality impacts than what is currently documented in the PHA.
• We estimated annual average concentrations by assuming that military training exercises, over the course of a year, did not preferentially occur during any specific meteorological conditions. This approach is sensible, given that the Navy scheduled the dates of its training exercises months in advance. A modeling approach that considered "emissions punctually" would likely generate only marginally different answers. However, marginal differences in the modeling predictions would not change our overall conclusions, given that the estimated annual average concentrations were orders of magnitude lower than levels of health concern.

Overall, we believe our modeling analysis was scientifically sound and provided reasonable upper-bound estimates of air quality impacts. We have not made any changes in the PHA in response to this comment.

"ATSDR (page 47) points out the following observation: 'sources other than those related to Navy training exercises contribute to actual ambient air concentrations of metals during military training exercises.' If this is true, why hasn't ATSDR evaluated this situation?"

The PHA was prepared to address a specific community concern about potential air quality impacts from the Navy's military training exercises at Vieques. Accordingly, the PHA appropriately focuses on this emissions source. When evaluating air exposures, however, ATSDR recognizes that releases from numerous emissions sources contribute to exposure. Unfortunately, a detailed emission inventory is not available for Vieques. However, ATSDR implicitly evaluated the contributions from other sources in our review of PREQB's ambient air monitoring data for particulate matter. Those data reflect the contributions from all local sources and show that residents are not being exposed to particulate matter at levels of health concern.

"Dr. Massol in the 'Ecological assessment of heavy metals in Vieques, Puerto Rico' found: 'excessive levels of Pb and Cd were found in plant species of agricultural production such as squash, chili pepper, pigeon peas, pineapple and yucca; only the plants of guamá and mango trees demonstrated acceptable levels for these toxics. The most affected species were those with shallow root systems, such as chili peppers, pigeon peas, squash and pasture grass. This is consistent with the thesis that heavy metals might have been deposited in the civilian area through air dispersion by windblown dust from the bombing zone' and 'based on ATSDR methodology, our studies found that five seeds of the pigeon pea will provide enough Cd to surpass the oral health guideline for children.' This data is not in accordance with ATSDR findings based en la Wind Rose from US Naval Station Roosevelt Roads."

The comment refers to a study that evaluated levels of contamination reported for plant tissues in the residential areas of Vieques. ATSDR reviewed the original report (Massol and Diaz 2001), which is a study conducted by researchers from Puerto Rico. The study, which has not been published in the peer-reviewed literature, reports elevated levels of several metals in the tissues of several plant species collected from the residential areas of Vieques.

ATSDR critically reviewed this study when preparing our PHAs. Given the implications of the study, we asked Dr. Rufus Cheney to review the study as well (Cheney 2002); Dr. Cheney is a premier research agronomist with the U.S. Department of Agriculture who has dedicated much of his professional career to examining plant uptake of environmental contaminants. Both ATSDR and Dr. Cheney have expressed serious concerns about the quality of the data reported in the plant tissue sampling study cited in the comment. At the heart of the concern is the fact that Dr. Massol's study presents no information on quality assurance (QA) and quality control (QC) procedures. Laboratories that routinely collect and analyze plant tissue samples typically report numerous QA/QC findings to convince the reader that the sampling results are valid and free from bias. Examples of such findings include results from duplicate samples and analyses of Standard Reference Materials. Without any such information documented in Massol's study, ATSDR can not be assured that the analytical methods have been applied correctly, and Dr. Cheney concluded that the study is "of suspect quality and all concerns based on these analyses must be rejected until proper analysis of appropriate samples are available."

In addition to our concerns about the lack of QA/QC information, ATSDR has serious questions about the underlying premise of the study–that air contaminants from the bombing range are reportedly being found in plant tissues more than 9 miles downwind from the source. The study does not clearly indicate the mechanisms by which the reported uptake is occurring (i.e., by root uptake or by deposition of particles onto plant surfaces). Researchers investigating these mechanisms would typically collect soil samples to investigate the matter further; yet Dr. Massol's study apparently did not include a soil sampling component, making the results (if one assumes they are correct) difficult to interpret. More importantly, ATSDR emphasizes that the available soil sampling data provide no evidence of contaminants from the bombing range accumulating in the soils at locations far from the source (ATSDR 2002). The fact that there is limited evidence of considerable environmental impacts in the residential area of Vieques raises further questions about the validity of the data in Massol's study.

"The study shows significant limitations such as: time of sampling, data related to PM2.5, information about other sources of pollution."

The comment identifies three potential limitations in the PHA. Our responses to comments #8 and #11 address the topics of PM2.5 and other sources of pollution on Vieques. As those responses indicate, we view neither aspect as a significant limitation in the PHA.

