Proceedings
WORKSHOP ON EXPOSURE ASSESSMENT
OF
HAZARDOUS CHEMICALS
Sponsored by:
U.S. Environmental Protection Agency
Industrial Environmental Research Laboratory
Environmental Monitoring and Systems Laboratory
Environmental Sciences Research Laboratory
Research Triangle Park, North Carolina
EPA Contract No. 68-02-3171, Work Assignment 2
Coordinated by:
Radian Corporation
1 April 1980
'.0. 3ox 9948 / Austin, Texas 73766 / (512)454-4797
-------
CONTENTS
1. Introduction 1
2. Summary of Recommendations 3
2.1 Emissions Estimating Committee 3
2.2 Atmospheric Modeling. Committee 4
2.3 Population Estimating Committee 4
2.4 Statistics Committee 5
3. Workshop Summary: Emissions Estimating. ... 6
3.1 Workshop Recommendations. 6
3.2 Recommended Procedure for Evaluating
and Improving Estimated Emissions Data. . 8
3.3 Obtaining and Documenting Emissions
Data 15
3.4 Closing Remarks 17
4. Workshop Summary: Atmospheric Modeling. ... 18
4.1 Recommendations 18
4.2 Critique of Methods Used in Previous
Studies 20
4.3 Committee Comments and Discussions. ... 21
4.3.1 Cost Effectiveness 21
4.3.2 Toxicity/Carcinogenicity 22
4.3.3 Documentation 23
4.3.4 Time Resolution of Emissions
Data 23
4.3.5 Other Media, Occupational
Exposures 24
-------
5. Workshop Summary: Population
Exposure Estimating 26
5.1 Summary of Recommendations 26
5.2 Recommendations for Population
Exposure Assessment 28
5.2.1 Population Data 28
5.2.2 Activity Pattern and Time
Budget 31
5.2.3 Population Mobility 31
5.2.4 Concentration/Population
Data Interface 32
5.2.5 Cut-Off Point 33
5.2.6 Population Exposure Parameter. . . 34
5.3 Uncertainties in Estimating
Population Exposure 35
5.4 Research Needs 36
6. Workshop Summary: Statistics 38
6.1 Recommendations 39
6.2 Statement of the Problem 40
6.3 Recommended Approach 43
6.4 The Three Levels of Analysis 44
6.5 Statistical Problems in Exposure
Assessments 45
Appendices
A Workshop Committees. • 46
B Workshop Attendees 49
C Exposure Assessment Documents Reviewed .... 50
D Individual Comments From Atmospheric
Modeling Committee Members 52
E Regular Users of Census Data 67
ii
-------
FIGURES
FIGURE PAGE
1 Decision Tree for Estimated Emissions Data. . . 10
2 Comparison of Estimate Quality vs.
Level of Effort 21
3 Exposure Level Not Within Hazardous
Range 41
4 'Exposure Level -Within .Hazardous Range 41
5 Combination of Data to Estimate Hazard 43
Dl Model Selection and Implementation Procedure. . 57
iii
-------
TABLES
TABLE PAGE
1 Recommended Level of Sophistication in
Each of the Six Elements to be Considered
in Estimating Population Exposure 29
Dl Regional Annual Average Climatology
for Level I Modeling Estimates 54
D2 Average Annual Climatology for Urban
Area Sources 55
D3 Potential Sources of Error 61
LV
-------
SECTION 1
INTRODUCTION
The purpose of this workshop was to review the current
methods of conducting an exposure assessment and to suggest any
possible improvements. The result of an exposure assessment is
an exposure estimate. An exposure estimate is an approximation
of the number of people expected to be exposed to various con-
centrations of the substance being investigated. It consists of
three major portions: (a) estimating the emissions; (b) deter-
mining the dilution due to atmospheric transport; and (c) esti-
mating the number of people exposed to the various concentration
levels.
The workshop was divided into four expert working groups,
one for each of the above subtopics and a statistics group.
Initially, a statistician met with each of the groups. During
the last day, the statisticians met as a group to suggest how
confidence limits could be placed on the population exposure
estimates. This document consists of reports obtained from each
expert working group and a summary of recommendations prepared
by Radian.
The exposure assessment workshop resulted when EPA's Office
of Air Quality Planning and Standards (OAQPS) requested that the
Office of Research and Development (ORD) review the methods cur-
rently used in developing an exposure assessment. Three EPA
laboratories, the Environmental Monitoring and Systems Labora-
tory - Research Triangle Park (EMSL-RTP), Environmental Sciences
-------
Research Laboratory - RTF (ESRL-RTP), and the Industrial Envi-
ronmental Research Laboratory - RTF (IERL-RTP), were charged with
with performing this evaluation. William Petersen (E'SRL) , David
Mage (EMSL), and William Baasel (IERL) were selected to coordi-
nate the evaluation. Radian'Corporation was selected by the
laboratories to conduct a workshop where non-EPA experts in the
field would evaluate the current exposure assessment procedures
and make suggestions for their improvement.
Three levels of exposure assessment were evaluated by work-
ing participants. Level I assessments are essentially screening
studies, which assist in identifying chemicals that may warrant
further investigation.. Level II studies are used to support a
listing decision and source prioritization for regulation under
Section 112 of the Clean Air Act. Level III assessments employ
more sophisticated modeling techniques and require very accurate
emissions and population data. Regulations for new and existing
sources of listed chemicals are based on Level III assessments.
Level III assessments form the basis for determining the neces-
sary degree of emission control.
Prior to attending the workshop, each participant was asked
to review the thirteen reports listed in Appendix D. With the
exception of the report on chloroprene, which was a Level I
report, all were Level II reports. At the meeting, the draft
report of the modelling section of a Level III report was pro-
vided for participants' review.
The workshop was held December 11-13, 1979, at the Ramada
Inn-North in Durham, North Carolina.
-------
SECTION 2
SUMMARY OF RECOMMENDATIONS
Some of the major recommendations from each of the work-
shop committees are summarized in this section.
2.1 EMISSIONS ESTIMATING COMMITTEE
• Complete emissions results should be provided in each
exposure assessment report. The emissions data base should in-
clude, when possible, the sampling and analytical procedures
used to obtain the data. Additionally, the report should in-
clude a description of plant process conditions, emission con-
trol equipment, and production capacity at the time of sampling.
The range and distribution of emissions data should be given.
• Industrial groups should be asked to comment on the
emissions data presented in each exposure assessment report.
This may be done either before the report goes to EPA for review,
or during the time in which OAQPS is reviewing the report.
Industry review should take place at all three levels of assess-
ment. Whenever specific plant data are used in a report the
data should be provided to the facility's manager for comment.
If the data are verified by plant personnel, less chance of
litigation exists.
-------
2.2 ATMOSPHERIC MODELING COMMITTEE
• The types of models recommended in the OAOPS Guideline
Series, "Guideline on Air Quality Models," should be used in
Level II studies.
• Level III studies should use the best quality atmo-
spheric models currently available.
The modeling committee was unable to agree on recommendations
for a Level I effort. Section 4 describes committee members1
differing ideas.
2.3 POPULATION ESTIMATING COMMITTEE
• MEDLIST (Master Enumeration District List) population
data should be supplemented by a population growth factor, U,S,
Geological Survey population data, other census population data
(e.g., the 4th count population data), and special population
survey data.
• Human activity information such as time spent indoors/
outdoors and physical activity levels (resting, working, or
exercising) should be incorporated in population exposure esti-
mating.
• Population mobility, such as daily migration from resi-
dence to a work place and seasonal or other periodic patterns
of movement should be considered in population exposure estimat-
ing.
• Concentrations of chemicals should be estimated at a
population centroid instead of the other way around.
-------
• Instead of restricting a population exposure study area
to an arbitrary distance of 20 km from a source, the study area
should be extended to a distance at which an annual mean concen-^
tration due to that source decreases to a diminutive level.
• Results of population exposure analyses should be re-
ported in a manner independent of an arbitrary assumption of a
particular dose-response function.
2.4 STATISTICS COMMITTEE
• Use a quantitative approach based on expected values and
95 percent confidence limits of the individual elements of the
analysis to estimate exposure levels and their associated con-
fidence limits.
• The precise definition of the three levels of analysis
should remain flexible to allow for the contractor to allocate
his resources in the most cost-effective manner.
• The definition of exposure should include continuous
and correlated space and time variations.
-------
SECTION 3
WORKSHOP SUMMARY: EMISSIONS ESTIMATING
This chapter contains the emissions estimating committee's
recommendations for improving emissions estimates in future EPA
exposure assessment studies. It also contains a recommended
procedure for evaluating emissions data used in exposure studies.
Additionally, this chapter lists potential sources of emissions
data-and the -kind of information that should be recorded.
3.1 WORKSHOP RECOMMENDATIONS
After reviewing and discussing available EPA exposure
assessment documents, the emissions estimating committee
developed the following recommendations.
Specific sources of emissions data should be given in each
report. The report should state if the data are from litera-
ture, regulatory inventories, or industrial records and if
these data are a result of estimates, emissions factors, mate-
rial balance calculations, or sampling. Such documentation of
emissions data will be useful if future studies require addi-
tional data.
When specific data obtained from plant sampling are used,
the analytical procedure, frequency of sampling, and process
conditions should be included, if possible.
-------
The year that emissions data were obtained and the plant
production capacity at the time of sampling should be stated.
The emissions data for each source should be listed as a
range of values in addition to a single numeric measure of cen-
tral tendency. The basis for the range should be given. The
range should reflect the precision of the available data, as
well as the possibility that all emission sources have not been
included in the available data base. A range of emission values
is necessary to quantitatively indicate the level of certainty
associated with the emissions data. Providing a range of values
may require an increased level of technical effort for determin-
ing emissions, may potentially involve personal judgments during
the initial level of effort, and may result in industry's con-
cern that only the high value of the range will be used. How-
ever, in addition to a measure of central tendency, a numerical
range is the best way to alert the users of the data to the
inaccuracies involved. If the final risk assessment indicates
that a greater accuracy in the emissions data base is required,
the improved accuracy would then be reflected in a narrowing of
the range.
At each level of technical effort, a decision must be made
concerning the adequacy of the emissions data when combined with
the precision of the atmospheric dispersion modeling effort and
the population exposure estimates. If emissions data are judged
inadequate, better emissions data must be collected. Emissions
data for a screening-level effort might be based on literature
and regulatory inventories data for generalized plants and in-
dustries. For technical efforts which may result in standards
development, emissions data should include site specific infor-
mation (e.g., individual stack parameters) which will require
an improved data base. For the most detailed atmospheric dis-
persion models and population estimates, the emissions data
should include daily and seasonal variations in emissions.
7
-------
As the accuracy of atmospheric dispersion modeling and pop-
ulation exposure estimates is increased, more accurate emissions
data may be necessary. Emissions data bases should be improved
by expanding the data base sources. For example, more extensive
literature searches, additional industrial input, and added
information from regulatory agencies may provide more data.
Industry data obtained under Section 114 of the Clean Air Act,
ambient monitoring data, or source sampling data may be neces-
sary to expand the emissions data base. The report at each
level of technical effort should include specific recommenda-
tions for improving the emissions data base.
The sources of emissions data must be thoroughly documented
to. withstand the pressures of litigation. This may require that
estimates and historical data be improved by requiring industry
to sample at specific sites.
At each level of technical effort, a copy of the data
should be provided to industry for review. An explanation of
its potential end' use should accompany the data. If industry
reviews the data, the accuracy of the data base may be improved,
and potential conflicts may be resolved early in the evaluation
process.
Additional review of emissions data by trade and profes-
sional associations should be considered.
3.2 RECOMMENDED PROCEDURE. FOR EVALUATING AND IMPROVING
ESTIMATED EMISSIONS DATA
The degree of uncertainty involved in an exposure assess-
ment cannot be defined before the assessment is started. In
some cases a wide range or broad confidence interval in the
estimated emissions data can be tolerated. Even with a wide
range, the results of a Level I study of exposure assessment
8
-------
may allow a clear decision regarding further work on a partic-
ular pollutant.
In many cases, however, refinement of the emissions data
will be required as the assessment is pursued. The repetitive
proces.s involved in this refinement has been illustrated in a
decision tree prepared by the emissions estimating committee.
The decision tree is illustrated in Figure 1.
In the initial stages of a screening level assessment,
several factors must be evaluated for any given pollutant.
(During the workshop, the screening level assessment was
referred to as a "Level I" effort.) These factors include:
1. The types of sources (point, fugutive,
storage, area, etc.) of the pollutant,
2. The physical characteristics of the sources
(height, velocity of emitted stream, temperature, etc),
3. The geographical location of the manufacturers
and users,
4. The location, in some cases, of individual
sources within a plant, and
5. The estimated emissions rates of the particular
pollutant of concern.
As shown in Figure 1, .estimated emission rates can be
obtained from numerous origins (block 1). Potential leads for
first estimates of rates include technical literature, indus-
trial data, EPA reports, and files of local, state and federal
regulatory agencies. The emission rates obtained must be quan-
tified. At a minimum, a mean and a range of values for the
emission rate'must be defined.
To understand why a given emission source may have a widely
varying composition requires an understanding of variables which
-------
Figure 1. Decision tree for Level I and II assessments
10
-------
affect the emissions. As shown in the decision tree, the esti-
mated emission or range of the emission rates must be judged for
adequacy after the emissions data source is first examined
(block 2). The estimated emissions data are inadequate if the
preliminary analysis of the data and/or data origins does not
provide, as a minimum requirement, the mean and the numerical
range of emission rates. In this situation, more data must be
obtained. This improvement can be accomplished by expanding the
data search to include additional resources (block 3). These
resources can include industry contacts, trade and professional
associations, files of additional government agencies, manufac-
turers and vendors of equipment and processes, and more detailed
engineering evaluations of the emissions source. This further
effort must result, at the very least, in a mean and a numerical
range for the estimated emissions before the assessment can pro-
ceed as shown in Figure 1.
