TECHNICAL REPORT DATA
(Please read Instructions on the reverse before complc
1. RFPORT NO. 2.
EPA-600/R-97-005
3.
4. TITLE AND SUBTITLE
Improving Emissions Estimates with Computational
Intelligence, Database Expansion, and Comprehensive
Validation
5. REPORT DATE
January 1997
6. PERFORMING ORGANIZATION CODE
7. AUTHOR(S)
J. G. Cleland, V. E. McCormick, H. L. Waters, J. R.
Youngberg, and J. A. Zak
8. PERFORMING ORGANIZATION REPORT NO.
RTI 81U-5388
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Research Triangle Institute
P. O. Box 12194
Research Triangle Park, North Carolina 27709
10. PROGRAM ELEMENT NO.
11. CONTRACT/GRANT NO.
EPA Cooperative Agreement
CR819542-01-0
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
National Risk Management Research Laboratory*
Research Triangle Park, NC 27711
13. TYPE OF REPORT AND PERIOD COVERED
Final; 8/92-1/94
14. SPONSORING AGENCY CODE
EPA/600/13
15. supplementary notes ^kerl project officer is P. Jeff Chappell, Mail Drop 62, 919/
541-3738. (*)The former Air and Energy Engineering Research Laboratory.
i6. abstract repQrt discusses an EPA investigation of techniques to improve methods
for estimating volatile organic compound (VCC) emissions from area sources. Using
the automobile refinishing industry for a detailed area source case study, an emis-
sion estimation method is being developed that uses advanced computational tech-
niques and updated, comprehensive, emissions-related information. New computa-
tional techniques contributing to the estimation method are fuzzy logic, neural net-
works, and genetic algorithms. This method development requires a thorough char-
acterization of the area sources, an analysis of current emission estimation methods,
the development and execution of a nationwide industry activity survey, and a compi-
lation and analysis of the survey results and other explanatory variables. Results
will be captured in a personal-computer-based emissions estimation system called
VOCEES (VOC Emissions Estimation System). VCCEES has been developed as a
dual-use tool that prepares VOC emissions inventories and analyzes the impact of
many factors on emissions. This methodology can be easily extended to other area
sources.
17. KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
b. 1DENTIF1 ERS/OPEN ENDED TERMS
c. cosati Field/Group
Pollution Organic Compounds
Emission Volatility
Estimating Painting
Artificial Intelli- Automobiles
gence Neural Nets
Computation Algorithms
Proving Genetics
Pollution Control
Stationary Sources
Computational Intelli-
gence
Database Expansion
Comprehensive Vali-
dation
13B 07 C
14 G 20 M
13H
13 F
06A 06D, 06P
12A
OnC
13. DISTRIBUTION STATEMENT
Release to Public
19. SECURITY CLASS (This Report)
Unclassified
21. NO. OF PAGES
125
20. SCCURITY CLASS (This page)
Unclassified
22. PRICE
EPA Form 2220-1 (9-73)
-------
NOTICE
This document has been reviewed in accordance with
U.S. Environmental Protection Agency policyand
approved for publication. Mention of trade names
or commercial products does not constitute endorse
ment or recommendation for use.
-------
FOREWORD
The U.S. Environmental Protection Agency is charged by Congress with pro-
tecting the Nation's land, air, and water resources. Under a mandate of national
environmental laws, the Agency strives to formulate and implement actions lead-
ing to a compatible balance between human activities and the ability of natural
systems to support and nurture life. To meet this mandate, EPA's research
program is providing data and technical support for solving environmental pro-
blems today and building a science knowledge base necessary to manage our eco-
logical resources wisely, understand how pollutants affect our health, and pre-
vent or reduce environmental risks in the future.
The National Risk Management Research Laboratory is the Agency's center for
investigation of technological and management approaches for reducing risks
from threats to human health and the environment. The focus of the Laboratory's
research program is on methods for the prevention and control of pollution to air,
land, water, and subsurface resources; protection of water quality in public water
systems; remediation of contaminated sites and groundwater; and prevention and
control of indoor air pollution. The goal of this research effort is to catalyze
development and implementation of innovative, cost-effective environmental
technologies; develop scientific and engineering information needed by EPA to
support regulatory and policy decisions; and provide technical support and infor-
mation transfer to ensure effective implementation of environmental regulations
and strategies.
This publication has been produced as part of the Laboratory's strategic long-
term research plan. It is published and made available by EPA's Office of Re-
search and Development to assist the user community and to link researchers
with their clients.
E. Timothy Oppelt, Director
National Risk Management Research Laboratory
i i i
-------
Abstract
The Environmental Protection Agency is investigating techniques to improve
methods for estimating volatile organic compound (VOC) emissions from area
sources. Using the automobile refinishing industry for a detailed area source case
study, an emission estimation method is being developed that uses advanced
computational techniques and updated, comprehensive, emissions-related
information. New computational techniques contributing to the estimation method
are fuzzy logic, neural networks, and genetic algorithms. This method
development requires a thorough characterization of the area sources, an analysis
of current emission estimation methods, the development and execution of a
nationwide industry activity survey, and a compilation and analysis of the survey
results and other explanatory variables. Results will be captured in a personal-
computer-based emissions estimation system called VOCEES (VOC Emissions
Estimation System). VOCEES has been developed as a dual-use tool that
prepares VOC emissions inventories and analyzes the impact of numerous factors
on emissions. This methodology can easily be extended to other area sources.
Acknowledgements
Sincere thanks are extended to P. Jeff Chappell, Project Officer, EPA/AEERL, who
conceived and directed the methodology development using computational
intelligence, inferential techniques, and a new broad set of continuously updated
databases.
iv
-------
TABLE OF CONTENTS
Abstract iv
Acknowledgements iv
List of Figures V1-
List of Tables vi i
1. INTRODUCTION 1-1
2. COMPLETED METHOD DEVELOPMENT PROCESS 2-1
2.1 Area Source Characterization 2-2
2.1.1 Refinishing Procedures 2-4
2.1.2 Environmental Concerns and Controls 2-6
2.2 Review of Current Estimation Methods 2-9
2.2.1 Emission Inventories 2-12
2.2.2. Compilation of Air Pollutant Emission Factors 2-14
2.2.3 Control Techniques Guidelines (CTG) 2-15
2.2.4 State Implementation Plans (SIPs) 2-16
2.2.5 EPA Study of Recent Methods 2-20
2.2.6 Accuracy of Emissions Inventories 2-22
2.3 Database Development 2-23
2.3.1 Surrogate Data: Sources. Variables. Availability. Accuracy .... 2-26
2.3.2 Prescreening of Variables by Statistical Analysis 2-34
2.4 Analytical Tool Set Selection 2-45
2.5 Configuration of the VOC Emission Estimation System (VOCEES) . . . 2-53
3. EXTENDED METHOD DEVELOPMENT 3-1
3.1 National and Intensive Local Area Surveys of Auto
Refinishing Shops 3-4
3.2 Use of Computational Intelligence (CI) 3-10
3.3 Use of VOC-Containing Product Manufacturers Data 3-21
3.4 Sampling and Analysis 3-27
3.5 Past Emission Estimation Data and GIS Interpretations 3-30
4. INTERIM RESULTS AND CONCLUSIONS 4-1
5. REFERENCES 5-1
Appendix A Survey Questionnaire A-l
Appendix B Sampling and Analysis Dispersion Model B-l
v
-------
List of Figures
Figure 2-1. National Trends in the Number of Body Shops 2-3
Figure 2-2. Body Shop Size Distribution by Number of Employees 2-4
Figure 2-3. Difference in Emission Estimates (emission estimates
using per employee emissions - emission estimates using
per capita emission factor) 2-11
Figure 2-4. National VOC Estimations by Different Methods 2-18
Figure 2-5. Information Sources for Emission Estimates and Data
Distribution 2-23
Figure 2-6. System Configuration in Context of Method Elements 2-54
Figure 3-1. Recommended New Method for Estimating Emissions 3-2
Figure 3-2. Representation of 5900 Samples and 6 Local Intensive
Surveys 3-6
Figure 3-3. Neural Networks in VOC Emissions Estimation 3-12
Figure 3-4. Membership functions for paint use and percent change
in VOC estimate 3-15
Figure 3-5. The 10 EPA Regions 3-16
Figure 3-6. Statewide Representation of 10 EPA Regions 3-17
Figure 3-7. Disclosure laws resulted in information being incomplete
or non-existent for the counties colored in light gray 3-18
Figure 3-8. Method Flowchart with Evolving Expert Systems 3-20
Figure 3-9. Mass Balance of Emissions Flowchart 3-29
Figure 4-1. Automotive Body Shops in San Francisco, as Displayed
in Maplnfo 4-3
Figure 4-2. Changes in County Emission Levels for California Due to
Regulatory Impact in 1988, 1993, and 1995 4-3
Figure 4-3. Differences in Three Explanatory Variables Used in
Generating Emissions Estimates 4-3
Figure 4-4. SIP Information Generated by VOCEES for the
Raleigh-Durham-Chapel Hill, NC Area 4-4
vi
-------
List of Tables
Table 1-1. Project Initiatives for New Method Development 1-3
Table 2-1. Contacts, Organizations, and Information Sources for the
Automobile Refinishing Industry 2-2
Table 2-2. Recommended Emission Factors for the Automobile
Refinishing Area Source 2-11
Table 2-3. National Emissions and Activity Levels for VOC from
Automobile Refinishing (Emission Factor Rating: C) 2-14
Table 2-4. Baseline VOC Emissions from Automobile Refinishing
by Facility Type (Athey, 1988) 2-15
Table 2-5. Baseline VOC Emissions from Automobile Refinishing
by Facility Type (Draft, 1991) 2-16
Table 2-6. Auto Refinishing Activity Factors 2-17
Table 2-7. Other Emission Estimation Methods for Automobile
Refinishing from State Implementation Plan 2-19
Table 2-8. Automobile Refinishing Variables from Recent Review 2-21
Table 2-9. Comparison of National Paint and Solvent Usage
Estimates for Automobile Refinishing 2-22
Table 2-10. Major Data Items and Sources 2-25
Table 2-11. Major Census Bureau Items by Geographic Level
and Industry 2-27
Table 2-12. Summary of State Regulations for VOC Emissions
from Automobile Refinishing 2-32
Table 2-13. Summary of Explanatory Variables and Their National
Totals 2-36
Table 2-14. Annual Summary of Explanatory Variables by State: 1987 2-37
Table 2-15. Dependent Variables for Regression Examples 2-38
Table 2-16. Receipts for SIC 7532 Predicted by 5 Independent Variables . . . 2-38
Table 2-17. Employment in 7532 as Predicted by 5 Independent
Variables 2-39
Table 2-18. Results from Motor Vehicle Accident Analysis 2-40
Table 2-19. Estimating SIC 5198 Sales 2-40
Table 2-20. Estimating SIC 5198 Employment 2-41
Table 2-21. Estimating SIC 7532 Total Painters (Single-Variable
Regression) 2-42
Table 2-22. Single Variables for Emissions Estimates from Number
of Painters: 48 States 2-43
Table 2-23. Single Variable Linear Regressions for 48 States:
Automobile Top and Body Repair and Paint Shops SIC
7532 Sales ($1000) from the 1987 Economic Census 2-43
Table 3-1. Estimated Accuracies of Solvent Use Data 3-6
Table 3-2. Example rules used to modify VOC estimate using
coatings usage 3-15
vi i
-------
Table 3-3. Estimation of Unknown Data for County Using Fuzzy
Weights and Center-of-Moment Method 3-18
Table 3-4. Manufacturer Product Category Questionnaire 3-23
Table 3-5. Manufacturer Geographic Distribution Questionnaire 3-24
Table 3-6. Auto Refinishing Air Concentrations of Organies 3-27
vi i i
-------
1. INTRODUCTION
Under the 1990 Clean Air Act Amendments, state air pollution control agencies
are required to prepare and maintain emission inventories for carbon monoxide
(CO) and precursors to ozone (03). Ozone is photochemieally produced in the
atmosphere when volatile organic compounds (VOCs) are mixed with CO and
oxides of nitrogen (NOx) in the presence of sunlight. The role of the emission
inventory is to identify the types of emission sources present in an area, the
amount of each pollutant emitted by each type of source, and any emission control
devices being employed. In developing these inventories, the state agencies may
either use emission estimation methods endorsed and provided by the U.S.
Environmental Protection Agency (EPA) or their own methods. New methods of
estimation are sought which will be more accurate, efficient, and cost effective
than current methods. Results of this work will either validate existing area
source emission estimation methods or recommend improved methods. The EPA
Air and Energy Engineering Research Laboratory (AEERL) has directed this
research effort.
Stationary sources of pollutant emissions are designated as either point sources
or area sources. Whereas point sources are inventoried on an individual basis,
area sources are processes, activities, or businesses that are too small or too
numerous to be practically tracked as individual emission sources. The distinction
between point and area sources is an annual emissions threshold, such as 10 tons
(907 kg) of volatile organic compounds (VOCs) per year per source.
The required components of an emission estimation methodology are: (1)
calculation of the emission estimates; (2) temporal and spatial allocation of the
emissions; (3) validation of the emission estimates, and (4) speciation of the
emissions. This project concentrates on the first three components. Tools are
available from EPA that can be used to develop a VOC inventory grouped into
reactivity classes suitable for modeling (VOC/PM, 1990). The criteria established
for this research effort include: reasonable accuracy and cost; dynamic and robust
behavior: use of readily available information; and ease of use.
There are two primary reasons to improve existing emission estimation
methods: (1) the current emission estimation methods require further development
to improve accuracy and ease of use; and (2) innovative tools have emerged which
may contribute significantly to improving estimation methods. Among the issues
and uncertainties are:
Most solvents used by any one solvent area source category are also used by
other industries. This prevents use of solvent manufacturer production
figures alone for estimating emissions from one solvent source category.
1-1
-------
There are many very different solvent area sources categories. Therefore,
reasonably accurate emission estimates require area source category-specific
estimation methods.
Current methods endorse the use of data sources whose published data are
from 2 to 5 years old and may misrepresent a significant segment of the
area source industry. Federal information derived from business payroll
records often provides no record of very small businesses. Also, data at the
county level for certain Standard Industrial Classification (SIC) codes are
often not disclosed.
Current methods do not consider the dynamic factors that impact area
sources such as local economics, changing technology, and regulatory
influence. Nor do they consider human behavior and consumption patterns.
New technologies, such as innovative computational techniques, offer the
opportunity to improve existing methods. This project investigates the
appropriateness of applying advanced, inference-based computational intelligence
techniques, such as neural networks, fuzzy logic, and genetic algorithms, to
emissions estimation. These techniques make it possible to determine and use
relationships between two domains {e.g., industry variables and emission levels),
and to augment or supplant traditional mathematical models. Also, since all
emissions have a geographic component in that each emission has a source, it is
helpful to capture and represent emissions-related data using a geographic
information system (GIS).
This research is based on three essential directives to ensure an innovative and
improved emission estimation method. The directives are:
1) Focus
The research concentrates on a case study of a single area source, automobile
refinishing. in order to demonstrate the feasibility of developing an improved
method for deriving VOC emission estimates from area sources. Automobile
refinishing was chosen as representative of VOC emitting sources based on
three criteria: relative environmental significance, national prevalence, and an
accessible information base representative of such sources. The new method of
source evaluation and emission estimation can be extended to many other area
sources.
2) Product Development
The project goal is to produce a prototype, personal computer-based, user-
interactive system and database that allows a novice, personal computer (PC)
user to estimate VOC emissions from automobile refinishing at the county level,
anywhere in the United States. Maplnfo, Inc.'s Maplnfo software has been
used as both the interface and database management system. Maplnfo's
1-2
-------
generic interface was customized using MapBasic, a BASIC language compiler,
to provide an interface specific to emissions estimation.
3) Innovation and Improvement
The project plan incorporates several new or expanded features for emissions
estimation. These are listed in Table 1-1.
Table 1-1. Project Initiatives for New Method Development
INITIATIVE
DESCRIPTION
Annual updating
The method will use information and logic that is continuously reviewed and revised.
Improved
validation
New validation activities include: (a) preparation of the most thorough nationwide
survev of the automobile refinishintr industry (i.e.. auto refinishing shops) to date.
The survey will obtain industry answers about activity levels (e.g., number and types
of employees or repair jobs), solvent usage, and emission control. In addition,
extensive product distribution data have been requested from all major automotive
naint manufacturers: fb) selective sampling and chemical analvsis of local sources has
been designed to establish actual mass rates of shop emissions: and (c) literature
review and consulting with exnerts to provide new insieht into maior influences on
emissions.
Intensive area
source
characterization
Characterization includes main industry variables influencing emission levels. In
addition to the national survey, the research team has held meetings with shop
managers and industry consultants, attended national conferences, contacted
associations, and analyzed the literature.
More extensive
data
Data analysis will extend down to the county level. Data from federal and state
agencies, trade associations, and industry sources have significantly increased the
EPA information base related to this area source.
Improved data
recovery and
structuring
One result is a systematic and efficient approach for using information, emphasizing
data sources which are regularly updated (e.g., annually) and which are readily
accessible at minimal cost to EPA and state agencies.
Improved data
correlations
This involves applying statistics, for prescreening of the best data relationships.
Application of
imprecise
information and
expert opinion
Such "artificial intelligence" techniques as fuzzy logic, genetic algorithms, and rule-
based expert systems can augment the use of pertinent information which had been
neglected to this point.
Regulatory policy
consideration
The method incorporates regulatory influence on emissions at local, state and national
levels.
Alternate
methods review
Methods (used by States) other than the standard EPA guidelines and emission
factors have been examined.
1-3
-------
2. COMPLETED METHOD DEVELOPMENT PROCESS
The area source emissions estimation method development process involves:
1) area source characterization, examining materials balances, process
operations and controls, and economic influences,
2) review of current emissions estimation methods,
3) database development, including retrieval and screening of the best
accessible data correlating with area source characteristics and emissions,
4) selection of analytical methods (the "tool set"), for calculating or inferring
final emissions levels and their distributions,
5) system configuration for data input computation and reporting, and
6) validation of results (for this case study, emphasizing a national survey of
automobile refinishing shops).
These main elements are necessary for essentially all area source emissions
estimation improvements, although specific techniques can vary. Steps 1) and 2)
are essentially complete, steps 3) to 5) are near completion, and a plan for step 6)
(covered in Section 3) is complete. The completed portions of the method
development are discussed in the following.
-------
2.1 Area Source Characterization
A detailed characterization of the automobile refmishing industry was
conducted in order to fully consider the issues and variables associated with the
industry's VOC emissions. Table 2-1 lists some of the contacts and companies
which make up this $23 billion per year industry.
Table 2-1. Contacts, Organizations, and Information Sources for the
Automobile Refmishing Industry.
Data Sources
Examples
Organizations
National Paint and Coatings Association (NPCA), Automotive Service
Association (ASA), Automotive Service Industry Association (ASIA),
Inter-Industry Conference on Auto Collision Repair (I-CAR), Society of
Collision Repair Specialists (SCRS), Motor & Equipment Manufacturers
Association (MEMA)
Paint
Manufacturers
BASF, E.I. DuPont de Nemours & Co., Nason Automotive Finishes, PPG
Industries, The Sherwin-Williams Co., Sikkens/AKZO Coatings, ICI
Autocolor, American Standox, Inc., and Spies Hecher, Inc.
Associated
Industries
Binks Manufacturing Company, Herkules Equipment Corporation,
Safety-Kleen Corporation, 3M Automotive Trades Division
Publications
I-CAR Technical Manuals, Mitchell Manuals, Chilton's Manuals, Body
Shop Business, Body Shop Tool & Equipment News, Automotive Body
Repair News, Motor, and local trade publications (e.g., Hammer and
Dolly, Midwest Autobody Magazine, Alabama Automotive Business)
Body Shops
An estimated 65,000 shops in 1993. A nationwide survey of 5900 body
shops has been proposed to obtain data for both method development
and validation.
Consultants
CCC Information Services, CARSTAR Automotive, Inc., AutoChex, Inc.,
Mitchell-ADP Xchange (MAX), and individual industry consultants
Conference
Attended 1992 National Autobody Congress and Exposition (NACE). At
NACE, numerous contacts were made with industry representatives and
consultants. NACE also provided a more complete understanding of the
scope and magnitude of the industry.
According to industry representatives, the impact of technological and
regulatory changes, along with economic factors, has provoked a steady decrease
in the number of body shops nationwide over the past twenty years. While there
may have been as many as 125,000 body shops in 1975 (Athey, 1988), there are
presently estimated to be 63,000 body shops in operation. Compared with an
estimated 83,000 (Athey, 1988) shops in 1988, and including estimates from
industry experts that there will be half that number by 1997 (and only 25,000 by
the year 2000), a nearly linear decrease in body shops with time is projected, as
shown in Figure 2-1.
2-2
-------
120
Industry Experts
80
ABI
60
County Business fctterra
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Year
Figure 2-1. National Trends in the Number of Body Shops
However, Figure 2-1 also shows that numbers from County Business Patterns
from 1975 to 1989 (County, Annual) contradict the drastic drop-off over the past
17 years. While EPA recommends using County Business Patterns for making
emissions estimates, industry estimates seem more realistic. Data obtained from
American Business Information (ABI) computer files shows a nearly flat growth
trend from 1981 to 1993, but agrees much more closely with industry experts on
the current number of shops.
Several trends drive the market. For example, although the number of U.S.
cars on the road increases each year (growing well ahead of the population curve),
that growth is not reflected in the number of accidents that take place because of
improved vehicle safety and more stringent enforcement of traffic laws {e.g.,
against drunk driving). Manufacturers' finishes have also become more corrosion
resistant. More efficient paint spray guns mean less paint is wasted, again
lowering consumption per paint job. Overall, these opposing changes have tended
to stagnate the volume of refimshing over the past decade. As cars become
smaller, there is less surface area to paint per accident, which lowers paint
consumption. Such trends have tended to stagnate the volume of refinishing over
the past decade.
Beyond the number of body shops which can be captured through industry
surveys, County Business Patterns, and the Yellow Pages as represented by ABI,
there exist many unlicensed "backyard" shops. Industry experts estimate that the
number of these shops has grown as increased regulatory costs and economic
conditions have reduced the number of licensed body shops. Estimates indicate
that these shops may represent an additional 25-40% more body shops than the
number which are on record. These "unseen shops" are potentially a large source
of VOC emissions since their control technology is more likely to be primitive and
2-3
-------
they arc more likely to disregard laws and regulations. Figure 2-2 shows the
preponderance (45%) of shops employing only 1-2 people.
2&
V)
a
20
W S
-a
O x
5 S
II
15--
ia
HPS
pi
fe--
H
fSS
11 ; gg
1-2 <45%) 2-3(20%) 3-7 (19%) 7-9 (11 %) 9-10(3%) 10+(2%)
Number of Technicians
Figure 2-2. Body Shop Size Distribution by Number of Employees
2.1.1 Refinishing Procedures
While there may be minor differences from shop to shop, the procedure for
refinishing an automobile can usually be thought of as a four-step process. These
four steps are: 1) vehicle preparation, 2) vehicle priming, 3) topcoat application,
and 4) equipment cleanup. The vehicle preparation consists of cleaning followed
by a solvent-based compound to remove grease, wax, and other contaminants.
These compounds are usually made entirely of VOC's and contain solvents such as
toluene and xylene.
Priming creates and makes up for deficiencies in the present finish or the
underlying surface. Materials used include primer-surfacers (nitrocellulose
lacquer, acrylic lacquer, and alkyd enamel). Solvents are prevalent in topcoats
because of the need to properly blend a repair with the surrounding finish
(blending requires the application of successively thinner coats, and more
solvents). The materials used in topcoats can be divided into lacquers, enamels,
and polyurethanes. Their relative usage in today's market is 34%, 54%, and 12%,
respectively.
When cars are manufactured, the paint is applied by machine and baked in an
oven. This is a relatively simple process since the car is essentially a metal frame
2-4
-------
at the time. This process has enabled manufacturers to use more polyurethanes
because cyanates in the polyurethane are not a direct hazard to employees. Auto
refinishers, however, cannot use high temperatures to cure their finishes since this
would ruin the interiors, plastic, and electronics present. They must use different
materials and techniques from the manufacturers while reproducing the same
finish. Drying time becomes important to a refinisher because shorter drying
times reduce contamination from dust and dirt and more cars can be finished in a
given time. Drying time is even more important for metallic paints.
According to the National Paint and Coatings Association (Connolly, 1990), over
36 million gallons of coatings (topcoats and primers) were sold in the U.S. in 1989.
That number has remained basically constant over the past ten years. Coatings
can generally be categorized as two major types: solvent-based coatings (which
contain VOCs), and water-borne coatings.
Solvent-Based. Coatings
Solvent-based coatings are by far the most common in automobile refinishing.
These coatings can be divided into groups that cure by three different methods:
crosslinking (two-component epoxies and urethanes, and baked acrylic and alkyd
enamels), noncrosslinking (thermoplastic acrylics and vinyls, and cellulosic
lacquers), and drying (some alkyds) (Connolly, 1990).
Solvent-based materials can also be divided into lacquers, enamels, and
urethanes. Lacquers were the first materials used for automotive refinishing.
Lacquer paints rely completely on solvent evaporation to transfer pigments and
therefore are have high solvent and VOC content. They are usually based on
acrylic resins and, in some cases, acrylics modified with nitrocellulose. Enamel
use by manufacturers began in the 1950's and by the refinishing industry in the
1960's. Enamels can be divided further into two categories, acrylic enamels and
alkyd enamels, which provide different appearance and durability characteristics.
Enamels still rely predominantly on solvent evaporation to provide pigment
transfer and adhesion, but also use chemical linkages formed during the curing
process. Since they rely less on solvent evaporation, they usually contain less
VOCs than lacquers. Urethanes are the newest materials used by automotive
refinishers. They rely more on chemical processes to form the bonds necessary for
adhesion and therefore potentially contain less VOCs than either lacquers or
enamels. While this makes them more attractive environmentally, they also
contain isocyanates which are potentially hazardous to the people applying the
coatings.
2-5
-------
Water-Borrie Coatings
Water-borne coatings can be divided into four groups: emulsions, latexes, water
solubles, and colloidal dispersions (Connolly, 1990). In emulsions and latexes, the
principal film resin is in a fully polymerized state and is suspended in the water
by the addition of a dispersing agent. These polymers are made by emulsion
polymerization. The resins used in water soluble and colloidal dispersions systems
have relatively low molecular weights and undergo further cross linking or
polymerizations during the curing stage. This category includes all electrocoat
systems.
Water-borne coatings are only now becoming popular in the reflnishing
industry. While car manufacturers have used them for years as undercoats and
primers, their long drying times have precluded their use by refinishers.
Advances in technology still have not been able to provide a suitable topcoat
formulation which will replace conventional topcoats in the near future.
Future Trends
Automobile original equipment manufacturers (OEMs), dictate the changes in
styling demands, increases in finish durability, and new multi-component
substrates. New paint styling will come in the form of specialty pigments and the
use of colored aluminum and colored micas in paint formulations. OEMs are
beginning to use tinted clearcoats, more trieoats, and super-smooth elearcoats that
the refinishers must be prepared to match. Increased corrosion resistance will
come through improvements in electro-deposition. Resistance to environmental
deposits will come through the use of fluorinated clearcoats (which, since they are
so durable and long-lasting, present a problem of proper disposal). There will also
be improvements made in resistance to sunlight (UV stability). Expanding use of
treated steel and advanced alloys, aluminum, plastics, and composites means most
refinishers must address the proper coating of plastics and composites (plastics are
especially sensitive to solvents, which break down their chemical structure).
The consequences of the changes which are taking place will be seen in four
areas: 1) larger shops will be necessary to maintain profitability and keep pace
with technology, 2) coatings will carry a higher apparent cost because of their
chemical makeup, 3) new finishes will become more and more durable, and 4)
more training will be necessary to keep up with advances and changes.
2.1.2 Environmental Concerns and Controls
Refinishing regulation and industrial controls can be divided into two
components ~ environment and health. Environmental regulations primarily
concern volatile organic compounds (VOCs). Most regulations require the
2-6
-------
reduction of VOCs through either the use of reduced VOC coatings or the control
of vent emissions. While automobiles themselves are the largest source of VOCs,
the refinishing of these autos accounts for only 3% of VOC emissions. The three
main approaches to control are 1) use of lower-VOC coatings, 2) use of enclosed
cleaning devices, and 3) increase of paint (to surface) transfer efficiency. There is
also a movement to require licenses to purchase coatings. This movement is
intended to allow coatings only to be sold to "legitimate" body shops and not non-
compliant, "backyard" shops.
Paint manufacturers have met some State requirements through the
development of low-VOC coatings and waterborne coatings. Low-VOC coatings
have forced the industry away from lacquer and enamel coatings and toward
urethanes.
Typical VOC contents range from 4.8 lbs/gal for acrylic urethanes to 6.8 lbs/gal
for acrylic lacquers. New laws in Southern California in 1995 will limit topcoat
VOCs to 3.5 lb/gal, primer/surfacer VOCs to 2.1 lb/gal, primer/sealers to 3.5 lb/gal,
and basecoat/clearcoat combinations to 4.5 lb/gal and less. Ultimately these
regulations will require primers with 2.1 lb VOC/gal and baseeoat/clearcoats with
3.5 lb VOC/gal. For the primers, this means the industry will be turning towards
water-borne products and some very-high-solids solvent-based products. Basecoats
will also became water-borne, while clearcoats will consist of water-borne and 2-
pack urethanes. Binders used in these products will move towards resins
(isocyanates in particular) trapped in a water thinner.
The move to low-VOC coatings will result in 1) better application conditions
(i.e., better airflow, better temperature control), 2) improved curing techniques
(i.e., use of downdraft booths and infrared heaters), 3) more training for painters
and jobbers, and 4) reductions in VOCs through increased paint transfer efficiency
(TE) from spray gun to car surface. While conventional spray guns have TEs of
25-50%, high TE guns are in the 50-75% range. Most high TE guns are high
volume, low pressure (HVLP) guns. While first generation HVLP guns generally
gave poor performance, current models produce acceptable results.