Regarding the time of sampling, ATSDR based its conclusions on all of the information we gathered for this site, which included valid, representative sampling data from July 2000 through the present. Although we would have preferred to base conclusions for the time when the Navy used live bombs on ambient air sampling data, the fact remains that limited data are available for that time frame. The data we did obtain from the 1970s showed no evidence of unusually elevated levels of air pollution, but those data are of questionable quality, because original documentation for PREQB's sampling studies is not available. While the absence of sampling data is unfortunate, it does not preclude us from making scientifically defensible conclusions about this site. In fact, ATSDR's own guidance manual indicates that modeling results should be considered in cases where sampling data are incomplete (ATSDR 1992). Therefore, the approach we took–using modeling analyses to estimate exposures for times when samples were not collected–is not only scientifically defensible, but consistent with our internal guidance.

"The isolated form of the presentation does not allow straightforward interpretation of the problem. It is important to present all the pathways integrated in the same health circle, to know better how is the integration of all information."

Early in the public health assessment process for this site, ATSDR decided that releasing a series of documents that address individual exposure pathways would provide the most timely release of information to the petitioner. We are currently compiling a summary document that integrates the findings from all of our previous PHAs (ATSDR 2001, 2003a, 2003b).

"Risk factor evaluation and their implications is the methodology that should be used in the pathways and the study is limited in that way."

ATSDR evaluated the public health implications of exposure using well-established methodologies outlined in our Public Health Assessment Guidance Manual (ATSDR 1992). This approach was developed and reviewed by highly experienced environmental health professionals not only from ATSDR, but also from several other public health and environmental agencies. The approach includes several measures to ensure that health assessors make defensible conclusions that are protective of public health. We do not believe following the approaches described in our Public Health Assessment Guidance Manual is a limitation for this PHA.

While we acknowledge that other approaches are available to assess the health implications of potential exposures, it is important to remember that the estimated air quality impacts from the military training exercises at Vieques were so low that even highly sensitive air sampling devices likely would not have been able to measure them (e.g., see Tables 8 and 9). We doubt that any sensible health assessment approach would find legitimate public health concerns based on the minimal exposure levels that we predicted.

"ATSDR states that the PHA was prepared pursuant to the Comprehensive Environmental Response, Compensation, and Liability Act (CERCLA) Section 104(i)(6) and in accordance with their implementing regulations, found in 42 CFR Part 90. These regulations clearly specify that when information necessary to perform a health assessment is not readily available from other sources, ATSDR may arrange for sampling or additional data gathering. (See 42 CFR Part 90.8c.) ATSDR did not follow its own implementing regulations by not gathering any additional environmental data and relying solely on insufficient information and statistics. Public health implications cannot be determined if there is insufficient information on concentrations, frequency and duration of exposure."

ATSDR strongly disagrees with this comment. When preparing the Vieques PHAs, we made extensive efforts to gather all existing environmental and health data from multiple parties, including PREQB, PRDOH, EPA, several researchers from universities and private organizations in Puerto Rico, and the Navy and its contractors. Additionally, we conducted an extensive fish tissue sampling study, helped organize an expert panel committee among international experts in cardiology and epidemiology, performed a rigorous air dispersion modeling analysis, and initiated air sampling events. We believe that the existing information gathered and supplemental information we produced were sufficient to reach scientifically defensible public health conclusions. Additional information on the available data follows:

• For exposures from July 2000 to the present, we based our conclusions largely on the results of several hundred valid particulate air samples that PREQB collected in the residential areas of Vieques. These samples measured the air that residents actually breathed. To date, not a single sample has revealed particulate matter levels of health concern.
• For exposures prior to July 2000, the available sampling data are more limited. We did identify three studies that measured air pollution levels during the time when the Navy used live bombs during military training exercises. The air concentrations reported in these studies are clearly not at levels of health concern. However, the quality of the sampling data from these studies is not known.

  The comment suggests that we should have conducted sampling to characterize exposures during this time frame. Practically speaking, this was impossible: the Navy had already stopped all live bombing exercises at Vieques before ATSDR was petitioned to evaluate this site. However, we remained committed to evaluate past exposures and conducted a rigorous dispersion modeling analysis. This was the most feasible and defensible approach to evaluate past exposures.

Overall, ATSDR strongly believes the PHA is based on the best available information on environmental contamination levels at Vieques. We gathered as much existing data as we could find and used defensible modeling techniques to fill data gaps, and none of the public comments identifies available data sets that were not considered.

"Another major deficiency with the PHA is in the analysis of past levels of air contamination. Air samples that were collected on Vieques when Navy used live bombs are poorly documented; therefore, no reliable measurements of past levels of air contamination are available."