If the emissions estimate can be judged to be adequate
enough to proceed (block 2), the emissions estimate should be
sent to industrial representatives for their review and comments
(block 4). These representatives should include those companies
or plants which emit the particular pollutant being studied.
The estimated emissions data should be sent to as many of the
sources as possible within the effort's budget and time con-
straints. Any company whose emissions data were used in de-
veloping the estimated emissions rate should have the opportu-
nity to comment on these data. It is also suggested that the
estimated emissions data be sent to related associations and
trade groups for review (e.g., the Technical Association of the
Pulp and Paper Industry - TAPPI - or the Electric Power Research
Institute - EPRI). Even for small sources, such as dry clean-
ers, trade associations may have useful comments on the data.
If documented industrial review results in the data's being
judged inadequate, all data should be carefully re-examined
11
-------
(block 5). The data should be reviewed with regard to the basis
and methodology used to develop them (material balance, sam-
pling, vendor specifications, etc.)- If a technical assessment
of all the data indicates the need for an adjustment in the .
estimated emissions rate, this change should be made. Addi-
tionally, if resources permit, industrial review of the revised
estimated emissions rate can be sought again. It should be
obvious that this repetitive process cannot be continued in-
definitely. Time and economic constraints will govern the
extent to which this process can be pursued.
The initial dispersion modeling effort could be performed
concurrently with the industrial review (block 5). Changes in
estimated emissions rates could easily be accommodated at inter-
mediate stages in the modeling (block 7).
After emissions estimates have been reviewed by industry
(and adjusted if necessary) they can be used in the atmospheric
dispersion model in conjunction with the population estimates
to yield the screening level exposure assessment (block 7).
The results of this initial assessment will indicate if
further refinements of estimated emissions are needed at the
screening level of effort (block 9). Three possibilities may
exist. With the existing precision of the atmospheric model,
population estimates, estimated emissions rate, and health
assessment, it may be obvious (even in the worst case) that the
exposure to the particular pollutant is below the harmful to
health level. In this case the pollutant can be eliminated
from further consideration. On the other hand, the exposure
assessment may show, even under best case conditions, a high
exposure and a high risk when considered with the health assess-
ment. The pollutant must then be assessed at a more intensive
level of technical effort which could be the basis for setting
12
-------
a standard. (During the workshop this type effort was generally
referred to as "Level II" effort.)
The third possibility is that the results of the screening
level of assessment are not accurate enough for a clear decision
regarding the further consideration of a particular pollutant
(block 13). In this case, a decision must be made on the
improvement of the accuracy of the estimated emissions, The
part of the assessment to be improved will generally be decided
on the basis of one or more of the following factors:
cost of improvement
required and available time
probability of success
degree of probable improvement in accuracy
However, a decision to improve the estimated emissions
data or concentration profiles may require that more sophisti-
cated techniques be used than were initially applied. These
could include the acquisition of more specific data from in-
dustry (through unofficial request or through Section 114 let-
ters) , the actual sampling of some representative emission
sources, and/or through ambient air monitoring (blocks 3 and 15).
When the data have been acquired, the estimated emissions
should be revised to reflect any-increase in accuracy gained
from the additional effort. The revised emissions estimate
should then, ideally, progress through the steps shown in the
.decision tree, starting with submission of the revised estimated
rates to industry for review. At some point in the repetitive
process, the precision of the exposure assessment should be suf-
ficient to justify a decision regarding the further considera-
tion of the pollutant being studied.
It should be clearly recognized that, in the extreme, the
recommended procedure for obtaining estimated emissions rate
13
-------
data for even a screening exposure assessment could require a
substantial funding and technical manpower commitment. In the
majority of cases, however, very few repetitions are antici-
pated for the screening assessment.
When the screening level of effort results in a more exten-
sive assessment for a pollutant, the adequacy of each part of
the exposure assessment must be evaluated as to its ability to
support standards. If the dispersion model and population
estimate are adequate, the adequacy of the estimated emissions
data must be determined (blocks 13 and 14).
If emissions data are determined to be inadequate, the
basis for estimating the emissions rate must be improved and
expanded (-blocks 3. and 15) . Data could be -obtained under Sec-
tion 114 of the Clean Air Act, or from an ambient sampling
source. If data from the screening effort are judged adequate,
they should be sent to industry for review. Potential conse-
quences of the assessment should be conveyed to industry at
this point (blocks 10 and 30), Industry review should also in-
clude review by appropriate professional and trade organiza-
tions, such as (but not limited to) American Institute of Chem-
ical Engineers, American Petroleum Institute, or American
Society for Testing and Materials (block 11). For generic
sources such as dry cleaners or gasoline service stations, it
will not be practical to send data for each to review. In that
case the use of trade organizations should be actively pursued.
After industry data review, EPA must again determine the
adequacy of the emissions data (blocks 12 and 24). If data are
judged inadequate, the data base must be improved (blocks 3 and
23). If data are judged adequate, they should be combined with
the dispersion model and the population estimate (blocks 6 and
18).
14
-------
When estimated emissions data are linked with population
estimates through dispersion modeling, the resulting exposure
assessment is combined with the health assessment to determine
the risk caused by the pollutant under study. At this point,
the results must be validated. Three possibilities exist. The
results may indicate that the health risk is below the value
which requires a standard for control. In this case, the study
is complete. Alternatively, the results may indicate that the
health risk is above the value judged to require a standard for
control. In this case, EPA may recommend regulation to the
Science Advisory Board.
If the results are judged inadequate to support a recom-
mendation to the Science Advisory Board, a decision must be
made-as to which .part of the study needs improvement. If the
dispersion model and population estimate are judged adequate
and the emissions data inadequate, improved data must be
obtained (blocks 3 and 15). If the emissions data are adequate,
either or both the model and population estimate must be im-
proved. This will involve a higher level study (blocks 13 and
25).
3.3 OBTAINING AND DOCUMENTING EMISSIONS DATA
The information presented in the reports reviewed during
the workshop did not provide an adequate basis for judging the
quality of the data. The following paragraphs contain sugges-
tions for obtaining and documenting data so that their adequacy
may be assessed.
The report from which the data were taken should be ref-
erenced. Additionally, the original developer and reporter of
the data should be noted, if different from the reference
report. Data included in the reports which are excluded from
15
-------
consideration should be noted. The reason for exclusion should
be given.
Information obtained should include the time period of the
base data; the time period for any projection; the capacity of
the process (design, nominal, actual); any sampling and monitor-
ing information, site visits, written and verbal communications,
and/or engineering estimates; the types of emission sources in-
cluded (point, fugitive, storage).
The emissions data sources listed below comprise a starting
list of sources and types of information which may be useful.
The extent to which these and other sources are used will depend
on the quality of data needed to provide compatible results with
the other parts of an exposure assessment. Sources of data
include:
trade/technical journals
industry publications
licensor specifications
federal, state, local and international regulatory
agencies
industry contracts
If possible, the data should include emission rates for point,
area, and fugitive emissions, as well as emissions from storage
and waste disposal. If data are not available from one or more
of these sources, the data gaps should be clearly identified.
Necessary physical characteristics of each source must be
included. A partial list of source characteristics includes:
chemical composition of the emission
physical state of the emission
16
-------
temperature, pressure, velocity, etc.
of the emission
height above grade of the emission source;
plot and plan views of surrounding structures.
The extent of source characteristic details to be used will
vary as the level of technical effort changes. For the Level I
efforts, general or representative characteristics can be used.
For Level III efforts (especially in coordination with detailed
atmospheric modeling efforts), individual site specific data will-
be required. When the emissions rate varies in a distinct pat-
tern (e.g., seasonal or daily,) these data should be presented
consistent with the detail of the dispersion modeling and popu-
lation estimates to be used.
3.4 CLOSING REMARKS
Emissions data, combined with estimates of population,
atmospheric dispersion, and health effects, must be used by EPA
to obtain an accurate risk assessment for pollutants under study.
Budget, time and other limitations will cause the accuracy of the
risk assessment to change for each pollutant. When the objective
is only to screen and give priority to a list of pollutants, the
quality of the estimated emissions data does not need to be the
same as that required when the objective is to provide a basis
for setting a standard. As a consequence, the amount of detail
and effort expended to estimate emissions varies.
For each pollutant being studied, there will be many deci-
sions required to obtain data of the desired quality. It should
be emphasized that when the effort may lead to the setting of a
standard, the estimated emissions rate must be documented so that
it can withstand the trials and pressures of litigation. The
emissions estimating committee developed its recommendations as
a guide to assist EPA in providing adequate data for future risk
assessment studies.
17
-------
SECTION 4
WORKSHOP SUMMARY: ATMOSPHERIC MODELING
The atmospheric modeling committee reviewed the exposure
assessment reports provided by EPA and discussed at length a
number of factors influencing selection of a modeling approach.
Members of the committee could not agree on the best modeling
approach for Level I studies. However, the committee was able
to recommend approaches for Level II and III efforts.
Section 4.1 of this chapter contains the committee's recom-
mendations for Level II and III exposure assessments. Addition-
ally, committee members' differing opinions as to Level I modeling
approaches are summarized. Modeling methods used in previous
exposure assessments are critiqued in Section 4.2. Section 4.3
summarizes committee discussion throughout the meeting. In-
depth comments by individual committee members are included in
Appendix D.
4.1 .RECOMMENDATIONS
The atmospheric modeling committee recommended atmospheric
modeling approaches for Levels II and III exposure assessments.
These recommendations are summarized below.
The types of models recommended in the OAQPS Guideline
Series, "Guideline on Air Quality Models," should be used in
Level II studies. This document describes various models suitable
for many situations. The selection should be based upon at least
18
-------
a brief study of the local situation. The meteorology to be
used in such a model application should be the best meteorolog-
ical data for the particular area which can be obtained by phone
and mail. Normally this would be the STAR from the nearest site
for which this compilation is available. However, if another
meteorological data base is more suitable, it should be used.
If the site has high terrain or is on a lake front or ocean
front or in a deep valley which perturbs the simple meteorology
normally used in these models, more complex meteorological data
should be obtained, if the budget allows.
Level III studies should use the best quality atmospheric
models currently available. The current "state-of-the-art" con-
sists of models somewhat more advanced than those described in
"Guideline on Air Quality Models." Level III studies should
consider site-specific meteorological data, and all of the com-
plexities of the interaction of plumes and structures.
The modeling committee could not agree on a recommendation
for a Level I modeling approach. Some members felt that a
simple, uniform meteorological approach should be taken. This
type of approach (but not necessarily the details of the
approach) is exemplified in the SRI International reports
reviewed during the workshop.. SRI used a set of dispersion
estimates which can be reduced to a single curve on a plot of
(concentration/emissions) vs. radial distance from the source.
Some of the group felt that such a plot could be computed for
average meteorology for the United States with several possible
lines indicating different stack heights. If this approach is
used, EPA should prepare the best possible single-estimating
nomogram or chart and request that it be used by all contractors
preparing Level I assessments.
Some committee members thought that limited site-specific
terrain and meteorology should be considered in addition to the
19.
-------
simple approach described above. If there is complex terrain
or extremely unusual meteorology at the site, the simple nomo-
gram of (concentration/emissions) vs. radius should be modified
to account for this.
Other members of the committee felt that under no circum-
stances should a simple meteorological dispersion approach ever
be used for any meteorological dispersion calculations. These
members felt that an approach similar to the Level II approach
recommended by the committee be used for Level I as well. This
view is explained in more detail in Appendix D.
4.2 CRITIQUE OF METHODS USED IN PREVIOUS STUDIES
During the workshop, the modeling committee reviewed 13
reports describing previous EPA exposure studies. These reports
are listed in Appendix C. This section contains the committee's
comments regarding the modeling approaches described in these
reports.
The committee felt that the modeling methods and assump-
tions were not adequately documented. This was particularly
true in the case of the SRI International reports which refer
the reader to calculations made by OAQPS. While those calcula-
tions may be of the higest quality, there was not enough in-
formation presented in either the reports or in any document
cited to allow an outside observer to understand in detail
what was done and form an intelligent judgement as to whether
it was done properly.
OAQPS identified some reports as being Level I reports;
others were identified as Level II reports. It appeared to the
committee that several Level II studies had dispersion modeling
suitable for a Level I study, while some assessments in Level I
used models suitable for Level II. Mr. Suta, of SRI
20
-------
International, indicated in his presentation that he had not
heard of the distinctions between Levels I and II before the
workshop. Thus, this is not a criticism—explicit or implied--
of his work or of the work in the SRI documents; rather it is
an indication that there should be agreement as to what will be
presented in future Level I and II studies.
4.3 COMMITTEE COMMENTS AND DISCUSSIONS
4.3.1 Cost Effectiveness
For each level of study, there will be distinct budget limi-
tations. EPA's objective is to select that level of study which
produces the most useful, reliable, and accurate information
within these budgetary constraints. The committee questioned how
much our information will improve as we go to increasing levels
of complexity. One panel member suggested that the accuracy of
the information as a function of the budgetary expenditure has
the shape shown in Figure 2.
en
01
u
CO
Cfl
W
14-1
o
CO
o*
Expenditure (manpower and financial)
Figure 2. Comparison of estimate quality vs level of effort.
If that is true, then extreme expenditures are probably not
justified. One committee recommendation for future studies would
21
-------
be that whenever a Level II study is done on a pollutant for
which a Level I study has been done, the Level I study should be
included as an appendix. The change in exposure estimate as a
result of going from Level I to Level II methodology should be
made clear in this report. Similarly, if a Level III study is
done for a chemical which has had a Level II study, a similar
comparison should be reported. This will not be of great help
in the particular set of modeling exercises, but may contribute
to our understanding of the shape of the curve shown in Figure 2.