Equipment cleaning is a significant contributor of VOC emissions. An
increasing number of body shops use special cleaning stations to remove coatings
from their spray guns and to allow reuse of solvents in cleaning, lowering
emissions.
Another method of reducing VOC emissions is the addition of control technology
on paint spray booths. The main purpose of a spray booth is to provide a safe and
clean environment, and is not necessarily to trap emissions. Booth exhaust fumes
can be controlled by 1) thermal incineration, 2) catalytic incineration, and 3)
carbon adsorption. Thermal incineration co-fires a fuel with the exhaust from the
2-7
-------
booth stacks to destroy VOCs. Destruction efficiency is limited only by the cost of
fuel. Catalytic incineration uses a catalyst to create the oxidation reaction at a
lower temperature than thermal incineration (approximately 900 ฐF compared to
1800 ฐF) and has a destruction efficiency nearly as high with much less fuel. In
carbon adsorption, the exhaust stream is passed through an activated carbon bed
where the carbon particles adsorb VOCs. Care must be taken to properly
maintain the bed and prevent saturation.
Besides the controls described, other body shop operations can be better
managed to reduce emissions. For example, control techniques include tight
fitting containers, reducing spills, rigid inventory control, tracking worker use
rates, mixing paint to need, providing operator training, proper cleanup methods,
and solvent recycling with in-house equipment or by lease agreements.
Health issues are a major concern within the industry. Isocyanates are quite
hazardous when improperly handled. At the Saturn automobile manufacturing
facility, for example, when urethane coatings are being applied, workers are not
allowed to enter the paint booths. If they do enter, the paint machines
automatically shut down. In addition to isocyanates, heavy metals such as lead
and chromate are being phased out. Some sacrifices have been made, especially in
coating corrosion resistance. Manufacturers are not barred from putting most
compounds in paints (they only need to label them properly), but mainly provide
their own motivation for excluding hazardous materials.
2-8
-------
2.2 Review of Current Estimation Methods
Most EPA-endorsed methods for estimating solvent emissions from area sources
have been derived from a methodology developed as part of the National
Emissions Data System (NEDS) (Myers, 1975). This methodology determines
national solvent emissions through material balance. Solvent use by an area
source is typically correlated with population or employment in a particular SIC to
create an activity level or activity factor (e.g., lbs. solvent/person). Activity levels
are combine with emission factors (solvent emitted/solvent used), which are based
on source-specific emission measurements to determine emissions by states and
counties. Emission factors are "developed from only a limited sampling of the
emissions source population for any given category, and the values reported are an
average of those limited samples and may not be statistically representative of the
population" (Rethinking, 1991).1 The EPA recommended emission factor for auto
refinishing is 2000 lb. emission/ton solvent used, or 1. The NEDS area source
emission estimation method has been used in several emission inventory efforts.
These efforts include the 1985 National Acid Precipitation Assessment Program
(NAPAP) emissions inventory prepared by EPA (Demmy, 1988), the Regional
Ozone Modeling for Northeast Transport (ROMNET), and the Area and Mobile
Source (AMS) Subsystem of the Aerometric Information Retrieval System (AIRS)
(Aerometric, 1992), which is planned to replace NEDS while using similar
estimating techniques.
There have also been detailed studies performed on individual area sources in
order to provide guidance on area source emission control technology. Methods for
estimating emissions are often included in these studies. A 1988 study of VOC
emissions from automobile refinishing area sources uses a rather complex
calculation dependent upon the final thickness of applied coatings, in thousandths
of an inch, while categorizing the entire industry into three types of establishment
-- small, medium and volume shops (Athey, 1988). This approach has been
criticized as being inappropriate for an industry which has a large variance in
operating characteristics (National, Nov. 1991).
The Emission Factor Rating (A through E, with A being the best) is a qualitative rating which "reflects the quality
and the amount of data on which the factors are based". (Compilation, 1985) Factors based on many observations
or on more widely accepted test procedures are assigned higher rankings {e.g., an A rating means that 10 or more
tests were conducted at different plants using a single valid reference measurement method or equivalent
technique). EPA also states that, "at most, a rating should be considered an indicator of the accuracy and
precision of a given factor used to estimate emissions from a large numher of sources." (Compilation, 1985;
Joyner, 1991)
2-9
-------
The basic approach in estimating emissions is by a simple calculation:
Emissions = Activity level x Emission factor x [1-(CE x RE. x RP)]
where: CE = Control efficiency pcrcent/100
RE = Rule effectiveness percent/100
RP = Rule penetration percent/100
CE is related to control technology, (e.g., carbon filtration of VOCs from auto
refinishing paint booths.) RE is an adjustment to reflect that air pollution rules
are not able to assure full compliance with regulatory requirements at all times.
RP is a measure of the extent to which a rule applies to a source category. EPA's
guidance stipulates use of a default value of 0.80 for RE unless data can be
collected to determine a "true" value. RP is calculated as a ratio of the controlled
emissions covered by a regulation applicable to a category to the total uncontrolled
emission for a category. The RP term may be appropriate to deal with "backyard
shops" engaging in auto refinishing. If such operations are not subject to air
pollution rules, estimates of uncontrolled emissions are needed to estimate RP.
This requires being able to account for emissions both from licensed operations
and backyard shops. RE and RP are of no concern of CE = 0, or if there are no
applicable regulations for a category. However, in more and more areas rules will
apply, so area source methods need to address RE and RP.
This study on improved estimation methods is focused almost entirely on
improved activity levels. Industry characterization, and most importantly a future
national survey, can also produce better evaluations for CE, RE, and RP.
Activity levels can be found in a number of EPA references. Table 2-2 contains
activity factors for automobile refinishing from AP-42 Compilation of Air Pollutant
Emission Factors (Compilation, 1985) and its Supplement D (Joyner, 1991), the
EPA State Implementation Plan (SIP) guidance document (Procedures; 1991), and
AIRS/AMS (Kimbrough, 1992)(Aerometric, 1992). In analyzing those data, it can
be noted that:
The AP-42 activity levels and National VOC emissions did not change for
more than ten years. They may not have changed since the AP-42 was first
published in 1968.
The SIP guidance document's per capita activity level is 21% greater than
that of the AP-42 document, while the per employee activity level is 48% less
than that of the AP-42 document.
2-10
-------
Table 2-2. Recommended Emission Factors for the Automobile
Reflnishing Area Source
Reference
National VOC
emissions
(tons per year)
VOC emission factors
(pounds/year)
U.S. population
(millions)
SIC code
Number of U.S.
automobile
refinishmg
employees
Year derived
Per capita
Per employee
AP-42
(Compilation,
1985)
199,000
1.9
5200
209.5 (cst'd)
76,500 (est'd)
1972 (est'd)
AIM 2
Supplement D
(Joyner, 1991)
199,000
1.9
5200
209.5
(est'd)
76,500
(est'd)
1972
(est'd)
SIP Guidance
Document
(Procedures,
1991)
281,00
(est'd)
2.3
3519
245.7
7532
163,000
(est'd)
1989
AIRS / AMS
(Kimbrough,
1992)
3519
7535, 7532
* For AP-42 and AP-42, Supplement D, data from 1972 most closely corresponds to the data given
in the table.
Examples of two problems with current methods are:
1) Differences in Emission Estimates. The two activity levels (per capita vs.
per employees) seldom produce the same emission estimate, as shown in the
example in Figure 2-3 for 26 California counties. These discrepancies arise
from the static nature of emission factors and the different characteristics of
individual geographic regions.
o
o
>
C/i
o
C
3
O
a
o
o
c
fi
3=
a
90G0C00
83C0003
70C0003
BOCOOOD
50C03GQ
4X0X0
3000000
2000000
1000X0
0
-1030X0
! |
|
|
i i
i
i
|
| Difference in SIP Guidance estimates
|
|
-
-
-
1
1
1
il
-
I
ฆ. _
1
1
1 1
1
I
ฆ
ฆ
i
!
|
.
!
I
CD
z
<
-------
2) Missing Data. County Business Patterns (County, Annual) exercises
disclosure protection for counties where revealing information may disclose
details about the operations of individual businesses. For example, in 1990,
Standard Industrial Classification (SIC) 7532 employment for 47 of North
Carolina's 100 counties was not disclosed. An emission estimation method
should utilize activity factors that are available for all areas required to
compile emission inventories.
The following sections provide additional detail on past and ongoing emission
inventory efforts and methods.
2.2.1 Emission Inventories
Some past and ongoing emission inventories efforts by EPA include the
National Emissions Data System (NEDS ~ still-prevalent emission estimation
method), the NEDS successor ~ the Aerometric Information Retrieval System
(AIRS), and its Area and Mobile Source (AMS) Subsystem emission inventories.
Emission inventories are used in photochemical air-quality models which
determine the emission reductions needed to achieve the desired air-quality
standard, such as the National Ambient Air Quality Standard (NAAQS). EPA air-
quality modeling guidelines specify the Urban Airshed Model (UAM) as the
recommended model for ozone concentrations over urban areas.
National Emissions Data System (NEDS)
The EPA established the NEDS for sources of airborne pollutants in 1971
(Myers, 1975). NEDS summarized annual cumulative estimates of source
emissions for the Clean Air Act's five criteria pollutants: particulate matter, sulfur
oxides, nitrogen oxides, VOCs, and carbon monoxide by Air Quality Control
Region, by state, and nationwide.
EPA has compiled emission factors for a variety of sources and activity level,
reporting the results since 1968 in AP-42 Compilation of Air Pollutant Emissions
Factors, for which supplements are issued regularly. However, "emission factors"
currently in use are developed from only a very limited sampling of the emissions
source population for any given category. The values reported are an average of
those limited samples and may not be statistically representative of the emissions
sources.
National Acid Precipitation Assessment Program (NAPAP)
The NEDS area source emissions estimation method has been used in other
emission inventory efforts. The 1985 National Acid Precipitation Assessment
2-12
-------
Program (NAPAP) emissions inventory prepared by EPA (Demmy, 1988), includes
emissions from the U.S. and Canada for 1985. The U.S. emissions data were
derived primarily using existing methodologies previously used by EPA
(Zimmerman, 1988; Demmy, 1988). The NEDS point and area source inventory
was used as a starting point for modification and refinement in the development of
the 1985 inventory. The NEDS inventory development and quality assurance
methods were refined under the Regional Ozone Modeling for Northeast Transport
(ROMNET), but as with NAPAP, the NEDS area source solvent inventory methods
remained essentially the same (work in progress, Battye, W., EC/R).
Aerometric Information Retrieval System (AIRS) Area and Mobile Source
Subsystem (AMS)
While NEDS has historically been the computer software system the EPA has
used to calculate, store, and retrieve area and mobile source emissions, EPA is in
the process of designing and developing a new data subsystem. Residing in the
Aerometric Information Retrieval System (AIRS), the new subsystem is called the
Area and Mobile Source (AMS) Subsystem (Aerometric, 1992).
AIRS is the current computerized database management system developed by
the Office of Air Quality Planning and Standards (OAQPS) to support EPA and
state needs for air quality, air emissions, and air compliance data. One
component of AIRS, the ATRS Facility Subsystem (AFS), replaces EPA's NEDS for
storage and retrieval of point source emissions data (Aerometric, 1992).
Once fully implemented, the procedure planned to be used by AMS to generate
emission estimates for area and mobile source categories for all areas of the U.S. ~
nonattainment and attainment ~ is similar to that used under NEDS. To some
extent, these methods were previously documented in Area Source Documentation
for the 1985 National Acid Precipitation Assessment Program Inventory (Demmy,
1988); however, the NAPAP report did not include certain initial data calculations.
In addition, over the years, numerous changes have occurred to the sources of the
data that "feed" these methodologies. The initial data calculations and source
data changes can be found in (Aerometric, 1992). The most recent review of
emissions computation procedures is contained in Documentation of AIRS AMS
National Methodologies (Kimbrough, 1992).
While the AIRS/AMS emission estimation method is not included in the official
State Implementation Plan (SIP) emission inventory guidance documents being
disseminated by EPA's Office of Air Quality Planning and Standards (OAQPS),
this does not preclude slate and local air agencies from using this method to
estimate their area source emissions.
2-13
-------
For AIRS/AMS, the reported national consumption of each solvent is extracted
from the Department of Energy's Petroleum Supply Annual (Petroleum, 1992) and
International Trade Commission's (ITC) Synthetic Organic Chemicals (Synthetic,
1991). The percentage of each solvent consumed by each solvent user category is
obtained from Chemical Products Synopsis (Chemical, 1990) and Chemical Profiles
(Chemical, Weekly). Total employment is obtained from County Business
Patterns. Point source employment is estimated using plant data from the
AIRS/FS and employment data from County Business Patterns. Solvent
consumption amounts used for surface coating are taken from the annual "Trends"
report (National, March 1991). County population is obtained from Current
Population Reports (Current, Annual).
2.2.2. Compilation of Air Pollutant Emission Factors
Table 2-3 contains national emissions and activity factor for VOC from
Automobile Refinishing as excerpted from AP-42 Supplement D.
The numbers in Table 2-3 have been applied to determine the population and
number of Automobile Refinishing (SIC 7532) employees at the time of the activity
levels' derivation. After reviewing the U.S. Census data, it was concluded that
AP-42 activity levels must have been derived around 1972 when the U.S. resident
population was 209,284,000 (Bureau, 1992) and there were 16,237 equivalent SIC
7532 establishments with 70,121 employees (Bureau, 1972). Therefore, emission
factors appropriate for 1972 are published for use in estimating VOC emissions in
a September 1991 document (Joyner, 1991).
Table 2-3. National Emissions and Activity Levels for VOC from
Automobile Refinishing (Emission Factor Rating: C)a
(Procedures, 1980; Lamason, 1981; End, 1979)
Emissions
Automobile
Refinishing
National:
Mg/year
181.000
Ton/year
199,000
Per Capita:
kg/year (lb/year)
0.84 (1.9)
g/day (lb/day)
2.7 (0.006)b
Per employee:
Mg/year (ton/year)
2.3 (2.6)
kg/day (lb/day)
7.4 (16.3)b
" All non-methane organics
b Assumes a 6-day work week (312 days/year)
2-14
-------
2.2.3 Control Techniques Guidelines (CTG)
The EPA has funded two recent studies to evaluate VOC emissions in
automobile refinishing. The first document (Athey, 1988) was finished in 1988,
with preparation by the Control Technology Center (CTC). As a part of this study,
methods were developed to estimate VOC emissions. The method, used in a 1988
study, is based on "broad assumptions" about three model automobile refinishing
shops.
The baseline nationwide VOC emissions from the motor vehicle refinishing
industry (Zimmerman, 1988; Athey, 1988) are presented in Table 2-4. The 1988
Athey report recommends that state and local agencies conduct a survey of shops
in their area to determine their baseline VOC emissions from automobile
refinishing operations.
Table 2-4. Baseline VOC Emissions from Automobile Refinishing by
Facility Type (Athey, 1988)
Facility Description
Number of shops
Total tons VOC/yr.
nationwide
Percent of total VOC
emissions
Small shop
33,200
42,200
15
Medium shop
(includes dealers)
41,300
149,900
52
Volume shop
(includes franchises)
8,600
95,600
33
Total industry
83,100
287,700
100
With the requirement of State Implementation Plans for reducing VOC
emissions, EPA began work on a revised Control Techniques Guideline (CTG) to
provide assistance for nonattainment areas. A 1991 draft of the CTG (Draft, 1991)
expanded on the 1988 work to include eight model shops to estimate automobile
refinishing VOC emissions in the U.S. Table 2-5 shows the results of this
emissions estimation method.
2-15
-------
Tabic 2-5. Baseline VOC Emissions from Automobile Refinishing by
Facility Type (Draft, 1991)
Model Shop
Number of Shops
Total tons VOC/yr.
nationwide
Percent of total VOC
emissions
A
6,300
6,300
2.6
B
8,820
16,141
6.7
C
3,150
5,733
2.4
D
6,300
19,908
8.3
E
24,570
84,767
35.3
F
4,410
21,609
9.0
G
3,150
12,537
5.2
H
6,300
72,891
30.4
Total
63,000
239,886
1.00
The emissions estimation technique is elaborate and depends on a great many
variables. In talking with automobile refinishing industry representatives, it was
found that many of these parameters have much too wide a variance to be simply
averaged or estimated. As an example, many auto refinishers state that the mils
thickness of coatings recommended by paint manufacturers is routinely ignored in
favor of applying enough paint to make the product "look right".
The "model shop" method has been criticized by both paint manufacturers and
auto refinishing organizations. Industry representatives contend it is impossible
to categorize body shops into types, and repair jobs into overall or partial jobs.
The industry also contends that many of the average numbers used in the
estimates were considerably high and did not represent realistic situations. After
repeated discussions with these organizations, EPA dropped the "model shops"
concept.
2.2.4 State Implementation Plans (SIPs)
The 1990 base year state emission inventories are being submitted to EPA
Office of Air Quality and Planning Standards (OAQPS) Emission Inventory
Branch (EIB). OAQPS disseminates the SIP guidance documents (not the AP-42)
to state agencies for the preparation of emission inventories. Volume I
(Procedures, 1991) of the SIP guidance documents contains the emission factors
and formulae for calculating inventories.
Every state is required to submit their VOC emission estimation method with
their estimates. Responses have fallen under four categories: "No Method", "Per
Capita", "Per Employee", and "Other Method".
2-16
-------
"No Method" means that the documentation supplied by the State agency with
their estimates is insufficient for EPA to determine the estimation method used
(Telephone, 1993). EPA's OAQPS Emission Inventory Branch may ask for
clarification from states designated under "No Method".
The per capita and per employee methods (Lamason, 1980) use activity factors
provided to the state pollution control agencies by the EPA in the SIP Guidance
Document. The automobile refinishing area source emission factors are listed in
Table 2-6.
Table 2-6. Auto Refinishing Activity Factors
SOURCE
CATEGORY
COATING
USE
(106 gal/yr)a
COATING USAGE FACTORS
(gallons/year)
VOC EMISSION
FACTORS (pounds/year)b
Nat'l
Emissions
ton/yr
Year
PER
CAPITA
PER SIC
EMPLOYEE
PER
CAPITA
PER SIC
EMPLOYEE
Automobile
refinishing
36
0.15
221
2.3
3519
241,000
1991
A1M2
Supplement D
-
-
-
1.9
5200
199,000
1991
Total Industry
(CTC)
-
-
-
-
-
287,700
1988
Total Industry
(CTG)
-
-
-
-
-
239,886
1991
New York
see Table 2-7
36
-
-
0.9
-
111,600
1990
California
Counties in
South Coast Air
Basin
.41
446
43,000
1990
a The volume and distribution of coating usage is derived from the Frost and Sullivan's Industrial Solvents report
(Industrial, 1989) and from Stanford Research Institute's (SRI) Chemical Economics Handbook (Chemical. 1992).
b Ninety-nine percent of the organic solvents used in surface coatings are classified as VOC.
Other Methods
Figure 2-4 is another representation of the type of estimation discrepancies seen
in Table 2-6. A national value for VOC emissions from auto refinishing has been
calculated from information provided by the 7 different sources shown in the
figure. The activity factor used in each case is based on population. The Auto
Refinishing column represents the SIP Guidance document emission factor/activity
level coefficients. The CTG represents factors from the Control Guidance
Document study referenced earlier. Note that the two state studies offer the
lowest estimates. They are most closely based on actual contacts with body shops
within specific regions of the country. The low values for California, provided by
the California Air Resources Board may reflect the impact of stringent regulation
2-17
-------
in that state. Without further validation, it is not possible to proselytize for one
method or the other.
BASIS
California (U.S., 1990)
New York (Everette, 1993)
Paint Market Analysis
(Connolly, 1990)
AP-42 (Joyner, 1991)
Auto Refinishing
(Draft, 1991)
CTG (Draft, 1991)
CTC (Athey, 1988)
VOC 1000 tons per year in U.S.
Figure 2-4. National VOC Estimations by Different Methods
States that chose to use an alternative VOC emission estimation method
("Other Method") are required to submit their method with their estimates. Table
2-7 contains a description of the methods used by the States of Alabama, Georgia,
New York and the State of Indiana's Marion County, the only "Other Methods"
submitted to date.
Although the SIP guidance document recommends use of the per employee
factor for automobile refinishing (Lamason, 1980), as mentioned in Section 2.2, the
number of automobile refinishing employees may not be available. This was the
case for the State of Georgia. For the counties listing employment for SIC 7532 in
the CBP, the State of Georgia used the number of accidents per county to derive
an SIC 7532 employee-per-accident multiplier. Using this multiplier, Georgia
extrapolated SIC 7532 employment in the counties for which the CBP invoked
disclosure protection. The combined SIC 7532 employment was multiplied by the
per employee factor to determine the State's emission inventory.
2-18
-------
Table 2-7. Other Emission Estimation Methods for Automobile
Refinishing from State Implementation Plan (SIP)
Submissions.
STATE
AREA
ESTIMATION METHOD
Alabama
Jefferson County
A 1988 survey of all body shops yielded a
308 tons/year estimate for the county.
While VOC emissions for individual shops
ranged from 0-33.2 tons/year, 89% of the
shops emitted 4 tons or less.
Georgia
Nonattainment counties: Cherokee,
Clayton, Cobb, Coweta, Dekalb, Douglas,
Fayette, Forsyth, Fulton, Gwinnett, Henry,
Paulding, and Rockdale
For counties with number of SIC 7532
employees in CBPฎ, derived number of
accidents'/employee. From this determined
number of SIC 7532 employees in each
nonattainment county. Multiplied
employees by 3519d pounds per employee-
year emission factor to get annual county
emissions; divided by 260 workdays per
year to get county workday emissions.
Attainment areas: All other counties in
Georgia
Used the accidents/employee number
derived above to determine the number of
SIC 7532 employees.
Indiana
Marion County (Indianapolis)
A survey, conducted in 1987 with a 59%
response rate, yielded a 141 tons VOC/year
estimate for the county. A representative
random sample taken in 1990 indicates
that VOC emissions have not changed
significantly since the 1987 survey.
New York
All counties except those in the New York
City Metropolitan Area (NYCMA)
Derived new per capita emission factor
from 111,600 tons VOC per US-yearฎ and
247,920,500 persons per US - 0.9 lbs VOC
per capita-year.
NYCMA counties: Bronx, Kings, Nassau,
New York, Orangeb, Queens, Richmond,
Rockland, Suffolk, and Westchester
Using VOC limits on coatings0, and rule
penetration and effectiveness1*, derived
VOC emissions = population *0.9 lbs
VOC/person * [1-(0.19 *0.53 *0.8)].
a Derived from national coating usage (Industrial, 1989) and New York State's VOC/gal. limit for touch-up coatings of 6.2
lb. VOC/gal., as sprayed. (Everette, 1993)
b Only portions of Orange County are nonattainment
" 6.2 lbs VOC/gallon on partial jobs (10 ft2) and 5.0 lbs. VOC/gallon on full jobs (100 ft2)
A From SIP Guidance document (Lamason, 1980)
* Census Bureau's County Business Patterns (County, Annual)
From Georgia Department of Public Safety accident reports
2-19
-------
Board. Report
In a 1991 report, the State of California Air Resources Board (CARB) estimated
1987 total organic gas (TOG) emissions from automobile reflnishing to be
15,786.84 tons. These TOG estimates were updated from 1983 estimates using
growth factors, which were updated from 1982 estimates using a growth factor
based on the increase in vehicle registration in the United States.
California's fraction of the nationwide usage was determined by the ratio of the
vehicle registration in California to the vehicle registration in the United States.
Each county was apportioned its fraction of California's total by its percentage of
the State's population. A 1979 composite emission factor of 5.275 pounds VOC per
gallon was used to determine the TOG emission levels. It was assumed that the
amount emitted was equivalent to the amount consumed.
2.2.5 EPA Study of Recent Methods
In a recent analysis of current methods for estimating VOC emissions from area
sources, it was recommended that the best method for estimating VOC emissions
from automobile refmishing is the use of a paint market analysis, updated every 4
years (Connolly, 1990), and the 1988 EPA CTC study (Zimmerman, 1988; Athey,
1988). It was also stated that "given the absence of data to discredit the
employment data set or to recommend another data set, the availability of data
favors the employment-based method" (work in progress, Battye, W., EC/R). Table
2-8 contains a list of variables that have been examined, with comments on the
solvent use correlations, and the strengths and weaknesses of each.
The employment-based method appears to be as good as any other allocation
method in the absence of other evidence. However, there is little to support the
recommendation to use two national data sources whose national usage estimates
disagree by over 100% (572 billion vs. 264 billion pounds of solvents per year), as
seen in Table 2-9. As a result of conversations with industry representatives, the
use of the paint market analysis estimates is recommended.
2-20
-------
Table 2-8. Automobile Refinishing Variables from Recent Review (work
in progress, Battye, W., EC/R)
Variable
Solvent Use
Correlation
Strengths
Weaknesses
Paint usagea,b,c
Requires average
pounds VOC/gallond
Accurate on national
level
No State or county
consumption data
Solvent usagea,b
Direct
Accurate if correctable
with multiple sources
Uncorrectable across
sources; No State or
county data
Employees by four-
digit SIC and county*
Requires pounds
VOC/SIC 7532
employee-year
Includes number of
establishments and
employees and payroll
by shop employment
Census does not
disclose details of
individual firms
Employees by four-
digit SIC and latitude
and longitudef
Includes address,
latitude, longitude,
phone number
No other information
on operations
Accidents
Requires pounds
VOC/accident-year
Annual county data at
State level2
Accident definition,
severity, and repairs
vary with regional
characteristics
Annual State data at
National level11
No county data; (same
as above)
Fatal Accidents
Requires pounds
VOC/fatal accident-year
Annual county data at
State levelg
Vehicles in fatal
accident are least
repairable due to
damage severity
Annual State data at
national level*1'1
No county date; (same
as above)
Receipts by four-digit
SIC and county'
Requires pounds
VOC/SIC 7532 $ receipt
Includes number of
establishments and
employees and payroll
by shop employment
Census does not
disclose details of
individual firms;
repair costs vary
regionally
(Athey, 1988)
(Connolly. 1990)
(Commerce, 1987)
5.5 pounds VOC/guIlon is a conservative nationwide
estimate from Midwest Research Institute (MRI) study
for National Paint and Coating Association (Athey,
1988).
(County, annual)
Computerized telephone directories
State Department of Transportation
National Safety Council
Insurance Institute for Highway Safety
2-21
-------
Table 2-9. Comparison of National Paint and Solvent Usage Estimates for
Automobile Refinishing
Automobile Refinishing Material
EPA Control Technology Center
Study (Athey, 1988)
Paint Market Analysis (Connolly,
1990)
Paint usage (106 gallons)
41
36
Solvent usage (106 pounds)
201
162
In paint as purchased
118
102
Thinning solvents
118
102
Other solvents
253
--
Total solvent
572
264
2.2.6 Accuracy of Emissions Inventories
The accuracy and precision required of an emissions inventory is dictated by its
intended use. For example, the Urban Airshed Model input data require data that
have accurate and precise spatial and temporal distributions and speciations.
Such requirements are quite rigorous when compared with those needed for
emission trend analysis.
It has been stated by a National Academy of Sciences report that due to high
cost and the absence of "proven measurement techniques to test the accuracy and
precision of emission data" (Procedures, 1991), the uncertainty of emission
inventories have only been subjectively estimated. That is, the methods used to
estimate emissions have not been adequately checked by comparison with field
measurements (Procedures, 1991).
Emission estimation accuracy is best addressed by acquiring actual emission
data. The validation method recommended here includes a nationwide survey and
several intensive surveys, the collection of nationwide materials distribution data,
and site emission sampling and analysis.
2-22
-------
2.3 Database Development
The success of any new method to estimate area source emissions is limited by
the existence and accessibility of information. The objective is to identify data
that are updated at least annually and represent county-level activity. Records
should exist for at least the past five years and into the foreseeable future. The
data must be statistically defensible, representative, "universal" (representative of
national distributions and/or variations), and result in accurate emissions
estimates. The data are then related to the primary area source variables of both
a) solvent use and b) emissions. These variables may be normalized, for example,
on the basis of per capita or per employee or per operation.
Figure 2-5 illustrates the information sets of interest in emissions estimation.
The figure shows how the data are distributed over time and space under the
method being developed. The axes also demonstrate that the data can have
resolution down to the county level (or, for example, a 30x30 km grid size). Data
as far back as 15 years is of interest for developing and validating the method
(considering the past 15 years for major industry trends, the past five years for
state/regional distribution figures and general validation, the past 2 years for
detailed county level information, the past one year for detailed validation data,
and the next 2 years for detailed projections).
2 years future,
ฃ prasoni^
^.1 M5 years
15 years past
ASSEMBLED DATA
For the United States: 60 variables over 12 years
For 51 states: 25 variables over 12 years
For 3126 counties: 11 variables over 22 years
For 64,524 establishments: 10 variables in 1993
Figure 2-5. Information Sources for Emission Estimates and Data
Distribution
2-23
-------
The database that has been assembled has both geographic, "spatial"
components (e.g., nation, state, county, or city) and temporal components (e.g.,
year or month). The current data set assembled for this study includes
60 variables over 12 years for the United States
25 variables over 12 years for 51 States (including D.C.)
11 variables over 12 years for 3126 counties
10 variables for 1993 for 64,524 automobile refinishing establishments
Table 2-10 summarizes several sources and variables. Each of the variables has
several dimensions.
The database development step in the method has been conducted as follows:
1) List the broadest scope of information sources containing variables that
might influence or correlate with auto refinishing emissions.
2) Select a large subset of variables from these sources based on expert opinion
(e.g., from paint manufacturers, emissions experts, auto refinishing
managers, statisticians) of which variables are likely to influence or
correlate with emissions.
3) Further reduce the number of variables based on the accessibility and
frequency of updating of the data.
4) Correlate variables with one another to show similarities in predicting
trends and variables in regions of the United States, or variations from
year-to-year. New variables are also correlated with those being generally
applied in emissions estimation, i.e., population or number of SIC
employees.