ATSDR used standard, defensible approaches to evaluate past levels of air contamination. For the Vieques site, we were asked to evaluate an emissions source that was no longer operating (i.e., military training exercises using live bombs). In cases such as this, we first try to gather existing sampling data and then use other approaches, as necessary, to fill data gaps. Although it is unfortunate that more sampling data were not collected at Vieques before ATSDR was involved with the site, we do not view the lack of extensive sampling data as a "major deficiency." Rather, our modeling results and the limited available sampling data both suggest that the Vieques residents were not exposed to air pollution at levels known to be associated with adverse health effects.

"Also, the PHA distinguishes between times of activity (decades) for purposes of evaluating exposure. Archive research conducted by ATSDR could not produce an accurate date of the first air-to-ground military exercise in Vieques. On page 16 of the PHA, the Agency states that 'none of the reports ATSDR has obtained documents exactly when first air-to-ground exercises took place on Vieques.' Therefore, exposure evaluation based on this research is also inaccurate because ATSDR cannot conclusively determine whether harmful effects occur if a determination or assumption regarding contacts with hazardous substances is based on this information."

The comment correctly points out that we were unable to determine the first date on which air-to-ground military exercises occurred on Vieques, but the comment incorrectly infers that this is a critical limitation in the PHA. Our health conclusions for the time when the Navy used live bombs at Vieques are based on the outputs of our modeling analysis (i.e., the concentrations in Tables 8 and 9). These concentrations are our estimates of chronic exposure levels, based on range utilization statistics for the time frame when the Navy's live bombing exercises at Vieques were most intense. We believe these concentrations are reasonable estimates of exposures between the early 1970s and 1999, which is the time frame with the most extensive use of the bombing range (see Section III.D).

For years prior to the early 1970s, we cannot make precise estimates of exposure concentrations. However, multiple accounts we reviewed indicate that the frequency and intensity of military training exercises at Vieques was much lower prior to the early 1970s than during the time frame we considered for the modeling, primarily because the Navy previously conducted much of its training exercises on the island of Culebra. Because training exercises were far less frequent prior to the early 1970s, it is reasonable to conclude that air emissions and any associated exposures prior to the 1970s were notably lower than those that occurred for the time frame we evaluated. In other words, exposures that occurred prior to the early 1970s were likely much less than the exposures we predicted in our modeling (i.e., the concentrations in Tables 8 and 9), and therefore also were far lower than levels of health concern. Therefore, while we cannot pinpoint the exact date when potential exposures might have begun, we are confident that the magnitude of potential exposures was considerably lower than levels associated with adverse health effects.

"If air-to-ground bombing accounts for the greatest proportion of high explosives used at Vieques, and none of the reports ATSDR has obtained documents exactly when the first air-to-ground exercises took place on Vieques, then ATSDR cannot present an accurate assessment of potential risks to human health pursuant to CERCLA 104(i)(6). On a similar note: complete statistics of range use are only available for exercises conducted between 1983 and 1999. Statistics for exercises conducted by foreign countries are also unavailable. If the type of ordnance used during military training exercises is one of the most important indicators of the amount of contaminants released into the air, and these statistics are not available for a considerable period of time, then exposure evaluation is in effect deficient."

The comment correctly indicates that statistics on the amount of ordnance used during military training exercises is a critical input to the air exposure pathway evaluation; however, ATSDR does not agree that the available information are insufficient for performing a public health assessment. Following is ATSDR's response to specific issues raised in the comment:

• Reliance on data from 1983 to 1999. The comment correctly states that the air modeling analysis in our public comment release PHA is based on a review of range utilization statistics from 1983 to 1999. ATSDR has since consulted reports that document range utilization statistics back to 1974 (TAMS 1979; Ecology and Environment 1986). The statistics ATSDR has obtained for earlier years are consistent with the statistics used in our modeling analysis in the public comment release PHA. For instance, the annual average range utilization statistics for 1974-1998 indicate that, on average, 441 tons of high explosives were used at Vieques per year. ATSDR's modeling analysis examines air quality impacts for a base year in which 458 tons of high explosives were used. Therefore, ATSDR's modeling analysis likely does not systematically understate annual average concentrations over the long term. Several statements in these reports (TAMS 1979; Ecology and Environment 1986) also confirm that the range utilization statistics we considered for evaluating short-term exposures are representative of activities that occurred at Vieques in the 1970s and 1980s.
• Date of the first exercise. This comment also suggests that our lack of knowledge of the exact date when air-to-ground bombing commenced is a critical data gap in the PHA. ATSDR again disagrees. As our response to the previous comment indicates, we are confident that the magnitude of potential exposures prior to the early 1970s was considerably lower than levels predicted by our models, even though we cannot pinpoint the exact data when air-to-ground exercises began. ATSDR is confident in this statement because our modeling analysis is based on the time frame when air-to-ground bombing activity was most intense. Finally, ATSDR notes that none of the reports we reviewed or public comments we received indicate, or even suggest, that emissions from the Navy's military training exercises at Vieques prior to the early 1970s were greater than the emissions considered in our modeling analysis. For these reasons, ATSDR considers the lack of information on the first day of bombing activity to have virtually no bearing on the conclusions of this PHA.
• Ordnance used by foreign countries. The comment suggests that our evaluation did not consider ordnance used by foreign countries at Vieques. The reports ATSDR recently accessed provide some perspective on this matter. These reports note, for example, that Germany, Great Britain, Canada, and Holland all have conducted ship-to-shore exercises at Vieques in the past. One report lists the specific ordnance types that these countries used, and several of these ordnance types are also accounted for in the range utilization statistics tabulated in the reports (e.g., TAMS 1979). Therefore, the past reports suggest, though do not clearly state, that the range utilization statistics do include some portion of the ordnance used by foreign countries. Furthermore, the reports state that foreign ships previously accounted for only 8% of the days with training exercises (TAMS 1979). Therefore, even if the statistics do not include every ordnance item used by foreign countries, this omission would account for a very small proportion of the range utilization statistics and would not have a strong influence on our overall predictions, especially considering that the estimated air quality impacts for most contaminants were orders of magnitude lower than levels of potential health concern.