4.3.2 Toxicity/Carcinogenicity
EPA has indicated that future exposure assessment studies
will be conducted using the most conservative dose-response
relationship for a carcinogenic sub-stance. The committee assumed
that the probability of an excess cancer is proportional to the
lifetime inhalation or ingestion of the material, independent of
dose rate. This is obviously a "worst case" assumption which
has interesting consequences for modeling because it indicates
that there is no concern whatever for short-term, high-concentra-
tion dosages. This is the exact reverse of the situation with
criteria pollutants, which have a toxicological type assumed
response. Our recommendations for models generally direct the
attention to annual average concentrations whi'ch are suitable for
this type of assumption. However, we recommend that, at least
in Class III studies, the short-term concentration be determined
if it is convenient.
A natural consequence of this worst-case dose-response assump-
tion is that there is no concentration level so low as not to be
of interest. We are inherently in the situation where long-range
transport is of significance. Any arbitrary choice such as "we
will terminate the dispersion calculations and exposure calcula-
tions at 20 miles" or "we will terminate the exposure concentra-
tion to 1/100 ng/m3" is not consistent with this dose-response
22
-------
assumption. Such arbitrary cut-off points may result in grossly
underestimated exposure concentrations.
The committee felt that efforts should be made to estimate
removal mechanisms and removal rates for potential airborne
carcinogens. Based on these efforts, decay constants for each
chemical should be factored into the dispersion model. Use of
such a constant will allow modeling to extend to boundaries at
which concentration-exposure levels become negligible compared
to those close to the emission source.
4.3.3 Documentation
The committee strongly urged that in any studies which will
be used for regulatory purposes, the models and meteorological
data used must be adequately documented. Although proprietary
models whose details are not revealed to the public may be use-
ful for other applications, the committee felt that any modeling
whose quality may be examined by a court of law or challenged
before a regulatory agency should not be done by such a model.
This implies a strong preference for models such as those listed
in the EPA Guideline Series whose details are widely known and
understood. However, if a contractor is willing to document a
proprietary model, then it should be considered.
4.3.4 Time Resolution of Emissions Data
If it is possible to obtain emission estimates which are
resolved by time of day and season of the year, that would be
most helpful. It is an observational fact that in many areas of
the country the wind directions are strongly biased by season
and time of day. (For example, in Salt Lake City during most of
the year, the wind is from the north during the daytime and from
the south at night.) Our current level of modeling sophistication
23
-------
is adequate to take advantage of information about time resolu-
tion of emission data if those data are available.
4.3.5 Other Media, Occupational Exposures
The completed exposure assessment reports reviewed by the
committee suggest that EPA intends to consider only exposures
to the populace from airborne pollutants which are inhaled.
That may be the best approach for the present. However, exam-
ples are known in which pollutants enter drinking water and food
by passing through an airborne state; for example, lead which
enters drinking water and food in particulate fallout from the
atmosphere. Certainly for any Level III study, the question of
whether the atmospheric removal processes for this material will
cause it to enter drinking water or food should be examined.
Similarly, it is possible that air pollutant control mea-
sures intended to minimize atmospheric exposure will have the
consequences of increasing occupational exposure. Normally,
both are decreased by such measures, but examples can be imag-
ined in which reducing atmospheric exposure would result in
more tightly enclosed factories with the result that the indoor
concentrations would increase, etc. This possibility should
be considered in any study leading to detailed regulations.
Finally, the possibility exists that in -attempting to solve
an air pollution exposure problem, we may create a water pollu-
tion problem. The converse is also true. For example, the
reports on cadmium show that municipal incinerators are a major
source of atmospheric cadmium. There we are solving the solid
waste problem and creating an air pollution problem. Similarly,
it seems that with limestone scrubber sludge, we are solving an
air pollution problem and creating a solid-waste problem. In a
study leading to detailed regulation, the technological
24
-------
possibility of such intermediate transfer of pollution should be
considered.
25
-------
SECTION 5
WORKSHOP SUMMARY: POPULATION EXPOSURE ESTIMATING
The population exposure estimating committee discussed the
exposure assessment methodology, as well as the primary elements
that are needed to adequately quantify exposures of .the general
population to a given toxic substance. The committee reviewed
the thirteen reports of population exposure assessment provided
by the U.S. Environmental Protection Agency.
After long', intense deliberation, the committee has arrived
at several recommendations and caveats for population exposure
assessment. These are discussed in the following sections. Sec-
tion 5.1 contains a summary of the recommendations made by the
committee. A more detailed description of each of the recommen-
dations is given in Section 5.2. Sources of uncertainties in
the population exposure estimating are discussed in Section 5.3;
in Section 5.4, further research needs for improving population
exposure estimates are described.
5.1 SUMMARY OF RECOMMENDATIONS
The population estimating committee recommends that the
following six improvements be incorporated in population expo-
sure assessments made in the reports reviewed.
26
-------
MEDLIST (Master Enumeration District List) population data
should be supplemented by a population growth factor, U.S. Geo-
logical Survey population data, other census population data
(e.g., the 4th count population data), and special population
survey data.
Human activity information such as time spent indoors/out^-
doors and physical activity levels (resting, working, or exer-
cising) should be incorporated in population exposure estimating. .
Population mobility, such as daily migration from residence
to a work place and seasonal or other periodic patterns of move-
ment should be considered in population exposure estimating.
Concentrations of chemicals should be estimated at a popula-
tion centroid instead of the other way around.
Instead of restricting a population exposure study area to
an arbitrary distance of 20 km from a source, the study area
should be extended to a distance at which an annual mean concen-
tration due to that source decreases to a diminutive level. Here,
a diminutive level is a level of chemical-specific concentration,
life-time exposure to which does not pose any significant health
risk to the public.
Results of population exposure analyses should be reported
in a manner independent of dose-response functions. To make the
results usable for a nonlinear model, as well, the numbers of
people exposed to various annual concentrations should be reported
in either a histogram (i.e., for each concentration interval) or
a cumulative distribution (i.e., at or above various concentra-
tion levels).
27
-------
5.2 RECOMMENDATIONS FOR POPULATION EXPOSURE ASSESSMENT
Our discussion of population exposure assessment identified
six areas where additional efforts appeared to bring about a
significant improvement in either population exposure estimates
or the utility of estimated population exposure for a final
health-risk assessment. These areas are:
1. Population Data
2. Activity Pattern and Time Budget
3. Population Mobility
4. Concentration/Population Data Interface
5. Cut-Off Point
6. Population Exposure Parameter
This section discusses how each of the six elements affects esti-
mated population exposure to a given toxic substance and provides
some recommended methods for improving estimates of population
exposure as well as estimates of each element. Table 1 lists
recommended improvements as a function of the sophistication
level of the exposure assessment.
5.2.1 Population Data
The census data tapes called "MEDLIST" provide a convenient
population data base, indicating where and how many people reside,
down to the enumeration district/block group (ED/BG) level. An
enumeration district is the smallest census statistical area and
is usually covered by a single census surveyor. But, even though
MEDLIST provides a reasonable spatial coverage of population dis-
tribution all over the United States, its population estimates are,
by now, nearly 10 years old and need to be updated. The commit-
tee does not believe that MEDLIST alone can provide adequate
information on either the current (instead of 1970) population
or the dynamic (instead of stationary resident) population.
28
-------
TABLE 1. RECOMMENDED LEVEL OF SOPHISTICATION IN EACH OF THE SIX
ELEMENTS TO BE CONSIDERED IN ESTIMATING POPULATION EXPOSURE
Element
Level I Study
Level II Study
Level III Study
Population
MEDLIST (census)
Estimate of cur-
rent population,
national growth
factor.
U.S.G.S. improve-
ments, other pri-
vate sources
(state or county
level). Local
estimated growth
factors.
4th count (tract
data (population
mix)
Activity
Type (micro-
environment) NO
Mobility
(displace-
ment)
Concentration/
Population
Interface
Cut-Off Point
Exposure
Parameters
NO
Concentrations to
be interpolated
into ED/BG
Centroids
Up to a distance
at which concen-
trations .due to
a given source
decrease to a
diminutive level
1. (people x cone.)
2. No. of people
exposed to
specified cone.
interval
YES (general)
YES (general)
Concentrations to
be interpolated
into ED/BG
Centroids
Same
YES (detailed)
YES (detailed)
Distribute popula-
tion within ED
where concentra-
tion gradients
are high.
Same
., 2. and
3. Sum of cone.
from different
sources before
computing expo-
sures to a
given pollutant
1., 2., and 3.
29
-------
Population estimates are updated annually by the Bureau of
the Census by taking random samples from a few percent of the
entire population. Results are summarized down to the number of
people residing in each county and by the so-called population
growth factor, i.e., the current population relative to the 1969-
1970 census-year population. The committee considers estimates
of current population in a study area made by taking a product
of the 1970 population and the national growth factor adequate
for Level I studies.
For Level II and III studies, however, the committee recom-
mends use of a more region-specific growth factor for estimating
the current population. Because population growth takes place
nonuniformly over a study area, a search of spatially resolved
growth factors or current population estimates is encouraged.
Potential sources of such population data include local planning
agencies, local Chambers of Commerce, local/state real estate
boards, the U.S. Geological Survey (U.S.G.S.), the Bureau of
Economic Analysis, and the Bureau of the Census. A great deal
of assistance can sometimes be obtained from regular users of
the census population data whose names and addresses are ,listed
in Appendix E.
It should be noted that the U.S.G.S. population estimates
may be more appropriate in rural areas in which the Census Bureau
estimates of population do not have adequate spatial resolution
to accurately locate the rural population. For Level III studies,
it might be advisable to use the 4th count census data in order
to obtain detailed population characteristics by age, sex, etc.
It could also be used to perform specific site surveys which
would provide an up-to-date distribution of population in the
vicinity of a given set of sources.
30
-------
5.2.2 Activity Pattern and Time Budget
Population exposure analyses conducted in the thirteen
reports reviewed by the committee are based on the implicit
assumption that people are exposed to ambient air 24 hours a day
at their residence location. This 24-hour ambient-exposure
assumption greatly simplifies a population exposure analysis.
However, it may not be appropriate when concentrations of a study
chemical are expected to be different between indoors and out-
doors or when diurnal patterns of human activity levels and
ambient concentrations are correlated.
Because many people both live and work indoors, information
on human activity pattern and time budget (i.e., how long and
what time of the day people stay indoors or outdoors) is needed
to make a realistic population exposure estimate. For a Level I
study incorporation of this information may not be necessary.
But it should be incorporated into a more elaborate population
exposure analysis, particularly for the Level III study.
5.2.3 Population Mobility
Population mobility also poses a problem for a population
exposure analysis. The population in any given area is not
static, and its* size and spatial distribution exhibit consider-
able diurnal and seasonal variations. Many people commute to
the place of their employment, go to school, or visit a resort
area. At these places they may be exposed to pollutant con-
centrations different from those at their residence.
In a large metropolitan area, the diurnal shift of the
population from residence to work place is quite large. With a
sharp gradient of pollutant concentrations over the metropolitan
area, consideration of population mobility becomes quite impor-
tant for correctly estimating exposures of the population to any
given chemical. Cyclic movements of the population occur not
31
-------
only within a day but also from season to season and year to
year.
For Level II and III studies this mobility consideration
should be incorporated into the population exposure analyses.
In particular, such a consideration'becomes essential for esti-
mating exposures of employed adults who commute to places where
air quality is very different from that at their residence.
5.2.4 Concentration/Population Data Interface
To obtain a population exposure estimate requires a proper
interface of population and concentration data in space and, for
a more elaborate analysis, in time, too. Concentration is con-
tinuously distributed over space and its value at a given loca-
tion can be estimated by air quality simulation models or from
nearby air quality monitoring values. Population, however, is
distributed in a discrete manner and is incorporated into a
statistical area. Furthermore, without knowing a detailed
housing pattern over a study area, it is not possible to cor-
rectly estimate the population in a particular area from popu-
lation figures of nearby statistical areas. Population esti-
mation differs in this way from concentration estimations
which can be accomplished by using data from adjacent areas.
With .these Limitations in mind, the committee recommends
estimating a concentration at a population centroid into which
local population is aggregated, rather than the other way around.
When MEDLIST data are used, a concentration at the population
centroid of each enumeration district should be estimated. Then,
according to the concentration value, values of population expo-
sure parameters should be calculated. If the population data are
aggregated into a larger area (a rectangular or segmented con^-
centric area is often used), a concentration at the population
centroid of that area should be used for population exposure
32
-------
calculations. Using a concentration at the population centroid
is preferred over using a concentration at the geographical cen-
ter.
5.2.5 Cut-Off Point
In computing a population exposure parameter, a problem
arises as to where such calculations should be determined.
Without natural sinks for a toxic substance, the concentration
never decreases to absolute zero. Therefore, the calculated
number of people exposed or the product of exposed population
and concentration keeps increasing as the study area is enlarged.
This problem of monotonically increasing results in population
exposure estimates comes from the basic assumption of a linear
dose-response function for toxic substances.
In all the reports reviewed, the area in which measureable
population exposure takes place is bounded by an arbitrary dis-
tance, e.g., 20 km from the source. This cut-off criterion ap-
pears to be too arbitrary and may result in missing a signifi-
cant portion of the population who are exposed to a low but not
negligible level of concentration. Suppose a source is located
25 km upwind of New York City. Then, the exposed population or
the population-concentration product of New York City may ex-
ceed that of the 20 km study area around the source.
To avoid the potential problems outlined above, the commit-
tee strongly recommends exploring better criteria for the cut-
off point. One possible criterion may be a distance at which a
concentration due to a given set of sources drops to a diminu-
tive level, e.g., 0.1 ng/m3 for chemical i. Here the diminutive
level should be chemical-specific and should be determined by
considering the toxicity of that chemical. A diminutive level
may be defined as a low concentration at which level an average
person can withstand exposures to the chemical throughout his
33
-------
life without increasing any measurable health risk.