The desired dependent data, area source solvent use and VOC emissions, is
essentially nonexistent. Therefore, final correlations and variable selection must
wait on the availability of validation data (see Section 3).
Following variables selection and final arrangement of the desired databases,
the analysis and estimation system can be tailored for optimal emission
predictions, but not before. Therefore, the current information remains surrogate
data until validation is provided. Nevertheless, some interesting preliminary
results and selections have come from the evaluations to date.
The surrogate data is summarized next (Section 2.3.1), followed by a description
of data prescreening using statistical analysis (Section 2.3.2).
2-24
-------
Table 2-10. Major Data Items and Sources (Fatal, 1991; General, 1991;
Highway, 1992)
Source of
Nationwide Data Variable
Finest Level of
GeoqraDhic Aggregation
Update
Freauencv of
Measurement (est)
U.S. Bureau of the Census
Economic Censuses
- Bv Standard Induslrial Classification:
Number of establishments and firms
County
Quinquennial
Quinquennial
Employment
County
Quinquennial
Quinquennial
Payroll
County
Quinquennial
Quinquennial
Sales, receipts, or value of shipments
County
Quinquennial
Quinquennial
Operating expenses
Varies by SIC
Quinquennial
Quinquennial
Value added
Varies by SIC
Quinquennial
Quinquennial
Net income produced
National, some SICs
Quinquennial
Quinquennial
Capital expenditures
National
Quinquennial
Quinquennial
Inventories
State, some SICs
Quinquennial
Quinquennial
Specialization
National, SIC 2851
Quinquennial
Quinquennial
Countv Business Patterns
- Bv Standard Industrial Classification:
Number of establishments and lirms
County
Annual
Annual
Employment
County
Annual
Annual
Payroll
County
Annual
Annual
Current Peculation Rfioorts
Median income of households
State
Annual
Few Months
Percent of persons below poverty
State
Annual
Few Months
Total population
County
Annual
Few Months
Land area
State
Decennial
Few Months
Bureau of Labor Statistics
Civilian labor force
State
Annual
Monthly
Consumer price index
National
Monthly
Monthly
Producer price index
National
Monthly
Monthly
Bureau of Economic Analysis
Rfinional Economic Information Svstem
Personal income
County
Annual
Monthly
Disposable personal income
County
Annual
Monthly
Population
County
Annual
Monthly
Total wages and salaries
County
Annual
Monthly
Wage and salary employment
County
Annual
Monthly
Average wages per job
County
Annual
Monthly
Census journey to work
County
Decennial
Decennial
Gross state product
County
Annual
Monthly
Full and part time employees
County
Annual
Monthly
Regional economic profile
County
Annual
?
Transfer payments
County
Annual
?
U.S. Department of Transportation
Federal Hiahwav Administration
Non-fatally injured persons
State
Annual
Monthly
Public road and street mileage
State
Annual
Monthly
Licensed drivers
State
Annual
Monthly
Registered automobiles
State
Annual
Monthly
Registered motor vehicles
State
Annual
Monthly
Vehicle miles of travel
State
Annual
Monthly
National Hiahwav Traffic Safetv Administration
Fatal accidents
State
Annual
Monthly
National Safety Council
Motor vehicle accidents, on road
State
Annual
Monthly
Motor vehicle dealhs
State
Annual
Monthly
Insurance Information Institute
Motor vehicle accidents, on and off road
State
Annual
Monthly
Motor vehicle injuries
State
Annual
Monthly
American Chamber of Commerce Researchers Association
Cost of living index
Selected MSAs
Annual
Annual
2-25
-------
2.3.1 Surrogate Data: Sources. Variables, Availability. Accuracy
The following is a description of major relevant data sources and their
advantages and disadvantages.
Census Bureau Published Data for SICs
The level of industry emissions is likely to be proportional to the amount of
business activity (e.g., measured by receipts, payroll, or employees). However,
identifying the amount of business activity in the automobile reflnishing industry
is not simple. The most complete coverage of industrial data across geographic
regions of the United States is available from the Census Bureau. The Census
Bureau has two programs that gather and publish data by Standard Industrial
Classification (SIC) code:
1) The Economic Census collects information every five years directly from
business establishments. 1987 is the most recent year for which data are
available. In 1987, 3.9 million questionnaires were mailed to businesses
across the country. Each business receives only one type of more than 500
industry-specific types of forms. All businesses receiving the forms are
required by Title 13 of the United States Code to complete the forms and
return them. The 1992 Economic Census forms will be mailed to more than
3.5 million companies in early 1993 and preliminary results will be
available in 1994.
2) County Business Patterns gathers most of its data from administrative
records of the Federal Government every year. The Census Bureau states
that "County Business Patterns is the only series that provides annual
sub-national data by two-, three-, and four-digit levels of the Standard
Industrial Classification (SIC) system" (County, Annual). However, while
County Business Patterns is published annually, the Economic Censuses
present data addressing more specific SICs.
Census Bureau publications aggregate information by SIC for geographic
regions (see Table 2-11). The information includes: (a) the number of
establishments and firms, (b) the number of employees, (c) payroll, (d) sales,
receipts, or value of shipments, (e) operating expenses, and (f) other categories,
usually with less detail. Four SIC codes related to the automobile refinishing
industry can be useful for the estimation or validation of VOC emissions:
1) 7532 - Services: Automotive top and body repair and paint shops
2) 5198 - Wholesale Trade: Paints, varnishes, and supplies
3) 2851 - Manufacturers: Paints and allied products
4) 5511 - Retail Trade: New and used car dealerships
2-26
-------
Table 2-11. Major Census Bureau Items by Geographic Level and Industry
(Guide, 1990)
ECONOMIC CENSUSES
Service Industries
Manufacturers
Wholesale Trade
Retail Trade
Every 5 Years
Auto Bodyshops
Paints & Allied Products
Coatings Distribution
Auto Dealerships
2-4 year Laa-Time
SIC 7532
SIC 2851
SIC 5198
SIC 5511
Number of Establishments and Firms
Establishments with payroll
N, S, M, C
N, S, M, C
N, S, M, C
N, S, M, C
Individual proprietorships
N, S, M, C
N.S
N
N, S, M, C
Partnerships
N, S, M, C
N, S
N
N, S, M, C
Firms
N
N, S
N
N
EmDlovment
All employees
N. S, M, C
N, S, M, C
N, S, M, C
N, S, M, C
Production workers
N, S, M, C
- Hours
N, S, M, C
- Wages
N, S, M, C
Employment size of establishment
N
N, S, M, C
N
N
Employment by principal activity
N
Payroll
All employees, entire year
N,S, M,C
N, S, M, C
N, S, M, C
N, S,M, C
All employees, first quarter
N, S, M, C
N,S, M.C
N, S
Production workers
N, S, M, C
Supplemental labor cost
N
N, S
N
N
Sales. Receiots or Value of Shinments
Establishments with payroll
N, S, M, C
N, S, M, C
N, S. M, C
N, S, M, C
Specific product, line, or type of const.
N, S
N,(S,M)
N,S, M
Class of customer
N
N
N
Sales size of establishment
N
N
N
N
Ooeratlna Excenses
Total
N, S
N
Cost ol materials
N.S, M,C
N
N
Specltic materials consumed
N
N
- Quantity
N
N
- Cost
N
N
Cost of fuels
N
N.S
N
N
Energy consumed
N, S
- Quantity
N, S
- Cost
N.S
- Cost of electricity
N
N
N
Products bought or sold (resales)
N, S
Other
Value added
N, S, M, C
N
N
Net income produced
N
N
Capital expenditures, total
N
N, S
N
N
- New
N
N, S, M, C
N
N
Inventories (end of 1986,1987)
N, S
N,S
N
Specialization by type of manufacturing
N
COUNTY BUSINESS PATTERNS
Service Industries
Manufacturers
Wholesale Trade
Retail Trade
Every Year
Auto Bodyshops
Paints & Allied Products
Coatings Distribution
Auto Dealerships
2-3 Year Laa-Time
SIC 7532
SIC 2851
SIC 5198
SIC 5511
Number of Establishments & Firms
Establishments with payroll
N, S, M, C
N.S
N,S,M, C
N, S, M, C
Emolovment
All employees
N, S, M, C
N,S
N,S, M, C
N, S,M, C
Employment size of establishments
N, S, M, C
N, S
N, S,M, C
N, S, M, C
Payroll
All employees, entire year
N, S, M, C
N.S
N, S, M, C
N, S, M, C
All employees, first quarter
N, S, M.C
N, S
N, S, M, C
N, S, M, C
GEOGRAPHIC LEVa
N- National
S-Slate
M-Metropolltan Statistical Area
CtCounty
SIC CODE
2651 - Paints and allied products
5198 - Paints, varnishes, and supplies
5511 - New and used car dealerships
7532 - Top and body repair, paint shops
2-27
-------
The SIC system does not reveal the amount of business for a particular activity,
such as automobile refinishing, but classifies each establishment based on its
primary activity. For example, automobiles may be refinished at auto body shops,
auto paint shops, or car dealerships. At the same time, establishments in the
same primary SIC may engage in other activities that contribute to their total
business to varying degrees. For SIC 7532, there is no way to single out what
percent of the business is specifically automobile refinishing.
The Census does not include establishments with no payroll, but these small
auto refinishing operations may make up 1/3 of all shops (I-CAR, 1992). Some
shops may also not maintain the proper licenses or registrations to be identified by
government agencies.
Census Bureau "Microdata" for SICs
Some detailed information collected from individual establishments, referred to
as "microdata", is kept confidential by the Census Bureau. In addition, disclosure
protection requires the Census Bureau to suppress numbers in published tables
that might reveal an establishment's level of operations through logical inference.
Disclosure often prevents publication of data in areas where a small number of
firms are dominant in one industry classification.
"Microdata" can be accessed by approval and use at the Bureau's site, but the
benefits are not substantial. Many area sources are service industries, rather
than manufacturers, which are often classified as point sources. The Census
Bureau does not collect any information about the consumption of materials such
as paint or solvents by the service industry or the source of receipts by type of
industrial activity. The form does, however, ask each establishment to show
distribution of total receipts for labor performed, parts installed, tire services,
retreading, and sales, automobile fuels and lubricants, sales of other merchandise,
and all other receipts. The same form (Commerce, 1987) is sent to all
establishments that are classified as "motor vehicle service shops, including tire
retreading". No information can be presented that might indicate the operations
of a single establishment or business in any part of the country. The "microdata"
does not provide information for auto refinishing and most other area sources that
is likely to provide better correlation than other, more accessible, variables.
Census Population Data
The Census Bureau gathers population data from households, including the
number of people in the household, their relationship to one another, household
income, and the number of people employed. It also includes other demographic
information about the age, race, education, national origin, and citizenship of the
2-28
-------
residents of the household. The information provided by the individual households
is privileged information of the Census Bureau, just like the Economic Censuses.
The Census Bureau is the primary source of population data and its most
significant population data collection effort is the Decennial Census conducted
every ten years (most recently in 1990). The Census Bureau also conducts a
Current Population Survey every few months to provide updates of the population.
This survey only collects information from a representative sample of the United
States and provides no characteristics of the population for smaller geographic
regions such as counties and cities.
In this study, different measures of the population are being examined that
better capture the variation in consumption of automobile refinishing for different
regions of the country. Examples include the number of licensed drivers and
civilian labor force.
Product or Service Data
The first requirement for automobile refinishing demand is the existence of an
automobile, owned by a private citizen, the government, a rental car agency, or
other company. Therefore, automobile sales or accident data could be related to
the level of auto refinishing. However, since cars of any age may need refinishing,
auto sales must be integrated over time and are difficult to relate to refinishing
levels in a single year. The primary source of demand for automobile refinishing
is accident damage. To help lessen the financial burden of collision repair,
motorists carry collision insurance (in addition to liability insurance) on
approximately 66.5% of all vehicles (Economic, 1992). However, not all accident
damage is repaired. Several factors might explain why damage to a vehicle may
not be repaired, including:
1) Cost of repair
2) Cost of replacement
3) Amount of damage
4) Type of damage
5) Age of car
6) Condition of vehicle before accident
7) Automobile insurance coverage
8) Financial status of owner
According to the technical report by the National Highway Traffic Safety
Administration (NHTSA) (Economic, 1992), there were more than 28 million
vehicles damaged in 1990. Approximately 4 million of these vehicles were
damaged in accidents resulting in an injury to either the occupants of the vehicles
2-29
-------
or a pedestrian. About 78 percent of accidents involving an injury are reported to
police, while the reporting rate for property damage only vehicles is 52 percent.
Some states require motorists to report accidents if an injury is involved while
other states require accidents to be reported if the property damage is expected to
exceed a certain dollar amount. NHTSA reports non-fatal injuries on an annual
basis, but state-supplied data may be inconsistent. NHTSA updates data by
contacting the states submitting inconsistent numbers, also.
NHTSA also provides the Fatal Accident Reporting System (FARS). These data
are more accurate than the measures of non-fatal injuries, but these accidents,
due to their severity, are less likely to generate demand for automobile
reflnishing.
Additional, travel-related data have been collected that might influence the
demand for automobile reflnishing. These variables include the number of
registered vehicles, the number of vehicle miles traveled, and the number of miles
of highways.
Neither sales, accidents, nor travel data have shown the best correlation with
auto reflnishing (relative to other selected variables discussed later).
General Economics Data
Several tables with economic data are reproduced in the Statistical Abstract of
the United States (Bureau, 1992). These tables come from a number of sources,
including the Bureau of Economic Analysis and the Bureau of Labor Statistics. It
is usually necessary to contact the agencies directly to obtain historical data or
explanation of the variables included.
The Department of Commerce operates the Economic Bulletin Board (EBB) that
allows access to press releases and data files using a modem. The EBB includes
statistical data from the Bureau of Economic Analysis, the Bureau of the Census,
the Bureau of Labor Statistics, and other Federal agencies. Some of the
information provided by the EBB includes employment data, personal income
statistics, price indices, major economic indicators, current business statistics,
industry statistics, and regional economic statistics. The data content is provided
to the EBB by different agencies.
The Regional Economic Measurement Division of the Bureau of Economic
Analysis produces a CD-ROM that includes annual income and employment
information covering 1969 to 1990 for states, counties, and metropolitan areas.
This source has been used to verify state populations, but additional review is
needed to determine what specific economic variables are available. This Regional
2-30
-------
Economic Information System (REIS) CD-ROM is useful for accessing information
at the county level.
EPA is in the process of developing an Economic Growth Analysis System (E-
GAS). This is a system that uses economic growth forecasts to calculate growth
factors for each of the counties in designated ozone nonattainment areas. The
forecasts of economic growth that are inputs for the E-GAS model may be useful
for producing forecasts of automobile refinishing business. The outputs from the
E-GAS model for specific counties may also be useful, not only as a check against
this study's estimates of automobile refinishing, but as possible inputs to our
model. One of the major components of the E-GAS model is vehicle miles
traveled.
The economic data that has been analyzed in this study is the civilian labor
force data (CLF, both CLF "total" and CLF "employed") from the Bureau of Labor
Statistics. This data provided good correlations with auto refinishing business
activity and with estimated auto painting employees.
While there is no immediate indication that other economic data will provide
the most representative variables for emissions estimation, economic data deserves
further analysis.
The "economics" category was addressed last in this study, and insufficient time
has been available for complete review. The best variable candidates should be
cost-of-living by metropolitan area, and annual income and employment at county
levels. For such major indicators as consumer price index (CPI), the values are
available only on national or state levels, and the state data is not at all
representative of the often significant variations in CPI from county to county.
Economic data files often have gaps in both time and spatial distributions,
especially for county-level data. The REIS data files appear most useful for
correlation with area source validation data.
Direct marketing information
There has been continuing growth of direct marketing {e.g., "junk" mail,
telemarketing) to consumers in the U.S., accompanied by voluminous and detailed,
information files on the population. The problem with direct marketing data is
that it should be focused on the product under consideration to be useful in
assessing an area source. Direct marketing almost always implies marketing of
high-volume, retail consumer goods and not the products or services used or
offered by most area sources. The other potentially useful direct marketing
information is the demographics relating to sales of related products {e.g.,
automobiles or refinishing paint). However, tracking auto sales rather than
trends predicted by marketing data makes more sense for automobile refinishing.
2-31
-------
The research under this study addresses marketing data by obtaining paint
manufacturer information.
Direct marketing information usually includes density of a population, age,
occupation, affluence, consumption patterns, and so on. Direct marketing
information is not indicated to be more useful than census and business pattern
data. Direct marketing requires specific information for specific addresses, but
this case study does not, nor does more generic system development for emissions
estimation. However, direct marketing data might relate better to such area
sources as dry cleaning or printing. The emissions estimation model needs more
general and area-source-specific data. Demographics data could potentially be
related to types of vehicles being refinished, but relates no better to volume of
VOC use than other variables.
State Regulations
Regulations are influencing the VOC content of paints and solvents and even
the types of equipment used in shops. The actual influence of regulations on
emissions can only be evaluated qualitatively. Current regulations outside of
California do not impose VOC content significantly different from national
averages. The limited number of pertinent state regulations confirmed by survey
is summarized in Table 2-12.
Table 2-12. Summary of State Regulations for VOC Emissions from
Automobile Refinishing
STATE
REGULATIONS
(ranges of control levels; regulations vary by type of vehicle and by regions
within States)
California
VOC < 2.8 to 6.5 lb/gal (current); VOC < 2.1 to 4.5 lb/gal (1995); or 85%
transfer efficiency.
Require electrostatic or HVLP painting, or 65% transfer efficiency.
Require closed cleanup when using organics.
Cannot use more than 5% of coatings or 7 lb/gal VOC for specialty coatings.
In some areas: must use equipment to control 90% of VOC (mass basis);
maximum 1.67% VOC per lb of cleanup solvent; reduction of 95% VOC by
weight; organic vapor pressure < 45 mmHg; must use approved spray booth.
New Jersey
VOC < 4.4 to 6.0 lb/gal
New York
VOC < 5.0 to 6.2 lb/gal
Texas
VOC < 1.4 to 6.2 lb/gal
Closed cleanup; recycle all wash solvents
California has the strictest air quality regulations as exemplified by the
regulations under the South Coast Air Quality Management District (SCAQMD),
introduced in 1988, and the Bay Area Air Quality Management District
2-32
-------
(BAAQMD). New Jersey's State Implementation plan was enacted in 1982, and is
periodically updated. In New York, the first State Implementation Plan was
prepared in 1973. It has undergone subsequent updates in 1979, 1982, 1984, and
1988. Texas first enacted specific regulations for the automobile refinishing
industry in 1987, revised in 1988.
National Regulation
There is no present federal regulation governing VOC emissions. Work has
recently been curtailed on a revision to the EPA's Control Techniques Guideline, a
document intended to assist states with their SIPs. One alternative is the
establishment of a national rule on the VOC contents of coatings. Paint
manufacturers have thus far shown a willingness to support such a rule as long as
they are given ample time to distribute present supplies and develop the new, low-
VOC coatings. Eventually, the burden on manufacturers would be lighter than it
is at present since there would be less variation in VOC content requirements.
Occupational Safety and Health Regulations
Since the mid-1980's, the Occupational Safety and Health Administration
(OSHA) has required users of hazardous products (such as paints and solvents) to
maintain a Material Safety Data Sheet (MSDS) on location for each toxic material
used. MSDS contains information provided by the product manufacturer, such as
hazardous ingredients, level of health hazard, fire and explosion, and special
protection data. Users, such as auto refinishing shops, found to be in non-
compliance with the law are subject to fines. Manufacturers, distributors, and
retailers are required by law to provide a MSDS to the purchaser with the first
shipment. Following the initial transaction, some companies elect to provide
MSDS with every delivery, while others do so only upon request. MSDS do not
provide any quantitative information on paint and solvent volume or geographic
distribution.
OSHA also regulates ventilation spray booth design and fire hazards, such as
solvent containers. However, the numerous shops and OSHA personnel
limitations result in only occasional enforcement.
Miscellaneous Data
Other data of possible interest include weather and climate, social trends, or
major national events. For instance, it has been speculated that a higher rate of
accidents may occur during abnormal weather conditions, such as a snowstorm in
the South. However, data provided by the University of North Carolina at Chapel
Hill Highway Safety Research Center indicate that between 1986 and 1989,
almost two thirds of the accidents in North Carolina occurred on clear days.
2-33
-------
Drivers may be more cautious during inclement weather and drive less carefully
on sunny days. In order to establish a correlation between weather conditions and
accidents, data on precipitation, sunshine, cloudiness, and fog would have to be
available for every county. A "normal weather" benchmark would also be needed
to evaluate any unusual weather conditions. In North Carolina, only five
metropolitan regions and the Outer Banks are covered by the National Weather
Service "first-order" stations. Weather data for more than 80 of the 100 counties
within the State would have to be obtained from scattered reporting site locations.
Most weather information available appears to be city-specific, rather than eounty-
or state-specific.
On average, between 80% and 90% of the repair jobs handled by auto
refinishers consist of cars involved in collisions. The remaining 10%-20% consist
of cars affected by corrosion, vandalism, and various other factors. Automobile
corrosion is often caused by salt, either due to proximity to the ocean, or extensive
use of salt on snowy roads. In order to correlate corrosion and VOC emissions,
data describing the proximity of a region (e.g., county seat) to saltwater, and the
location and frequency of salt use on roads would be needed. This may, however,
require an extensive data search to describe only a small portion of the auto
refinishing business.
Other factors, such as crime and vandalism, or national political and
socioeconomic trends may also be important. An example of such a trend was the
OPEC oil embargo which caused a strong decline in the auto refinishing industry,
along with a general economic downturn in transportation. Nationwide, increased
environmental concern has also had an influence, mainly reflected in regulatory
policy, as discussed. To have a measurable impact on emissions, these trends
should be national rather than local. Local trends would be captured by SIP
inputs.
2.3.2 Prescreening of Variables by Statistical Analysis
Data subsets were first selected based on availability, frequency of revision, and
expert option on relation to auto refinishing emissions.
To provide working (but unvalidated) variables and data for the emission
estimation method, several data subsets were further screened using regression
analysis. Annual data collected at the state level have proven to be the most
useful for analyzing trends and regional variations.
This working surrogate database organizes annual state-level data, and
calculates totals for each of the 9 Census divisions and 4 Census regions of the
United States. Examples of the explanatory variables in the database, along with
their national totals, is shown for 1980 and 1985 to 1991 in Table 2-13. Database
2-34
-------
completeness and availability were very important in this selection. The
distribution of a single variable across the states can also be displayed over time.
Table 2-14 shows data for all states in 1987. The database contains several
measures of:
1) economic health (unemployment, gross State product)
2) population (resident population, civilian labor force)
3) automobile travel (vehicle miles of travel, public road and street mileage)
4) quantity of traffic (licensed drivers, motor vehicle registrations)
5) traffic accidents (accident injuries, fatal accidents, accident deaths)
Most variables were considered as independent, or explanatory. A few variables
-- i.e., number of auto refinishing employees, receipts for auto refinishing SIC
(from CBP), and sales for the paints and allied products SIC (Census data) were
selected as most closely representing area source activity. Using these as
surrogates for solvent use, the independent variables were correlated. Scores of
multivariable regression analyses have been conducted for several variables
selected from statistical analysis of more than 20 original variables.
Six examples of data regressions are included in the following. The dependent
variables in the examples are shown in Table 2-15. Independent variables are
included with each of the tables of results for each example.
The selection procedure selects the variable that explains the greatest amount
of the variation in the dependent variable. Once a variable is added to the
analysis, an equation is generated, Y = b0 + bjXj + b2X2 + b3X3 + .. . + bjXj, and
statistical checks are performed. The procedure may select all, some, or none of
the predictor variables for the final model.
2-35
-------
Table 2-13. Summary of Explanatory Variables and Their National Totals
Name
lExDlanatorv Variable
Definition
Units
1980
1985
1986
|Ye a r
1987
1988
1989
1990
1991
ACC_FTL
Fatal crashes (accidents)
number
45,284
39,196
41,090
41,438
42,130
40,741
39,836
36,895
ACC_NF
Non-fatally injured persons
number
3,029,693
3,358,177
3,452,766
3,555,788
3,626,777
3,639,084
3,536,707
3,434,330
AREA_LND
Land area
sq. miles
3,539,289
3,539,289
3,539,289
3,539,289
3,539,289
3,539,289
3,536,346
3,536,346
CLF_EMP
Civilian labor force, employed
thousands
99,500
107,363
109,652
112,490
115,036
117,367
117,121
116,875
CLF_TOT
Civilian labor force, total
thousands
107,153
115,691
117,894
119,903
121,740
123,903
124,603
125,303
CLF_UN
Civilian labor force, unemployed
thousands
7,625
8,327
8,241
7,423
6,701
6,529
6,876
8,424
DEATH_MV
Motor vehicle deaths
number
53,476
46,159
48,140
48,290
49,391
45,203
44,419
41,314
GSP_RPR"
Gross state product, automobile repair
1982 dollars
24,810
28,672
29,006
28,405
28,319
28,814
0
0
HWY_RUR
Public road and street mileage, rural
miles
3,331,000
3,171,000
3,178,000
3,163,838
3,131,669
3,122,724
3,122,788
3,139,435
HWY_URB
Public road and street mileage, urban
miles
624,000
691,000
701,000
710,188
739,474
753,777
757,363
749,865
LIC_DRV
Licensed drivers
thousands
145,295
156,868
159,487
161,818
162,853
165,555
167,015
168,995
POP_RES
Resident population
thousands
227,255
237,951
240,162
242,325
244,536
246,819
249,466
252,181
REG_AUTO
Registered automobiles
thousands
121,724
132,108
135,431
137,324
141,252
143,081
143,550
142,956
REG_MV
Registered motor vehicles
thousands
155,796
171,655
176,191
179,043
183,396
189,261
188,798
188,372
VMT_RUR**
Vehicle miles of travel, rural
millions
672,030
730,221
747,780
780,450
817,534
847,225
868,878
883,621
VMTJOT
Vehicle miles of travel, total
millions
1,527,295
1,774,179
1,834,872
1,921,204
2,025,962
2,096,456
2,144,362
2,172,214
VMT URB"
Vehicle miles of travel, urban
millions
855,265
1,043,958
1,087,092
1,140,754
1,208,428
1,249,231
1,275,484
1,288,593
-------
Table 2-14. Annual Summary of Explanatory Variables by State: 1987
keg**1- Drvarar, 1
and State ACC_FTl.
number
i ' 3 4 I 6 7 6 9 10 11 12 13 14 15 16 17
ACCJf AREA_LND CLF.EMP CLF_TOT CIF_UN DEATH_MV GSPRPR HWY_RUR HWY_URB UC_0RV POP_RES RฃG_AUTO REG_MV WT_RUR VMT_TOT VWT.URB
ragnber sq mites thousands thousands thousands nunbar 1982 Dofcn miles miies Iwxnds thousands thousands thousands mflfcons mllions mrtions
to
I
CO
3
U.S. Tot*
41.438
3.555.788
3.539.289
112.490
119,903
7.423
48.290
28,406
3.163.838
710.188
161.818
242.325
137.324
179.043
780,450
1,921,204
1.140.754
NORTHEAST
6.480
776.216
162.745
23.820
24.930
1,121
7.513
6.283
241.494
128.802
32.732
50.306
27.597
32,152
107.121
331.977
224.856
MIDWEST
8.067
884.810
752.093
27.858
29.845
1,966
10.647
6.366
1.139.341
176.427
39.258
59.027
34.064
44.332
211.303
459.571
248.268
SOOTH
18,834
1.220.635
873,006
37.819
40.579
2,756
19,416
8.616
1.086.406
266.664
56,173
83.221
47,384
64,411
325.013
713,937
388.924
WEST
9.257
674.127
1.751,448
22.993
24.549
1.560
10.714
7.140
696.596
138.295
33.655
49.769
28.259
38.148
137,013
415.719
278.706
New England.
1,(43
184,797
63.012
6,601
8,818
229
1,855
1,713
68,485
41.931
8,997
12,953
8,098
9,397
32.346
98,271
69,932
Mane
212
18.558
30.995
561
578
26
217
132
19.550
2.414
871
1.185
695
828
7.845
10.766
2.921
New Hampshire
162
7.872
8.993
573
588
15
174
158
12.220
2.391
786
1.054
688
874
5.622
9.167
3.545
Vermont.
103
7.359
9.273
265
296
11
112
61
13,101
970
402
540
330
442
3.886
5,039
1.153
Massachusetts
642
89.519
7.824
2.987
3.066
99
748
831
13.203
20.604
3.944
5.936
3.393
3.887
7.929
42.306
34.376
Rhode Island
108
10568
1.055
500
519
20
140
108
1,472
4.380
648
990
544
654
935
6.003
5.068
Connecticut..
418
50.881
4.872
1.696
1.752
58
464
423
8.949
10.772
2.347
3.248
2.449
2.612
6.129
24.996
18.869
Mddle Atlantic.
4.837
891,458
99,723
17,219
18,112
892
5.858
<570
172.999
87,271
23,735
37,355
19,500
22.755
74,775
233.699
158,924
NewYo*.
2.119
288.350
47.377
8.078
8,490
412
2.496
2.079
72.967
37.364
10.029
17.871
8.361
9.593
28.953
98.002
69.049
New Jersey
929
151.694
7,468
3.806
3.967
160
1.094
1.101
11.628
22.413
6.022
7,672
4.956
5.520
10.597
57,071
46.474
Pennsytvana
1.789
151.415
44.888
5.335
5655
320
2.069
1.390
88.414
27.494
7.684
11.812
6.181
7.G42
35,225
78.626
43.401
East North Central....
8,163
882,521
243,981
19,293
20,780
1,488
7,354
4,339
444,583
121,144
27,806
41,597
23,852
29,932
131,686
314,458
182,772
Otto
1.588
201.197
41.004
4.887
5.254
367
1,831
1.089
81.041
31.113
7.402
10,762
7.002
8.521
33,999
79.1S7
45,158
tndana
966
73.497
35,932
2.577
2.752
175
1.127
594
73.833
17,702
3.590
5.474
2.748
3.708
21.657
43.643
21.986
iinois
1.483
186.400
55.645
5.328
5.751
423
1.835
1,315
103.842
31.468
7.186
11.393
6.242
7.662
23,769
75,756
51,987
Michigan...