In summary, ATSDR considered a large volume of range utilization statistics for the years 1983-1998 when preparing the public comment release PHA, and we have since acquired supplemental data that document ordnance usage from the early 1970s through the early 1980s. All data we have compiled to date indicate that the ordnance usage rates used in our modeling analysis are representative of actual conditions at Vieques during the years when military training exercises occurred most frequently. ATSDR believes that the available data are sufficient to make public health conclusions.

"On page 26, the PHA says that contaminants will disperse greatly over the 7.9 miles that separate the LIA and residential areas. These contaminants including heavy metals will then accumulate in food items such as vegetables and fruits. In fact, recent studies indicate that high levels of metals are present in plant tissues examined from hills bordering Navy property (Camp Garcia) in Eastern Vieques. By addressing primarily on the issue of direct inhalation exposure to air contaminants, the PHA is limiting the exposure pathway analysis by not adequately considering indirect exposure (accumulation on other media or bio-accumulation) to air contaminants."

The comment raises the possibility of Vieques residents being indirectly exposed to air contaminants (i.e., the residents possibly being exposed to contaminants that settle from the air onto soils and plants). ATSDR considered this possibility in our other PHAs, which examined levels of contamination in groundwater, soil, fish, and shellfish (ATSDR 2001, 2003a, 2003b). We have yet to see any evidence of indirect exposure to air contaminants occurring in appreciable amounts. Moreover, as our response to Comment #12 discusses in far greater detail, we have not seen any reliable evidence of heavy metals accumulating in plant tissues in the residential areas of Vieques, as the comment suggests. To address the general concern regarding consideration of exposures to contaminants in multiple media, we are currently preparing a brief summary report that presents our final evaluation of contamination levels in all environmental media we have studied to date. We expect to release the summary report before the end of 2003.

"A cumulative effect should be considered. The National Ambient Air Quality Standard (NAAQS) does not specify if the parameters are for one event or for years of exposure. Even though the data available does not exceed the NAAQS, it should be accounted the fact that the Vieques population has been exposed to these concentrations for many years."

The comment suggests that the PHA does not consider the possibility of effects resulting from many years of exposure. ATSDR disagrees. When evaluating the public health implications of exposure to air pollutants, we considered both chronic and acute exposure scenarios. None of the concentrations we estimated were at levels that would raise concern for exposure, whether over the short term or the long term. In fact, the estimated concentrations for the majority of pollutants were orders of magnitude lower than levels of health concern.

When mentioning the NAAQS, the comment specifically addresses our interpretation of particulate matter levels at Vieques. In response, ATSDR notes that all measured and modeled data compiled to date quite clearly indicate that ambient air concentrations of particulate matter at Vieques have been considerably lower than levels of health concern. Moreover, the concentrations are not unusually higher or lower than what we would expect to see for communities with similar population levels. We do not believe that prolonged exposure to the particulate matter concentrations typically observed at Vieques (and at locations throughout the United States) should be viewed as a public health hazard.

"A comprehensive study that includes all types of airborne material that could affect the Vieques community should be analyzed. Taking into consideration, all together, PM10, TSP, PM2.5, metals, and explosives. The breathable airs received by the community members include a mix of all the contaminants present."

The comment suggests that a comprehensive study considering a wide range of contaminants should be conducted. We believe our PHA was such a study. For the last 3 years, we have evaluated the air quality impacts of nearly 100 contaminants that the military training exercises likely released to the air. The estimated air concentrations for most every contaminant was so low that even highly sensitive air sampling devices would likely not have been able to measure them. These evaluations have shown that the residents of Vieques were not exposed to contaminants from the military training exercises at levels that are associated with adverse health effects. We do not believe conducting a supplemental study is warranted.