5.2.6 Population Exposure Parameter
Population exposure to a toxic substance is usually re-
ported in two quantities: the number of people exposed, and a
product of people exposed x concentration. Basically, these
two parameters appear to be adequate for quantifying population
exposure. However, values of the two parameters should be com-
puted for various concentration intervals rather than for an
entire range of concentrations.
The linear dose-response model leads to the practice of ex-
pressing exposures as the-..product-of ,annual, average concentra-
tion times population. A nonlinear model, however, would not
necessarily be compatible with this measure. It is therefore
recommended that all exposure analyses also report the numbers
of people exposed to various intervals of annual average concen-
tration or, alternatively, the number of people exposed to con-
centrations at or above various concentrations (i.e., a cumula-
tive distribution). These values are independent of any partic-
ular dose-response model and may be used with any future dose-
response models that may be developed.
For source-related population exposure estimates, people
may be counted more than once: first as exposed to a concentra-
tion due to one source; second as exposed to a concentration due
to another source; and so forth. Such double or triple count-
ings of people do not make any difference as long as a linear
dose-response function holds. To make the population exposure
estimates useful for a more general dose-response function,
total exposure of the population to a combined concentration due
to several sources or types of sources should also be computed.
34
-------
5.3 UNCERTAINTIES IN ESTIMATING POPULATION EXPOSURE
In estimating population exposure to a toxic substance, many
assumptions and simplifications have been employed and will also
be used for future studies. However, uncertainties involved in
estimating population exposure must be explicitly stated. Sources
of uncertainty in the estimation of exposed populations include
the following parameters:
1. Population Estimations - Data from the Census Bureau
and from private sources make available ways of esti-
mating the uncertainties involved in the population
estimations. These uncertainties are known and are
not very large.
2. Growth Factor - This is a parameter also furnished
by the Census Bureau and .other sources. The uncer-
tainties involved in the growth-factor estimations
are relatively higher than the ones involved in the
population estimations.
3. Mobility - This parameter estimates the daily move-
ment of a population in and out of an enumeration
district. Uncertainties are difficult to estimate,
yet they should be explicitly considered because,
in certain cases, they may be significant.
4. Indoor/Outdoor - This is a field researched by few.
Indoor (non-work-place) sources of hazardous sub-
stances should be investigated as contributors to
total pollutant exposures. Exposures from such
sources may constitute a major portion of total
population exposure (e.g., perchlorethylene inside
of dry-cleaning establishments, .polycyclic organic
matter (POM) from cigarette smoke). If indoor
sources are not considered, the relative contribu-
tion of outdoor sources (and the benefits from
controlling such sources) may be greatly overesti-
mated. The uncertainties involved are high.
5. Time Budgets - Time allotted to the indoor environ-
ment has been studied by sociologists. Uncertainties
involved in estimating time budgets are known rela-
tively well for certain populations, but are not
known well for the general population.
35
-------
6. Cyclic Variation - This parameter estimates seasonal
or periodic populations (such as the 4-year cycle in
the Washington, D.C., area). There is a moderate
range of uncertainty in estimating cyclic variation.
It is case-specific and, on certain occasions, may
be significant. Therefore, it should be addressed
in all studies.
7. Uniform Distribution - The assumption of population
uniformly distributed within an ED involves large
uncertainties, specifically in certain areas. Urban
populations, for example, are not uniformly distrib-
uted in either the horizontal or vertical (high-
rise apartments or offices) directions. Population
distribution is of importance close to major pollutant
sources and in rural areas where each district is large,
The uncertainties involved must be studied indepen-
dently, and the population distribution is recommended
as -a -research -area.
8. Population Grid Points - The approach of allocating
all population to the geometric center (grid point)
of a district as opposed to the centroid (population
weighted center) introduces uncertainties that relate
to the air pollution patterns. These uncertainties
can be calculated and should be addressed.
5.4 RESEARCH NEEDS
Extensive data bases exist indicating where people reside,
down to the enumeration district/block group (ED/BG) level. For
Level I and possibly Level II efforts these data are probably
sufficient. For greater accuracy, more information is needed in
three areas:
1. How people are distributed within the ED/BG.
2. How people move about and where they go during the
course of a day, a week, or over the period of a year.
3. How to better measure the actual dosages they
receive.
36
-------
The distribution within an ED/BG can become critical close
to emission sources, where emission levels are high; or in thinly
populated areas, where the size of an ED/BG can be large. In
either case, significant errors may be encountered in assigning
the correct concentration if-there is a -significant concentration
gradient over the ED/BG.
More research is needed in the second area to avoid possible
errors in exposure estimation from failing to take into account
the movement of people on a daily, weekly (in particular, week-
day/weekend), and seasonal basis. Since, for purposes of esti-
mating cancer risks, there is an interest in lifetime exposures
it may also be desirable to characterize permanent moves of people
from one area to another (e.g., general shift of elderly people
from northern industrial zones to the sun-belt region) . In con-
nection with the movement of people, it will also be important
to determine whether they are indoors or outdoors. There is a
need to determine how the movement is to be characterized for use
in an overall model and the most appropriate way to develop
a usable data base. ' ... • •, - _
Health effects are a function of the amount of hazardous
material absorbed by the body. This absorption is a function of
both activity level (breathing rate, level of exertion) and expo-
sure. Activity levels may vary with time of day, season, and
location. The correlation of activity levels with exposure pat-
terns shoul-d be investigated as a potentially significant factor
for future risk assessment studies.
37
-------
SECTION 6
WORKSHOP SUMMARY: STATISTICS
The ultimate objective of the entire assessment is to make
a best estimate of the number of people exposed at various con-
centration levels to the potentially hazardous chemical. To do
this, we need to know the emissions, the atmospheric dispersion,
and the population distribution data. If we assume that these
factors interact with one another, then the uncertainty or
variance of the final estimate is a complex function of the
variances of the individual parts of the calculation. If the
variances of the three parts of this calculation decrease in a
particular order such as a2dispersion >
this ranking of uncertainties should aid in deciding how
resources could be allocated in attempting to improve our knowl-
edge and attempting to make the best estimates of exposure.
Similarly, it should be a guide to understanding the expected
confidence intervals of the estimates .
When we consider the exposure part, we want to know what
would be indicated by individual personal samplers worn by
a large and representative sample of the populace. Very
little work has been done in getting people to wear such
samplers, but we are ultimately attempting in this kind of
study to predict the answer which would be found if a large
group of the populace did indeed wear such samplers. The read-
ings of such samplers would be influenced by the emission rate
and its temporal variation, the variations in concentration at
various points caused by the meteorological dispersion of the
38
-------
emissions that variation caused by indoors-outdoor concentration
differences, and the variations caused by the movement and
activities of the individual. In the ultimate calculation
scheme, a statistical distribution of exposures would be obtained.
The purpose of the statistics committee was to aid EPA in
developing realistic confidence intervals for use in exposure
assessment studies. During the EPA Exposure Assessment Work-
shop, members of the committee observed committees in atmospheric
modeling, emissions estimation, and population estimation. In-
formation gathered from these other groups was applied in the
development of confidence intervals for exposure assessment
studies.
6.1 RECOMMENDATIONS
The recommendations of the statistics committee are summa-
rized below.
A quantitative approach based on expected values and 95
percent confidence limits of the individual elements of the
analysis could be used to estimate exposure levels and their
associated confidence limits. Each of these limits should be
accompanied by a detailed description of the process used to
estimate it.
The definition of exposure should include continuous and
correlated space and time variations. Since error is intro-
duced, details of the process and/or calculations used to esti-
mate exposure should be included in the contractor's report.
The precise definition of the three levels of analysis
should remain flexible to allow for the contractor to allocate
his resources in the most cost-effective manner. The contractor
should be required to explicitly define and document his
39
-------
approach to the allocation of resources early in the course of
the work. Proposals should include a description of the con-
tractor's plans in this regard.
The procedure should not require accuracy levels to be
specified in advance for any pollutant, nor should comparable
accuracy levels be required for the three component areas of any
one study. A decision may be reached on the basis of data having
limited accuracy. Each initial study of a pollutant should deter-
mine the accuracy of the data and decide if further accuracy is
necessary.
6.2 STATEMENT OF THE PROBLEM
For statistical purposes, it is convenient to consider an
exposure assessment as an estimate of population exposure in
ppb-person-years plus a measure of the accuracy of this estimate.
A convenient and useful form for this information is a set of
confidence limits and their associated probability, e.g., we are
95 percent confident that the result is between a and b ppm-
person-years. These numbers may be calculated if each element
in the process of producing the exposure estimates is associated
with a confidence level of this form.
This information may then be used to determine whether the
exposure estimate exceeds the value H (e.g., the desired cut-
off point for Level I listing) at a specified level of confidence.
Thus the exposure estimate Hx and its associated upper and lower
control limits UCL^ and LCL^ are compared to H . If, as in the
case of Figure 3, the UCLjj is less than HQ, we would accept the
hypothesis that the true exposure to this chemical is below H
and thus decide not to list it.
40
-------
w 1
LCLg
1 1
HTTfT
1 •^T
H
o
Figure 3. Exposure level not within hazardous range.
On the other hand, if as in Figure 4, the lower confidence
limit exceeds H , we would decide that the true exposure is in
excess of HQ and thus list the pollutant. The third case,
obviously, is where the confidence limits straddle HQ; in this
case it is necessary to undertake further study to refine (nar-
row) the limits by taking more (or more accurate) measurements.
Such refinement may still be within the framework of a Level I
study, or may require accuracies normally regarded as belonging
to Level II.
H.
LCLH;
Figure 4." Exposure level within hazardous range.
It should be noted that, due to the two-sided nature of the
95 percent confidence interval, the probability of making a wrong
decision (e.g., not listing a pollutant which should be listed or
vice versa) does not exceed 2% percent.
It should be clear from Figures 3 and 4 that it may be pos-
sible to reach a decision regarding a pollutant (i.e., to list
or not to list) even when limited precision in one of the three
component areas causes the confidence interval to be quite large.
41
-------
Obviously, very dirty or very clean sources may be easily recog-
nized without substantial study. However, no attempt should be
made to require the same precision in different studies, or in
the components of a single study. An initial analysis, based on
readily-available data, should be used to decide what statistical
level of accuracy is sufficient. If this level is not reached
in each study, efforts can be made to improve the accuracy of
the assessment. Cost effectiveness may be increased consider-
ably if the "adaptive approach" described above is used.
The preceding paragraphs explain in general terms what type
of information is desired and how it will be used. To be more
specific the estimate of exposure which we have referred to as
"H" above may be expressed as:
H
all space time
sources
Q(s,t) *(s,t), p(s,t) dtds (1)
where :
i|» =»• meteorological dispersion function
Q «* emissions rate
p » population distribution function, and
s and t are indices of space and time.
Equation (1) summarizes, in a very general way, the roles of
the three component parts (emissions, atmospheric transport, and
population distribution) . Again, it should be emphasized that
no uniformity of precision for the three components should be
sought in the initial analysis, since it is possible that valid
conclusions may be drawn from imprecise data. Precision in one
area may* compensate for lack of precision in another. It is most
important that the precision of the data in each component be
well known and documented.
42
-------
It should be noted that continuous and correlated space and
time variations are included in this definition of exposure.
This is an important concept. This type of definition is recom-
mended for exposure estimates. Upper and lower confidence limits
should account for errors created by simplifications, assumptions,
and uncertainties in the data and modeling procedures used to
estimate H. Details of the process and data used to calculate
these limits are to be considered an essential part of the report-
ing process.
6.3 RECOMMENDED APPROACH
The approach to calculating H that we recommend is to esti-
mate an expected value and upper and lower confidence limits on
each element used in the calculation of H. The confidence limits
are not necessarily symmetrical but are defined to provide a 95
percent probability that the true value falls within the interval
as shown:
Prob (LCL < X < UCL ) = 0.95
xi xi
(2)
Where X^ is Element i used in calculating H.
Each of these limits should be accompanied by a detailed descrip-
tion of the process used to estimate it. By combining these
estimates in accordance with the schematic in Figure 5,
H
e
E£
Q
IISSIONS
3TIMATES
D
E
*
ISPERSION
STIMATES
PO
E
P
PULATION
STIMATES
Figure 5. Combination of data to estimate hazard.
H and its associated confidence limits not necessarily at 95 per-
•cent probability may be estimated.
43
-------
The estimation of the confidence limits of H is not neces-
sarily the result of a straightforward convolution of the error
distributions assumed for each element of the calculation.
Rather, correlations between emissions, dispersion, and popula-
tion movement must be accounted for, thus increasing the complex-
ity of the problem. We have chosen to leave the development of
an approach to this problem to the contractor because it is a
nontrivial problem for which an efficient solution may best be
developed on a case by case basis..
Depending on the error distributions used for each of the
elements of the calculation, the confidence limits of H may be
estimated in several different ways. Monte Carlo techniques
provide a straightforward, albeit possibly expensive, method
of estimating the error distribution of H. Stratification of
the variables into seasonal, day vs. night, indoor vs. outdoor,
and/or other categories may help to reduce the multicolinearity
problems to the point where straightforward convolution of the
distributions may provide good results. A third possibility is
the development of an analytical approach to the derivation of
the distribution of the product of Q, tp, and p.
6.4 THE THREE LEVELS OF ANALYSIS.
The three levels of analysis discussed at the meeting of
11-13 December correspond roughly to good, better, and best;
these are largely qualitative categories bounded by the levels
of effort suggested for each. The precise definition of each
level and the confidence limits which may be expected of the
results are the subject of the individual committee reports.