1.425
160.509
56.954
4.159
4.529
369
1,730
921
91.120
26.683
6.379
9.189
5.481
6.945
31.038
75.706
44.668
Wisconsin
711
60.918
54.426
2,342
2.494
152
831
420
94.747
14.178
3.248
4.779
2.378
3.096
21.223
40.196
18.973
West North Cental...
2,704
202.289
908,132
8.965
9,085
800
3.293
2,027
694,788
55.283
11.492
17.430
10,232
14,400
79.617
145,113
65.496
Mmesota
466
42.019
79.548
2.139
2.261
122
601
532
118,776
14.067
2.471
4.236
2,466
3.172
16.531
3S.167
18.636
Iowa
443
27.074
55.965
1.369
1.449
80
508
275
103.840
8.632
1.843
2.767
1.937
2.699
13.216
20.606
7,592
Missouri
928
67.750
68.945
2.426
2.590
164
1.06S
634
104.385
15,297
3.471
5.057
2.667
3.712
20.922
43.379
22.457
North Oekota.
90
5,119
69.300
314
331
17
113
57
84.496
1.747
434
661
383
6S0
4,175
5.601
1.506
South Dakota
107
6.221
75.952
340
355
15
156
65
71.743
1.726
485
696
410
674
4,996
6.426
1,428
Nebraska
256
21.917
76.644
772
812
40
296
192
87.445
4.956
1.069
1,567
867
1.306
8,279
13091
4.812
Kansas
415
32.189
81.778
1.205
1.267
G2
534
272
124,073
8.858
1.678
2.446
1.502
2.188
11.496
20.561
9.065
South Atlantic
8.632
(79.712
266.910
19,591
20.649
1,058
9,714
4.487
387.779
111,771
28.540
41,625
25,182
32.835
156.024
352.504
196.480
Delaware
131
9.090
1.932
321
332
10
152
72
3.793
1.548
463
637
383
490
2,777
6.066
3,309
Maryland
729
85.265
9.837
2.303
2.406
102
815
565
16.129
11,836
3.009
4.566
2.719
3.309
12,293
36.493
24.200
District of Columbia.
51
13.935
63
310
331
21
82
86
0
1,102
389
637
251
268
0
3.368
3.368
Vi/ginia ..
906
80.115
39.704
2.869
2.996
126
1.075
606
51.631
14,494
4.053
5.933
3.614
4.628
25.093
54,834
29.741
West Virprva
418
27.805
24.119
6GG
750
81
503
99
32.202
2.971
1.303
1.858
821
1,194
9.896
13.742
3.846
North Caroina.
1.416
114.674
48.843
3.130
3276
146
1.577
613
74.928
18.306
4.319
6.405
3.492
4.870
29.636
54.600
24 964
South Can**
968
38.058
30,203
1.543
1.634
91
1.087
278
54,104
9,316
2.195
3.381
1.783
2.366
19,394
30.224
10,830
Georgu
1.441
94,941
58.056
2.888
3.056
167
1.618
709
86.594
20.173
4.215
6,209
3.597
5.026
27.649
60.293
32.644
Florida
2.572
215.829
54.153
5.558
5.870
312
2.805
1.439
68.396
32.025
8 593
11.999
8.522
10.684
29.206
92.864
63.578
East South Cental....
3.502
186,240
178,824
8,501
7,067
566
4,178
1,360
269,070
44,481
10,044
15,073
8.984
12,055
71,729
130,058
58.327
Kentucky
768
49.768
39.669
1.536
1.664
148
883
327
62.078
7.551
2.338
3.684
1.811
2.720
17.604
30.320
12.716
Tennessee
1.101
68.602
41.155
2.182
2,336
154
1.296
502
68,559
15.132
3.157
4.784
3.196
4.027
20.483
42.126
21.643
Alabama
978
42.670
50.767
1.748
1.895
147
1.185
353
73.493
14.673
2,774
4.016
2.633
3.547
20.120
37.437
17.317
Mississippi
657
25.200
47.233
1.035
1,152
117
812
178
64.940
7.125
1.776
2.589
1.345
1.761
13.522
20.173
6.651
West Soufi Central..
4,700
354,683
427,271
11,727
12.863
1,134
5,524
2,789
429.599
110,412
17.589
26,523
13,218
19,921
97,260
231,377
134.117
Askansas
550
20.206
52.078
1.003
1.092
88
658
215
69.457
7.630
1.658
2.343
936
1.445
11,829
18,306
6,477
Louisiana.
730
72,962
44.521
1.715
1.949
234
867
322
46.061
12.211
2.614
4.345
1.953
2.891
16.724
30,279
13555
Oklahoma.
540
34.620
68,655
1.449
1.565
115
668
359
96.748
12.334
2,163
3.211
1.931
2.887
16.192
31.606
15.414
Texas
2.880
226,896
262.017
7.560
8.257
697
3.330
1.893
215,293
78.237
11.153
16.624
8.398
12.296
52.515
151,186
98.671
Mountain..
2.881
190.480
855,194
8,965
8,417
484
3,209
1,981
444,227
42.155
8.980
13,147
7,356
10,736
56,281
116,448
60,167
Montana
198
8.442
145.388
373
403
X
223
81
69.396
2.416
604
805
416
660
5.949
8.074
2.125
Idaho......
242
10.712
82.412
435
473
38
263
108
69.253
2.386
601
985
579
947
5.672
8.119
2.447
Wyomhg
111
4.906
96.909
219
240
21
107
47
38.211
1.864
339
477
282
478
4.070
5.367
1.297
Colorado.
516
39.870
103.596
1.564
1.694
130
615
424
65.653
11.077
2.310
3.261
2.221
3 033
10.497
26.968
16.471
New Mewco
493
25.780
121.335
622
682
61
525
148
49.157
4.592
1.038
1.479
803
1,285
8.234
15.116
6.882
Arizona
811
63.278
113.508
1,511
1.612
101
925
456
66.517
11.206
2.273
3.438
1,712
2.417
13.367
31.729
18,362
Utah
271
21.492
82.073
709
757
48
307
180
44.532
5.369
1.006
1.678
754
1.114
4.819
12.679
7.860
Nevada
239
15.965
109,894
522
566
35
244
143
41.509
3.245
719
1.024
589
812
3.673
8.396
4.723
Pacific
1,376
483,662
96,252
17.038
18,132
1,096
7,508
5,559
252.368
96.140
24,673
36.622
20,903
27,412
80.732
299,271
218,539
Washington
691
67.665
66.511
2.065
2.255
171
865
606
63.399
16.110
3.157
4.533
2.702
3.828
14,799
38.520
23,721
Oregon
553
38.850
96.184
1.301
1,387
88
642
345
85.388
8,527
2,028
2.701
1.657
2.243
13.676
23.332
9.656
Cafdomia
4.935
360.699
156.299
12.938
13.729
792
5.774
4.391
90.546
68.386
18,563
27.781
15.715
20.294
47.932
226.301
178,369
Alaska
70
4.963
570.833
221
248
27
90
72
10.360
1.722
300
539
225
357
2.123
3.900
1.777
Hawaii....
127
11,465
6.425
493
513
20
134
145
2.675
1.395
628
1.068
604
690
2.202
7,218
5.016
-------
Table 2-15. Dependent Variables for Regression Examples
Example
Dependent Variable
1
Receipts for SIC 7532 - Top/Body Repair & Paint Shops
2
Employees for SIC 7532 - Top/Body Repair & Paint Shops
3
Sales for SIC 5198 - Wholesale Trade: Paints, Varnishes, and Supplies
4
Employees for SIC 5198 - Wholesale Trade: Paints, Varnishes, and Supplies
5
Non-Fatal Accidents by State
6
Estimated Number of Refmishing Painters by State and County
In the regression analysis tables for each example below, the t-ratio is
calculated for each variable included in the model. The t-ratio indicates whether
the coefficient (b,) of the variable (Xj) is statistically different from zero. The
statistical level of significance is coded at the bottom of each table. A significance
level of 0.05 suggests there is 95 percent certainty that the coefficient of the
variable is not equal to zero, while 0.01 suggests 99 percent certainty.
R2 measures the strength of the linear relationship between the dependent
variable and the predictor variable(s). The value indicates what percent of the
total variation in the dependent variable can be explained by the model. An R2 of
zero represents no correlation while an R2 of one represents a perfect fit. There is
no predetermined value of R that represents a strong relationship. Interpretation
of the strength of the relationship has the greatest meaning when comparing
similar regression models using the same data.
Example 1: Auto Reftnishing Receipts
The first example results for correlations with refmishing receipts are given in
Table 2-16. Licensed drivers correlates best and the multiple regression R2 is
0.963.
Table 2-16. Receipts for SIC 7532 Predicted by 5 Independent Variables
Dependent variable ซ REC_7532
Variables
t-ratio
Licensed Drivers
9.470**
Highways (Rural)
-3.075**
Accidents (Non-fatal)
-4.549**
Accidents (Fatal)
-3.652**
Highways (Urban)
-2.572*
R2 =0.963001 Adjusted R2 =0.958597 Cp = 39.32060
* Statistically significant at the 0.05 level. ** Statistically significant at the 0.01 level.
2-38
-------
Example 2: Auto Refinishing Employment
Employment correlates slightly better than receipts with the independent
variables. The same explanatory variables serve to predict county refinishing
employment levels, as shown in Table 2-17.
Table 2-17. Employment in 7532 as Predicted by 5 Independent Variables
Dependent variable - EMP_7532
Variable
t-ratio
Licensed Drivers
9.937*
Highways (Rural)
-2.831*.
Accidents (Non-fatal)
-4.465*
Accidents (Fatal)
-3.154*
Highways (Urban)
-2.858*
R2 =0.972707 Adjusted R2 =0.969457 Cp = 31.74646
* Statistically significant at the 0.01 level.
The results shown in the first two examples are very similar. The stepwise
selection chose the same five variables to predict the quantity of receipts and the
number of employees in SIC 7532. More than 95 percent of the variation can be
explained by these five variables. The stepwise regression also selected licensed
drivers as the first (and most significant) variable for SIC 7532, explaining more
than 91 percent of the variation in receipts and more than 93 percent of the
variation in employees.
Example 3: Motor Vehicle Accidents
For this example, a previous explanatory variable, Non-fatal Accidents
(ACC_NF) was selected as the dependent variable to help explain the variation in
the business activity from SIC 7532. Conversations with experts from the
automobile refinishing industry have revealed that most automobile refinishing is
performed as the result of collision damage. Therefore, it is likely that emissions
from automobile refinishing is proportional to the number of accidents.
The regression takes three years of data collected for each of the contiguous
states. These accidents (as opposed to fatal accidents) are more likely to generate
demand for automobile refinishing. The explanatory variables in this regression
were chosen because they do not exhibit multicolinearity (redundant variables
which can skew the results of statistical analyses).
Table 2-18 shows that there is a strong linear relationship between the number
of accident injuries and the explanatory variables, as expressed by the relationship
2-39
-------
Non-fatal Accidents = bn + b, (Licensed Drivers) + b2(Urban Highways) + b:j (Rural Highways)
Table 2-18. Results from Motor Vehicle Accident Analysis
Variable
t-ratio
Constant, b0
1.715
Licensed Drivers
16.278*
Highways (Urban)
1.744
Highways (Rural)
-2.678*
R2 = 0.949 Adjusted R2 = 0.948
* Statistically significant at the 0.01 level.
The t-ratios and R2 indicate that:
1) As the number of licensed drivers and miles of urban highway in a state
increases, the number of accidents is expected to increase.
2) As the number of miles of rural highway increases, the number of accidents
is expected to decrease.
3) Nearly 95 percent of the variation in accidents can be explained by these
three predictor variables.
Example 4: Sales in SIC 5198 as Predicted by Labor Force. Total Vehicle Miles,
and Population
The sales for SIC 5198 should provide the best guess at sales of auto refinishing
paints using readily available data (see Table 2-19 the best data source would be
the paint manufacturers see Section 3.3).
Table 2-19. Estimating SIC 5198 Sales
Dependent variable =
Sales from SIC 5198
Independent Variable
t-ratio
Civilian Labor Force Employment
4.114**
Total Vehicle Miles Traveled
-2.979**
Resident Population
-.2.387*
R2 = 0.923707 Adjusted R2 = 0.918125 Cp = 22.27208
* Statistically significant at the 0.05 level.
** Statistically significant at the 0.01 level.
These results are not as impressive, but that was expected because SIC 5198 is
defined as a wholesale trade industry. The sales and employees associated with
the industry are not always located in the state where the auto refinishing is
2-40
-------
performed, and a significant portion of the paint products are likely to be used for
other applications, such as marine painting.
Example 5: SIC 5198 Employees as Predicted by Labor Force and Rural Highway
Miles
The labor force and rural highways were the best two variables for this
correlation. The strong negative t-ratio for rural highway miles would seem to
imply that most SIC 5198 employment is in non-rural, metropolitan areas.
Despite these characteristics, 92 percent of the variation in sales and almost 96
percent of the variation in employment could be explained using the variables
selection. The number of persons employed in the civilian labor force was the
most significant variable with SIC 5198, explaining 89 percent of the variation in
sales and 95 percent of the variation in employees (see Table 2-20).
Table 2-20. Estimating SIC 5198 Employment
Dependent variable = EMP_5198
Independent Variable
t-ratio
Civilian Labor Force Employment
29.192**
Highways (Rural)
-26.27*
R2 = 0.957546 Adjusted R2 = 0.955525 Cp = 28.83186
* Statistically significant at the 0.05 level.
** Statistically significant at the 0.01 level.
Example 6: Auto Refinishine Painter Estimation
Establishment-level data was acquired from American Business Information
(ABI), including all establishments with business in SIC 7532-01 (Automobile
Body Repairing and Painting). This list reported 53,890 establishments with
primary SIC 7532 (Automobile Top and Body Repair and Paint Shops) and 10,634
with other primary SICs (Total: 64,524). Approximately 90% of the listing
includes a measurement of the exact number of employees.
For those establishments without employment figures, the shop was assumed to
have 4.5 employees. This is the average number of employees for the other
establishments with primary SIC 7532.
To estimate the number of painters (see Table 2-21), it was assumed:
(1) One-third of employees at an establishment are painters (BodyShop,
1992)
(2) No more than four painters per establishment (Meeting, 1993).
2-41
-------
To estimate VOC emissions from the painters, it was assumed:
(1) 5.72 lbs. VOC/coating gallon as sprayed (Meeting, 1993).
(2) 72% to coatings (i.e., According to one MRI study (Athey, 1988), 28%
of emitted VOCs come from surface preparation and
cleanup. Therefore, 5.72 lbs. VOC/coating gallon was
divided by 0.72 to account for surface preparation
and cleanup emissions, resulting in 7.94 lbs.
VOC/gallon.)
(3) 7.94 lbs. VOC for all gallons
(4) 0.125 gallons per painter hour (Mitchell, 1992)
(5) Painters work 8 hours per day, 250 days per year (Meeting, 1993).
From these assumptions, emissions from the automobile refinishing industry
are estimated to be 185,763,471 lbs. (or 92,882 tons) of VOC for the United States
in 1993.
As a rough estimate, these figures can be compared to:
(1) A recent study (Everette, 1993) indicating emissions of 106,700 tons.
This figure does not include surface preparation or cleanup.
(2) The EPA-CTG which estimates emissions at 286,000 tons.
Table 2-21. Estimating SIC 7532 Total Painters (Single-Variable
Regression)
Variable = 7532 PNTRTOT
Independent Variable
t-ratio
R2
Licensed Drivers
121.270
0.9878
Civilian Labor Force (Total)
114.903
0.9864
Registered Motor Vehicles
103.438
0.9833
Resident Population
101.655
0.9827
Motor Vehicle Fatalities
52.627
0.9380
All variables are statistically significant at the 0.01 to 0.05 levels.
In addition to the multiple regressions just shown in Example 6, emission
estimates were also derived per state using the same assumptions. Single
variable linear regressions were executed using the derived emission estimation
from estimated painters and 1991 explanatory variable values from the 48
contiguous states. The explanatory variables included in the regressions in order
of their correlation rankings (in terms of their R2 value) are shown in Table 2-22.
2-42
-------
Table 2-22. Single Variables for Emissions Estimates from Number of
Painters: 48 States
Regression Results
Explanatory Variable
Coefficient
Constant
R2
Civilian labor force, total
0.6918
131.45
0.97963
Licensed drivers
0.5213
101.58
0.97961
Civilian labor force, employed
0.7465
120.14
0.97931
Resident population
0.3376
163.96
0.97488
Registered motor vehicles
0.4783
59.451
0.96990
Registered automobiles
0.6078
126.48
0.96951
Vehicle miles traveled, total
0.0411
74.348
0.95944
Vehicles miles traveled, urban
0.0536
494.11
0.93508
Non-fatal accidents
0.0237
243.44
0.91846
Fatal accidents
2.4101
82.473
0.91197
Road miles, urban
0.1119
183.02
0.90994
Vehicle miles traveled, rural
0.1287
-433.33
0.75825
Road miles, rural
0.0221
481.70
0.20368
Therefore, the best explanatory variables for use in estimating the derived
state-level emission estimates are total civilian labor force, licensed drivers,
employed civilian labor force, and resident population
The same sort of single variable linear regression was also performed for SIC
7532 Sales. The relative correlations, as reflected by R2, Table 2-23, are seen to be
less strong, and with a somewhat different priority ordering for the explanatory
variables. However, the results support a preliminary conclusion that good
estimates can be obtained from annually accessible, local data.
Table 2-23. Single Variable Linear Regressions for 48 States: Automobile
Top and Body Repair and Paint Shops SIC 7532 Sales ($1000)
from the 1987 Economic Census
Explanatory Variables
R2
Vehicle miles traveled, urban
0.9328
Licensed drivers
0.9116
Civilian labor force, employed
0.9092
Civilian labor force, total
0.9046
Registered automobiles
0.9040
Resident population
0.8943
Registered motor vehicles
0.8811
Vehicle miles traveled, total
0.8729
Non-fatal accidents
0.8411
Fatal accidents
0.8220
Road miles, urban
0.7285
Road miles, rural
0.7081
Vehicle miles traveled, rural
0.5200
2-43
-------
In many cases, the observations (data) are likely to exhibit different variances
(square of the standard deviation) heteroscedasticity because the observations
represent individual states that vary considerably in magnitude of data (California
has more than 60 times the population of Wyoming). Histograms for most of the
variables show that many states have relatively small measures while a few states
have relatively large measures for the different variables. This issue, though
relevant, is corrected using statistical methods such as taking the natural log of
the explanatory variables before running the regression model. The results of this
transformation showed the same direction of correlation between the explanatory
variables and the number of accidents, but with even greater statistical confidence
(R2 = 0.973, Adjusted R2 = 0.972).
A few state-level variables (e.g., Licensed Drivers, Civilian Labor Force
Employment) appear to be highly correlated to the auto reflnishing industry. This
type of data is collected at regular intervals, available to the public, and is
available for small geographic regions, such as counties and cities. If these same
variables show a high correlation to auto reflnishing solvent usage, national
emission figures may be allocated to smaller geographic regions using these
predictor variables.
2-44
-------
2.4 Analytical Tool Set Selection
The previous Section demonstrates that data for emissions estimation exist and
that they are readily available within acceptable limits. Therefore, to improve
emissions estimation, the next question was how to assess the variable
relationships and how to more accurately predict solvent use and emissions. The
typical multiplicative calculation for emissions was shown in Section 2.2 to be
Emissions = Activity level x Emission factor x [1-(CE x RE x RP)]
However, such an algorithmic approach may not deal well with such
information as expert opinion, linguistic descriptions of variables (e.g., "very high
solvent use"), very large numbers of variables, or large heterogeneous data sets.
Validation of solvent usage in the auto refinishing industry may show that such
complications are not actually important. However, until validation results are
available (see Section 3), the estimation should provide for using techniques that
can manipulate the best data to produce the most accurate results. This means
the development of a complete analytical tool set, using some new techniques that
are proving themselves to be efficient and capable of extending the range of
analysis in many research areas.
This Section describes the basic tools and Sections 2.3.2, 2.5, 3.2, and 4 describe
the planned utilization of the tool set once validation data is available to optimize
the analysis.
The current toolset includes Computational Intelligence tools, statistical tools,
and a geographic information system. Computational Intelligence (CI) is a term
adopted by the Institute for Electrical and Electronics Engineers (IEEE) for
innovative computational techniques like artificial neural networks, fuzzy logic,
and genetic algorithms. CI techniques have advantages where:
Information is imprecise or qualitative. Heuristic rules can analyze and
interpret this information.
The amount of data is very large.
Tool set characteristics are described in more detail in the following. Tools will
be used to
select the best data,
incorporate the best computational forms for using data,
incorporate expert opinion and linguistic descriptions, and
provide optimal information displays and information transmittal.
2-45
-------
Statistics
Simple statistical techniques are being employed as the preliminary step for
analyzing variable relationships and selecting a priority variable set. Multivariate
regression analyses (linear correlations) have been applied and correlation
indicators derived. Results obtained so far have encouraged development of a
comprehensive database of easily attainable variables that can later be used to
predict emissions. There appears to be a high correlation between the automobile
refinishing industry and several of the explanatory variables have been included
in the database, as discussed in the previous Section.
Analysis has been performed within Lotus 1-2-3 by examining 1) the
distribution of the variables across different regions of the country, 2) the change
in the variables over time, and 3) the ratio of one variable to another. These
relationships have been examined in both spreadsheet form and graphical form.
As described in Section 2.3.2, regression analysis has been applied to examine
the relationships between the variables. Econometric Software's LIMDEP a
flexible software package for estimating regression models most frequently
analyzed with cross section data has been used as the tool to examine the
county- and state-level, annual data. A first set of regressions predicts the level of
business (in terms of receipts, sales, or number of employees) in automobile
reflnishing-related SICs.
Expert Systems
An expert system is a computer-based system that contains human expertise or
reasoning capabilities. Traditionally, expert systems have been developed to
capture the experience and knowledge of a human expert(s). An expert system
may be developed because the expert may be leaving his organization or
profession, or cannot meet the many requests for his knowledge. More often than
not, an expert system may play an advisory role as a decision support system.
This use of an expert system is more readily accepted than replacing the "man in
the loop."
This study has examined over 30 expert systems which are related to
environmental control, such as GEOTOX (Alberta, 1990), a program with
knowledge base and inference engine designed to recommend hazardous waste
disposal methods. The programs can be generally classified under 1) diagnosis, 2)
planning, or 3) prediction. Almost all deal with highly specific, well-researched
problems and extend one or more previous solutions to similar cases. Most of
these expert systems are experimental and not used in actual practice.
2-46
-------
A true expert system approach for VOC emissions estimation would imply that
someone is now performing emissions estimates correctly and their knowledge can
be incorporated into rules that will produce consistently accurate estimates. This
is not the case. Therefore, a new expert system must evolve if one is to be applied
(see Section 3.2).
Expert opinion is considered to be useful in such areas as the following:
1) Explaining discrepancies between receptor models and other emissions
estimates.
2) Assessing potential impact of controls and operating practices in the
workplace to reduce emissions.
3) Characterizing shops most likely to use controls and responsible shop
practice.
4) General factors relating solvent use to population, employee numbers and
sales.
5) Rules of thumb for how much paint a shop will use.
Reliance on environmental experts has been deemphasized because of possible
bias toward previous estimation methods. Also, effort under this study has been
able to gather most of the substantive and creative information from past
emissions estimators during discussions in the first part of the case study and
from the literature.
Further study (Section 3) should emphasize industry experts, and possibly an
expert in the area of source dispersion. The latter should have an overview of the
influences on emissions by 1) types of equipment in use, 2) VOC chemical/physical
properties as they affect time and space allocations associated with individual area
sources, and 3) control and pollution prevention techniques at area sources. This
expertise may be strongest in the industry itself. The directive for the research at
present does not encompass atmospheric chemistry and specific precursor
potential for the VOC constituents from auto refinishing. A fuzzy logic expert
system has been chosen to augment estimation of VOC emissions from the
automobile refinishing area sources. The fuzzy logic concept is discussed next.
Fuzzy Logic
The rules derived from industry and emissions experts thus far have been, at
best, very general. Since "crisp-logic", rule-based, expert systems do not handle
"approximate reasoning" well, fuzzy logic expert systems are being considered.
Fuzzy logic is an approximate reasoning technique used in processing inexact
information. While a typical expert system may be thought of as defining "true or
false" conditions, fuzzy systems allow for varying degrees of truth, or "shades of
gray," more like human reasoning.
2-47
-------
Typical artificial intelligence (AI) expert systems chain through rules,
inferencing through a decision tree. These trees usually employ only part of the
non-interactive rules, the rules are crisply interpreted as true/false (0,1), and the
knowledge is structured but stored and processed symbolically, outside of
numerical computation. Fuzzy systems use structured knowledge in a numerical
framework, use weighted "principles" instead of true/false rules, and fire every
interactive rule with every inference (AIM, 1993). The numerical framework and
interactive approach allows analysis of the association of "clumps" of output spaces
with "clumps" of inputs, so that individual rules can act like partial derivatives, as
used to solve engineering problems.
Climate might be a factor in emission levels, and could be classified as dry,
moderate, or rainy. Areas where emissions occur might also be loosely classified
as: rural, suburban, or urban. A CI rule, based on expert opinion, may be
expressed as, "IF the climate is dry AND the area is rural THEN emissions are
low." Another rule may state "IF the climate is wet AND the area is urban then
emissions are high." These rules describe the increased likelihood of accidents and
auto reflnishing in a congested area with poor weather and vice versa. This type
of reasoning can augment more rigid evaluation computations.
Fuzzy logic will probably supply a set of secondary emission factors (VI, V2,
V3,....Vn) based on qualitative or uncertain influences to augment the best
quantitative data correlations between emissions and independent variables.
These would contribute by such a relationship as
Total Emission Factor = EFquantitation x VI x V2 x V3...
where a specific example might be
if(county=suburbanj and (winter=averagej then (VOC emissions factor V3=moderate)
Quantitative data are simple to fuzzify. If, for instance, a county has a given
population and a given number of farms, it may have a suburban membership of
0.3 (on a 0 to 1 scale) while its rural membership may be 0.7, based on
membership functions. A membership function in this case might be based on
population per unit area. In another case, a county's snowfall this year might set
membership for the winter as severe, average, or mild.
Fuzzy logic can be applied in the computational part of the estimation method
programmed rules and a resident database or using information supplied by the
interactive user while involved in completing local emission estimations. Of
course, an actual rule set will cover a wide range of possible parameters such as
accidents, refinishing regulations, equipment and types of coatings used, and
economic data. As the number of inputs grows, the number of possible rules may
2-48
-------
grow into the hundreds. To verify these rules, it is necessary to have extensive
validation data. This data should come from the nationwide survey (see Section 3)
which asks questions that are specifically geared to the variables considered
essential. Further clarification of fuzzy logic is presented in the next Section.
Neural Networks
An artificial neural network (ANN) is an analysis tool that is modeled after the
massively parallel structure of the brain. It simulates a highly interconnected,
parallel computational structure with many relatively simple, individual
processing elements or neurons.
Neural network paradigms are capable of learning or extracting a relationship
between two domains, i.e., two sets of information. Neural networks are best
applied where there are no known rules (expert systems) or algorithms
(statistically-based methods) that accurately relate the independent and dependent
variables of interest, but there is an abundance of data resulting from the
relationship.
Mathematically speaking, neural networks are non-linear differential equations
that parameterize themselves through various learning algorithms. One of the
most commonly used learning algorithms is call the backpropagation algorithm.
The name originates from the fact that the neural network learns from errors that
are propagated backward from the network's output to its input. Neural networks
are able to generalize. That is, a trained network will correctly classify an input
that it has not seen before. This has proven to be a powerful capability when the
training data is incomplete.
Advantages of Neural Nets: 1) The neural network does not have to be
programmed. It will learn from example; as in statistics, the more data points,
the better. 2) A neural network's ability to generalize will prove beneficial in
estimating emissions since disclosure rules may limit the availability of some
useful explanatory variable data points. 3) Once a network has been trained, it
provides an instantaneous output for each set of inputs. There are no iterations
required in generating an output.
Disadvantages of Neural Nets: 1) Since the network learns from example, if it
learns from bad data, its answers will be of equivalent quality. 2) The
backpropagation learning algorithm does not allow a network to train on
additional data points without being retrained on the original data set. This can
be a time consuming price to pay if one wishes to regularly update the network.
3) Exactly what relationship the network has learned is not readily apparent.
There is no built-in explanation capability. Since the relationship learned by the
2-49
-------
network is non-deterministic, it is difficult to quantify the reliability of a network's
output without exhaustive testing.
An important distinction between two important techniques in this study is that
"neural networks sum throughputs, and fuzzy systems sum outputs"
(ModelWare, 1992). Neural networks can adaptively generate fuzzy rules per
the technique of product-space clustering (ModelWare, 1992). This is an
attractive technique for handling a large body of unstructured knowledge {e.g.,
from experts in auto refinishing or computer files from different paint
manufacturers). Experts who generate input-output data do not need to articulate
the fuzzy rules. They only behave as experts. Rules are made equivalent to the
neural net synaptic vector clusters which track how experts associate responses
with input stimuli.
Polynomial Networks
Another possible technique considered for estimating emissions is polynomial
networks (AIM, 1993). The polynomial network tool tested is called AIM. AIM is
a numerical modeling technique that automatically synthesizes a mathematical
model of relationships among data. AIM then implements the resulting model as
a standard computer program. AIM excels at modeling numeric parameters such
as expert judgements, probabilities, fuzzy values, prices, costs, sensor readings,
and control settings. Models are not implemented as rules, but as mathematical
models called polynomial networks.
Rules (if7then rules) are very effective at modeling discrete symbolic knowledge.
AIM can develop functional models for complex numerical relationships that are
generally more compact and faster to execute than if/then rules.