"A detailed accounting of all munitions and management activities at the Bombing Range (LIA), Eastern Maneuver Area (EMA) and Atlantic Fleet Weapons Training Facility (AFWTF) must be provided. This historical assessment should also include area of training activities that occurred at non-range locations. This information could reduce the travel distance for some contaminants. The assumption that the main concentration of contaminants has to travel more than 7.9 miles is based on actual operations not on the whole history of the facility."

When gathering information about past military training exercises at Vieques, the one concern repeatedly communicated to ATSDR was regarding inhalation exposures to air emissions from the air-to-ground bombing exercises. Accordingly, we focused on characterizing air emissions from the time frame with the Navy's most extensive use of the bombing range on Vieques. For this time frame, we estimated emissions and air quality impacts based on a large volume of the munitions and management activities at the bombing range: annual range utilization statistics from 1983 through 1999, daily summaries from a 6-year time frame, and a highly detailed inventory for 1998. We believe these statistics form an adequate basis for characterizing air quality impacts for the years when military training exercises were most frequent and intense.

In the PHA, we acknowledge that limited information is available on the exact quantities of munitions used between 1941 and the early 1970s. However, numerous site documents we reviewed report that the air-to-ground bombing activity prior to the early 1970s was far more intense on the island of Culebra than on the island of Vieques. Because the frequency and intensity of air-to-ground bombing on Vieques increased during the 1970s, it is reasonable to assume that the air quality impacts and any associated exposures prior to the early 1970s were lower than the concentrations we predicted in our models. In other words, the concentrations we predicted in the modeling (for years 1970s through 1999) are upper-bound estimates of air quality impacts for earlier years.

When preparing the PHA, we considered the fact that some training exercises have been conducted in areas outside the Live Impact Area. In recent years, for example, the Navy operated a small arms firing range in the Eastern Maneuver Area. However, the air emissions associated with such operations are truly minimal when compared to those that occur as a result of the air-to-ground bombing, and the shorter travel distance for the small arms firing range has limited effect on overall estimates of air quality impacts. Although other training activities have also occurred at locations outside of the Live Impact Area, these activities were limited in scope and duration and generally involved dramatically lower levels of high explosives. Consequently, the air quality impacts of such activities that occurred outside the Live Impact Area are believed to be limited.

"A comprehensive chemical composition must be determined for each type of munitions. This is to include explosives, propellants, and pyrotechnics. The chemicals, firing by-product compounds, and breakdown product compounds cited in the report must then be compared to the health-based standards."

ATSDR's modeling analysis is based on a highly detailed inventory of the munitions the Navy used at the Live Impact Area during a representative year of activities. We considered statistics on the total amount of different types of ordnance used and composition data for both the ordnance casings and high-explosive content. The inventory included the entire range of ordnance used in air-to-ground, ship-to-ground, and land-based exercises. When estimating emissions, we considered nearly 100 different contaminants that could be released to the air, including chemicals in the high explosive charges, by-products of the explosions, and metals from the casings. We compared the estimated ambient air concentrations to health-based standards and health-based comparison values, just as the comment suggests.

"Knowing that the US Navy has been using the LIA for more than 50 years, and having already established that wind speed and direction, munitions constituents and activities characteristics play an important role in the dispersion of chemical constituents in the waste, how is this represented in the average ambient air concentration number?"

The measured and modeled ambient air concentrations both reflect actual wind speeds and wind directions. Specifically, our modeling is based on meteorological data collected from US Naval Station Roosevelt Roads, which suggests that winds in the area blow from east to west approximately 75% of the time. The concentrations we estimated, however, also reflect the large distance that separates the Live Impact Area from the residential area of Vieques. Our modeling found, for instance, that chemicals emitted from the Navy's military training exercises dispersed to extremely low levels over the 7.9 miles that separate the emissions source from the nearest residential location. Therefore, even though the winds in the area primarily blew emissions from the Live Impact Area toward residents (a fact that we explicitly accounted for in our modeling), a more important determinant of the air quality impacts is the distance that separates the emissions source from the receptors.

"Does ATSDR have evidence that the process by which multiplying the concentrations from soil samples by the 'average' of PM10 or TSP is a conservative or even real approach?"

In the PHA, ATSDR estimated the highest possible ambient air concentration of metals that would have resulted if the only source of pollution on Vieques was wind-blown dust from the Live Impact Area. In this highly ideal scenario, we estimated concentrations of metals by multiplying the measured particulate matter levels by the amounts of levels found in soils at the Live Impact Area. The results of this calculation were used only to demonstrate that it is not possible for the one emissions source considered (i.e., wind-blown dust from the Live Impact Area) to cause air pollution to reach levels of health concern. This is an entirely defensible approach to evaluate the potential contributions from this one source.