The concern of the statistics, committee in this regard is to leave
the contractor sufficient latitude in the definition of "levels"
so that his resources may be allocated to the area where the
most cost effective improvements in the estimate of H are to be
found. Thus, the contractor should be required to evaluate
early in his efforts an appropriate allocation of resources with
44
-------
respect to narrowing the confidence limits of H. An early inter-
mediate result should be an analysis of the potential for improv-
ing each element of the estimation of H, and its cost effective-
ness. The final reports at. each level should conclude with
similar information pertaining to the potential for improvement
of the estimates with further work. We would expect that the
contractor's proposal discuss his plans for dealing with these
issues.
6.5 STATISTICAL PROBLEMS IN EXPOSURE ASSESSMENTS
The estimation of numerical confidence limits on quantities
for which there is a paucity of data is a task which many investi-
gators will approach with a substantial and understandable degree
of trepidation. The process vmay require at times that educated
j
guesswork be substituted for measurement. For the present purpose
a key point to remember is that professional judgment in these
matters is an improvement over using estimates without confidence
limits. The members of the statistics committee feel comfortable
with this type of information provided that it is accompanied
with a clear statement of the basis for the estimates whether
they come from measurement, engineering judgment, surrogate emis-
sions measurements, etc. However, it should be recognized that
if the precision estimates are not well-based, the conclusions
will be correspondingly less reliable. The discussion would allow
the user to understand the kind and quality of information in-
cluded in the report and to act accordingly.
45
-------
APPENDIX A
WORKSHOP COMMITTEES
ATMOSPHERIC MODELING COMMITTEE
Noel deNevers, Chairperson
University of Utah, 3062 MEB
Department of Chemical Engineering
Salt Lake City, Utah 84112
Richard -Porter
Texas Instruments, MS 3949
P.O. Box 225621
Dallas,Texas 75265
Elmer Robinson
Air Pollution Research
Washington State University
Pullman, Washington 99164
Walter Dabberdt
Atmospheric Science Division
SRI International
Menlo Park, California 94025
Richard Shultz
Trinity Consultants
100 N. Central Expressway, Suite 910
Richardson, Texas 75080
Steve Hanna
ATDL/NOAA
P.O. Box E
Oak Ridge, Tennessee 37830
EMISSIONS ESTIMATING COMMITTEE
Kenneth Baker, Chairperson
Greene & Associates, Inc.
One Energy Square
Dallas, Texas 75206
46
-------
Macon Shepard
Environmental Consultants, Inc.
P.O. Box 1178
Clemson, South Carolina 29631
Thomas Kittleman
E.I. DuPont de Nemours and Co, Inc,
1352 Louviers Building
Wilmington, Delaware 19898
Robert G. Wetherold
Radian Corporation
P.O. Box 9948
Austin, Texas 78766
Ralph White
Hydroscience, Inc.
9041 Executive Park Drive
Knoxville, Tennessee 37919
POPULATION ESTIMATING COMMITTEE
Yuji Horie, Chairperson
Technology Service Corporation
2811 Wilshire Blvd.
Santa Monica, California 90403
William F. Biller
68 Yorktown Road
East Brunswick, New Jersey 08816
Demetrios Moschandreas
Geomet Inc.
15 Firstfield Road
Gaithersburg, Maryland 20760
Anton Chaplin
Teknekron
2118 Milvia Street
Berkley, California 94704
Richard Londergan
TRC
125 Silas Deane Highway
Wethersfield, Connecticut 06109
47
-------
STATISTICS COMMITTEE
Richard Pollack
868 Gellston Place
El Cerrito, California 94530
M.R. Leadbetter
Department of Statistics
University of North Carolina
Chapel Hill, North Carolina 27514
Terence Fitz-Simons
EPA/EMSL
Research Triangle Park, North Carolina 27711
48
-------
APPENDIX B
WORKSHOP ATTENDEES
EPA EXPOSURE ASSESSMENT WORKSHOP ATTENDEES LIST
December 11, 1979
NAME
Gerald E. Anderson
William D. Baasel
Kenneth Baker
William F. Biller
Anton Chaplin
Bob Coleman
Walter Dabberdt
Noel deNevers
James L. Dicke
Mike Dusetzina
Tom Feagans
Terence Fitz-Simons
Steve Hanna
Yuji Horie
Tom Kittleman
Ross Leadbetter
Barbara Lee
Richard Londergan
David Mage
Justice A. Manning
Jack McGinnity
Craig Miller
D.J. Moschandreas
Nancy B. Pate
Bill Petersen
Richard L. Pollack
Richard A. Porter
Elmer Robinson
Harold Sauls
George J. Scheve
Richard H. Schulze
Macon Sheppard
Ben Suta
Bob Wetherold
Ralph White
ORGANIZATION
Systems Applications, Inc.
EPA/ORD/IERL
Greene & Associates, Inc.
Consultant
Teknekron Research
EEA, Inc.
SRI International
University of Utah
EPA/OAQPS
EPA/SASD
EPA/OAQPS
EPA/EMSL
ATDL
Technology Service Corp.
DuPont
Statistician, U. of N.C.
Radian Corporation
TRC
EPA/ORD/EMSL
EPA/OAQPS/SASD
EPA/OAQPS/SASD
EEA, Inc.
Geomet, Inc.
USPHS/EPA/OAQPS
EPA/ESRL
LBL
Texas Inst. Inc.
Washington State Univ.
EPA/EMSL
NOAA/EPA/OAQPS
Trinity Consultants, Inc.
Env. Consultants, Inc.
SRI International
Radian Corporation
Hydroscience
PHONE NO.
(415)
(919)
(214)
(201)
(415)
(919)
(415)
(8011
(919)
C919)
(.919)
(919)
C615)
(213)
(302)
(919)
(512)
(203)
(919)
(919)
(919)
(703)
(301)
(919)
(919)
(.415)
(214)
(509)
(919)
(919)
(214)
(803)
(415)
(512)
(615)
472-4011
541-2815
691-3500
257-0164
548-4100
471-2506
326-6200
581-6024
541-5381
541-5355
541-5355
541-2792
526-1237
829-7411
366-4718
929-5172
454-4797
563-1431
541-2231
541-5345
541-5204
528-1900
948-0755
541-5202
541-4564
486-6292
238-5635
335-1526
541-3123
541-5391
234-8567
654-5410
326-6200
454-4797
690-3211
49
-------
APPENDIX C
EXPOSURE ASSESSMENT DOCUMENTS REVIEWED
1. Anderson, Gerald, Rinnan Exposure to Atmospheric Concentra-
tions of Selected Chemicals. EPA Contract No. 68-02-3066.
SAI No. EF-156, Systems Applications, Incorporated, 950
Northgate Drive, San Rafael, California 94903, December
1979.
2. Coleman, R., et al., Assessment of Human Exposures to Atmo-
spheric Cadmium, EPA Contract No. 68-02-2836, Tasks 3 and
6, Energy and Environmental Analysis., Inc., Arlington,
Virginia, June 1979.
3. Coleman, R. et al., Sources of Atmospheric Cadmium, Energy
and Environmental Analysis,Inc., Arlington, Virginia, n.d.
4. H.E. Cramer Company, Inc., Dispersion Model Analysis of the
Air Quality Impact of Emissions from Benzene_Storage and
Loading Facilities, EPA Contract No. 69-02-2507, Salt
Lake City, Utah, February 1979.
5. Mara, Susan J., and Shonh S. Lee, Assessment of Human Expo-
sures to Atmospheric Benzene, Center for Resource and
Environmental Systems Studies Report No. 3OR, EPA Contract
Nos. 68-01-4314 and 68-02-2835, SRI Projects EGU-5734 and
CRU-6780, SRI International, Menlo Park, California, May
1978.
6. 'Mara, Susan J., Benjamin E. Suta, -and Shonh S. Lee, Assess-
ment of Human Exposures to Atmospheric Perchlorethylene,
draft final report, Center for Resource and Environmental
Systems Studies Report No. 73, EPA Contract No. 68-02-2835,
SRI Project CRU-6780, SRI International, Menlo Park, Cali-
fornia, January 1979.
7. Schewe, George J., "Modeling Analysis of Maximum Ambient
Arsenic Concentrations Due to Primary Copper Smelters,"
Paper for Environmental Protection Agency, Model Applica-
tion Section, n.d.
50
-------
8. Suta, Benjamin E. , Assessment of Human Exposure to Atmospheric
Acrylonitrile, final report, Center for Resource and Envi-
ronmental Systems Studies Report No. 100, EPA Contract No.
68-02-2835, Task 20, SRI Project CRU-6780, SRI International,
Menlo Park, California, August 1979.
9. Suta, Benjamin E., Human Exposures to Atmospheric Arsenic,
Resource and Environmental Systems Report No.50, EPA
Contract Nos. 68-01-4314, 68-01-2835, SRI Projects EGU-
5794 and CRU-6780, SRI International, Menlo Park, Cali-
fornia , September 1978.
10. Suta, Benjamin E., Human Population Exposures to Coke-Oven
Atmospheric Emissions, final report, Center for Resource
and Environmental Systems Report No. 27, EPA Contract Nos.
68-01-4314 and 68-02-2835, SRI Projects EGU-5794 and CRU-
6780, SRI International, Menlo Park, California, October
1978 (rev. May 1979).
11. Suta, Benjamin E., Human Population Exposures to Coke-Oven
Atmospheric Emission Under Attainment of OSHA Worker
Standards, final report, Center for Resource and Environ-
mental Systems Studies Report No. 65, EPA Contract No.
68-02-2835, Task 10, SRI Project CRU-6780, SRI Interna-
tional, Menlo Park, California, Rev. April 1979.
12. Suta, Benjamin E., Assessment of Human Exposure to Atmo-
spheric Ethylene Bichloride^final report,Center for
Resource and Environmental Systems Studies Report No. 82,
EPA Contract No. 68-02-2835, Task 17, SRI Project No.
CRU-6780, SRI International, Menlo Park, California,
May 1979.
13. Systems Applications, Inc., Human Exposure to Atmospheric
Chloroprene, preliminary report — technical progress
narrative, Office of Air Quality Planning and Standards,
SAI Project No. 179-55, San Rafael, California, April 1979
51
-------
APPENDIX D
INDIVIDUAL COMMENTS FROM ATMOSPHERIC MODELING
COMMITTEE MEMBERS
Dl.O PROBLEMS ENCOUNTERED USING THE SINGLE-NOMOGRAPH APPROACH
IN LEVEL I ASSESSMENTS
Dl.l Level I Nomographs — Richard Porter
It has been suggested that nomographs be used by the con-
sultant for the Level I survey of pollutants. It is important
that a sufficiently detailed set of nomographs be developed by
EPA for the contractors' use to insure that pollutants be ranked
in the proper order. It is not possible to design a single nomo-
graph that can characterize all pollutants for all conditions.
For instance, the use of a nomograph designed for elevated
releases of buoyant plumes in level rural areas for a pollutant
that is emitted from roof top vents in industrial valley situa-
tions could result in the pollutant being ranked in the lower
20% of the list instead of the upper 20%. Separate nomographs
should be provided to the contractor for each major source
release height and terrain situation.
The following are situations for which distinct nomographs
should be provided.
1. Rural flat terrain-release heights: roof top,
30 m, 75 m, 100 m, 200 m.
2. Urban flat terrain-release heights: roof top,
30 m, 75 m, 100 m, 200 m.
52
-------
3. Rural complex terrain-release heights: roof top,
30 m, 75 m, 100 m, 200 m.
4. Urban complex terrain-release heights: roof top,
30 m, 75 m, 100 m, 200 m.
This suggests four nomographs with five Xu/Q curves each.
In the case of a pollutant that is only emitted from roof top
vents in urban complex terrain, a simple nomograph would be used.
In the case of a nationally distributed pollutant emitted from
urban sources, the nomographs for complex and flat terrain could
be used in proportion to the distribution of the sources.
D1.2 Input Meteorology for Modeling Activities — Elmer Robinson
In assessing the potential impact of toxic chemical materials,
it is useful to use various modeling or simulation techniques.
These techniques can be carried out at several levels of sophisti-
cation with concurrent savings in cost and effort but at a sacri-
fice in output specificity. The several levels of modeling and
sophistication have been denoted as Levels I, II, and III for this
particular toxic material evaluation.
In Levels I and II, certain savings in effort are available
by using more-or-less standard meteorological factors in the
dispersion estimates. While this procedure will always reduce
the precision of the resulting dispersion, there are some readily
recognizable situations where an answer can be grossly in error.
Since a number of these suspect situations can be easily recog-
nized and dealt with, it seems worthwhile to consider this prob-
lem so that meteorological input data can come from a "modified
standard" scheme rather than a single set of standard values
such as have been described and used by Youngblood.
For the dispersion modeling program under Levels I and II
approaches, the most important meteorological factors are wind
speed and stability. Wind direction frequencies for prevailing
53
-------
winds may also be important in certain situations. It is recog-
nized that under Levels I and II that site-specific information
will probably not be available within the constraints of cost
and time. Level I studies will use some sort of national annual
average input while Level II studies will seek out local climato-
logical data, such as STAR tabulations, for use in the analyses.
For Level I studies, it is proposed that the meteorological
factors be based on regional or national data with categories
selected from a tabulation of climatological data. For example,
readily available maps of wind climatology could be used to
develop a table of wind speed factors for specific sites in
various regions. For some sources, similar data could be developed
based on conditions averaged over the largest urban areas in the
country (maybe 25?). Table Dl. is an example matrix for a
regional table of annual average wind speed and prevailing direc-
tion frequency, and Table D2. is an example of the average urban
area situation.
TABLE Dl. REGIONAL ANNUAL AVERAGE CLIMATOLOGY FOR
LEVEL I MODELING ESTIMATES
Annual Average Persistence
Region Windspeed Factor
NE Coastal
NE Inland
SE Coastal
SE Inland
Appal. Val.