Universal Process Modeling
Another technique being considered for use in estimating emissions involves the
use of universal process modeling (UPM). The UPM tool being used for this
analysis is called ModelWare (ModelWare, 1992). The UPM algorithm is an
empirical modeling technique. It requires a set of example data, known as the
Reference Library, which describes how the system or process behaves under
known operating conditions. ModelWare predicts system behavior from the
stored Reference Library, making a prediction for all system variables from each
set of input system observations. As with all other techniques, ModelWare's role
in the prototype system will be determined by its prediction accuracy with respect
to the emissions validation data.
2-50
-------
Genetic Algorithms
Genetic algorithms (GAs) are an optimization technique based on Darwin's
"survival of the fittest" theory. GAs use three processes found in natural genetics
to attempt to optimize a given problem: reproduction, crossover, and mutation.
The parameters of the problem are converted into a binary string. For example a
problem with four parameters may have a string which looks like
|10010|11000|10010|11101|
**4
Each variable has a given range over which one is trying to find the optimum
value. This range is linearly mapped to give the binary string as shown above.
A population of these strings is then created randomly and becomes the gene
pool of initial possible solutions. Each string is then evaluated for its "fitness" to
determine how well it solves the problem (e.g., emissions estimation). Based on
the fitnesses of all the strings, a new population is generated. Strings with high
fitnesses will be given a proportional number of copies in the gene pool, while
strings with low fitnesses may die out. The members of the new generation are
then randomly matched for crossover, where a point along the string is randomly
chosen and the "tails" of the two strings beyond this point are exchanged as shown
below.
Parents: 100101100010010 | 11101
010100101011101 | 00101
Children-. 100101100010010 | 00101
010100101011101 | 11101
After crossover, each bit is flipped (mutated) using a low probability
(somewhere around 0.0005). Mutation is used to assist in keeping diversity in the
population. This sequence of events is continued over several generations until all
the members of the gene pool converge towards the same string. This string
should contain the optimum values for the desired parameters.
Geographic Information System
Emissions are characterized by their levels and by their distribution in space
and time. Maplnfoฎ is a geographic information system (GIS) being used to assign
emissions-related explanatory variable values, such as number of employees and
annual sales, to the actual location of an individual business. These values can
2-51
-------
then be aggregated to combine values within a zip code, county, nonattainment
area, metropolitan statistical area (MSA), State, EPA Region, or nation. For State
Implementation Plan (SIP) emission inventory purposes, for example, area source
emissions would be aggregated to county and to nonattainment area boundaries.
The next subsection on system configuration and Section 4 better illustrate the
advantages of GIS.
2-52
-------
2.5
Configuration of the VOC Emission Estimation System (VOCEES)
The best screened data and analytical tools must be combined into an
integrated computation system that handles database management, data retrieval,
computation and analysis, and output of results. This system should obviously be
computerized but simple and affordable. This system has been designed and
demonstrated with surrogate data and tools as part of the current effort.
Developing the computational system for emissions estimations consists of
selecting the best explanatory variables and survey data and processing this with
the assembled tool set. Computational intelligence tools have been specified to
help meet the specified method development criteria. The tool set's statistical
components have been used to prescreen the best data for use in development.
System Tools
A personal computer-based VOC Emission Estimation System (VOCEES)
automates the data management, computation, and displays/reports under the
new method. The main components of the system are: 1) the essential explanatory
variable database; 2) basic algorithms; 3) the supplemental fuzzy logic expert
system; and 4) the GIS-based user interface.
Figure 2-6 shows the system paradigm to which the research team has worked.
This model is adaptive. The final format for emission estimation will depend on
the final, validated data correlations and variable selections.
VOCEES is a fully integrated PC-based software package. Since the system is
intended to present data in both map and graph oriented fashions, it was designed
to work under Microsoft Windows. The heart of the system is an emissions
estimation engine derived both in-house and through the use of commercially
available software. The core of this engine is written in BASIC, as is the user
interface. As additional components become available for the engine (i.e., fuzzy
logic or neural networks), they will be written as BASIC code and incorporated
directly into the core code or can be written as C or Pascal direct link libraries
(DLLs) and connected through Windows.
Maplnfoฎ is a geographic-oriented, information display software that provides a
natural choice for the VOCEES system. It incorporates the two major components
necessary for this type of project: database management and mapping capability.
Maplnfo provides a full-featured Structured Query Language (SQL) database
management system that can directly use either dBASE files or files written in
Maplnfo's native format. Also, Maplnfo can import Lotus 1-2-3 and Microsoft
Excel spreadsheets, and, since it is a Windows based product, can use cut and
2-53
-------
Dual Products
VOCEES
Computerized
Analysis
State
Implementation
Plan Report(s)
Expert
Analyses
Displays
Comp
So
utational
tware
Resident
Datafiles
Analytical Tools
Data base Managers
Statistics, GIS
Al: Expert Systems,
Neural Networks,
Genetic Algorithms
Selection of Best
Computational Technique
& Data
County / State / National
Data
Figure 2-6. System Configuration in Context of Method Elements
paste to create Maplnfo files from spreadsheets not directly supported. Adding to
the list of features which made Maplnfo the basis for VOCEES is its companion
product, MapBasicฎ. MapBasic is a BASIC language compiler which provides
additional commands specifically designed for Maplnfo. MapBasic was used to
provide the customized Windows interface which transformed Maplnfo from a
generic mapping software program to an emissions estimation software program.
The geographic coordinates of source locations can be stored in the GIS and
analyzed and displayed along with county boundaries and even roads, if desired.
Road networks, with attributes indicating the type of road, are available from the
Census Bureau's TIGER data (Topologieally Integrated Geographic Encoding and
Referencing). The TIGER data also has zip code and street address information
which allow locations to be geographically located from their street address. The
TIGER files and other data bases which can be displayed with the GIS system can
represent all 3300 counties as polygons with representative coordinates. The past
2-54
-------
emissions estimates by county can be represented using the GIS system to
highlight any geographic trends or biases. The same can be done with vehicle
registrations, accidents, distribution of auto body shops, and nonattainment areas.
System Integration and Function
Using the object libraries available through MapBasic, a Windows-based
Graphical User Interface (GUI) was created. This interface was designed with
ease of use as a primary consideration. The interface follows the standard
Windows interface in both look and feel. This is necessary to maintain the concept
that all Windows applications should behave and appear the same to increase
their usability. The menu bar is the main control center of the interface. From
the menu bar, the user can open windows containing information about different
geographic regions, edit this information, calculate emissions, and graph or map
the results. As with any Windows program, the application can be minimized to
allow work on other applications or maximized for full screen use.
VOCEES was designed with a maximum flexibility in the display of
information. Data can be presented as maps, graphs, or tables. Users have a
choice of examining counties, nonattainment areas, states, or EPA regions. Also,
since VOCEES can be used to examine emissions for different years, temporal
changes can easily be observed. This capability is especially useful in the
examination of the impact of regulations: local, state, and possibly national. One
of the advantages of VOCEES is its incorporation of regulatory impact of VOC
content limitations and the enforcement of these regulations. Maps, graphs, and
tables can display the explanatory variables or the emissions calculated using
these explanatory variables. Maps have a wide range of combinations, including
counties in a Nonattainment Area (NAA), counties in a state, NAAs in a state, and
so on. Since emissions can be calculated using any of these explanatory variables,
the user has the opportunity to graphically display variations in emissions
distribution using these variables. The ability to display information in many
formats can also enhance SIP generation. Many different display modes enhance
the presentation of data, and this fact is key to VOCEES functionality. Mapping
provides the ability for the user to get an immediate idea of emissions distribution
at the county, NAA, or state level, something that cannot be done through the use
of tabular data.
VOCEES can examine temporal changes in explanatory variables, regulations,
and emissions. The database used to generate these emissions can be modified by
the user. From the main menu, the user can choose to update any information at
the county level. These changes will be used to update any NAA information (if
the county belongs to a NAA), state level information, and EPA region
information. Changes cannot be made above the county level since that would
mean an aggregation of county level data would not match the corresponding
2-55
-------
NAA, state, or EPA region data. Included with the system is default data for each
year so if data is accidentally lost a reasonable replacement value can be obtained.
Regulatory information is given in the form of VOC content limits for the various
auto reflnishing products. These limits can be updated and the enforcement of
these regulations can be modified (from 0% to 100% enforcement).
File structure is kept as simple as possible. There are 4 input file types
containing county, NAA, state, and EPA region level data. Each year of data
would contain each of these four types of files. These files are in Maplnfo's .TAB
format. In addition there are 4 files that contain the information to map the
counties, NAAs, states, and EPA regions. Separating the maps from the data can
greatly reduce the amount of disk space needed since only one set of maps is
needed for many years worth of data. Output can be done in a number of ways.
Information changes will be kept in the aforementioned files. Also, any tables
created to display infoimation in VOCEES can be saved as either an ASCII text
file or a dBASE file. Finally all information displayed on the screen can be sent to
any printer through the use of Windows.
2-56
-------
3. EXTENDED method development
Five areas of new method development have been begun under this case study,
but could not be completed within the confines of time available and the original
scope of work. Based upon analysis, preliminary results and EPA
recommendation, these five areas should be extended to completion in future. The
areas, in suggested order of importance, are
1) Nationwide Survey for Validation of Emissions and Verified Correlations
with Explanatory Variables - this activity is essential to the project's
success, providing confirmation of solvent use and variable correlations.
2) Application of Computational Intelligence (Fuzzy Logic, Expert Systems,
Neural Nets and Genetic Algorithms) to Emissions Estimation Using
Validated Data ~ these tools allow new kinds of information to be used,
accelerate the data selection process, and provide more accurate estimates.
3) Re-examination and Potential Negotiations of the Use of VOC-Containing
Product Manufacturers Data for Validation and Database Updates -- very
important to provide an accurate estimate of geographic distribution of
product categories and to compare with survey data, and extend survey
results to an annual validation update.
4) Sampling and Chemical Analysis of Selected VOC Species at a Limited
Number of Area Source Sites ~ important to improve credibility of
emissions predictions and to improve the VOCEES extrapolation of product
use data to actual annual volume of emissions of VOC.
5) Graphical Interpretation of Past Emission Estimation Data important for
comparison with VOCEES estimates, to illustrate needs for estimating
improvement, and to provide a baseline perspective of techniques to date.
A Method for applying the techniques is shown in Figure 3-1. This figure
diagrams the recommended New Method for area source emissions estimates.
Under the method, a national survey of auto refinishing shops and input from
area source experts combine to provide validation data. "Industry experts" would
ideally include data from paint manufacturers on distribution of parts and
solvents by geography and type. Expert opinion from industry consultants,
managers, and sales representatives will also contribute to rules that make up the
knowledge base and inference engine for computational intelligence analysis.
Validation data consists mainly of area source VOC use by region.
This information is matched against the explanatory variables like those used
as surrogate data in the first part of this study. Part of the data will be used to
3-1
-------
DUAL PRODUCT
SIP
= Complete
COUNTY I STATE / NATIONAL
NEW DATA
Factors
Weights
Selected
Data
VOCEES
Rules
INDUSTRY
EXPERtS
EXPERT
ANALYSIS
GIS / OUTPUT
VALIDATION
DATA
NATIONAL SURVEY
OF SHOPS
ALGORITHMS
FUZZY INFERENCE
ENGINE
NEURAL NETS
GENETIC
ALGORITHMS
Figure 3-1. Recommended New Method for Estimating Emissions
train a neural network to establish optimum weights and function modes to
accurately predict VOC use. Trained emission estimation performance will be
compared to the remainder of the validation data. Genetic algorithms and
statistics will also be used to select the best variables for estimation and optimize
estimates. With such validated variables and database selection, surveys need not
be performed annually. The main variations in VOC use from year to year should
be predictable by expert rules for such factors as new regulations and the
distribution of use of various types of paints (as qualitatively assessed by industry
consultants and paint manufacturers). A resident database is derived and
formatted using the best explanatory variables and the best data for the period(s)
under evaluation. Under the planned method, the resident data would be updated
by the State's using local data. This information would be filtered, selected, re-
correlated, assigned improved rules, and so on by centralized processing by EPA or
its agent.
Selected variables, weights, correlation coefficients, relational algorithms, fuzzy
logic membership functions and a rule set are then combined to complete the
calculation paradigm and inference engine. It is anticipated that expert opinion
and fuzzy logic will be applied where linguistic descriptions of behavior pertain, as
well as data uncertainty, qualitative assessment, and complex, and nonlinear
relationships.
3-2
-------
The influence of regulations, estimated enforcement, general economic trends,
prevalent shifts in paint mixtures, swings in insurance legislation (e.g., increased
no-fault insurance) could possibly all be best evaluated through inference methods.
Examples are given in the following discussions of neural nets, genetic algorithms
and fuzzy logic.
Once the tools have been applied to the resident database to generate emissions
estimates, then the results can be reported and further examined. As described in
Section 2.5, this is accomplished through the MapInfo-based VOCEES data
handling. A dual product has been devised: first, a State Implementation Plan
type of report can be generated showing emission estimates for nonattainment
counties and map displays of same. Second, considerable analysis of emission
estimates can be done over the U.S. geography and over any period of time for
which data exists. Examples of those products are given in Section 4.
The shaded areas in the modules on Figure 3-1 give an estimate of the level of
completion of each. Completion of the national survey of auto refinishing shops
should allow final resident database selection, inference methods, final design, and
training and exercising of the entire method with validated data. Accuracies of
the method can be closely assessed and compared with the current approaches
discussed in Section 2.2. The generation of expert rules, resident database
updates, and low-cost, selective validation should be part of a continuing process.
Further attention to items 3) and 4) above could provide EPA a very cost-effective
means for insuring validation and high accuracy for its area source emission
estimations.
Therefore, the new method covers the basic standard emission estimation
elements: activity level, emission factor, control factor, and spatial and time
allocation. Speciation can be included, but has not been addressed in this study.
For auto refinishing, the emission factor level is straightforward, since all VOCs
are either emitted or controlled (e.g., recycled, incinerated, charcoal filtered).
Therefore, emission factor = 1. Control has only a small impact on auto
refinishing emissions at present.
The five areas of method completion are summarized next.
3-3
-------
3.1 National and Intensive Local Area Surveys of Auto Refinishing
Shops
The only obvious sources for obtaining VOC-use validation data are from the
user, from the distributor, and from other agencies. While OSHA requires
Material Safety Data Sheets (MSDS), no quantities of the amount of product
purchased are included (although it might be possible to correlate MSDS mailings
and sales forms at the wholesale or retail levels). No Federal Census microdata
{i.e., County Business Patterns (CBP)) provides quantities of materials used by
area sources, nor do "Yellow Pages" files. Documentation of Census of Sales by
paint and solvent suppliers does not adequately focus on auto refinishing to enable
estimation to better than 50% in a region. Even then, more widespread
distribution of the product from each source to other regions is not known.
Therefore, validation data must be generated through users or manufacturers.
Manufacturers are reluctant because of 1) proprietary and marketing
considerations, and 2) the level of effort estimated to be required to document.
The latter source should not be neglected (see Section 3.3), but it was decided to
move ahead with a user survey approach - to take a verifiable snapshot of VOC
use by the approximately 60,000 shops in the U.S. A preliminary survey plan and
questionnaire had been prepared. Focus group meetings have been held with
managers of auto refinishing shops and a preliminary endorsement for the survey
has been provided by the Automotive Services Association (ASA), the main trade
association for the auto refinishing shops. The planned survey has been named
ARSUS - for Automobile Refinishing Solvent Use Survey.
The planned Automobile Refinishing Solvent Use Survey (ARSUS) is the first
survey of its kind and almost certainly the most important area source emissions
estimation validation undertaken to date. It is also essential for proving the merit
of the new estimation methods under development. Survey and statistical experts
will conduct ARSUS.
Other important features of ARSUS are
Results will be statistically defensible, based on random probability sets,
with the results represented by statistically correct accuracy estimations
and confidence levels.
Proper survey techniques will be applied to assure a high percentage
response, assuring that results are representative and defensible.
Results are expected to increase available validation data by 3 orders of
magnitude and improve accuracy by at least 20 to 200 percent.
3-4
-------
The survey is designed based on a detailed knowledge of the industry built
on numerous contacts with shop operators, paint manufacturers and
association representatives.
This validation will allow detailed examination of geographic (spatial)
influences on solvent use and will provide a more representative data base for
area source solvent emissions. Results are being displayed and analyzed utilizing
the geographic information system.
ARSUS is a national survey of auto body repair shops, and will also include
local-area intensive surveys of 6 high-population areas. The national survey data
will be divided into two independent sets with probability proportional to
population, each containing 30 areas or Primary Sample Units (PSUs). One of the
samples will be assigned for use in estimating model parameters (training the
technique).The second national sample will be used to develop a comparison
variable (validating the technique or model for emissions estimations) with an
unbiased estimate of the difference between the model estimate and the true
value. This information should improve current emission estimates accuracy by
20% to 200%.
The survey contains 5900 field samples for the national aggregate survey, about
3400 of which will be allocated to the 6 intensive local-area surveys (a subset of
the national sample). Local areas for intensive survey will be selected randomly
and all sampling will be probability-based. Examples of such "local areas" are the
1) four county region comprising most of the Boston Primary Metropolitan
Statistical Area (PMSA) with a combined population of approximately 2.8 million
persons or 2) the Central Piedmont region of North Carolina comprising 12
contiguous counties and three MSAs (Greensboro area; Raleigh-Durham; and
Burlington) with a combined population of about 2 million.
The map shown in Figure 3-2 is a hypothetical example of the scope of coverage
that would be provided by the 5900 samples and the local-area intensive surveys
of 6 high-population areas. The survey data is divided into two independent sets
with probability proportional to population, each containing 30 areas or Primary
Sample Units (PSUs). One set is assigned for estimating model parameters
(developing the method), while the second set is for developing a comparison
variable (evaluating and validating the method).
The PSUs are counties or groups of contiguous counties. About 60 to 75
completed interviews can be expected for each PSU or local area. Since this small
a subset is insufficient for precise local estimates, much larger samples are to be
used for the 6, randomly selected, locally intensive surveys. About 400-550
responding firms are expected in each local intensive area. These figures reflect
the fact that the survey experts associated with this study typically obtain a high
3-5
-------
. =PSU
- local Intensive study
Figure 3-2. Representation of 5900 Samples and 6 Local Intensive Surveys
response rate from 70% to 95%. The estimated accuracies of solvent use data from
the surveys are shown in Table 3-1 below.
Table 3-1. Estimated Accuracies of Solvent Use Data
SURVEY
SAMPLE SIZE
FIELDED
PRECISION
ESTIMATE
DIFFERENCE
ESTIMATE
NATIONAL
5900
7%
10%
6 LOCAL
INTENSIVES
3400
10%
14%
Information services (e.g., American Business Information (ABI) or Dun and
Bradstreet) and computerized "Yellow Pages" will be used to retrieve information
abstracts on firms with either primary or secondary SIC code 7532 for auto
refinishing in the counties selected. Primary and secondary SIC 7532 includes all
top/body repair and paint shops and all car dealerships with body and paint shops
(covering more than 95% of all listed refinishing establishments). Codes for auto
repair such as 7533, 7536, 7538, and 7539 are not representative of refinishing.
This file will be stratified by SIC and number of employees and a probability
sample of organizations in each of the sample PSUs will be constructed. A
detailed file for each sample organization will then be retrieved with the names
and addresses for the mail survey and auxiliary data for use in the final statistical
analyses. These files constitute a "frame" that assures complete coverage of the
3-6
-------
industry and statistically valid estimates (unbiased estimates with known
precision).
The questionnaire for the auto refinishing industry has been prepared. This
questionnaire is found in Appendix A. The questionnaire is being further updated
following review at the ten EPA regional offices and at national trade associations.
The survey's design and data collection are summarized as follows:
Survey experts will collect the data from the selected auto refinishing shops
using a combined mail and telephone survey approach to maximize response rates,
minimize respondent burden, and complete the data collection efficiently. A lead
mailing is sent to each selected establishment, including a cover letter, letters of
endorsement from State agencies and trade associations, the questionnaire,
instructions, a postage-paid return envelope, and a toll-free number to request
additional information or arrange an immediate interview. Two weeks after the
surveys are received, those establishments which have not returned the data
collection instruments or called to provide the data or arrange an appointment will
be called. Telephone Survey Unit (TSU) interviewers will prompt for the
completed questionnaire and will offer to record the data over the phone. Another
call may be made at a time more convenient for the respondent. Results will be
entered using bar-coded ID labels and event codes which indicate the pending or
final status of the case. Batches of responses are sent to Data Entry (DE) for
processing. All documents will be double keyed for quality control, and sent to
statisticians for analysis.
A preliminary survey of 9 randomly selected local auto refinishing shops has
been performed. This is the maximum number allowed by the OMB without fall
survey approval. Results for the most part are confidential. So few samples
cannot provide any association of solvent use with the explanatory variables.
They do offer some limited indications of shops practice, and projected survey
success, such as:
Response in general was thorough and prompt. Only one shop declined any
response.
The correlation between the number of employees and number of jobs or
amount of VOC use may not be as strong as anticipated.
Revenues and number of employees are difficult to correlate with a shop's
use of more advanced equipment or paint products. A high ratio of
revenues/ employees may indicate "high-end" shops refinishing more
expensive vehicles, and utilizing higher grade (and lower emission) products
and tools (e.g., HVLP spray guns).
Knowledge and data gained from interaction with regulatory and industry
representatives from the six local intensive areas will also be used to expand the
3-7
-------
knowledge base of the emission estimation expert system, to more accurately
assess influences of climatic, demographic, regulatory, and economic diversity.
The survey data are expected to help weigh the contribution of complementary
activity levels and their emission factors. This is analogous to tuning the system
with shop-level data. Complementary, regionally-specific survey data will be
integrated in a Geographic Information System (GIS) data base format with U.S.
Census data and other activity level data and geographic information. Once in the
GIS format, these data can be manipulated, analyzed, and visualized to support
and better validate the expanded emission estimation method.
Results from a survey by the trade publication, BodvShop Business, are already
quoted in this report. Two other surveys of auto refinishing have been reviewed
which provide some perspective of the industry, although little environmental
analysis. These survey results are discussed below.
Twin Cities 1990 Autobodv Survey
In 1990, Research Innovations of Minneapolis, MN conducted a survey of
autobody shops in the Minneapolis - St. Paul area for the Automotive Service
Association (ASA). Although not directed towards VOC estimates, it does assist in
characterizing the body shop industry. Research Innovations purchased a list of
480 body shops in the Twin Cities metro area. Initially, 51 responses were
received (10.6%). A telephone follow-up yielded another 154 responses for a total
of 205 (43%). Survey results indicated that approximately 80% of all body shops
in the metro area were independently operated with the other 20% being
dealerships.
The survey found that 40% of the body shops had 1-4 employees, 20% had 5-7
employees, 40% had 8 or more employees, and the average number of employees
was 7.5. As the number of employees increased, so did the shop's profitability. An
increase in personnel was usually characterized by an increase in support
personnel and not painters or bodymen. This indicates that using support
personnel to do paperwork enables paint and bodymen to be more productive and
therefore make the business more profitable. Also, higher profitability for larger
shops lends credence to the observation that although the number of body shops
are decreasing, those that remain open are becoming larger.
Technology influences are also beginning to show in modern body shops. Of the
205 respondents, 30% indicated that they had a HVLP spray gun. Of shops that
had over $1 million in sales (the highest sales volume category), this number was
nearly 60%. Computerized management systems have also become popular,
especially with larger shops. While only 3% of shops with sales of less than
3-8
-------
$250,000 have a computerized management system, that number jumps to 56% for
shops with sales of over $1 million.
Cook ! DuPage. County 1992 Autobodv Market Study
In the summer of 1992, Research Innovations also did a study of the
Cook/DuPage County, Illinois area for ASA. Using a list of 1,139 body shops,
Research Innovations used 144 responses (12.6%) in their analysis. Of the
respondents, 89.6% were independently owned, 6.4% were franchises, and 4.0%
were dealership owned.
The average body shop in the Cook/DuPage area had 9.89 employees, with 2.3
refinishers on average. As was true in the Twin Cities area, larger shops had the
highest gross profit. Of the total shops with gross profit margins of 40% or more,
nearly 40% were shops with sales volumes of $1 million dollars or more.
Technology changes are also becoming more apparent in the Cook/DuPage area.
54.4% or respondents indicated they presently own a HVLP gun with another
11.2% indicating they plan to purchase one within a year. Computerized
management systems are also becoming more popular with 43.1% of shops
presently owning a system and 10.1% planning on buying one. Of the
computerized management systems in use, 22.6% provide the ability to track
VOCs through information on paint consumption.
3-9
-------
3.2 Use of Computational Intelligence (CI)
As described in the introduction to this Section, Computational Intelligence will
be used to
1) Select the best initial data matrix to become the VOCEES resident datafile,
by assessing estimates from data and assigning weights to the multiple
explanatory variables. CI techniques such as neural networks can prioritize
variables represented by complex data that may have little apparent
correlation on face value.
2) Continuously update the best data by a training function that reassigns
data based on its ability to most accurately predict validated VOC levels.
3) Verify and/or improve statistical correlations between validation data and
explanatory values.
4) Quantitatively interpret qualitative responses from the nationwide survey
or from other sources (an example is provided for fuzzy logic analysis later
in this section).
5) Interpolate to fill in data gaps in the survey or independent, explanatory
data sets. For example, a neural net or genetic algorithms can learn the
pattern of response to a certain question about shop operation as the
response relates to other responses to other questions. The missing
response can then be predicted. The same interpolation function can also be
performed for data other than that derived from the survey questionnaire.
6) Develop rules. Produce if/then scenarios that allow the integration of expert
opinion and the quantitative relationships for uncertain inputs, as
mentioned above in 4.
7) Assess geographic and time dependent influences to establish trends that
can be incorporated into future estimates.
8) Provide a structured technique for adding and evaluating new factors
influencing emissions, as other creative means of relating data evolve.
Neural nets and genetic algorithms will be emphasized for the tasks of
selecting, weighting and ordering data. Rules and inferential reasoning are
planned to augment and improve on the algorithmic relationships and
correlations. These are usually linear or polynomial relationships with coefficients
and weights determined by statistics or weighting in a neural net scheme.
3-10
-------
For example, statistics and the neural net analysis will produce a relation such
as
EE = Ct x (# licensed drivers) + C2 x ($$ sales for refinishing)2,
where EE is emission estimate in tons of VOC/year.
This emission estimate could then be improved where additional expert opinion
is available or where interpretation of qualitative information or complex
relationships have been ascertained though other computational intelligence
methods. For example, an improved emission estimate, IEM, might become
IEM=EE x(Control factor)x(Regulatory faetor)x(Paint type factor)
Fuzzy logic would be used, for example, to determine that a high average ratio
of shop revenues to shop employees in a county should mean a high transfer
efficiency and a resulting control factor of 0.87. Similarly, fuzzy rules derived
from expert opinion could say that the moderately low incidence of environmental
litigation in a state should result in a regulatory factor of 1.05. A very high
annual sales of new automobiles over the past three years could imply more
refinishing with reduced VOC paints and a paint type factor of 0.93. These are
only unsubstantiated examples without validation results but do illustrate how
computational intelligence can provide a "value-added" function.
For more inferential knowledge, interaction with the paint manufacturing and
auto refinishing industries has allowed the formulation of questions which can be
put to experts for further developing a knowledge base. The next steps are
therefore to
1) select the 2 or 3 experts from paint manufacturing and auto refinishing and
complete preparations for their review of VOCEES requirements and of the
current state of the industry, to be followed by interview sessions, both
singly and collectively, and
2) complete the list of questions (using on regression analyses results, industry
comments, and EPA review).
Resident Database Selection
Emission levels can be calculated from material use data collected from a
sample of the bodyshops nationwide through a survey or through analysis of state-
by-state material distribution. The emissions data are then used, along with other
readily available state-level, explanatory variables, to teach a neural network the
relationship between the explanatory variables and the emission levels. In the
3-11
-------
training process the network learns which variables are most important to
determining emission levels.
The preprocessed survey results and explanatory variable data can be presented
to a neural network as illustrated in Figure 3-3. Initially, half of the available
data will be used to train the neural network while the other half will be used to
evaluate that which the network has learned. Genetic algorithms will be used
during this phase of development to identify the optimum neural network
architecture, learning algorithm, and other critical parameters. Examples of using
GA would be in the training of the neural network or development of membership
functions. In both cases, a test set would be necessary so GA could properly
determine the fitness of each string. This fitness function is the most difficult part
of GA to construct and is essential to its performance. To properly optimize a
system, GA must know what constitutes good performance. In the emissions case,
this would be a knowledge of what emissions are like for a given set of variables.
101
z
o'
EXPLANATORY
VARIABLES
Resident
Population
Licensed
Drivers
Registered Motor
Vehicles
Civilian Labor
Force, Total
Motor Vehicle
Deaths
in'- -
|r
o,
z
1
Use of Control
Equipment
Network learns inter-variable relationships
without mathematical models
UNKNOWN
VOC |
Emissions^
Survey data will be used to train the network ฆ
Figure 3-3. Neural Networks in VOC Emissions Estimation
Finally, the entire database will be used to train the network, to continue to
identify and minimize the input variable set which best estimates emission levels.
Analysis of the trained network will identify the input variables which are most
important in estimating emission levels, reducing the number of input variables
required. A sensitivity analysis will also be performed to determine the error
introduced by this minimized artificial neural network.
3-12
-------
Emissions Estimation Computation Module
As stated earlier, emissions computation will first be tried with coefficients and
weights tailored to the resident database and validation data. Polynomials will be
used as prescribed by the neural net and statistical correlations.
The inferential system module under development will use fuzzy logic. The
neural network can also be structured into the computational module of VOCEES
to perform its role in relating dependent and independent variables. Genetic
Algorithms will be used to optimize membership functions for fuzzy logic analysis.
To select the fuzzy variables of interest, the statistical correlations are applied
along with the analysis of survey data, and the set of questions for experts. Data
and expert input will establish the design of the fuzzy sets and rules. A fuzzy
system with less than 5 variables and 20 rules is anticipated to be necessary for
any area source. A larger model would require fine-tuning beyond the accuracy of
available information and would place an excess burden on system users and
upgraders.