We used the estimated concentrations to evaluate this one emissions source. The calculation approach, as stated above, represents a highly ideal scenario and the concentrations we estimated may not represent actual air pollution levels. If the air contamination levels at Vieques are found to be higher than the levels we estimated, then we would investigate the issue further. The most logical explanation for why observed levels would be higher than what we estimates would be that other emissions sources at Vieques are contributing to air pollution. ATSDR will evaluate this issue further if we receive a detailed account of PREQB's air sampling results for metals (see our response to the following comment).

"ATSDR should re-evaluate the conclusions regarding the concentrations and presence of metals once PREQB supplies the samples analysis."

PREQB has been collecting particulate samples at Vieques since July 2000 and analyzing many of these samples for concentrations of metals. ATSDR previously tried to obtain the metals sampling results from PREQB (see Appendix C.1 of the public comment release PHA); however, PREQB did not release the sampling results. ATSDR will review the health implications of the sampling data once PREQB makes the raw data and associated quality assurance information publicly available.

"When will ATSDR have a complete health assessment including all the studies made in Vieques?"

ATSDR is currently preparing a document that summarizes our complete findings on exposures to contaminants in drinking water, soil, air, fish, and shellfish. That document should be available before the end of this calendar year.

"As said by ATSDR, the cancer registry will be analyzed, but also the health history including asthma episodes and the heavy metals concentrations in the blood of Vieques residents should be researched and studied."

ATSDR remains committed to evaluating health and environmental information that is released relevant to Vieques. The cancer registry has not yet been officially published, nor are we aware of official publications on the prevalence of asthma episodes among island residents. Regarding concentrations of heavy metals in the blood of Vieques residents, ATSDR made efforts to conduct an exposure investigation to address this exact issue back in 2001. We allocated funding and conducted extensive planning to collect hair, blood, and urine samples from the residents of Vieques. Further, we solicited the support and assistance of several Vieques physicians. However, the physicians expressed reservation about whether the results will be reflective of actual exposures and were unwilling to proceed. While the offer to collect hair, blood, and urine samples is still available, the project has been indefinitely postponed until the physicians are willing to assist. We will be happy to review results of any relevant and thoroughly documented health studies that are published for the island.

"Page 5, Section I: What is the particle size and specific composition of chaff? Under what category will it fall (TSP, PM10, metals, or PM2.5)?"

As the public comment release PHA indicates, chaff fibers are typically 25 microns (µm) thick and between 1 and 2 centimeters long. After being released from aircraft, the chaff fibers might break into smaller sizes before reaching the Earth's surface. In its health evaluation of chaff, ATSDR assumed that all airborne chaff fibers that fall to the ground are small enough to be completely respirable. This assumption caused us to overestimate potential exposures, because some portion of airborne chaff is likely too large to be respirable. Regarding chemical composition, chaff is aluminum-coated fiber glass fibers, and the primary components are aluminum and silicon oxides.

"Page 9, Section III, Line 22. 'Vieques is several miles removed from sources of air pollution on any other island in the Caribbean Sea.' Even though there is a significant space separating the other islands this should not be a determining factor. Africa is considerably farther and the 'African dust storm' still affects Viequenses and Puerto Ricans."

Our primary goal when preparing the PHA was to respond to the concerns of the petitioner and community members. These concerns focused almost entirely on air emissions from the military training exercises. Accordingly, our PHA focused on this source of air emissions. However, we implicitly considered the results from other sources as well, by basing most of our conclusions on the results from ambient air monitoring data. The monitoring data reflect the contributions from all nearby sources, including those associated with the military training exercises and those found on other islands in the Caribbean Sea. PREQB's monitoring data quite clearly show that the combined effect of all emissions sources on and near Vieques does not cause concentrations of particulate matter to reach levels of health concern. Therefore, we do not believe that air emissions sources on other islands in the Caribbean have notable effects on air quality at Vieques.

"Page 11, Section III, Line 24. 'The overwhelming majority of ordnance impacts the LIA, but some bombs and surface . . . have landed in the waters.' Taking into consideration that for many years some targets in the LIA were in the water, and the fact that the Eastern Tip of Vieques is very narrow, the word some should be replaced by the word several of for a more meaningful word."

The public comment release PHA states that "some bombs and surface fire projectiles have landed in the waters near the LIA." We believe this description is correct and appropriate, especially given that there are no detailed accounts of the exact percentage of ordnance that has landed in the waters, and no changes have been made in the PHA in response to this comment. Moreover, it is worth noting that the percentage of bombs landing in the water (rather than on land) has no bearing on the conclusions drawn in the air pathway PHA, because our air modeling evaluation assumes that every bomb hits the land and explodes upon impact. This modeling approach leads to higher emission rates than actually occurred, because the emissions resulting from bombs landing in the water are lower than those associated with bombs exploding on land.