Upper Midwest
3 mph
2
3
2
2
4
60%
30
60
30
60
30
NW Coast 4 60
54
-------
TABLE D2. AVERAGE ANNUAL CLIMATOLOGY FOR URBAN AREA
SOURCES
Annual Average Persistence
Urban Size Windspeed Factor
Over 1,000,000 pop. 3 mph 40%
over 500,000
over 100,000
over 50,000
Although the- data listed in the above tables are only
illustrative guesses and are not valid numerical values, a
range of at least a factor of 2 should be expected in these
values. They would transfer directly to similar factors in
terms of dispers.ion and. pollutant concentration .estimates . Thus
since the use of this improved climatological input data does
not involve significant additional effort in the preparation of
a Level I summary (once the factors have been derived), it is
suggested that this scheme of regional climatic factors be
developed for this type of report.
Stability could also be summarized on a regional basis
dependent on whether the source had a diurnal cycle. For
example, an 8-5 urban source might be characterized by C stabil-
ity while a rural, 24-hour source may have D stability.
In Level II studies, some more-or-less local climatology in
the form of STAR tabulations will be used in the analysis. These,
however, should reflect reasonable changes in climatology from
the observation site, usually an airport, to the source site.
An investigation using the STAR data should be encouraged to
consider modifying the STAR frequencies to determine whether
modified conditions would be important in the concentration
and population exposure indices. The sort of changes to be
considered include (1) channelling of winds at a valley site
55
-------
but not in open, upland sites, (2) more stable and more limited
wind variability at coastal sites compared to inland locations,
(3) less stable and lower wind speeds for urban areas compared
to a local rural airport site, etc. The STAR-type of evaluation
has also been found to over-predict the frequency of D stability
in many areas; i.e., arid western sites. If such a shift to
unstable and/or stable conditions relative to neutral conditions
would influence the exposure estimates significantly, then such
changes should be considered.
D2.0 SPECIFIC MODEL RECOMMENDATIONS - Walt Dabberdt
The recommended, generalized procedure for selecting and
implementing dispersion models is outlined schematically in
Tigure Dl. The procedure applies equally to any of the three
levels of analysis. However, the rigor and specificity of the
process and hence the results vary considerably among the three
levels. Accordingly, the guiding premise for each level should
first be recognized:
LEVEL SOURCE OF ANALYSIS
I Objective is to provide a first-order ranking of
the various substances and source categories.
Approach employs a simplistic methodology (i.e.,
nomographs) to relate annual-average normalized
concentration to fundamental classes of source
configuration (i.e., plume height and spatial
extent of source) and emission features (i.e.,
daytime vs. continuous emissions) using a default
set of meteorological data.
II Objective is to identify specific sources and pol-
lutants that may require controls. Approach employs
available, site-specific emissions, meteorological,
and monitoring data as inputs to a comprehensive
dispersion model (normally of the steady-state
Gaussian variety) that is of the type referenced
in the OAQPS modeling guidelines, UNAMAP series,
or equivalent..
56
-------
Source
Configuration
Pollutant
Characteristics
Terrain
Features
Available
Data
Analysis of
Concentration Data
Selection
Averaging or
Exposure Time
Receptor
Characteristics
Specification
of Model
Inputs
Modifleattoi
Calibration
of Model
Reaction Rates
Meteorology
Source Features
Physical:
Wake Effects
Plume Rise
OftaAii U«l«i*l
stacx neiuju
Source Layout
Figure Dl. Model selection and implementation procedure,
57
-------
LEVEL SOURCE OF ANALYSIS .
Ill Objective is to generate the best set of concentra-
tion estimates necessary to determine the degree of
control required of specific sources or source
categories. Approach employs models that match in
a mathematical sense the physical characteristics/
features of the site (i.e., source configuation
and rate, meteorology, terrain, pollutant trans-
formations/sinks, exposure time, and receptor loca-
tions) . Analysis is to be supported by site-
specific meteorological data and the model results
are to be evaluated using representative or sur-
rogate monitoring data.
Having determined the pollutant or substance of interest and
the level of analysis that is necessary, the first major step in
the analysis procedure is MODEL SELECTION. In the case of Level
I analysis, the modeling should already be available in the form
of a short series of nomographs that relates normalized concen-
tration (annual average) to source height and extent, and the
"schedule" of emissions (i.e., daytime only vs. continuous).
With Levels II and III, the model selected must meet the require-
ments imposed by: . source configuration, pollutant characteristics
(inert vs. reactive, gas vs. particle), terrain features, and the
exposure or integration time. Whereas the emphasis of the EPA
to date is only on annual-average concentrations, it is the
strong recommendation of the modeling committee that discrete
events (such as accidents) also be considered. While the
integration time may be short, the subsequent dosage received
may often be 'the-maximum for all -sources/time periods. That is
to say, the corresponding risk to health may be greatest for
sub-annual releases and exposures. Level II analysis will nor-
mally employ steady-state Gaussian plume (or equivalent) models,
whereas more sophisticated models may be required for Level III
(e.g., puff, grid, or particle-in-cell methods).
The second major task is the SPECIFICATION OF MODEL INPUTS.
Levels II and III modeling seek to provide maximum realism in
58
-------
the results, consistent with the quality of the inputs. In the
case of Level II, available site-specific or site-representative
data are to be used whereas Level III demands that data be used
that properly and thoroughly represent all important aspects of
the dispersion problems. For example, Level II may employ air-
port Service A wind data while Level III may require construction
of mesoscale flow patterns and the generation of time-dependent
curvilinear traj ectories.
In operating the model to simulate annual-average concentra-
tions, it is important to recall that the basic objective is to
calculate annualized population exposures (APE). Symbolically,
the APE is the cross-product of emissions (Q), dispersion (D),
and population (P)integrated over a one-year period, or:
APE = f I E Qtj D^ P^ (1)
where i = 1, 24 (hours per day) and j = 1, 365 (days per year).
In performing the modeling (i.e., the D or QD calculations), it
is imperative that the actual computed APE be equivalent to what
would be obtained through a strict adherence to the formulation
given in (1) above.
59
-------
D3>.0 ESTIMATION OF ERRORS - Richard Pollack
Atmospheric modelers seldom, if ever, have sufficient moni-
toring data to validate a modeling technique for more than a few
locations. In addition, the assumptions required by a model are
never perfectly satisfied in a given application; as a result,
it is difficult to specify confidence limits on the accuracy of
any given model. The most desirable method for assessing.the
validity of a model in a given application is to compare it with
a portion of the monitoring data not used in the calibration
and/or adjustment of the model. The recommendations presented
below are based, where possible, on this approach. For the
cases where modeling is to be performed in the absence of data,
we recommend considering the use of a default option; i.e.,
confidence limits which are sufficiently wide to account for any
reasonable possible error.
The details of the process of estimating confidence limits
where data are available and for setting a default option or other
procedure where it is not should be the subject of an EPA guid-
ance document to be supplied to the contractors. The modeling
group did not feel that addressing these issues in full detail
was within the scope of the brief meeting.
The modeling group recommends that, given the guidance
referred to above for each level, a modeling exercise should be
accompanied by numerical estimates of confidence limits and a
qualitative discussion of the potential sources of error. These
sources include, but are not necessarily limited to, those factors
listed in Table D3. The recommendation for analysis at specific
levels is as follows:
Level I: The development of the nomograph should include
validation exercises to develop estimates of expected
errors by comparing predictions and observations in a
variety of situations. The developer of the nomograph
60
-------
should consider the option of providing error estimates
stratified by the input variables; for example,separate
estimates for urban and rural cases, low level and
elevated sources, flat vs. complex terrain, etc.
Level II: Error estimates should be based on site-
specific data if available. Otherwise a default option
(or alternative approach as recommended by the study
suggested above) is recommended. The report should
include a discussion of the processes and factors (e.g.,
Table D3.) which are (un)accounted for in the model.
EPA guidance for developing confidence levels should
be followed.
Level III: For Level III, site-specific data should
be required for model validation. EPA guidance for
developing confidence levels should be followed.
TABLE D3. POTENTIAL SOURCES OF ERROR
Use of non-site-specific meteorology of source geometry
Inadequate representation of atmospheric process or
other factors influencing concentrations such as:
terrain
removal processes
wake effects
source height
plume rise
entrainment
The quantity to be estimated is ideally that which would
be obtained by "hourly summing" considering fully temporality
varying emissions, meteorology, and population. In fact, our
approach will not attempt to estimate this directly but rather
estimate either the product of the annual averages of E, x/Q. P
or a "slightly stratified" version of this, e.g., day, night, and
seasons. Therefore, there are two quantities at error:
61
-------
1. Our errors in estimating the annual averages or
"slightly stratified" variables -- This type of
error is discussed in the previous pages.
2. The error introduced by using annual averages or
"slightly stratified" data as a surrogate for the
full temporal case.' This error would require a
separate effort to address because it depends
largely on covariance terms; i.e., emissions and
x/Q correlate both seasonally and diurnally, popula-
tion movement, x/Q» and wind direction are all cor-
related. This is why the concept of convolving the
distributions won't work.
In other words, if we could estimate all annual averages
perfectly, how far off would we be from actual human exposure?
The multiple convolution may work reasonably well, strati-
fied by day, night, and season.
D4.0 DETAILED MODELING RECOMMENDATIONS -- Richard Schulze
For Level II studies, the committee recommends the use of EPA
Guideline models with routinely recorded meteorological data.
Where plume chemistry or terrain is important, the model used
should incorporate adjustments for these factors.
For annual concentrations in urban areas, model such as CDM
and TCM appear to be the most appropriate of the Guideline models,
but both have several shortcomings. For example, both models
include buoyancy plume rise and ignore momentum plume rise and
the effects of stack and building downwash. These models use
rural-derived formulas for estimating vertical plume speed and
then adjust the observed stability category to make a rough
adjustment to urban conditions. The vertical dispersion param-
eter for stack emission is adjusted to such a great extent that
it is probably not appropriate to locate receptors within 500
meters of stacks with heights of 20 meters or less. The models
do allow some rough estimates of the effects of daytime and night-
time emission.
62
-------
The area source algorithm in CDM requires extensive computer
time and some experience in correctly selecting the density of
the sampling points. In TCM the user must exercise judgement as
to the number of grid squares across which emissions from one
grid sequence are to be spread. Both user's guides caution that
best results from the area source algorithm are obtained when
there is relatively limited, 50 percent or less, variation in
area source emission rates from adjacent grid squares.
In rural areas a model such as VALLEY -- with no terrain
adjustments— or AQDM can be used. These models do not make
the adjustments for calculating the vertical dispersion coeffi-
cients for urban stability categories and the initial plume speed
contained in TCM and CDM. The VALLEY model is also suggested
-for analyzing concentrations in rough terrain after making adjust-
ments to include the effects of entrainment from plumes from
point sources.
Meteorological data should be carefully selected. In hilly
terrain and shoreline areas, the airport observations of wind
speed and direction are often not typical of those found at
plant sites. There appears to be a tendency for observed wind
directions to be skewed at some locations, especially those
operated by the Armed Forces. One must also carefully specify
STAR data. For CDM and TCM, a "day-nite" STAR should be used,
while a regular six-category STAR should be used with VALLEY
and AQDM.
For Level III studies, the committee recommends that the
use of refined models be considered, particularly if changes
in hazard exposures of greater than 20 percent from the use of
Level II models are identified.
For example, if hour-by-hour emission rates and population
exposures in specific grid squares are available, then it would
63
-------
be appropriate to use a sequential model, such as CRSTER, RAM,
or ISC to calculate annual averages. If sources and exposures
in hilly or mountainous terrain are analyzed, the use of site-^
specific meteorology is suggested.
The investigator for Level III studies is urged to critically
examine the assumptions in the models suggested for the Level II
analysis and evaluate the benefits of refinements. These include
trajectory models in hilly or mountainous areas, increased sophis-
tication in analyzing plume chemistry and plume removal processes,
studies of the effects of sea breezes in shoreline areas, the
effects of building downwash, and the use of monitored data to
adjust modeled results to realistic levels.
D5.0 MODELING CONSIDERATIONS - Steve,Hanna
D5.1 Importance of Short Term Studies
The EPA emphasis on annual averages is understandable
because of the dependence of carcinogenicity on dosage. The
modeling done so far has been concerned strictly with routine
emissions. However, accidental releases also can result in high
dosages, but over a short period of time. Possible accidents
include explosions, truck and train wrecks, and industrial
spills. In these cases, employees or nearby residents could
receive dangerous dosages within a few minutes. I recommend
that models be developed for handling accidental releases.
These models would need to have the following capabilities:
1. Ability to calculate plume rise or sink for dense
gases or highly momentum-dominated plumes,
2. Ability to account for source effects such as the
initial dilution imposed by the presence of a
railroad car,
3. Ability to calculate plume trajectory in a complex
terrain situation.
64
-------
4. Ability to calculate instantaneous puff diffusion.
D5.2 Use of Monitoring Data
If .good monitoring data exist, then they should be used to
improve the modeling technique. In general, high-quality monitor-
ing data should have precedence over model predictions. It is
recognized that high-quality data depend greatly on intelligent
site selection and competent maintenance of the instrument.
Level I studies are intended to be simple screening pro-
cedures to separate the innocuous chemicals from the potentially
dangerous ones. Even at this level, good monitoring data can be
used to "calibrate" simple diffusion models. For example, if
the average model prediction by an urban box model is 10 yg/m3,
and a properly located monitoring station indicates an average
of 2 yg/m3, then a factor of 0.2 should be applied to all future
model predictions for different scenarios.
In Level II studies, the monitoring data can be compared
with model predictions (UNAMAP series) to develop regression
equations between observed and predicted concentrations. For
critical cases under Level III, monitoring data can be used to
adjust internal model assumptions. Empirical relations between
observed concentrations and meteorological parameters and source
parameters should be developed. For example, this analysis may
show that high concentrations usually occur with a certain wind
direction and stability. Or, the analysis may suggest that an
emission source has been missed in the area around a certain
monitor. In all cases, the basic philosophy is to use all the
good data to the fullest extent possible.