In a fuzzy logic-based expert system, fuzzy rules reside in the "knowledge base"
and the membership functions and defuzzification (fuzzy-set definition, encoding,
decoding) are the techniques comprising an "inference engine." This knowledge
base will be built and the engine coded in C language.
An adaptive fuzzy system can use sample data and neural or statistical
algorithms to choose the coefficients for weighting each output membership
function and then to define the fuzzy system over any period of time. This process
of weighting and redefinition (new rules) can be a continuous ones with upgrades
made in VOCEES on an annual basis. If it is considered more efficient, the hybrid
system will be abandoned and all information incorporated in the form of fuzzy
logic.
Fuzzy logic system development will consider quantitative versus qualitative
inputs. Expert input may be either 1) an uncertain estimate (e.g., about 2 gallons
of paint used per employee per day) or 2) a qualitative assessment expressed
linguistically (e.g., small shops are much less likely to use closed gun cleaning
stations than large shops) where quantitative values are assigned through fuzzy
sets, and the outputs defuzzified and summed. The latter, qualitative assessment
must be valued on the basis of related, quantitative data (e.g., small means less
than 5 employees), much less likely means only 10% on a relative basis. These
values must be transferrable from one expert opinion to another, with a
reasonable expectation of consistency. Such consistency is established through the
statistical and neural data analysis.
3-13
-------
As stated, instead of using fuzzy logic to make the actual estimate of auto
refinishing VOCs, fuzzy logic is probably best suited to use auxiliary information
to refine estimates determined statistically or to fill in gaps of information not
available for a specific region. Fuzzy logic would be able to adjust estimates based
on local information that may not play a role in determining the initial estimate.
To properly create rule bases and membership functions, this auxiliary
information must come from experts or survey data must be fuzzified.
As an example of how fuzzy logic may be used, consider the types of coatings
used in auto refinishing. As mentioned in Section 2.1, coatings can be roughly
divided into three categories: lacquers, enamels, and urethanes. Lacquers are
high in VOC content, enamels are more moderate, and urethanes are low. If an
estimate of paint use by type (% of all coatings used that are lacquers, enamels,
and urethanes) can be made for a given region (county, nonattainment area, etc.),
this information can be used to adjust a VOC estimate that was based statistically
on a variable such as body shop employment. Not only does this refine the
estimate using local information, it also is able to include information about local
regulations since the type of paints to be used may be dictated.
Figure 3-4 shows an example set of membership functions for paint use. Since
lacquers are quite high in VOC content, the lacquer usage membership function is
biased towards lower percentages belonging to linguistically larger fuzzy sets.
Enamel usage fuzzy sets are more evenly distributed, while for urethanes there is
a bias towards belonging to the linguistically smaller fuzzy sets. The reason for
this is given in the fuzzy rules shown in Table 3-2. The rules show that if lacquer
usage is large or very large, it will tend to push revising the VOC estimate
upwards, while if urethane usage is high, the VOC estimate will be reduced.
The rule base shows why the aforementioned fuzzy set distributions have been
defined as they are for the membership functions. Since lacquers are much higher
than the other types in VOC content, enlarging the fuzzy sets for the "large" and
"very large" variables makes it easier for high usage of lacquer to result in a
higher percentage increase in VOC estimate. The reverse is true for urethane
usage, where smaller fuzzy sets for "large" and "very large" make it more difficult
to obtain a higher percentage reduction in VOC estimate.
Of course, the exact size and shape of these membership functions and the
structure of the rules can only be determined by an expert or with the type of data
that would come from the nationwide survey.
3-14
-------
Lacquer
VS. S . M
0 10 20 30 40 50 60 70 80 90100
% In County
VS = very small; S = small; M = medium; L = large;
VL = very large
Enamel
vs
VL
0 10 20 30 40 50 60 70 80 90100
% In County
= very small; S = small; M = medium; L = large;
= very large
Urethane
vs
VL
0 10 20 30 40 50 60 70 80 90100
% In County
= very small; S = small; M = medium: L = large;
= very large
NL .NS
-30 -20 -10 0 10 20 30
% Change in VOC Emissions
NL = negative large; NS = negative small; ZK = zero;
PS = positive small: PL = positive large
Figure 3-4. Membership functions for paint use and percent change in
VOC estimate
Table 3-2. Example rules used to modify VOC estimate using coatings
usage
Lacquer
Use in County
Enamel
Use in County
Urethane
Use in County
Change in
Emissions
Estimate
Very Large
Small
Very Small
Positive Large
Large
Medium
Small
Positive Large
Small
Medium
Small
Zero
:
Small
Medium
Large
Negative Small
Very Small
Very Small
Very Large
Negative Large
Another possible application of fuzzy logic would be in the fuzzification of
geographic locations. This would have a two-fold purpose, 1) to use general rules
that apply to specific regions and 2) to assist in estimating data for regions in
which the data was unavailable. Geographic locations, for example, can be
3-15
-------
partitioned into regions based upon their latitude and longitude. A state such as
Virginia would no longer be confined solely to such regions as Mid-Atlantic or
Southeast but would have a membership in both. Figure 3-5 illustrates the 10
U.S. EPA Regions. Each region can have overlap into adjacent regions (the gray
shading). If information is readily available for a given region or regions, it can be
used in the estimation of surrounding regions. Figure 3-6 shows how this may
look on a statewide basis. Using fuzzy sets of this type would enable the use of
such rules as "If the body shop is in the Northeast and the regional composite
consumer index is high, then VOC emissions are larger since more cars are
painted following accidents." Since the Northeast region is fuzzified to include
parts of other regions, this rule also applies to bordering regions (although less
than 100%).
Figure 3-5. The 10 EPA Regions
Also, since it is often difficult to obtain complete data for geographic regions,
fuzzy partitions provide a means to approximate what data is missing.
Fuzzification of boundaries between counties or regions in a state may be the most
useful application of this technique. This would provide the ability to "fill in the
blanks" when using County Business Patterns, where disclosure laws often
prevent the publication of data. Figure 3-7 shows an example using North
Carolina counties and nonattainment areas (in dark gray). Disclosure laws
resulted in information being incomplete or non-existent for the counties colored in
light gray. However, using information that is available for nonattainment areas
3-16
-------
Figure 3-6. Statewide Representation of 10 EPA Regions
and surrounding counties, a fuzzy estimator can be used to generate numbers for
the other counties. In deference to the importance of the nonattainment area, its
information could have a higher weight {i.e., membership) than the other counties.
This weight could be based on population, geographic distances, a combination of
the two or other means. This would be an ideal method of modelling the influence
of large metropolitan areas on surrounding areas. Using the above example,
assume the distribution of lacquers, enamels, and urethanes was known for a
nonattainment area and three surrounding counties. Table 3-3 shows the
calculation of the distribution for a fourth county using the center-of-moment
method, the method most commonly used in the defuzzification of an output from
a fuzzy system.
3-17
-------
Figure 3-7. Disclosure laws resulted in information being incomplete or
non-existent for the counties colored in light gray.
Table 3-3. Estimation of Unknown Data for County Using Fuzzy Weights
and Center-of-Moment Method
Area
Fuzzy Wt.
Lacquer %
Enamel %
Urethane %
Nonattainment
Area
1.0
10
20
70
County 1
0.6
20
30
50
County 2
0.4
40
30
30
County 3
0.3
40
40
20
County 4 Usage
Lacquer = {10(1.0)+20(0.6)+40(0.4)+40(0.3)}/{1.0+0.6+0.4+0.3} = 21.74 %
Enamel = {20(1.0)+30(0.6)+30(0.4)+40(0.3)}/{1.0+0.6+0.4+0.3} = 26.96 %
Urethane = (70(1.0)+50(0.6)+30(0.4)+20(0.3)}/{1.0+0.6+0.4+0.3} = 51.3 %
System of National and State Emissions Experts, and Local Data Experts
No expert or expert approach exists upon which to base an expert system, i.e.,
no state agency, EPA office, or EPA support group is consistently providing
reliable, accessible, continuously available emission estimates. (California has
gone beyond others in regulating, but there is no evidence of an exemplary
capability or system for emissions estimation, nor even a better data base on VOC
usage.) No obvious surrogate system is known which performs the same function
for other types of values. Therefore, an expert system would have to be built
through new expertise and knowledge and then captured in software.
3-18
-------
The study began with the concept of developing a knowledge-based/rule-based
expert system as the main emission estimation technique. It has been found that
this concept is a useful element in the method, but is too restrictive as a single
format. However, it is suggested that an "expert system" concept could be
embodied in the method in the form of EPA area source exports.
If area sources are to be individually regulated, it is obvious that each source's
individual characteristics warrant expert attention and analysis. EPA has experts
for such point sources as coal-fired power plants, for example. This is especially
true where regulatory intentions and control guidelines are ultimately aimed at
pollution prevention. Pollution prevention implies intimate knowledge of process
operations, materials and internal controls which can eliminate or reduce
emissions without added control technology. A system of experts could incorporate
this knowledge in EPA planning and regulation and continuously update the
knowledge to minimize emissions and costs (in terms of size, region, and economic
demographics).
The current VOCEES design, can incorporate an expert system. The process of
area source emissions validation can confirm an ever-improving technique,
resulting in a highly accurate system of rules.
The expert-system-supported estimation method might be developed in the
context of the overall emission inventory method, if possible. The organizational
structure for using VOCEES could take several forms. Figure 3-8 shows a
possible organization under which VOCEES could be applied.
It is suggested that at least part of the Area Source Emissions Estimation
expertise would reside with an "Area Source Officer" within EPA. The
information supplied from national data sources (and updated computational
techniques) could be transferred by electronic or computer storage media (CD
ROM, magnetic disk, electronic mail) and directly incorporated into VOCEES for
estimation at state/county levels.
3-19
-------
Nation?! I
Emissions I ^
Analytical
Results
Dual
Product
Environmental
Protection Agency
Area Source
Emissions Experts
VOCEES
Guidelines
n
State or Local
Air Pollution
Control Authority
VOCEES
1
Local / National Data / Information
Figure 3-8. Method Flowchart with Evolving Expert Systems
The State of local authorities own the emissions estimation data they supply to
EPA. However, EPA must approve the method and resulting implementation
plan. EPA also supplies activity and emission factors which may be used, or
better methods may be substituted by the State or local authority. An EPA expert
system could continually improve the default factors and method. Part of this
could be an incorporation of the best techniques by State or local experts. An
interactive system should provide an optimal knowledge base.
3-20
-------
3.3 Use of VOC-Containing Product Manufacturers Data
Materials data accumulation has emphasized retrieving auto refinishing
product information from the major paint manufacturers in the U.S. These are
AKZO Coatings, American Standox, Inc., BASF Corp., DuPont Automotive
Products, NASON (DuPont), PPG Industries, and Sherwin-Williams Company.
ICI and Glidden are also moving into the market. The National Paint and
Coatings Association (NPCA) has become involved in the negotiations, but NPCA
does not represent all manufacturers. The pertinent industry group is Special
Purpose Coatings, SIC 28513-13, which includes automotive, other transportation,
and machinery refinish paints and enamels, including primers.
The Census Bureau publishes information on product sales by Special Purpose
Coatings, SIC 28513, under Coatings Manufacturers in the Census of
Manufacturers. Information is reported for every 5 years with an annual survey
but only at national levels. The main data are the value and quantity in gallons
of shipments. County Business Patterns can provide data on distribution to the
first level (Wholesale Warehouses and Jobbers for Coating -- SIC 5198) to state
levels but only for "Sales" of ah paint products.
The results of the annual Survey of Manufacturers for SIC 2851, Paint and
Allied Products, mentioned in the preceding paragraph, are published in the
Census Bureau's Current Industrial Reports (CIR). National, annual consumption
figures for SIC 2851 and its constituent codes (e.g., 28513-13) are also published
every 4 years in Stanford Research Institute's (SRI) paint market report
(Connolly, 1990). The CIR data are derived from estimates based on a survey of
selected producers. SRI's paint and coatings data are based on communication
with industry, published information, and SRI estimates (Chemical, 1992). SRI
claims that differences between the two sets of estimates are "due to variances in
definitions and estimating techniques" (Connolly, 1990). It is readily apparent
that the CIR estimates are much more volatile than those from SRI, changing in
subsequent years by as much as 14% for SIC 28513 and 37% for SIC 28513-13.
The relative accuracy of each source is indeterminate and other mechanisms have
been pursued through which actual quantity distribution can be determined.
The most important information is the annual distribution of the entire
spectrum of automobile refinishing coatings throughout the U.S. However, other
characteristics of the area sources' VOC-containing materials are important
besides their distribution in time and space. For example, the chemical
constituents of the paint and solvent products are important to determine their
VOC content, relative volatility, potential toxicity, and tendency to contribute to
ozone formation. The type of product is important in order to know how it is used
to auto body shops (e.g., as primer, topcoat, cleaner or surface preparation). The
area of application and associated equipment determines evaporation or recycle.
3-21
-------
Types of products and marketing or regulatory trends perceived by the
manufacturing industry are also important.
The annual sales and geographic distribution of auto refinishing products
(paints, solvents, sealers) can provide a very good idea of the amount of VOC
emissions from associated body shops. The VOC content of almost all products is
well-documented and can be obtained through the MSDS forms that are required
for every product. As noted earlier, environmental control regulations are
increasingly emphasizing limitations on the VOC content of these products, and
therefore documentation of and attention to VOC content is increasing.
Acquiring distribution of products information in space and time is not simple.
Preliminary negotiations have been held to obtain a disclosure of such information
with the paint manufacturers to finalize a one-time reporting standard. This
reporting must not compromise companies' marketing information nor impose a
significant burden on their personnel or systems. The best information accessible
is sought (e.g., resolution to county-size areas and records over 5 years). This can
be a difficult goal for companies. The matrix of auto refinishing products is
typically in the hundreds for large companies. When compounded by over 3000
U.S. counties and coverage of 5 years (during which the product line changes), the
amount of data can reach millions of records. The product data formatting is not
standardized across the industry and is typically not well-categorized by
geographic areas. However, in response to short questionnaires provided under
this study (see Tables 3-4 and 3-5) the majority of major manufacturers
responding said that obtaining data from the first level of distribution would be
easy and the second level of distribution provides regular information, also.
However the convenience of the filing and data retrieval systems for the
distributors was not addressed.
One compromise is to obtain at least one year of information on products under
groupings to be agreed upon by the manufacturers. The minimum level of
geographic resolution would be to the first level of distribution, i.e., major
distribution warehouses or centers. However, because different companies have
varying distribution networks, this approach results in uneven coverage for the
different products. For example, PPG has about 26 major warehouses serving
3100 to 3200 distributors (jobbers) while DuPont has some 200 major distribution
centers (wholesaling warehouses), and AKZO uses only 4 major distribution
centers which ship to "wholesalers." This means that an averaging distribution
approach must be devised based on areas of coverage around warehouses, and
dividing these areas into uniform grids. This can almost certainly be done best by
utilizing GIS capabilities.
It is also known that complete records of sales and products are kept down
through the jobber level. Almost every paint retailer is on a computer network.
3-22
-------
Table 3-4. Manufacturer Product Category Questionnaire
POSSIBLE PRODUCT CATEGORIES FOR YOUR GEOGRAPHIC DISTRIBUTION DATA
RTI recognizes that geographic distribution data covering your auto-refinishing products is voluminous. To reduce the information burden, we
are trying to use product groups -like the categories below. We realize that some of these overlap. Which categories would you consider "good"
to organize your product data and estimate the distribution of auto-refinishing VOC's in the U.S.? We need to cover all auto refinishing products
and estimate average VOC content in each product group. Add other categories you prefer.
CO
i
DO
CO
GOOD CATEGORY FOR CLASSIFYING YOUR PRODUCT DATA?
Surface
Prep
Primer
Surfacer
Filler
Precoat
Sealer
Topcoat
Clear-
coat
Cleaning
Solvent
Other
Solvent
Based
Product
Water
Based
Product
Ure thane
Lacquer
Enamel
Yes
No
Other Categories
We are also considering another approach using only a few simple categories based on shop operations and major differences in VOC content by
group. These categories are:
1) Surface Prep products, 2) Sprayed products, 3) Cleaning products, 4) Water-Based products, and 5) Other
Would these work just as well? Yes ~ No ~
Comments:
-------
Table 3-5. Manufacturer Geographic Distribution Questionnaire
We need an idea about how much geographic distribution information on auto refinishing products is accessible in your central datafiles and how much trouble it
is to obtain. Please circle or fill in below the appropriate answers for your company.
CO
CO
~fc.
Level 1:
Distribution
Centers/
Warehouses
HQ:
Headquarters
Level 2:
Jobbers
Estimated
Number of
Outlets at
Level 1&2
Ease of
Access2
Typical
Rate of
Access3
Past Years of
Accessible
Back Data
Info in
Computer
Datafiles?
For which
regions do
you have
info?4
Number
EZ12 EZ1
DF1 DF12
0
M A
0
1 2 5
years
Y N
NE NC
SA SC
M P
Number
EZ12 EZ1
DF1 DF12
0
M A
0
1 2 5
years
Y N
NE NC
SA SC
M P
Level 3:
Individual
Customers
Comments
o
EZ12 = Your HQ receives info regularly from levels 1 and 2; EZ1 = Your HQ receives info regularly from level 1; DF1 = You will
have to make a special request to level 1; DF12 = Special request to levels 1 and 2; O = Other (please specify)
Q
M = Monthly; A = Annually; O = Other (please specify)
4 NE = Northeast; NC = North Central; SA = South Atlantic; SC = South Central; M = Mountain; P = Pacific
-------
Most retailers or jobber level wholesalers deal mainly with one or two paint
manufacturers and use similar computer records systems. The main problem is
not that records are not kept or are completely unavailable. The problem is one of
dealing with non-uniformity across manufacturers.
The other major consideration in this area of data retrieval is that of
confidentiality, and avoiding even an appearance of providing one manufacturer
advantage over another in terms of disclosure of information. This means that a
completely generic form must be devised for reporting data and results obtained
from manufacturer's data. It also means there will be no disclosure to EPA or
anyone else of any manufacturer or product names associated with data. This
confidentiality is not at all a disadvantage in completing the method and
emissions estimation system. More important is the generic disclosure format, so
that results cannot be "reverse engineered" by any company to obtain information
providing them an unwarranted competitive edge. A format that is proposed is
A computerized format for each paint manufacturer reporting total annual
shipments to distribution centers and wholesale outlets. Products would be
characterized in 6 to 12 major paint and solvent categories. Companies may
alternatively perform internal calculations and provide VOC volumes
shipped rather than product volumes. It is recommended that software for
data entry be developed in coordination with the manufacturers. The data
list would be entirely confidential and handled by a third party other than
EPA or any manufacturer. Shipments of new types of paint products {e.g.,
introduced with in the last 18 months) could be excluded from disclosure if,
for example, they make up less than 10% of total sales. This doubly
protects the competitive position and proprietary information associated
with new products, and also allows for a period in which their actual
acceptance and performance can be assessed in the market.
Data received will be tagged by zip code for the local distribution centers,
with all company or center names, addresses, and other specific references
removed from the data file. Data will be composited to multi-county regions
and averaged over the counties by population, with the minimal size for any
region being dictated by the territory served by the counties largest
distributing centers for which data is reported. Regions will be within a
state or within contiguous states served by the same distribution center.
The data will be entered into a validation database used in confidence by
EPA or its agent for method validation only, and no emissions will be
estimated solely on the basis of this data. Manufacturer data will not be
provided to states. Other explanatory variables will be used in the
VOCEES which best predict emission levels by geographic region.
Published results by EPA will not refer to manufacturers' data. EPA will
3-25
-------
neither receive nor publish any information on specific companies or
products.
Under this system, product distribution by specific companies could not be
related to area source emission estimates any more accurately than to many other
variables available to companies. This is especially true since the results from
emission estimation are normally reports for several area and point source
categories combined. Therefore, the only remaining issue should be that of the
inconvenience to companies compiling the information. It can be shown by survey
experts that this reporting would constitute a smaller burden than that of a
nationwide survey, for example. An accurate assessment of the real burden on
manufacturers to report such information is needed (it is speculated that normal
production inventories and marketing surveys would allow this data to be readily
available in computer format).
It is estimated that, using annual reporting by 5 to 7 manufacturers, VOC
distribution by county could be consistently validated to -10%, +20% accuracy by
this method. Some county validation could be within ฑ 5% using data for smaller,
localized distribution centers. This validation would serve to verify the predicted
level of accuracy for the larger regions where data is averaged and then applied to
counties.
3-26
-------
3.4 Sampling and Analysis
This study has examined the possibility of representative sampling of a few
auto body establishments to confirm emission concentrations and estimates of
pollutant loss rates from individual establishments. The literature shows that
typically measured constituents for auto refinishing (outdoors) are methyl-ethyl-
ketone (MEK), toluene, ethylacetate, xylene, and methylisobutylketone. The most
important constituents are MEK and toluene/xylene (TX). Total hydrocarbons is
also a good measure of combined TX and acetate plus ketones/acetones. Samples
can be taken using 100 mg charcoal tubes or diffusive samplers. Very good
correlations have been obtained between the analyses of samples taken using both
these methods at the same sites. Gas chromatographic analysis is sufficient for
accurate determination of concentrations.
A few previous ambient air sampling studies have been done, but they vary
considerably in their measurements of auto painting and refinishing emissions.
Table 3-6 lists some summary results. For measurements at a single shop,
outdoor adjacent concentrations of toluene were about 2-20 mg/m3, but less than 1
mg/m3 two blocks away. Measurements at nearby houses were about 1/3 to 1/2
levels at the body shop. At two blocks away, the levels were down to 10% of the
samples taken near the shop.
Table 3-6. Auto Refinishing1 Air Concentrations of Organics
COMPOUND
Auto Refinishing Workplace Ambient Concentrations (mg/m3)
(Verhoeff, 1987)
(Jayjock, 1984)
(de Medinilla,
1988)
(Veulemans,
1987)
Toluene/xylene
2 - 20
7 - 105
(40ฑ50%)
30-80
Isocyanates
Diisocyanates
.004 - 4
0.15
Acetates/ketones
/ethers
2 - 10
2 - 40
There are other published indicators of auto refinishing emissions. For
example, GM car manufacturing emissions (Sexton, 1983) totalled about 230 gal/hr
of hydrocarbons, BTX and aromatics, and about 220 gal/hr of oxygenated
hydrocarbons such as methanol, acetone, acetate and esters. These numbers can
be compared to production figures for coating losses per car.
OEM painting conditions are, of course, very different from body shops. It has
been estimated that, for refinishing surface coating, about 60 to 100 grams of
paint are required per square meter of area coated (about 1 to 2 pounds per 100
3-27
-------
square feet) (Hughes, 1975). Polyurethane coating requires 80 g/m2 and xylene
thinner is 119 g/m2. The lowest OSHA Threshold Limit Value (TLV) for glycol
ethers is 16 mg/m3. The TLV for toluene is 375 mg/m3.
It is recommended to first obtain the cooperation of at least 3 shops in a single
metropolitan statistical area (MSA). Sampling should be conducted at the
boundaries of the shop premises (in the prevailing wind direction) or at a similarly
nearby location. At least two 8-hour samples would be taken per day over a
period of at least 3 days. Cartridges would require a flow rate to allow
measurements down to 1 mg/m3. A flow rate of 25 cc/min of ambient air should
allow capture of at least 1 mg levels of species of interest.
TX, MEK and total hydrocarbons are the species to be analyzed. Total
hydrocarbon measurements may be adequate depending on background levels.
Sampling will begin and end with employee work shifts where possible. Rooftop,
main vent, or doorway exit samples will also be measured over two 8-hour periods
if indications are that boundary concentrations are significantly different from
"shop vent" levels. Standard gas chromatographic columns and analysis
techniques are available for these species. It is anticipated that this work can be
completed with less than 250 man-hours of effort. This will produce the first
reported results correlating a solvent usage inventory and other shop operating
variables with emissions measurements for auto refinishing.
Sampling data can be correlated with all relevant shop variables, from size to
operating conditions to products types. Figure 3-9 shows typical modules involved
in the solvent mass balance for auto refinishing shops. These areas, and their
operating variables, are all being considered and incorporated (where appropriate)
into the emission estimation method. A brief description is provided in Appendix
B of an approximate model derived under this study for relating near-shop
ambient samples to total shop VOC emissions.
3-28
-------
PRODUCT ACCRUAL
Jobs vs. Inventory
Orders
Shipment & Delivery
Spray Booths
Vents
Controls
Recycle/Disposal
DISPERSION
Product Storage
Cleanup
Mixing
Spray
Recycle Operations
Drying and Surface
Devolatilization
Waste Storage
Prep
EVAPORATION
Figure 3-9. Mass Balance of Emissions Flowchart
3-29
-------
3.5 Past Emission Estimation Data and GIS Interpretations
Emissions inventory data for auto refinishing are available for 1982 and 1987
from the EPA Office of Air Quality Planning and Standards (OAQPS). While
inherent inaccuracies in the data do not allow strict validation, they can be useful
for comparison with preliminary estimates using prescreened explanatory
variables. Major discrepancies may highlight bad assumptions. The data
medium, format and accessibility have not yet been firmly established. Most of
the emission estimates are based on population figures. The data review thus far
has included the States' approaches to assessing number of employee hours per
year, work days per week and per year, effect of point source emissions on area
source estimates, and obvious variations in "per capita" and "per employee" factors
used.
Future data analysis will include graphing the trend of emissions estimates by
county and nonattainment area for the past 1982, 1987, and some 1990, SIP
responses for all 3300 counties. If efforts proceed on the first two to four items
covered under this section, then there will be less need for revisiting past SIP-
related estimates. Perhaps only 1990 data should be given a thorough review.
3-30
-------
4. INTERIM RESULTS AND CONCLUSIONS
The results of the case study cannot be complete until validation data becomes
available and emissions estimates can be better substantiated. Once the data
from the validation survey comes in, the analytical tools will be tested and
optimized and the system configuration finalized, as described in the last section.
However, a number of steps toward an improved emission estimation method have
been taken.
Results of Current Study
A comprehensive information search and development of a database (see
Sections 2.3 and 3.3);
Assessment of current emission estimation methods related to the area source
and their limitations (see Section 2.2);
A PC Windows-, GIS-graphics-based system with computational techniques
which provide reasonable examples of system function and output (see Section
2.5);
Demonstration of the VOCEES system (see following examples) using existing
and projected regulations, past and estimated emissions levels, geographic and
time (annual) comparisons, extrapolation of emissions estimates to ppm
concentrations within nonattainment areas, and other analytical results that
might be prescribed by EPA;
An examination of computational intelligence and recommendations for
incorporation into the overall estimation method (see Sections 2.4 and 3.2);
Preliminary screening of explanatory variables using statistical regressions (see
Section 2.3.2);
Preparation of a survey (questionnaire, sampling population and data storage
and analysis system), and completion of a pre-survey (see Section 3.1);
Development of preliminary source sampling recommendations (see Sections 3.4
and Appendix B);
A few final examples illustrate the present capability of the new method to
produce results. The VOCEES part of the new method handles a resident
database, makes emissions estimation calculations, and displays results for further
analysis. One advantage of VOCEES is its ability to present information in
4-1
-------
a new, easy, and quickly understandable way. Figure 4-1 illustrates information
about automotive body shops in San Francisco obtained from American Business
Information in the form of an ASCII text file. This text file was imported into
Maplnfo, and each business was assigned a geographic location using a process
called geocoding. Geocoding of this file was done using the street addresses of the
body shops. Using a file containing all the street locations and block numbers for
San Francisco, CA, Maplnfo sequentially processed all the body shop addresses
and assigned latitude and longitudes to them using the street and block file. This
process is completely automated; Maplnfo stops only to query the user if a block
number or street name cannot be matched.
The ability to display these locations is not the main feature of Maplnfo,
however. Even more useful, is its ability to show information related to these
locations. As shown in the legend, the size of San Francisco's body shops ranges
from one-person operations to the S & C Ford Truck Headquarters with 230
employees. To show the size of each body shop, its location is marked with a star
that corresponds to its size. These graduated symbols show that, while there may
be a few large shops, the majority are small operations usually with five or less
employees.
VOCEES can also illustrate temporal changes in emissions estimates.
Temporal changes in emissions are affected by regulatory changes. For example,
in California local air quality management districts (AQMDs) have implemented
different emissions regulations over the past decade and will soon introduce new
ones. Figure 4-2 illustrates some of the changes in emissions in California's 58
counties. While estimates for each year were based on population, the estimates
do not grow as population grows in some of California's most populous counties.
In fact, there is a reduction in emissions in all of California's counties between
1988 and 1995. In 1988, 15 counties were above the average county emissions of
249 tons/yr; while in 1995, only 10 counties are projected to be above this average.
The VOCEES maps quickly and easily enable the end user to track the progress of
each AQMD and examine the impact of regulations.
The ability to examine different explanatory variables is another capability
built into VOCEES. Figure 4-3 illustrates the use of three different explanatory
variables to generate emissions estimates: population, American Business
Information (ABI) SIC 7532 Employment, and County Business Patterns (CBP)
SIC 7532 Employment. Different variables can produce profoundly different
estimates. Burlington County, for example, has an estimate of 200 tons/yr using
population as an explanatory variable and 285 tons/yr using ABI employment, a
43% increase. Cape May County, on the other hand, has an estimate of 49 tons/yr
using population as an explanatory variable and 19 tons/yr using ABI
employment, a 61% reduction. VOCEES enables the user to instantly and
4-2
-------
is
rate
Location of some of San Francisco's 165 auto
body shops using Maplnfo and American
Business Information data
it
Figure 4-1. Automotive Body
Shops in San Francisco, as
Displayed in Maplnfo.
Sfitd yitw ฃ<*!ป oil
ilry Etta) S|>' ฃnซ'yป>t
1995
CASXrOS.CAปปW Mlป T
-| UIปCOtWMM>PMlป
. VOCEES Expert Analysis Example
: County emission levels for California
considering regulatory impact in 1988.
.1993. and 1995.