"Page 32, Section V, Line 3: '. . . measured in 43 soil samples . . .' When, who, where, and how were these samples taken? What were the QA/QC?"

The soil sampling in question was conducted by Navy contractors in June 2000 (CH2MHILL 2000). In this study, surface soils at the Live Impact Area were sampled from selected targets and from storm drainage areas and other low lying areas. The soils were collected following standard sampling procedures and were analyzed according to EPA Methods 6010, 8330, and E314. Extensive quality assurance and quality control information are available on the sampling data, including results from duplicate samples, matrix spikes, and matrix spike duplicates. Further, the sampling report fully documents critical aspects of the sampling program, including photographs of sampling locations, soil logs, chain-of-custody forms, and data validation narratives. ATSDR critically reviewed the entire soil sampling package and concluded that these soil sampling results are of a known and high quality and can be used to support our public health conclusions.

"Page 34, Section V, Background Information on Particulate Matter. Following EPA evidence that links inhalation of PM2.5 to an adverse health effect, will ATSDR be conducting air monitoring for PM2.5?"

ATSDR recommends sampling when additional data are needed to make public health conclusions. Unfortunately, there is no sampling we can recommend now that would quantify past exposures. However, as our response to Comment #8 indicates, we do not believe that past exposures to PM2.5 were at levels of health concern. If PM2.5 sampling were to start now, it would characterize current and potential future exposures at Vieques. We do not believe it is necessary to initiate PM2.5 sampling now, primarily because such sampling would not reveal any insights on air contamination levels originating from Navy lands. The primary air emissions source from the Navy lands currently is wind-blown dust, which typically takes the form of TSP and PM10, not PM2.5.

"Page 38, Section V, Line 9: '. . . metals in wind-blown dust on Vieques do not represent a public hazard.' This conclusion is incorrect since it is already proven that concentrations of arsenic (in an ideal scenario) exceed the health-based comparison numbers."

The comment addresses our conclusion regarding metals in wind-blown dust, particularly our finding regarding arsenic. The comment correctly indicates that the arsenic concentration we estimated in the public comment release PHA was higher than conservative health-based comparison values. However, the comment incorrectly infers that these arsenic levels are a potential health hazard. To the contrary, the PHA points out how one should interpret health-based comparison values: "Ambient air concentrations lower than their corresponding comparison values are generally considered to be safe and not expected to cause harmful health effects, but the opposite is not true: ambient air concentrations greater than comparison values are not necessarily levels of air pollution that could present a possible public health hazard. Rather, chemicals with concentrations higher than comparison values require further evaluation." Therefore, the fact that the estimated arsenic concentration was marginally higher than a health-based comparison value does not suggest that a hazard exists.

We have not changed our conclusion regarding arsenic and view the arsenic levels as not of health concern. Specifically, we concluded "that the estimated concentration is within the range of ambient air levels of arsenic reported for remote areas in the United States and is lower than the ranges reported for rural and urban settings." The comment also notes that our estimated arsenic concentrations represent an ideal scenario, as our response to Comment #26 also indicates. We would have preferred to base this conclusion on PREQB's metals sampling data, but these data were not provided to us. ATSDR will review the health implications of the sampling data once PREQB makes the raw data and associated quality assurance information publicly available.

"Page 40, Section V, Line 21-23. Were the concentrations of particulate matter compared to the type, quantity, climate during the drop off, and if it was air-to-ground or surface-to-ground activity?"

The comment addresses our evaluation of particulate sampling data collected on days when the Navy conducted military training exercises using practice bombs. ATSDR compared the level of air-to-ground bombing activity on these days with the measured concentrations and found no relationship between the intensity of the bombing activity and the levels of air pollution. We did not conduct multi-variate statistical analyses to examine how other factors (e.g., climate) contribute to the data trends, or absence of trends. We have updated our statistical analyses of air sampling results based on data PREQB has reported up through the beginning of 2003, but these analyses still do not suggest any clear relationship between the intensity of practice bombing exercises and levels of air pollution.

ATSDR notes that we included the statistical analyses of the sampling data in the PHA to offer perspective on whether the practice bombing exercises contributed to the measured concentrations. The bigger picture among the sampling data, however, should not be lost: PREQB has collected more than 400 valid, representative air samples on Vieques since July 2000, and not one sample collected to date has indicated particulate matter levels of health concern. Our conclusions reflect this more important overall trend among the sampling data.

"Page 50, Section V, Line 17. How can ATSDR use the word 'confident' if the only sampling and study done was to run an ideal model and it is based on 'several assumptions' and 'inherent uncertainties.'"