65
-------
D5.3 Importance of Including Transport and Interaction With
Water, Land, and Biota
Potential carcinogens can follow pathways through several
media. For example, a substance released from a smokestack into
the air may fall out or deposit onto a certain watershed. It
then will be transported by ground water and streams or may enter
biota. Some of it may be resuspended into the air or may vola-
tilize. In other words, a multimedia model is needed for several
of the substances that will be studied. Emphasis must be on
transfer across interfaces, such as dry and wet deposition,
resuspension, or volatilization. These effects can be treated
in a Level I study using current techniques. Level II and III
studies would require extensive literature searches and possibly
more research.
D6.0 INDOOR/OUTDOOR EXPOSURE -- Richard Schulze (1st paragraph)
and Noel deNevers (2nd paragraph)
The committee assumed that outdoor air pollutant levels were
typical of population exposures. In fact, most people spend the
majority of their time indoors where concentrations are frequently
only a fraction of levels found outdoors. The committee made no
attempt to develop recommendations relating indoor to outdoor
pollutant levels or population exposure levels to outdoor con-
centrations .
Although indoor and outdoor concentrations are not the same,
if our health effects data are based on epidemiology, with con-
centrations measured outdoors (as in the studies which led to
the particulate, S02 and N02 ambient air quality standards),
then the real assumption in using outdoor air quality to esti-
mate health damage is that the relation of indoor to outdoor
air contamination will be the same in the future as it was during
the period of the epidemiological study. This is a much more
plausible assumption than the one that indoor and outdoor air
quality are the same. ,>.
-------
APPENDIX E
REGULAR USERS OF CENSUS DATA
U.S. BUREAU OF CENSUS REGIONAL OFFICES
ALABAMA
Graduate Program in Hospital
and Health Administration
Attn: Dr. Tee H. Hiett
School of Community and Allied
Health
University of Ala. in Birmingham
•Room 205 SCAB
Birmingham, Ala. 35294
(205) 934-5223
*Alabama Development Office
Attn: Gil Gilder
State Capitol
Montgomery, Ala. 36130
(205) 832-6400
C, 1-7, 10
*Alabama Public Library Service
Attn: Anthony Miele
6030 Monticello Drive
Montgomery, Ala. 36130
(205) 277-7330
C, 1-7, 10
**University of Alabama
Attn: Dr. Carl Ferguson
Center for Business and
Ecomonic Research
Box AK
University, Ala. 35486
(205) 348-6191
C, 1-7, 10
ALASKA
Institute of Social and Economic
Research
Attn: Mr Lee Huskey
University of Alaska
Anchorage, Ak. 99504
(907) 278-4621
or
Attn: Mr. Gary Lu
Fairbanks, Ak. 99701
(907) 479-7436
B, 1-5
ARIZONA
*Northern Arizona University
Attn: Dr. Ron Gunderson
College of Business and Research
Flagstaff, Ariz. 66011
(602) 523-3657
C, 1-10
**Arizona Department of
Economic Security
Attn: Mr. Richard A. Froncek
1717 West Jefferson
P.O. Box 1623 - 045Z
Phoenix, Ariz. 85005
(602) 255-5984
C, 1-10
67
-------
*Department of Library, Archives
and Public Records
Attn: Sally Hronek
Federal Documents Section
Capitol, Third Floor
1700 West Washington
Phoenix, Ariz. 85007
(602) 255-1121
C, 1-10
Resource Consultants, Inc.
Attn: James T. Kirk
P.O. Box 7132
Phoenix, Ariz. 85011
(602) 265-1161
B, 1-5, 8-10
*Arizona State University
Attn: Glenda Rauscher
Coordinator of Research
College of Business Administration
Tempe, Ariz. 85281
(602) 965-3961
C, 1-10
*University of Arizona
Attn: Dr. Lee B. Jones
Dean of the Graduate College
Administration Bldg., Rm. 501
Tucson. Ariz. 85721
(602) 626-4032
C, 1-10
ARKANSAS
**Industrial Research and
Extension Center
Attn: Dr. Forrest Pollard
University of Arkansas
P.O. Box 3017
Little Rock, Ark. 72203
(501) 371-1971
C, 1-3, 5, 6, 10
*0ffice of the Governor
Attn: Joan Roberts
State/Federal Relations
Little Rock, Ark. 72201
(501) 371-2611
C, 1-3, 5, 6, 10
CALIFORNIA
Urban Decision Systems, Inc.
Attn: James A. Paris
2032 Armacost Avenue
P.O. Box 25953
Los Angels, Calif. 90025
(213) 826-6596
A, 1-5, 8, 9
Research Systems, Inc.
Attn: Gerald J. Jansen
365 South Meadows Ave.
Manhattan Beach, Calif. 90266
(213) 372-8838
D, 1-3, 5-6
Allstate Research and Planning Center
Attn: Nicholas Gannam
Allstate Insurance Company
321 Middlefield Road
..Menlo Park, Calif. 94025
(415) 324-2721
D, 1-5, 7, 9
Decision Making Information
Attn: Ron Hinckley
2700 N. Main Street, Suite 800
Santa Ana, Calif. 92701
(714) 558-1321
C, 1-5, 10
Demographic Research Company
Attn: Joseph J. Weissmann
233 Wilshire Blvd.
Santa Monica, Calif. 90401
(213) 451-8583
D, 1-10
Speron, Inc.
Attn: Edward J. Skowron
14621 Titus Street
Van Nuys, Calif. 91401
(213) 873-4114
B, 1-5, 8, 9
R-marketers
68
-------
COLORADO
DISTRICT OF COLUMBIA
Business Research Division
Atten: C.R. Goeldner
Graduate School of Business
Administration and School of
Business -Administration
University of Colorado
Campus Box 420
Boulder, Colo. 80309
(303) 492-8227
B, 1-6
Bureau of Business and
Public Research
Attn: Dr. William Duff
School of Business
University of Northern Colorado
Greeley, Colo. 80631
(303) 351-2080
B, 1-4
CONNECTICUT
Urban Decision Systems, Inc.
Attn: James A. Lunden
21 Charles St.
P.O. Box 551
Westport, Conn. 06880
(203) 226-7367
A, 1-5, 8, 9
DELAWARE
**0ffice of Management, Budget
and Planning
Attn: Helen Gelof
Townsend Bldg.
P.O. Box 1401
Dover, Del. 19901
(302) 678-4271
C, 1-3, 5, 8-10
*University of Delaware
Attn: John Falcone, Director
Computer Center
Newark, Del. 19711
(302)" 453-6065
C, 1-3, 5, 8-10
Applied Urbanetics, Inc./Equal
Employment Opportunity Data Systems
Attn: Laura DeLong
Third Floor
1701 K Street, N.W.
Washington, D.C. 20006
(202) 331-1800
D, 1-10
Metropolitan Washington Council
of Governments
Attn: Frank Goodyear
1225 Connecticut Ave., N.W.
Washington, D.C. 20036
(202) 223-6800
A, 1-10
FLORIDA
Census Access .Program
Attn: Ray Jones
University of Florida Libraries
Library West
Gainesvilla, Fla. 32611
(904) 392-0361
D, 1-3, 6, 10
Regional Information
Coordinating Center
Attn: Steven A. Logan
Tampa Bay Regional
Planning Council
3151 Third Avenue N., Suite 540
St. Petersburg, Fla. 33713
(813) 898-0891
B, 1, 3-6, 9
Applications Programming Group
Attn: Brent Dorhout
Florida State University Computing
Center
Tallahassee, Fla. 32306
(904) 644-3860
C, 1-5, 7-9
69
-------
GEORGIA
*0ffice of Computing Activities
Attn: Dr. Margaret Park
Director of Information Services
University of Georgia
Athens, Ga. 30602
(404) 542-3106
C, 1-5, 7, 9, 10
**0ffice of Planning and Budget
Attn: Tom Wagner
State of Georgia
270 Washington Street, S.W.
Atlanta, Ga. 30334
(404) 656-2191
C, 1-5, 9, 10
R-Gov't
HAWAII
Department of Budget and ••Finance
Attn: Frances J. Santos
Electronic Data Processing Div.
P.O. Box 150
Honolulu, Hawaii 96810
(808) 548-5910
B, 1, 3
R-St. gov't
IDAHO
Center for Research, Grants
and Contracts
Attn: Dr. Emerson C. Maxson
Boise State University
1910 University Drive
Boise, Idaho 83725
(208) 385-15 71
B, 1^10
ILLINOIS
Chicago Area Geographic
Information Study
Attn: Edwin N. Thomas
Department of Geography
University of Illinois at
Chicago Circle
Box 4348
Chicago, 111. 60580
(312) 936-3112 or 996-5274
B, 1-5, 7-10
Dept. of Sociology, Anthropology,
and Social Work
Attn: Vernon C. Pohlmann
Illinois State University
Normal, 111. 61761
(309) 438-2387 or 436-7667
C, 1-6, 8-10
Concordia Teachers College
Attn: Dr. William Kammrath
7400 Augusta
River Forest, 111. 60305
(312) 771-8300, Ext. 299
D, 1-5, 7-10
INDIANA
Research Associates, Inc.
Attn: John J. Carter
P.O. Box 44640
Indianapolis, Ind. 46244
(317) 266-6926
B, 1-5, 8, 9
IOWA
University of Iowa
Attn: Chia-Hsing Lu
Laboratory for Political Research
Scheffer Hall, 321A
University of Iowa
Iowa City, Iowa 52242
(319) 353-3103
B, 1-5, 10
KANSAS
Center for Public Affairs
Attn: Fred A. Cleaver
607 Blake Hall
University of Kansas
Lawrence, Kans. 66045
(913) 864-3700
C, 1-5, 7
70
-------
KENTUCKY
Urban Studies Center
Attn: Dr. James M. Brockway
University of Louisville
Gardencourt, Alta Vista Rd.
Louisville, Ky. 40205
(502) 588-6626
B, 1-5, 7, 8, 10
LOUISIANA
*Experimental Statistics Dept.
Attn: Dr. Nancy Keith
173 Agriculture Administration Bldg.
Louisiana State University
Baton Rouge, La. 70803
(504) 388-8303
C, 1, 3, 5, 7, 9, 10
*Reference Department
Attn: .Blanche Cretini
Louisiana State Library
P.O. Box 131
Baton Rouge, La. 70821
(504) 342-4913
C, 1, 3, 5, 7, 9, 10
**State Planning Office
Attn: Leigh Harris
4528 Bennington Avenue
Baton Rouge, La. 70808
(504) 925-1584
C, 1, 3, 5, 7, 9, 10
Bureau of Business Research
Attn: Dr. Charles 0 Bettiner, III
College of Business Administration
Northeast Louisiana University
Monroe, La. 71201
(318) 372-2123
B, 1, 3-5, 10
*Division of Business and
Economic Research
Attn: Vincent Maruggi
University of New Orleans
Lake Front
New Orleans, La. 70122
(504) 283-0248
C, 1, 3, 5, 7, 9, 10
Louisiana Tech. University
Attn: Dr. Edward J. 0'Boyle
Research Division, CAB
P.O. Box 5796
Ruston, La. 71270
(318) 257-3701
B, 1-3, 5, 6, 9, 10
MARYLAND
BRC Associates, Inc.
Attn: Belur K. Radhakrishnan
4336 Montgomery Avenue
Bethesda, Md. 20014
(301) 656-2996
E, 1-5, 8
Systems Sciences, Inc.
Attn: Miles Letts
4720 Montgomery Lane
Bethesda, Md. 20014
(301) 654-9343
D, 1, 2-5, 7, 9
Data Services Division
Attn: Thomas E. Jones
Westat, Inc.
11600 Nebel Street
Rockville, Md. 20852
(301) 881-5310
D, 1-9
MASSACHUSETTS
Urban Data Processing, Inc.
Attn: Robert G. Coyne
20 South Avenue
Burlington, Mass. 01803
(617) 273-0900
D, 1-4, 6-9
MICHIGAN
Inter-University Consortium for
Political and Social Research
Attn: Dr. Jerome M. Clubb
P.O. Box 1248
Ann Arbor, Mich. 48106
(313) 764-8508 or 763-5010
D, 1-3, 5, 10
71
-------
Computing Services Center
Attn: Barbara B. Wolfe
5950 Cass Avenue
Wayne State University
Detroit, Mich. 48202
(313) 577-4777 or 577-4778
B, 1-3, 5, 7-9
Data Coordination Division
Attn: Patricia C. Becker
Planning Department
City of Detroit
3400 Cadillac Tower
Detroit, Mich. 48226
(313) 224-6389
B, 1, 3-6, 8-10
Michigan State University
Attn: Anders C. Johanson
Computer Laboratory •
East Lansing, Mich. 48824
(517) 355-4684
B, 1-5, 7-10
Tri-County Regional Planning
Commission
Attn: Gerald J. Burger
2722 E. Michigan Avenue
Lansing, Mich. 48912
(517) 487-9424
B, 1, 3-6. 8-10
Oakland County Advance
Programs Group
Attn: Jeffrey A. Kaczmarek
1200 North Telegraph Rd.
Pontiac, Mich. 48053
(313) 858-0732
A, 1, 3-6
MINNESOTA
Minnesota Analysis and Planning
System
Attn: Thomas A. Ehlen
Agriculture Extension Service
University of Minnesota
415 Coffey Hall
St. Paul, Minn. 53108
(612) 376-7003
D, 1-6, 9, 10
MISSISSIPPI
Department of Sociology
Attn: Ellen S. Bryant
Mississippi State University
P.O. Drawer C
State College, Miss. 39762
(601) 325-5024
B, 1, 3-6, 10
The Center for Population Studies
Attn: Dr. Max W. Williams
The Institute of Urban Research
The University of Mississippi
Bondurant Hall, 3W
University, Miss. 38677
(601) 232-7288
B, 1-6, 10
MISSOURI
Public Affairs Information Service
Attn: Ed Robb
University of Missouri-Columbia
10 Professional Building
Columbia, Mo. 65201
(314) 882-8256 or 882-8367
C, 1-7, 9, 10
Office of Administration
Attn: Ray Harrington
Division of Budget and Planning
P.O. Box 809
Room 129, Capitol Bldg.