OOB so
ซป1ป 00
JMป n M*1*
I MJWCOJ.KJ 1IPC P U m }*1ป
1 NJMiCOSJUIMBiUM 1*1*
IUC8PC0B.NJ1K8P Mm
*
-
By Population
By ABI Employment
By CBP Employment
Ml-
/cert H1ซ Ml.
Zm*
i,.uu.L:lJ
VOCEES Expert Analysis Example
County emissions for New Jersey In 1993
using 3 different primary variables
:
-1 Lป|t*
MJ in3 Enปn >>
ฆ <>ซป
D
Q acปni
H
H ปซ:>
ฆ XMaXO
ฆ
WW !! I !
~
i i
Figure 4-3. Differences in
Three Explanatory Variables
Used in Generating Emissions
Estimates
-------
graphically visualize the distribution throughout a state and determine which
variable is the best choice.
Finally, one of the major accomplishments of VOCEES is the ability to produce
documents which could be incorporated into state implementation plans (SIP).
Figure 4-4 illustrates the use of VOCEES to generate all the information
necessary for the Raleigh-Durham-Chapel Hill, North Carolina nonattainment
area. VOCEES can generate maps, graphs, and tables. This presents the user
with a great deal of flexibility for presenting data or adhering to a predetermined
standard without reformatting. In addition, this data must be easy to arrange,
print, and incorporate into other documents. Since VOCEES is a Windows-based
application, much of the input/output difficulty has been removed. Tables can be
saved as ASCII or dBASE, and graphics are easily created and printed through
the use of layout pages. As illustrated in the figure, the layout window allows the
page to be arranged as necessary and permits the inclusion of maps, graphs, and
tables.
nalA'ฃM)of>JL'v^}upซi M 0. NC NortftSfcrifMni Arซa
Figure 4-4. SIP Information Generated by VOCEES for the Raleigh-
Durham-Chapel Hill, NC Area
Conclusions and Recommendations
The automobile refinishing industry is representative of other major area
sources associated with VOC use: average number of employees is small; direct
customer interaction is important; interest in maintaining environmental
compliance is legitimate if only to avoid penalties in some instances; the
industry is affected more by major economic cycles than demographics; and
materials suppliers (e.g., paint manufacturers) very much determine the mix
and VOC content of materials used by the industry. Therefore, demographics
data for consumers are not always useful for this particular area source,
because advertising and customer preferences are not as closely related to the
4-4
-------
materials used as for many retail operations not associated with VOC
emissions. Except for very localized areas smaller than counties, industry
experts have not defined strong correlations between demographics and types of
cars (and related paints), or the likelihood that accident repairs will include
refinishing.
The auto refinishing industry expects to be using lower-VOC materials and to
expand the use of high transfer efficiency equipment like HVLP paint guns.
Vent emission controls may not become prevalent because of cost, but this is a
policy issue not yet definable. Paint product manufacturers say that regulation
may possibly put some small shops out of business or cause consolidation into
larger shops, but that environmentally non-compliant, backyard shops could
flourish until high-VOC paints and older model cars become a small fraction of
the market. No final conclusions are possible, but more sophisticated paint
mixes, as dictated by the OEM automotive manufacturers, are very difficult to
match with low-cost, unsophisticated application and drying techniques.
Difficulties exist for obtaining product distribution data from automotive paint
manufacturers because of "potential competitive advantage to be gained by one
manufacturer over another by 'reverse engineering' of distribution data to
obtain proprietary market and sales information" (from discussions with
industry representatives August 1993). The nationwide survey is an excellent
alternative approach to validation.
Some discrepancies in current emission estimation methods exist for auto
refinishing (see Section 2.2). These are mainly the result of unsubstantiated
activity factors and of fundamental inaccuracies in activity factor data (e.g.,
numbers of individual shop employees). The implications are that current
emission estimates should be liberally interpreted when planning regulations,
assigning nonattainment areas, and setting control priorities for area sources.
PC Windows and Maplnfo GIS with computer graphics displays are a good
combination for an accurate, easy to use, low-cost emission estimation system.
Data are easily input and displayed, and modifications are simple.
No expert or expert approach exists upon which to base an expert system, i.e.,
no state agency, EPA office, or EPA support group is consistently providing
reliable, accessible, continuously available emission estimates. No obvious
surrogate system is known which performs the same function for other types of
values. Therefore, an expert system must be built through new expertise and
knowledge and then captured in software. One model for this is described at
the end of Section 3.2.
Once validation data are obtained for VOC use associated with a representative
population of an area source, genetic algorithms and neural networks (see
Sections 2.4 and 3.2) should be efficient for completing selection and weighting
4-5
-------
of the best explanatory variables, and for training the system to optimally
integrate new information.
Fuzzy logic is appropriate for manipulating rules to apply inferential estimates
in augmenting the correlation of VOC usage variables. Expert opinion can be
advantageously used in predicting the influences on auto refinishing emissions
of such variables as economic factors, control equipment and emissions
abatement operation practices, actual impact of regulations, influence of new
products on emissions, and changes in OEM products. Experts can provide
insight into the general correlation of emissions with shop size, shop sales per
employee, and location in the United States (see Section 3.2).
State and county databases should be used as activity factor data wherever
possible. The best ones are typically updated annually. These databases are
usually readily accessible to state agencies that prepare state implementation
plans. They do not require the central processing and distribution needed for
national data. These data (e.g., licenses, registrations) are used for fee
collection, taxation, and personal identification related to other financial
transactions, so they are likely to be very reliable. EPA would continuously
refine the techniques and tools for applying these databases and should be
responsible for centralized validation of the method. However, local retrieval of
raw data will make the method most efficient.
Preliminary indications show that licensed drivers and registered vehicles are
better explanatory variables for auto refinishing emissions than the currently
applied variables of human population or the number of employees estimated to
be working in local auto refinishing SIC businesses (see Section 2.3.2).
Population data may correlate as well as most other variables but are not
updated as often as other choices. The number of employees is suspect in a
fluctuating industry and with inadequate data collection and disclosure by
Federal surveys.
Data related to the area source appear to be best for emissions purposes,
including data related to materials volumes and product users' levels of activity
(e.g., registered vehicles).
An estimate of the impact of regulations and standards, and of their level of
enforcement requires more accurate emissions estimation and prediction.
A system for tracking solvent distribution, from point of manufacture to point of
retail sale should be developed (see Section 3.3). Manufacturers and OEMs
(e.g., auto manufacturers for refinishing, clothing manufacturers for dry
cleaning) control the product mix for the most part. A tracking system is the
most direct way to determine VOC distribution, and is probably the easiest way
to collect the information. This is especially true for auto refinishing, where
sales records of about seven large companies (and their distribution centers)
could characterize more than 90% of the VOC distribution. The fact that
4-6
-------
Material Data Safety Sheets are already distributed to all purchasers indicates
that such a system is feasible.
Validation of data and variable relationships using industry responses is
essential to completion of a new estimation method. The two best ways to do
this are through a national survey of the users of the pertinent VOC-containing
materials or a mass balance distribution of the materials obtained from the
manufacturers and/or their distribution centers. The latter is preferred for
more continuously updated and state-wide/nation-wide accuracy while the
former is preferred because of better cooperation and more local accuracy.
4-7
-------
-------
5. REFERENCES
Aerometric Information Retrieval System (AIRS): "Short List" of AM S SCCs and
Emission Factors. U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle Park, NC. July 1992.
AIM User's Manual: AIM-PC Version 1.1. AbTech Corporation, 508 Dale Avenue,
Charlottesville, VA. 1993.
Alberta Research Council, Environmental Software. Vol. 5, No. 4. 1990.
Athey, Carol; Charles Hester; Mark McLaughlin; Roy Neulicht; and Mark Turner.
Reduction of Volatile Organic Compound Emissions from Automobile Reflnishing.
EPA-450/3-88-009 (NTIS PB89-148282), U.S. Environmental Protection Agency,
Office of Air Quality Planning and Standards, Research Triangle Park, NC.
October 1988.
BodyShop Business, 1992 Annual Industry Report. Babcox Publications, Akron,
OH.
Bureau of the Census, U.S. Statistical Abstract of the United States: 1992 (112th
Edition), Washington, DC. 1992.
Bureau of the Census, U.S. County Business Patterns. 1972: U.S. Summary.
CBP-72-1, Table lb.
Chemical Economics Handbook. SRI International, Menlo Park, CA. 1992.
Chemical Products Synposis. Mannsville Chemical Products Corporation,
Cortland, NY. 1990.
"Chemical Profiles" in Chemical Marketing Reporter (Weekly). Sehnell Publishing
Company, New York, NY.
Commerce, U.S. Department of. Bureau of the Census. Form CB-7503. "1987
Census of Service Industries - Motor Vehicle Service Shops, Including Tire
Retreading."
Compilation of Air Pollutant Emission Factors, Volume I: Stationary Point and
Area Sources, AP-42 (GPO 055-000-00251-7), Fourth Edition. U.S. Environmental
Protection Agency, Office of Air Quality Planning and Standards, Research
Triangle Park, NC. September 1985.
-------
Connolly, E.M., E. Linak, A. Jebens, R. Bradley, L. Fujise, and M. Peters. The
U.S. Paint Industry; Technology Trends. Markets. Raw Materials (Formerly NPCA
Data Bank Program). Prepared by SRI International for the National Paint and
Coatings Association, Washington, DC. 1990.
County Business Patterns (Annual). U.S. Department of Commerce, Bureau of
Census, Washington, DC.
Current Industrial Reports - Paint and Allied Products (MA28F). U.S.
Department of Commerce, Bureau of the Census, Washington, DC. Annual.
Current Population Reports: Local Populations Estimates (Annual). U.S.
Department of Commerce, Bureau of the Census, Washington, DC.
de Medinilla, J. and M. Espigares. "Contamination by Organic Solvent in Auto
Paint Shops," Ann. Occup. Hvg.. Vol. 32, pp. 509-513. 1988.
Demmy, J.L., W.M. Tax, and T.E. Warn. Area Source Documentation for the 1985
National Acid Precipitation Assessment Program Inventory. EPA-600/8-88-106
(NTIS PB89-151427), U.S. Environmental Protection Agency, Air and Energy
Engineering Research Laboratoiy, Research Triangle Park, NC. December 1988.
Draft Automobile Refinishing Control Techniques Guideline. U.S. Environmental
Protection Agency, Chemicals and Petroleum Branch, Research Triangle Park, NC.
September 27, 1991.
"Economic Cost of Motor Vehicle Crashes, 1990, The." U.S. Department of
Transportation. DOT HS 807 876. September 1992.
Everette, Valerie S., Radian Corporation. Correspondence to Jeff Chappell, Air
and Energy Engineering Research Laboratory, U.S. Environmental Protection
Agency, Research Triangle Park, NC. February 4, 1993.
Fatal Accident Reporting System 1990: A Review of Information on Fatal Traffic
Crashes in the United States. DOT IiS 807 794. U.S. Department of
Transportation, National Highway Traffic Safety Administration. December 1991.
General Estimates System 1990: A Review of Information on Police-Reported
Traffic Crashes in the. United States. DOT HS 807 781. U.S. Department of
Transportation, National Highway Traffic Safety Administration. November 1991.
Goldberg, D.E. Genetic Algorithms in Search Optimization and Machine Learning.
Addison-Wesley Publishing Company, Reading, PA. 1989.
5-2
-------
Guide to the 1987 Economic Censuses and Related Statistics, EC87-R-2. U.S.
Department of Commerce, Bureau of the Census. January 1990.
Highway Statistics. HPM-40/10-92(3.8M)P. U.S. Department of Transportation,
Federal Highway Administration. October 1992.
Holland, J.H. Adaptation in Natural and Artificial Systems. University of
Michigan Press, Ann Arbor, ML 1973.
Hughes, T.W., D. Horn, C. Sandy, and R. Serth. "Source Assessment:
Prioritization of Air Pollution from Industrial Surface Coating Operations." EPA-
650/2-75-019a (NTIS PB 243423), U.S. Environmental Protection Agency, Air and
Energy Engineering Research Laboratory, Research Triangle Park, NC. February
1975.
"I-CAR Annual Report: Fiscal 1993-1997." Inter-Industry Conference on Auto
Collision Repair, Rolling Meadows, IL. 1992.
Industrial Solvents - Winter 1989. Frost & Sullivan, Inc., New York, NY. 1989.
Jayjock, M.A. and L. Levin. "Health Hazards in a Small Automotive Body Repair
Shop," Ann. Occup. Hvg.. Vol. 28, pp. 19-29. 1984.
Joyner, W.M. Supplement D to Compilation of Air Pollutant Emission Factors,
Volume I: Stationary Point and Area Sources, AP-42, Fourth Edition (GPO 055-
000-00391-2). U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle Park, NC. September 1991. pp 2,
4.2.1-1.
Kimbrough, E.S. Documentation of AIRS AMS National Methodologies. EPA-
600/R-92-001 (NTIS PB92-132869), U.S. Environmental Protection Agency, Air
and Energy Engineering Research Laboratory, Research Triangle Park, NC.
January 1992.
Lamason, W.H. "Technical Discussion of Per Capita Emission Factors for Several
Area Sources of Volatile Organic Compounds." Technical Support Division, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards,
Research Triangle Park, NC. March 15, 1981. Unpublished.
Lamason, W.H., II and T. Lahre. Procedures for the Preparation of Emission
Inventories for Volatile Organic Compounds. Volume I. Second Edition, EPA-
450/2-77-028-2 (NTIS PB81-120164), U.S. Environmental Protection Agency, Office
of Air Quality Planning and Standards, Research Triangle Park, NC. September
1980.
5-3
-------
Meeting of the National Paints and Coatings Association (NPCA) with the EPA
Office of Air Quality and Planning Standards, Research Triangle Institute, and
Radian Corporation, Durham, NC. July 1993.
Mitchell Automobile Repair Guide. Mitchell International, San Diego, CA. 1992.
ModelWare and ModelWare Professional User's Manual. TERANET IA
Incorporated, Nanaimo, British Columbia, Canada. Idaho Falls, ID. 1992.
Myers, J.P. and F. Benesh. Methodologies for Countywide Estimation of Coal, Gas,
and Organic Solvent Consumption, EPA-450/3-75-086 (NTIS PB 259909). U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards,
Research Triangle Park, NC. December 1975.
National Air Pollutant Emission Estimates, 1940-1989. EPA-450/4-91-004 (NTIS
PB91-168559), U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle Park, NC. March 1991.
National Air Pollution Control Techniques Advisory Committee, Meeting Minutes,
Volume 2. U.S. Environmental Protection Agency, Office of Air Quality Planning
and Standards, Research Triangle Park, NC. November 19-21, 1991. p. 1008.
Ostojic, N. End Use of Solvents Containing Volatile Organic Compounds. EPA-
450/3-79-032 (NTIS PB80-124423). U.S. Environmental Protection Agency, Office
of Air Quality Planning and Standards, Research Triangle Park, NC. May 1979.
Petroleum Supply Annual. DOE/EIA-0340, U.S. Department of Energy, Energy
Information Administration, Washington, DC. 1992.
Procedures for the Preparation of Emission Inventories for Carbon Monoxide and
Precursors Of Ozone, Volume I: General Guidance For Stationary Sources, EPA-
450/4-91-016 (NTIS PB92-112168), U.S. Environmental Protection Agency, Office of
Air Quality Planning and Standards, Research Triangle Park, NC. May 1991. pp.
4-24, 4-47.
Rethinking the Ozone Problem in Urban and Regional Air Pollution. National
Academy Press, Washington, DC. 1991.
Sexton, K. and H. Westberg. Ambient Hydrocarbon and Ozone Concentrations
Near a Refinery. EPA-600/3-83-025 (NTIS PB83-195958). U.S. Environmental
Protection Agency, Environmental Sciences Research Laboratory, Cincinnati, OH.
pp. 329-332. June 1983.
5-4
-------
Synthetic Organic Chemicals, United Stales Production and Sales (Annual).
USITC Publication 1745, U.S. International Trade Commission, Washington, DC.
1991.
Telephone conversation between Charles O. Mann, Air and Energy Engineering
Research Laboratory, U.S. Environmental Protection Agency, and Hal Waters,
Center for Digital Systems Engineering, Research Triangle Institute. March 3,
1993.
U.S. Paint Industry Data Base. National Paint and Coatings Association.
Prepared by SRI International, Menlo Park, CA. 1990.
Verhoeff, A.P., M.M.W. Wilders, A.C. Monster, and J.II. Van Wijnen. "Organic
Solvents in the Indoor Air of Two Small Factories and Surrounding Houses."
International Archives of Occupational and Environmental Health, vol. 59, pp.
153-163. 1987.
Veulemans, II., D. Groeseneken, R. Masschelein, and E. Van Vlem. "Survey of
Ethylene Glycol Ether Exposure in Belgian Industries and Workshops." Journal of
the American Industrial Hygiene Association. Vol. 48, No. 8, pp. 671-676. 1987.
VOC/PM Speciation Data System Documentation and User's Guide, Version 1.32a,
EPA-450/2-91-002. U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, Research Triangle Park, NC. November 1990.
Written communications between Bill Lamason and Charles Mann, Technical
Support Division, Office of Air Quality Planning and Standards, U.S.
Environmental Protection Agencv, Research Triangle Park, NC. October 1980 and
March 1981.
Zimmerman, D., W. Tax, M. Smith, J. Demmy, and R. Battye. Anthropogenic
Emissions Data for the 1985 NAPAP Inventory. EPA-600/7-88-022 (NTIS PB89-
151419), U.S. Environmental Protection Agency, Air and Energy Engineering
Research Laboratory, Research Triangle Park, NC. 295 pp. November 1988.
-------
APPENDIX A
SURVEY QUESTIONNAIRE
A-l
-------
RTI ID Number:
Firm Name: _
Mailing Address:
City, County, State, Zip:
Business Telephone Number:
Respondent's Name:
Respondent's Title:
OP TYPE
FOR RTI USB ONLY
A.
(II) I .ABEL BOX)
Auto Refinishing Questionnaire
Identification
Is the information printed on the above label complete and c
01 ... YES Go to Question 2. 02 . . . NO
lb. Please correct or complete information as necessary:
Firm Name:
Mailing Address:
City State
Street Location (If different from above mailing address):
Business Telephone Number: | )_
Respondent's Name:
Respondent's Titie:
Zip
I .e. Please indicate which type of operation is your primary function.
A. New Car Dealership
B. Used Car Dealership
C. Auto Refinishing Shop
D. Other (specify)
2. Do you provide autobody refinishing services at this location?
01 ... YES ซป CONTINUE to 3.1. 02 . . NO
What service do you provide0
Please Specify:
STOP. Thank you for your time. Return
the questionnaire in the envelope provided.
A-2
-------
la. How many total employees do you currently have working at this shop location?
Number of employees:
lb. How many employees use paints, primers, sealers, or other solvents?
Number of Body employees:
2.
How many years have you been in the auto refinishing business at this location?
Number of years:
3a. ilow many hours per day, and days per week is the pil
Total operating hours per day
Total operating days per week
in operation?
3b.
Please indicate the time of day when most, of the painting is done in your shop.
Circle one answer.
01
02
03
04
morning (8 - 12)
afternoon (12 - 6)
evening (6 - 12)
other (specify)
. no variance by time of day
3c.
Please indicate the part of the week when most, of the pa'nting is done in your shop.
Circle one answer.
0 1 Beginning (Men. - Wed.)
0 2 Middle (Tues. Thurs.)
0 3 End (Thurs. - Sat.)
0 4 Weekend (Sat. - Sun.)
0 5 Other (specify)
A-3
-------
Do you regularly service any fleet accounts at this location?
01 YES 02 NO Please go to Question 5.
Please indicate (/) what type of fleet accounts you service and what percentage of
your total business each represents:
Government % business
Car Rental % business
Automobile Dealers % business
Other, Specify % business
Consider the time period from to ; How many jobs were
completed at this location, m each of the following categories? Please report jobs
completed for individual and fleet accounts separately. Enter a zero (0) in any
category where no work was done.
Individuals Fleets
Painting after collision jobs jobs
Painting after vandalism jobs jobs
Repainting under warranty jobs jobs
Cosmetic or discretionary painting jobs jobs
Custom and restoration jobs jobs
Other', please specify jobs jobs
What paint/coating system(s) do you use most of the time9
Manufacturer
Paint/Coating System
Example: DuPont
Example: ChromaBase
A-A
-------
6b. Again, consider the time period from to Please provide the
name or specific idcntifieriydii number) of each product you used with the system
mentioned in 6a. For each product you list., indicate the number of gallons you used
during this time period. If you have more than one system in your paint shop,
indicate the manufacturer/systemmme with each product.
Category Name and/or specific identifier Gallons
EXAMPLE
BASF LM-LF1508 Candy Apple Red Acrylic Enamel
10 gal.
Surface Preps
i.
2
3.
Primers
1.
2.
3.
i 1 i
Surf acers/tillers
1""
1.
2.
3.
Sealers
1.
2.
3.
Topcoats
(single and
base/clear)
I
1.
2.
3.
i
Newly Purchased
Cleaning Solvent
1.
9
KJ .
3.
I
Other Solvent Based
Products
1.
9
.
n
U.
A-5
-------
8c. How did you determine the number of gallons of materials used from to
9
01 Materials inventory records
02 Number of jobs completed
03 Other, please specify
7a. Would you consider the number of jobs you completed from to
to be typical for this shop location?
01 . . YES Please go to Question 8a. 02 . . NO
7b. Was the number of jobs more or less than your typical month?
01 . . More 02 . . Less
7c. Why do you think your business was different during this time period0
9
8a. Does your business activity vary from month to month?
01 . . YES 02 . . NO Please go to Question 9.
8b. Which three months are typically your busiest.? Please enter a "B" next to those
months. Which three months are typically your least busy? Please enter a "LB" next
to those months.
January May September
February June October
March July November
April August December
8c. Please describe why you think the three months you marked with a "B" are your
busiest.
A-6
-------
8d. Please describe why you think the three months you marked with a "LB" are your
leasLbus^.
9a. Do you use a factory built spray booth?
01 . . YES 02 . . NO Please go to Question 9c.
9b. Please indicate the manufacturer and mocei number.
Manufacturer:
Model ง\
9c. Do you use a factory built prep station9
01 . . YES 02 . . NO Please go to Question 10.
9d. Please indicate the manufacturer and model number.
Manufacturer:
Model ง:
10. Please indicate which types of spray guns are used at this location. Check (~) all that
apply. Por each type of gun being used, please indicate the percentage of time it is
used applying primer, basecoat,, or topcoat.
% time applying:
Gun Type
Check (~)
if used
Primer
Topcoat
(single stage)
Topcoat (two stage)
Basecoat
Clearcoat
Siphon
Gravity Feed
Pressure
HVLP
flt'hrr
A-7
-------
11a. Docs your shop have factory built gun cleaning equipment in operation?
01 . . YES 02 . . NO Please go to Question 11c.
lib. Please indicate the manufacturer and model number.
Manufacturer:
Model jr.
ll.c Is this gun cleaning system open or enclosed0 Please check (~) your answer.
Open Enclosed
lid. Approximately, how many gallons of waste solvent are generated in gun cleaning per
month?
Please specify: gallons
Solvent -based materials may be used In various locations in the shop. [This may include
areas outside of the spray booths and enclosed gun cleaning stations.J
12a. How often is solvent based material used in areas other than spray booths or
enclosed gun-cleaning stations? Check (~) one answer.
)
Daily jf
\
2 or 3 times a week / Go to Question 12b.
a
Once a week ^
-j
Less than once a week f
Never
Go to Question 13.
12b. During a typical week, how many gallons of solvent-based material are used in areas
other than spray booths and enclosed gun-cleaning stations?
gallons
A-8
-------
13. Various equipment is available to improve ventilation and to control emissions and
vapors in work areas. Have you installed any control equipment (e.g., ventilation
equipment, recyclers, vapor retaining containers) in your shop9
01 . . YES 02 . . NO iฎ* Please go to Question 14.
Please identify or describe this equipment (include make and model):
14. How many gallons of waste which contain solvents are removed from this shop
location by a licensed carrier (e.g. SafetyKleen, Herkules) each month?
gallons
15. Do you expect your auto refinishing business to increase, decrease, or remain the
same over the next five years9 Circle one answer.
01 . . Increase 02 . . Decrease 03 . . Remain the same
16. What factors (e.g., car safety equipment, DWI laws, traffic patterns, economic growth,
changes in weather) did you use to predict your business patterns in question 15?
17a. What percentage of the auto refinishing businesses in your county are unlisted,
unlicensed or "backyard" operations9
%
17b. Approximately what percent of the auto refinishing jobs are done by these unlisted,
unlicensed or "backyard" operations?
<7
AO
A-S
-------
18.
Do you expect the use of solvent-based painting materials (including surface preps,
additives, primers, and sealer) to increase, decrease, or remain the same within the
next five years at this location9 Circle one answer.
01 . . Increase 02 . . Decrease 03 . . Remain the same
19. Hew do current local, state, or federal environmental regulations impact your
business (e.g. product selection, recycling, productivity, additional costs)?
20. How might changes in mixtures of solvent-based materials (e.g. low VOC paints), or
new equipment requirements (e.g. afterburners) impact your business0
21. How arc rags used in vehicle prep and clean up0
A-10
-------
APPENDIX B
SAMPLING AND ANALYSIS
DISPERSION MODEL
B-l
-------
Distribution Model
The basic equation typically used to generate air quality models is the diffusion
equation. This equation reads as follows:
dc - dc - dc -de d ,v dc. d dc, d dcx
+ u + v + w = (K) + (K) + (K,) + source terms
dt dx dy dz dx X3x dy y dy dz 1 dz
+ transformation terms
where, c = time-averaged concentration
x, y, z = Cartesian coordinates
u, v, w = components of time-averaged wind vector
Kj., Ky, Kz = diffusion coefficients in the corresponding directions
Several assumptions are made to simplify the solution:
Steady state diffusion
The time averaged wind vector is assumed to be oriented in the x-direction
only, neglecting any buoyancy or height effects on the wind direction (i.e., v
= w = 0)
No sink or source effects
No chemical transformations
Mass transfer due to bulk motion in x-direction is assumed to be far greater
than that due to mass diffusion, i.e.,
dc d dc,
> (K j
dx dx Xdx
Constant mass diffusion coefficients (Kx, Ky, Kz)
The diffusion equation thus reduces to:
- dc ir &c v &c
u = Kv + K,
3* 'dy* ^2
One appropriate representation of the concentration profile downwind from a point
source is given by the following general equation:
c(W) - Kexp
x
I 2 2 \
'>'2 + Z2
K K
\ y i}
u
4x
B-2
-------
K is a constant whose value depends on the boundary conditions of the specific
atmosphere problem. One such boundary condition for a point source at ground
level is:
/ป 4-eo /ป +00
Q = jo j _ ucdydz
where, Q is the source emission mass flow rate.
By defining:
2Kx . 2K x
a - a -
y z
U U
the concentration downwind from a ground level point source takes the form of a
double Gaussian function:
c(x,yX) ฎ exp
_ 1
f yl +
TZUO 0
2
0
y z
\ y i /
The coordinate system used for this equation assumes that the point source is
located at position (0,0,0), where the x-axis is pointed towards the direction of
wind flow; the z-coordinate indicates the vertical position relative to the ground.
In order to evaluate the emission concentration at a given location (x,y,z), several
values are required:
The emission strength, Q (e.g., kg/s): The mass flow rate is the product of
the volume rate, V, which can be measured on location, and the density, p,
of the air at the point source1.
The wind speed, u (e.g., m/s): Since the wind speed varies with height, a
mean value taken through the plume should be used.
The standard deviations, CL and a7 (e.g., m): The horizontal and vertical
standard deviations are a function of the downwind position x, as well as
the atmospheric stability conditions. Although no theoretical expressions are
available, several experimental measurements in the atmosphere have led
to the evaluation and correlation of these values. It is recommended to
select the correlation that best approaches the conditions at the test site
(e.g., urban versus rural area).
It is assumed that the small concentration of VOC's in the air will have a negligible effect on the air density.
B-3
-------
Auto Refinishing Shop Model
Auto refinishing shops are typically one-story buildings and are generally
classified as area sources. Although the formulas presented previously arc only
valid for point sources at ground level and under certain specific conditions, they
could be slightly modified to apply to an auto refinishing shop.
As with point sources, no universal model is available to estimate the
concentration downwind of an area source. However, general methods are used
and must be modified for every specific problem. For instance, area sources may
be subdivided into a series of point sources separated by a constant distance d.
The concentrations due to the various sources would then be superimposed to
generate an estimate of the concentration downwind of the source.
Unlike other area sources, such as landfills, which emit continuously over the
whole area, auto refinishing shops will typically have 3 or 4 localized sources (e.g.,
exhaust vent, door, and window). Each of these emission "outlets" can be modeled
as an individual point source. The total VOC concentration would then be
approximated by superposition of the different point sources. Plume rise effects
can be neglected in this case due to the low exit velocity of the VOCs. The low
height of the buildings also makes it possible to assume that the emissions are
generated from sources at ground level.
Another approach, which will be used in this case, is to model the area source as a
single point source located at the centroid. The total emission rate would then be
the sum of the emission rates from the various outlets.
Several measurements have been taken in the past to determine the values of av
and oz at various distances x from a point source. Since these values vary with
atmospheric conditions and location, the correlation best approaching the specific
topographic and atmospheric conditions of the body shop should be selected. The
most frequently used set of data is that of Pasquill-Gifford. These standard
deviations were determined for a low level source in open country. Other deviation
values have been determined for a low level urban area by McElroy and Pooler2.
However, since most auto refinishing shops in the Triangle area are located near
highways, far from the few tall buildings in the city, the open country
approximation would be acceptable. The Pasquill-Gifford data will therefore be
used.
2
McElroy, J.L., and Pooler, Jr., F. St. Louis Dispersion Study. Vol. IT: Analysis. Rpt No. AP-53
(NT1S PB190255). National Air Pollution Control Administration, Arlington, VA December 1968.