The comment addresses the terms we used to describe our modeling analysis. We used air dispersion modeling to characterize past exposures at Vieques. The modeling was performed because limited sampling data are available for the time when the Navy conducted military training exercises on Vieques using live bombs. We believe that our modeling analysis not only was rigorous and scientifically defensible, but also served a very useful purpose in filling a data gap.

The comment specifically questions our confidence in the modeling results, given the limitations and uncertainties associated with the analysis. When basing conclusions on modeling analyses, our internal guidance indicates that it is important to acknowledge uncertainties and limitations with fate and transport models, because these models can only estimate exposure conditions (ATSDR 1992). Consistent with this guidance, we have acknowledged that the estimated concentrations might not exactly equal past exposure levels. However, when one critically reviews the assumptions in the modeling analysis, it becomes clear that our modeling analysis offers a reasonable account of past exposures and that we truly can be confident that we have not failed to predict exposures at levels of health concern. Here are more specific reasons why we can have such confidence:

• When predicting emission rates for metals, we assumed that 100% of the metals in the bomb casings and explosive charge become airborne and available for downwind transport. We also assumed that 100% of the metals in the soils ejected during bombing events becomes airborne and available for downwind transport. These assumptions clearly overstate potential exposures, because we know that some fragments from the bomb casings (which contain metals) actually remain on the ground, and are not completely vaporized. Therefore, the emission rate we used for metals is clearly higher than the actual emission rate. Consequently, the air concentrations we predicted for metals also are higher than the actual incremental impacts from the bombing exercises.
• When predicting emission rates for explosives, we assumed that 10% of the explosive chemicals in the bombs becomes airborne and is not destroyed during the exercises, even though available data from the BangBox studies suggest that detonations typically consume more than 99% of the explosive charge. Our approach, therefore, likely overstates potential exposures to explosive chemicals. Moreover, as the PHA notes, even if we had assumed that 90% of the explosives in the bombs were emitted directly to the air (and not destroyed), the potential exposures would still be lower than levels of health concern. In other words, any reasonable estimate of destruction efficiency for the live bombs would not lead to exposure estimates that are at levels of health concern. It is for this reason that we can have confidence that our modeling analysis has not reached an erroneous conclusion.
• When modeling exposures to chemical by-products of explosions, we based emission rates using the OBODM model–the only relevant model published on EPA's Support Center for Regulatory Air Models clearinghouse site. Using the emission factors in the OBODM model, the concentrations we estimated for every chemical by-product of explosions were all at least 1,000 times lower than their corresponding health-based comparison values. Although it is likely that the exposure levels predicted by the model might be slightly higher or slightly lower than the concentrations that actually occurred, we seriously doubt that the model underestimates exposures by more than 3 orders of magnitude, which is why we reported having confidence in the conclusion.
• When modeling exposures to particulate matter, we found that the estimated exposures were lower than levels of health concern. Moreover, we were aware that PREQB conducted two particulate sampling studies on the island in the 1970s, and the Navy conducted its own study. All three studies did not find particulate matter concentrations at levels of health concern. The fact that three separate studies had findings consistent with our modeling gave greater confidence in the modeling results, though we acknowledge in the PHA that detailed documentation of the sampling events is lacking.

In short, for the reasons stated above, we continue to have confidence that our modeling analysis is an adequate basis for reaching public health conclusions, and we do not believe the modeling analysis failed to detect exposure levels of health concern.

"Page 81, Section VI. Was the roof of some residents sampled? If not, why?"

ATSDR reviewed several information resources to respond to community concerns regarding potential exposures to contaminants that might have settled on rooftops used to collect rain water for drinking water purposes. The consensus reached by multiple environmental and health agencies on this matter is that rain water can be a safe and reliable source of drinking water, provided that good sanitation practices are followed in collecting and treating the water. Our PHA reaches the same conclusion, and we encourage residents who use such rooftop collection systems to follow the good sanitation practices mentioned in the PHA. Following such practices will eliminate, or greatly reduce, potential exposures to contaminants that settle on rooftops from a variety of sources.

"Page 83, Section VI. Rainfall should be sampled in order to get a better understanding of the quality of the water that will be collected and therefore be able to give better recommendations to the people to use collection systems in Vieques."

ATSDR has no reason to believe that contaminants from the Navy's past military training exercises at Vieques would be found in rainfall at the island. The only mechanism by which contaminants can currently in the rainfall is through wind-blown dust from the Live Impact Area being deposited during rain storms. All sampling data collected to date suggest that the impacts of wind-blown dust from the Live Impact Area on air quality in the residential parts of the island are immeasurably small. As a result, ATSDR does not believe that rainfall sampling is an important public health action for this site.