Jefferson City, Mo. 65102
(314) 751-2073
C, 1, 3-6
R-gov't
Mid-America Regional Council
Attn: Peter S. Levi
20 West Ninth Street Bldg.
Third Floor
Kansas City, Mo. 64105
(816) 474-4240
C, 1, 3-9
72
-------
University of Missouri-St. Louis
Attn: John G. Blodgett
Computer Center
8001 Natural Bridge Road
St. Louis', Mo. 63121
(314) 453-5131
C, 1-5, 7-9
MONTANA
Research and Information
Systems Division
Attn: R. Thomas Dundas, Jr.
Montana Department of
Community Affairs
Capitol Station
Helena, Mont. 59601'
(406) 449-2896
B, 1-7, 10
NEBRASKA
Academic Computing Services
Attn: D.F. Costello
University of Nebraska
3835 Holdrege
Lincoln, Nebr. 68508
(402) 472-3763
C, 1-10
NEVADA
Central Data Processing Division
Attn: Gordon L. Harding
Department of General Services
State of Nevada
Carson City, Nev. 89701
(702) 885-4091
B, 1-5, 7-9
NEW JERSEY
Princeton- Rutgers Census
Data Project
Attn: Jane Wolin
Center for Computer and
Information Services
Rutgers University
Hill Center, Busch Campus
P.O. Box 879
Piscataway, N.J. 08854
(201) 932-2483
D, 1-10
Princeton-Rutgers Census
Data Project
Attn: Judith S. Rowe
Princeton University Computer
Center
87 Prospect Avenue
Princeton, N.J. 08540
(609) 452-6052
D, 1-10
NEW MEXICO
Bureau of Business & Economic
Research
Attn: Lee Brown
Institute for Applied Research
Services
Robert 0. Anderson School of
Management
University of New Mexico
Albuquerque, N.M. 87131
(505) 277-2216
B, 1-6, 8, 9
NEW YORK
Community Services Research
and Development Program
Attn: Frank Rens
Department of Social and
Preventive Medicine
State University of New York at
Buffalo
2211 Main Street
Buffalo, N.Y. 14214
(716) 831-5521
B, 1-9
CDP Marketing Information Corp.
Attn: Clifford W. Potanza
7 High Street
Huntington, N.Y. 11743
(516) 549-5801
D, 1-4, 8, 9
National Planning Data Corp.
Attn: Patricia Allard
P.O. Box 610
Ithaca, N.Y. 14850
(607) 273-8208
D, 1-7, 10
73
-------
Market Statistics
Attn: Edward J. Spar
633 Third Avenue
New York, N.Y. 10017
(212) 986-1800
D, 1-5
Tri-State Regional Planning
Commiss ion
Attn: Max Schwartz
World Trade Center
New York, N.Y. 10048
(212) 938-3323
C, 1-9
Technical Assistance Center
Attn: Gordon DeVries
State University of New York
Plattsburgh, N.Y. 12901
(518) 564-2214
B, 1-5
NORTH CAROLINA
*Institute for Research in
Social Science
Attn: Susan Clarke
University of North Carolina
Manning Hall 026A
Chapel Hill, N.C. 27514
(919) 966-2411
C, 1-6, 10
Systems Sciences, Inc.
Attn: Edgar A. Parsons
Box 2345
Chapel Hill, N.C. 27514
(919) 929-7116
D, 1-5, 7, 9
**North Carolina Department of
Adminis tration
Attn: Karan Bunn
Division of State Budget and
Management
116 West Jones Street
Raleigh, N.C. 27603
(919) 733-7061
C, 1-6, 10
*North Carolina Department of
Cultural Resources
Attn: David Sevan
State Library
109 East Jones St.
Raleigh, N.C. 27611
(919) 733-3343
C, 1-6, 10
OHIO
Northeast Ohio Areawide
Coordinating Agency
Attn: Frederick E.J. Pizzedaz-
1501 Euclid Avenue
Cleveland, Oh. 44115
(216) 241-2414
A, 1-3, 5-9
Census Processing Center
Attn: Michael Melton
Battelle-Columbus Laboratories
505 King Avenue
Columbus, Oh. 43201
(614) 424-7760
D, 1-9
**0hio Department of Economic
and Community Development
Attn: Jack Combs
Office of Research
30 East Broad Street
Columbus, Oh. 43215
(614) 466-2115
C, 1-7, 10
*0hio State University Libraries
Attn: Bernard Bayer
Mechanized Information Center
1858 Neil Avenue Mall
Columbus, Oh. 43210
(614) 422-3480
C, 1-7, 10
OKLAHOMA
University Computer Center
Attn: Eldean Bahm
Oklahoma State University
Mathematical Sciences Bldg.
Stillwater, Okla. 74074
(405) 624-6301
B, 1, 3, 5, 6
74
-------
OREGON
Bureau of Governmental
Research and Service
Attn: Dr. Robert E. Keith
University or Oregon
P.O. Box 3177
Eugene, Ore. 97403
(503) 686-5234
B, 1-5, 7-9
Center for Population Research
and Census
Attn: Edward Schafer
Portland State University
P.O. Box 751
Portland, Ore. 97207
(503) 229-3922
C, 1, 3, 5-7
PENNSYLVANIA
Robinson Associates, Inc.
Attn: Morris Olitsky
Bryn Mawr Mall
15 Morris Avenue
Bryn Mawr, Pa. 19010
(215) 527-3100
E, 1-8, 10
Central Management Information
Center
Attn: L. Jeanne Yingling
State of Pennsylvania
Building #33, Bomb Rd.
Harrisburg International Airport
Middletown, Pa. 17057
(717) 787-1648
B, 1-3, 5
R-gov't, ed., quasi-public
Contract Research Associates
Attn: Marjorie L. McCann
251 Harvey Street
Philadelphia, Pa. 19144
(215) 438-9391
C, 1-5, 9
Delaware Valley Regional
Planning Commission
Attn: Edward McNichol
Penn Towers Building
1819 John F. Kennedy Blvd.
Philadelphia, Pa. 19103
(215) 568-3211
A, 1-4, 6-9
K.H. Thomas Associates
Attn: Kenneth H. Thomas
University City Science Center
Suite 200
3508 Market Street
Philadelphia, Pa. 19104
(215) 382-2700
C, 1-6, 10
Innovative Systems, Inc.
Attn: Robert J. Colonna
341 Fourth Avenue
.Pittsburge, Pa. 15222
(412) 391-2364
C, 1-5, 8, 9
Southwestern Pennsylvania
Regional Planning Commission
Attn: Wade G. Fox
564 Forbes Avenue
Pittsburgh, Pa. 15219
(412) 263-3500
B, 1-9
RHODE ISLAND
Social Science Data Center
Attn: Prof. James M. Sakoda
Department of Sociology
Brown .University
Maxey Hall
Providence, R.I. 02912
(401) 863-2550
B, 1-5
TENNESSEE
Bureau of Business and
Economic Research
Attn: Paul R. Lowry
Memphis State University
Memphis, Tenn. 38152
(901) 454-2281
C, 1-6
R-gov't, quasi-gov't, nonprofit
75
-------
Regional and Urban Studies
Information Center
Attn: Dr. Andrew S. Loebl
Oak Ridge National Laboratory
P.O. Box X
Oak Ridge, Tenn. 37830
(615) 574-5966 or 574-5957
D, 1-5, 7
TEXAS
Institute of Urban Studies
Attn: Dr. Frank W. Anderson
The University of Texas at
Arlington
P.O. Box 19588
Arlington, Tex. 76019
(817) 273-3071
C, 1-5, 9, 10
Texas Natural Resources
Information System
Attn: C.R. Baskin
Systems Central Office
P.O. Box 13087
Austin, Tex. 78711
(512) 475-3321
B, 1, 3, 5-7, 9
National Planning Data Corp.
Attn: Lawrence B. Miller
Suite 208
4560 Belt Line Road
Dallas, Tex. 75234
(214) 980-0198
D, 1-7, 10
Houston-Galveston Area Council
Attn: Dr. Joe W. Pyle
3701 West Alabama
Suite 200
P.O. Box 22777
Houston, Tex. 77027
(713) 627-3200
B, 1-6, 8-10
UTAH
Population Research Laboratory
Attn: Dr. Yun Kim
Utah State University
Logan, Ut. 84322
(801) 752-4100 Ext. 7662
C, 1-5, 8, 10
Bureau of Economic and
Business Research
Attn: R. Thayne Robson
University of Utah
Room 404
College of Business Building
Salt Lake City, Ut. 84112
(801) 322-7274
B, 1-10
VIRGINIA
C.A.C.I., Inc. - Federal
Attn: Ronald C. Steorts
1815 N. Fort Myer Drive
Arlington, Va. 22209
(703) 841-7800
D, 1-5, 7, 10
**Tayloe Murphy Institute
Attn: Dr. Julie Martin
Box 6550
University of Virginia
Charlottesville, Va. 22096
(804) 924-7451
C, 1-5, 10
International Data and
Development, Inc.
Attn: J.C. Barrett
P.O. Box 472
Merrifield, Va. 22116
(703) 525-7806
D, 1-5, 7-9
R-Private Companies
**Division of Planning and Budget
Attn: Robert Griffis
445 Ninth Street Office Bldg.
P.O. Box 1422
Richmond, Va. 23211
(804) 786-7771
C, 1-5, 10
76
-------
Data Use and Access
Laboratories Inc. (DUALabs)
Attn: Deirdre Gaquin
1601 North Kent Street
Suite 900
Rosslyn, Va. 22209
(703) 525-1480
D, 1-5, 10
R-gov't, nonprofit
WASHINGTON
Population, Enrollment, &
Economic Studies Division
Attn: John R. Walker
Office of Financial Management
State of Washington HOB (AL-01)
Olympia, Wash. 98504
(206) 753-5617
B, 1-6, 10
WEST VIRGINIA
State of West Virginia
Attn: Daniel S. Green
Program Support Services
Governor's Office of Economic
and Community Development
State Capitol Complex
Building 6, Room 548
Charlestown, W.Va. 25305
(304) 348-4010 or 348-2246
B, 1, 3, 5, 6
WISCONSIN
**Wisc. Demographic Services Center
Attn: Donald G. Holl
Department of Administration
Room B-110, 1 West Wilson Street
Madison, Wise. 53702
(608) 266-1927 or 266-1067
C, 1-10
*Applied Population Laboratory
Attn: Dr. Doris Slesinger
University of Wisconsin
240 Agriculture Hall
1450 Linden Drive
Madison, Wise. 53706
(608) 262-1510
C, 1-10
WYOMING
Institute for Policy Research
Attn: G. Fred Doll
University of Wyoming
P.O. Box 3925
University Station
Laramie, Wyo. 82071
(307) 766-5141
B, 1, 3
77
-------
*SDC participating agency **SDC lead agency
Geographic Coverage Codes: A-county or groups of counties in one or more
States, D-all States, E-tapes obtained as required.
Service Codes: 1-tape copies, 2-special files/extracts, 3-print-outs,
4-analytical rpts./area comparisons and profiles, 5-consultation, 6-map copies,
7-computer mapping/graphics, 8-address matching/geocoding, 9-GBF/DIME
assistance, 10-training
Restrictions Code: R (followed by organizations served)
BUREAU REGIONAL OFFICES AND DATA USER SERVICES OFFICERS
The 12 regional offices maintained by the Bureau of the Census in cities
outside the Washington area offer a variety of services to users of Census
Bureau data. These offices are now staffed with Data User Services Officers
who can answer inquiries about census publications and other Bureau products,
assist users in the access to and use of census data needed for specific
applications, and make presentations to groups interested in the statistical
programs and products of the Bureau. Regional office addresses and the names
and phone numbers of Data User Services Officers are listed below. When
writing, include "U.S. Bureau of the Census" in the address.
Atlanta, Ga. 30309: 1365 Peachtree St., N.E., Room 638
Wayne Wall (404-881-2274)
Boston, Mass. 02116: 441 Stuart St., 8th Floor
Judith Cohen (617-223-0668)
Charlotte, N.C. 28202: 230 South Tryon St., Suite 800
Lawrence McNutt (704-371-6144)
Chicago, 111. 60604: 55 E. Jackson Blvd., Suite 1304
Stephen Laue (312-353-0980)
Dallas, Tex. 75242: 1100 Commerce St., Room 3C54
Valerie McFarland (214-767-0625)
Denver, Colo. 80225: P.O. Box 25207, 575 Union Blvd,
Gerald O'Doimell (303-234-5825)
Detroit, Mich. 48226: U.S. Federal Bldg. and Courthouse,
Room 565, 231 W. Lafayette St.
Timothy Jones (313-226-4675)
Kansas City Kans. 66101: One Gateway Center,
4th and State Sts.
Kenneth Wright (816-374-4601)
78
-------
Los Angeles, Calif. 90049: 11777 San Vicente Blvd.,
8th Floor
E.J. (Bud) Steinfeld (213-824-7291)
New York, N.Y. 10007: 26 Federal Plaza,
Federal Office Bldg., Room 37-130
Jeffrey Hall (212-264-4730)
Phildelphia, Pa. 19106: William J. Green, Jr.
Federal Bldg., Room 9226, 600 Arch St.
David Lewis (215-597-8314)
Seattle Wash. 98174: 915 Second Ave., Room 312
Larry Hartke (206-442-7080)
79
------- |