B-4
-------
Evaluation of the Standard Deviations
The horizontal and vertical standard deviation, Gy and Gz are functions of the
downwind position x and the atmospheric stability conditions. The Pasquill-Gifford
(or Turner) correlations will be used. These correlations have certain restrictions3:
Sampling time of approximately 10 minutes
Open country terrain
Lowest several hundred meters of the atmosphere
Preferably, distance of receptors from source less than 1 km
The atmospheric conditions were listed under 6 different categories4:
A = extremely unstable
B = moderately unstable
C = slightly unstable
D = neutral
E = slightly stable
F = moderately stable
The conditions are listed in Table 1:
Table 1: Guide to Pasquill-Gifford Stability Categories3
Surface wind
speed @10 m
(m/s)
Day, incoming solar radiation
Night, cloud cover
Class
Strong
(1)
Moderate
(2)
Slight
(3)
Mostly
overcast
(4)
Mostly clear
(5)
< 2
A
A-B
B
E
F
2-3
A-B
B
C
E
F
3-5
B
B-C
0
D
E
5-6
C
C-D
D
D
D
> 6
C
D
D
D
D
Wark. Kenneth ond Cecil F. Wnrner. "Air Pollution: Its Origin and Control," Harper & Row, Publishers,
[nr. NVw York, 197f>.
Weber. h)rich. "Air Pollution: Assessment Methodology and Modeling," Plenum Press: New York, 1982.
B-5
-------
Some examples of these various classes would be:
1. Clear skies, solar altitude greater than 60 degrees above horizontal, typical
of a sunny summer afternoon.
2. Summer day with a few broken clouds
3. Sunny fall afternoon, summer day with broken low clouds, or summer day
with clear skies and solar altitude from only 15 to 35 degrees above
horizontal.
4. Can also be used for a winter day.
The standard deviations can be approximated by the following equations:
2
-0.152
0.820
0.850
0.793
1.255
1.419
^3
0.147
0.017
0.005
0.002
-0.042
-0.055
B-6
-------
LOCATION OF CENTROID
If the locations of the various point sources, i, are given by the coordinates and
yi with respect to one of the building corners, then the centroid is located at (xg,
yg), where:
_ *1 + *2 Qz + "*3 ^3
= Qi''+ Q2 +
-------
NTIS
Information it our business.
NATIONAL MARINE FISHERIES SERVICE,
CHARLESTON LABORATORY VOLUNTEER HANDBOOK
(U.S.) NATIONAL MARINE FISHERIES SERVICE, CHARLESTON, SC
MAY 97
U.S. DEPARTMENT OF COMMERCE
National Technical Information Service
-------
^*TOFCo
* w \
O ^
*
\ฎ/
^rcs
-------
NOAA TECHNICAL MEMORANDUM
NMFS-SEFSC-400
"IMP
NATIONAL MARINE FISHERIES SERVICE
CHARLESTON LABORATORY
VOLUNTEER HANDBOOK
by
Wayne E. McFee
U.S. DEPARTMENT OF COMMERCE
William M. Daley, Secretary
NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION
D. James Baker, Administrator
NATIONAL MARINE FISHERIES SERVICE
Rolland A. Schmitten, Assistant Administrator for Fisheries
MAY 1997
This Technical Memorandum series is used for documentation and timely
communication of preliminary results, interim reports, or similar special-purpose
information. Although the memoranda are not subject to complete formal review,
editorial control, or detailed editing, they are expected to reflect sound
professional work.
/
-------
NOTICE
The National Marine Fisheries Service (NMFS) does not approve, recommend, or
endorse any proprietary product or material mentioned in this publication. No
reference shall be made to NMFS, or to this publication furnished by NMFS, in
any advertising or sales promotion which would indicate or imply that NMFS
approves, recommends, or endorses any proprietary product or proprietary
material mentioned herein or which has as its purpose any intent to cause directly
or indirectly the advertised product to be used or purchased because of NMFS
publication.
This report should be cited as follows:
McFee, W.E. 1997. National Marine Fisheries Service, Charleston Laboratory
Volunteer Handbook. NOAA Tech. Mem. NMFS-SEFSC-400, 37 pp.
Author's affiliations: McFee, W.E. NOAA, NMFS, Charleston Laboratory,
Charleston, SC 29412.
Copies may be obtained by writing the author or:
National Technical Information Service
5258 Port Royal Road
Springfield, VA 22161
(703) 487-4650 FAX: (703) 321-8547
Rush Orders: (800) 336-4700
-------
TABLE OF CONTENTS
About the NMFS Charleston Laboratory 1
Organizational Chart 2
What You Can Expect 3
What NMFS Charleston Laboratory Expects from You 4
VOLUNTEERING POLICIES 5
Background Checks 5
Attendance 5
Confidential Information 5
Public Relations 6
Equal Opportunity 6
Harassment 6
Job Descriptions 7
OTHER POLICIES 7
Computer Software 7
Personal Use of Government Property 7
Copy Machines 7
Personal Phone Calls 7
Parking 8
Smoking 8
Dress Code 8
Substance Abuse 8
Problems 8
Expense Reimbursement 9
Resignation and Exit Interview 9
STANDARDS OF CONDUCT * 9
Unacceptable Activities 9
Disciplinary Actions 10
Dismissal 10
WORKERS'COMPENSATION 11
SAFETY AND SECURITY 12
-------
ACKNOWLEDGEMENTS 13
APPENDIX A 14
SECURITY WORKSHEET FOR NON-EMPLOYEES
APPENDIX B 16
VOLUNTEER PROGRAM APPROVAL CERTIFICATE
APPENDIX C 17
WAGE CLAIM WAIVER
APPENDIX D 18
CHEMICAL HYGIENE PLAN
APPENDIX E 30
CHARLESTON LABORATORY EMERGENCY PREPAREDNESS PLAN
vซ
ii
- ft
-------
Welcome To NMFS Charleston Laboratory
Dear Volunteer:
Welcome to the National Marine Fisheries Service (NMFS)Charleston Laboratory. Thank you
for joining us! We want you to feel that your association with NMFS will be a mutually
beneficial experience.
You have joined an organization that is dedicated to the management and protection of marine
resources as well as the education of the public about our mission. We hope you will also find
satisfaction and take pride in your work to conserve our marine resources.
This Handbook provides answers to most of the questions you may have about the NMFS
Charleston Laboratory programs, as well as the policies and procedures we abide by. If anything
is unclear, please discuss the matter with your immediate supervisor. You are responsible for
reading and understanding this Handbook to ensure safety for you and co-workers and to adhere
to NMFS policies and procedures.
Occasionally the information contained in this Handbook may change. Every effort will be made
to keep you informed of policy and procedural changes.
Personal satisfaction of doing a job well and career development are two of the many reasons
people volunteer. The NMFS Charleston Laboratory is committed to doing its part to provide
you with a satisfying volunteer experience.
We extend to you our personal best wishes for your success and happiness at the NMFS
Charleston Laboratory.
Sincerely,
Wayne McFee
Volunteer Committee Chairman
Sylvia B. Galloway, Ph.D.
Laboratory Director
111
ฆ" *
-------
Purpose of this Handbook
This Handbook has been prepared to inform you of the National Marine Fisheries Service's
history, philosophy, policies, and expectations of you and the NMFS.
No volunteer handbook can answer every question, so we hope through regular conversations
between you and your supervisor we can continue to add to the Handbook as conditions warrant.
We hope this Handbook will help you feel comfortable with us. We depend on you in an ever
shrinking workforce - your success is our success. Please do not hesitate to ask questions. Your
direct supervisor or Volunteer Coordinator will gladly answer them.
We ask that you read this Handbook carefully, and refer to it whenever questions arise. The
NMFS policies, benefits and rules, as explained herein, may be changed from time to time as
business, volunteer legislation, and economic conditions dictate. If and when changes are made,
you will be made aware of the changes.
iv
-------
About the NMFS Charleston Laboratory
MISSION STATEMENT
The mission of the Charleston Laboratory is to provide scientific information required to resolve
management issues associated -with NMFS' agency goals. These goals are to:
Build sustainable marine fisheries
Maintain currently productive fisheries
Advance fishery forecasts and ecosystem models
Integrate conservation of protected species and fisheries management
Improve seafood safety
Protect living marine resource habitat
Improve the effectiveness of international fisheries relationships
Reduce impediments to U.S. aquaculture
BACKGROUND
The Volunteer Program at the NMFS Charleston Laboratory was initiated as a result of the 1995
Federal Government National Performance Review leading to a reduction in the federal
workforce. The loss of Full Time Employee (FTE) positions has resulted in a need to utilize
volunteers and to provide opportunities to students and interested professionals interested in
marine fisheries careers. The NMFS Charleston Laboratory provides opportunities to volunteer
in a number of marine fisheries disciplines (Figure 1):
Marine Ecotoxicology
Risk Assessment
Marine Biotechnology
Marine Forensics
Managed and Protected Resources
Marine Biotoxins
Seafood Technology and Regulation
Biomedical Test Materials '
Some duties volunteers have undertaken include: routine laboratory duties, water and oyster
sampling for water quality analyses, chromatography, assistance in the retrieval and necropsy of
stranded marine mammals, preparation of tissue and skin samples for pesticide, heavy metals,
and genetic analyses, marine mammal skeletal preparation, and others. Approximately 10-15
volunteers spread among the various programs work throughout the week at the Charleston
Laboratory.
1
*
-------
National Marine Fisheries Service
Charleston Laboratory
Reproduced from
best available copy.
-------
What You Can Expect from the NMFS Charleston Laboratory
The NMFS Charleston Laboratory will provide you with:
A clear j ob descripti on
Assignments that fit your skills, interests, availability, and training
Orientation, training, and supervision for specific duties, and an explanation for such
duties
A friendly, safe workplace that promotes a pleasant learning experience
Safety precautions regarding chemicals, field work, and potentially dangerous areas
Prompt and fair responses to on-the-job problems which may arise
Recognition and appreciation for a job well done
Skills and contacts to enhance career objectives and goals
Confidentiality of individual records
Personnel records to document experience, skills learned, training, evaluations, and
commendations
Consultation on job performance prior to resigning
-------
What NMFS Charleston Laboratory Expects from You
The NMFS Charleston Laboratory expects you will:
Act professionally and maintain a good attitude towards your job and your fellow
workers
Honor the work schedule you have arranged with your supervisor, and notify your
supervisor if your schedule changes, or you can not work due to unforeseen
circumstances (i.e., illness, car trouble, etc.)
Perform duties assigned to the best of your abilities
Ask questions if you are unsure of a job assignment or safety issue
Follow all safety precautions, policies, and rules outlined by your supervisor and/or
provided to you by the NMFS Charleston Laboratory
Hold in confidentiality anything you may hear at the Laboratory which may jeopardize
the security of data produced at the Laboratory
Obtain permission from your supervisor or Laboratory official to "tour" a friend or family
member around the NMFS Charleston Laboratory
ป Make this an enjoyable and rewarding experience that will help you in your future
endeavors!
4
-------
VOLUNTEERING POLICIES
As a new volunteer to the NMFS Charleston Laboratory, you may be overwhelmed with the
number of policies and safety issues you are exposed to. As a returning volunteer you may
notice changes to policies and procedures. This section is designed to answer any questions you
may have regarding policies and procedures.
Background Checks
The federal government requires all volunteers fill out a Wage Claim Waiver (Appendix A) and a
Security Worksheet for Non-Employees (Appendix B) if work assignment is for less than 60
days. If working over 60 days volunteers are required to fill out a Wage Claim Waiver, a
Security Worksheet and submit two sets of fingerprints for a security check. Failure to comply
with the above will result in a termination of volunteer service. Acceptance into the Program
shall be granted after a Volunteer Program Approval Certificate (Appendix C) has been signed
by all listed approving officials and review of all credentials and background have been satisfied.
Time Commitment
Volunteers are expected to establish a work schedule with their supervisor by stating the days
and number of hours per week to be worked. Requests for changes in time commitment should
be communicated to your supervisor at the earliest opportunity.
If you are unable to report to work, or expect to arrive late, please contact your supervisor so that
arrangements can be made to cover for your absence. Repeated absence or lateness may require
a change in schedule or termination of appointment.
Attendance
Volunteers are required to "sign in" in the designated log book at the front desk, or if you do not
have a time sheet in the book, you must sign in in the guest register at the front desk. You may
also be asked to sign in in your particular area of work.
Sign-in procedures are important for keeping track of number of hours worked, starting and
ending dates, for security measures, and for safety requirements in case of fire. It is also a
measure to recognize time spent in a program and for evaluating the volunteer program.
Confidential Information
As a volunteer in this federal government laboratory, you may view sensitive documents, be
present during discussion of sensitive information, such as law enforcement evidence analysis
5
* t
-------
results, or overhear staff opinions about politically sensitive topics. Because you may not have
knowledge of some of these issues, you shall refrain from discussion of any staff conversations
or viewed documents outside the room where such conversations or viewings have occurred.
Public Relations
Government agencies are constantly in the public's eye and issues related to marine species,
especially endangered species, can be delicate subjects. Integrity of samples for receiving and
shipping to other researchers need to be handled in a professional manner in order to deliver
quality samples for research as well as protect the reputation of the NMFS Charleston
Laboratory.
As a volunteer you shall not discuss privileged research information, law enforcement cases, or
any other information deemed confidential by the NMFS Charleston Laboratory to the general
public. Your supervisor will explain these items to you. As a volunteer you shall not answer
any questions from the media without the permission of your supervisor. Please refer questions
from the media to your supervisor regardless of whether you can answer the question or not.
From time to time you may be asked to assist in educational projects or participate in field
research which places you face-to-face with the public. Please communicate pleasantly and
respectfully to the public and refer questions to your supervisor.
Acting professionally and using common sense can save you and the Laboratory from
embarrassing situations.
Equal Opportunity
The NMFS provides equal volunteering opportunity for everyone regardless of age, sex. color,
race, creed, national origin, religious persuasion, marital status, or disability that does not
prohibit performance of essential job functions, with or without accommodations, as stated in the
job description.
Harassment *
The NMFS Charleston Laboratory intends to provide each volunteer with a pleasant, healthful,
and comfortable working environment, free from intimidation, hostility or other offenses which
might interfere with your performance as a volunteer. Harassment of any sort - verbal, physical,
visual, sexual - will not be tolerated. Any volunteer who becomes aware of harassment in any
form, or is harassed him- or herself, has an obligation to report the incident(s) to their supervisor
or a trusted employee immediately. The NMFS is obligated by law to investigate allegations of
any form of reported harassment.
-------
Job Descriptions
We maintain a job description for each volunteer position in the NMFS Charleston Laboratory.
Job descriptions may be very general or very specific depending on the needs of the particular
Project If your duties are changed, your job description will be updated. You may ask to see a
copy of your job description at any time.
OTHER POLICIES
Computer Software
Unauthorized duplication of software is a federal crime! U.S. Code, Title 17, Section 106 states
that "it is illegal to make or distribute copies of copyrighted material without authorization" from
the copyright holder.
At the NMFS Charleston Laboratory you may receive authorization from your supervisor and/or
computer analyst to use the Government computers and will receive a password to log-in to the
network system.
The NMFS Charleston Laboratory computers shall be used only by volunteers for carrying out
routine laboratory activities that are described in your job description and approved by your
supervisor (e.g., data entry). Personal computer software shall not be added to Laboratory
computers. The NMFS Charleston Laboratory computers are not for personal use (i.e,
typing letters, school work, resumes, etc.).
Personal Use of Government Property
At no time shall a volunteer use Government property for personal purposes.
Copy Machines
*
Copy machines are not for your personal use. You may be asked to make copies as part of your
job duties. These machines are located in the mail room near the front reception desk and in the
back hallway near the Biotoxins section.
Personal Phone Calls
Personal use of Laboratory phones is not permitted unless absolutely necessary. Long-distance
phone calls are not allowed except in the case of an emergency. Otherwise you must use your
own calling card and long-distance carrier. Local calls must be kept to a minimum.
-------
Parking
Please park in the front of the NMFS Charleston Laboratory along with other staff members. Do
not park in the designated "Visitor" spaces or "Handicapped" spaces (unless you are authorized
to do so). Please do not park on the grass in front of the flag pole or in the back of the
Laboratory by the loading dock. These areas are needed for emergency vehicles and deliver}'
trucks.
Smoking
Smoking is not allowed anywhere in the Laboratory buildings. If you must smoke, please do so
outside and use the cigarette receptacles.
Dress Code
Please see your supervisor for the appropriate dress required for your duty station. Safety
concerns may require specific dress conditions.
While the majority of Laboratory personnel enjoy the freedom of dressing casually, there are
instances when more formal dress is appropriate (i.e, meetings, public appearances, etc.). There
are also instances when you may be in the field where "throw-away" clothes and footwear may
be appropriate. Your supervisor should keep you informed as to the above situations.
Substance Abuse
Mind altering substances, such as alcohol or cocaine, will not be tolerated at the NMFS
Charleston Laboratory. The use of drugs has an impact on the image of the Laboratory and
personnel, as well as in our ability to achieve our objectives of safety and security.
The possession, sale or use of mind altering substances while volunteering shall result in
immediate termination of volunteer service. If you have knowledge of anyone participating in
the above while at the Laboratory please see your supervisor, personnel director, or Laboratory
Director. Your name will be held in confidentiality.
Problems
Your problems are important to the NMFS Charleston Laboratory. In order to run an efficient
laboratory we need to know when problems arise.
Be assured we will be receptive to your concerns regarding any situation which you believe
violates your civil rights, is discriminatory, or presents a feeling of discomfort in your duties.
This includes statements, attitudes, or opinions held by your supervisor or a fellow worker.
8
-------
If you sense a problem, please see your immediate supervisor first. If the problem is against your
supervisor please see the Volunteer Coordinator of the Laboratory. After a confidential
discussion regarding the issue, if you are still not satisfied with the outcome, you may brine your
problem directly to the Volunteer Committee. The problem will be discussed with the
Committee and a decision rendered as final. The purpose of this procedure is to give each
volunteer the chance to voice opinions regarding any problem situation.
Expense Reimbursement
You must have your supervisor's authorization to incur an expense on behalf of the NMFS.
Situations may arise in the field where you may have to purchase small items (such as ice) to
complete a job. To be reimbursed for all authorized expenses, you must produce receipts and
signed approval from your supervisor.
Resignation and Exit Interview
If you are planning to resign as a volunteer, the NMFS Charleston Laboratory would appreciate
notice as far in advance as possible to make arrangements for an exit interview with your
supervisor and to allow for time to staff your volunteer position. An informal exit interview with
your supervisor is recommended in order to obtain your ideas of ways to improve our program,
answer any questions you may have, and to see how you enjoyed your time with us.
STANDARDS OF CONDUCT
Because of the professional nature of business conducted at this Laboratory, stringent safety
precautions, and law enforcement issues, certain standards of conduct are expected of each
volunteer, as well as NMFS employees, to assure high quality products in a safe environment.
The purpose of these standards is not to restrict your rights, but rather to be certain that you
understand what is needed for you to meet these important responsibilities.
Unacceptable Activities
If you have any questions concerning any safety rule or any of the unacceptable activities listed
below, please see your supervisor for an explanation. Occurrences of any of the following
violations may result in immediate dismissal without warning:
Willful violation of any NMFS policy or rule, including actions that are obviously
detrimental to the NMFS.
Willful violation of security or confidentiality policies.
9
*
-------
Willful or repeated violation of safety policies or rules; failure to wear required safety
equipment; tampering with NMFS equipment.
Negligence or any careless action which endangers the life or safety of another person.
Being intoxicated or under the influence of controlled substance drugs while volunteering
or on Government property; me or possession or sale of controlled substance drugs in any
quantity while on NMFS premises.
Unauthorized possession of dangerous or illegal firearms, weapons, or explosives on
NMFS premises or while working in the field.
Engaging in criminal acts of violence, threats of violence, fighting, provocation of
fighting, or negligent damage to property while on NWS premises.
Insubordination or refusing to obey instructions deemed proper by your supervisor.
Theft of NMFS or Government property or the property of co-workers; unauthorized
possession or removal of NMFS or Government property, including documents,
manuscripts, journals, books from the premises; unauthorized use of NMFS or
Government equipment or property for personal use; using NMFS or Government
equipment or property for profit
Dishonesty; willful falsification of personnel records; altering of personnel records or
other NMFS or Government documents.
Engaging in behavior designed to create discord and lack of harmony; interfering with the
work of others within the Laboratory.
Immoral conduct or indecency on NMFS premises.
Unauthorized presence on NMFS premises after hours.
<*
Disciplinary Actions
Unacceptable behavior which does not lead to immediate dismissal may be handled with a verbal
warning, written warning, or suspension. Written warnings will include the reasons for the
warning and any supporting evidence. You will have the opportunity to contest the warning by
presenting your case to the Volunteer Coordinator and/or Volunteer Committee.
Dismissal
Volunteers who do not adhere to the rules of the agency or who fail to satisfactorily perform their
10
-------
assignment are subject to dismissal. Repeated absence or failure to honor a pre-arranged
schedule may also be cause for dismissal.
Volunteers may be dismissed without warning for just cause. The NMFS has the right to request
a volunteer leave immediately. Grounds for immediate dismissal may include, but are not
limited to:
Gross misconduct or insubordination
Being under the influence of mind altering drugs, including alcohol
Theft of property or misuse of NMFS equipment or materials
Lies or falsification of records
Illegal, violent or unsafe acts
Conflict of interest
Breach of security policies
Breach of confidentiality policies
Abuse or mistreatment of co-workers or animals
WORKERS' COMPENSATION
You are covered under Workers' Compensation title 5 U.S.C. chapter 81, relative to
compensation for injuries sustained during the performance of work assignments. Claims related
to injuries should be referred to the Office of Workers' Compensation Programs, U.S.
Department of Labor.
Should you become injured or ill while performing your volunteer duties at the NMFS
Charleston Laboratory please notify your supervisor immediately. Your supervisor will put you
in touch with the proper authorizing official to fill out a claim form.
11
-------
4
SAFETY AND SECURITY
The NMFS Charleston Laboratory has a Laboratory Safety Program, which includes a Chemical
Hygiene Plan (Appendix D) and an Emergency Preparedness Plan (Appendix E), that all
employees and volunteers must read and follow. Each Program at the NMFS Charleston
Laboratory has its own specific guidelines and policies for safety and security. These guidelines
and policies will be explained to you by your immediate supervisor after assignment.
While each Program may have slightly different safety policies and procedures, the Laboratory
as a whole believes that safety comes first and is everybody's business. If you are placed in a
situation that you are not comfortable in, and you feel the activity is a hazard to you or someone
else, please stop and notify your supervisor of your concern before proceeding. Please pay
attention^ all warning signs. All NMFS volunteers are required to view safety videos before
beginning work.
Maintaining security in each Program may also be slightly different, but the Laboratory does
have common practices. Maintaining security is also everybody's business.
Everyone is to be aware of new faces in the building. If you see someone in the building or on
the property who is suspicious or who you do not recognize, kindly ask them if you can help
them, or find a co-worker who may be able to identify that person. Everyone should sign into the
guest register when they come in the building. If you see someone who does not sign in when
they come in the building, kindly ask them to do so, or alert an Office Support Specialist at the
front desk.
All outside doors to the buildings are to remain locked at all times. After you leave an area
please lock the outside door if it requires a key. If you do not have keys or a magnetic card and
notice an outside door unlocked, alert your supervisor so that he/she can lock the door. Again,
each Program or Project may have other security measures for door access, computer access or
file cabinet access. You will be issued a magnetic access card which will be coded for the
appropriate areas you need access to in order to perform your duties.
12
-------
ACKNOWLEDGMENTS
I would like to thank the following people for their review and comments: NMFS Charleston
Laboratory Volunteer Committee (Ms. Nancy Davey, Ms. Pat Smallwood, Ms. Carol Preston,
Dr. Dan Bearden, Dr. Mike Fulton, Mr. Ron Lundstrom, Dr. Peter Moeller), Dr. Sylvia
Galloway, Dr. Malcolm Meaburn, Dr. Paul Comar. Dr. Pat Fair, and Ms. Marlene Wiggins. I
would also like to thank all of the volunteers both present and past who have contributed their
time and efforts to make all of our work more manageable, productive, and enjoyable.
13
-------
appendix A
SECURITY WORKSHEET FOR NON-EMPLOYEES
1. Full names of non-employee:
2. Other names and dates used:
3. Position:
v
w
4. Project Title:
5. Place of Birth: . ;
(Be sure to include the city, county, state, and country, if other than the U.S.)
6. Date of Birth:
7. Social Security Number:
8. Sex: Male Female
9. Has guest worked for DOC in the past: Y N
Location: Date:
10. Period of Visit: Beginning Date: Ending Date:
ARREST RECORD:
11. During the last 10 years have you ever forfeited collateral, been convicted, been
imprisoned/been on probation or parole? Y N
12. Are you now under changes for any violation? Y N
13. Have you ever been convicted by a military court-martial or received Non Judicial
punishment under the Uniformed Code of Military Justice? Y N
14. In the last five years have you ever possessed, used or manufactured illegal drugs?
Y N .
14
-------
*
HEALTH CARE:
15. Have you ever seen a health care professional for the treatment of an alcohol, drug,
mental or emotional disorder? Y N
If you answered yes to any of items 11-15, please explain your answer on a separate sheet of
paper.
Signature Date
This section to be completed by the requesting official:
1. Name:
2. Mailing Address:
3. Position or Title:
4. Will access to departmental facilities be restricted to normal office hours or under escort?
Y N
5. Furnish accounting data if visit is for more than 60 days.
Accounting data: __
6. If the visit is for less than 60 days, this form may be sent to security for Regional Security
Officer review.
7. Failure to forward this form, assumes Facility Manager permitting visit accepts full
responsibility and risk for the actions of the non-employee.
Date of Request Signature of Requesting Official Date
Approved Y N
Date Received Date Processed
Robert E. Dickson
Regional Security Officer
15
Date:
-------
ซ v
APPENDIX B
NOAA/NMFS
CHARLESTON LABORATORY
Volunteer Program Approval Certificate
After review of credentials, including past and present coursework, employment history,
internships (if any), volunteer experiences (if any), and present availability, the undersigned
approve/reject the request of , for a volunteer position at
the NMFS Charleston Laboratory.
Signed Approve Reject Date.
Supervisor
Signed Approve Reject Date.
Division Chief
Signed Approve Reject Date.
Prom. Mgmt. Spec.
Signed Approve Reject Date.
Volunteer Comm. Chair
16
-------
4
APPENDIX C
WAGE CLAIM WAIVER
Against
Federal Agencies
Under the
Student Volunteer Service Program
I hereby acknowledge that no Agency of the United States Government is responsible for the
payment of any wages to me because of any work performed as an Enrollee for the National
Marine Fisheries Service, U.S. Department of Commerce as part of the Student Volunteer
Service Program. I agree that I will not make a claim against the United States Government or
the National Marine Fisheries Service for wages for my services.
Signature of Enrollee Date
Street Address
City State Zip Code
Witness Date
Witness Date
17
-------
4 <ป
APPENDIX D
CHEMICAL HYGIENE PLAN
FOR
CHARLESTON LABORATORY
SOUTHEAST FISHERIES SCIENCE CENTER
NATIONAL MARINE FISHERIES SERVICE
NATIONAL OCEANIC AND ATMOSPHERIC ADMINISTRATION
DEPARTMENT OF COMMERCE
REGULATORY REQUIREMENT/INTRODUCTION
*
The Department of Labor. Occupational Safety and Health Administration (OSHA)
Standard 29 CFR 1910.1450, Occupational Exposures to Hazardous Chemicals in Laboratories,
became effective on May 1. 1990. The Federal Register announcement states that "employers
shall have completed an appropriate Chemical Hygiene Plan and commenced carrying out its
provisions by January 31, 1991". The new standard recognizes that laboratories typically differ
from industrial operations in their use and handling of hazardous chemicals and that a different
approach than that found in OSHA's substance specific health standards is warranted to protect
workers.
The standard applies to laboratories that use hazardous chemicals in accordance with the
definition of laboratory use and laboratory scale as provided in the standard. Laboratories are
obligated to maintain employee exposures to hazardous chemicals at or below the permissible
exposure limits (PELs) specified in 29 CFR 1910, subpart Z. The manner in which this
obligation is achieved will be determined by each employer through the formulation and
implementation of a Chemical Hygiene Plan (CIIP). The CHP must include the necessary work
practices, procedures and policies to ensure that employees are protected from all potentially
hazardous chemicals in use in their work area.
The OSHA Standard defines a Chemical Hygiene Plan as a written program developed
and implemented by the employer which sets forth procedures, equipment, personal protective
equipment and work practices that (I) are capable of protecting employees from the health
hazards presented by hazardous chemicals used in that particular work place and (ii) meets the
requirements of paragraph (e) of this section (of the standard). Paragraph (e) states that the CHP
shall be available to all employees and shall include the following elements: (I) Standard
operating procedures for the use of hazardous chemicals, (ii) Control measures to be
implemented to reduce employee exposure to hazardous chemicals, (iii) Specific measures to be
taken to assure that fume hoods and other protective equipment are functioning properly, (iv)
Provisions for employee information and training, (v) The circumstances under which a
laboratory activity shall require prior approval, (vi) Provisions for medical consultation and
18
-------
4
medical examinations, (vii) Designation of personnel responsible for implementation of the
Chemical Hygiene Plan, (viii) Provisions for additional employee protection for work with
particularly hazardous chemicals. Eleven components of a CHP as recommended by the
National Research Council are included in this plan and are listed in the Table of Contents
(page I).
This Chemical Hygiene Plan as well as more specific references relating to chemical
safety, and safety in general at this facility, are available on the shelves adjacent to the lobby of
the Charleston Laboratory. Included are:
(1) Charleston Laboratory Emergency Preparedness Plan - Instructions for evacuation of
the building for fire alarms or other emergencies and preparation/response for hurricanes and
other natural disasters.
fc
*
(2) A copy of the Hazard Communication Program for the Charleston Laboratory- It
includes an inventory list of hazardous chemicals used in this laboratory.
(3) Radiation Safety Manual - Rules and guidelines for those who work with radioactive
materials.
(4) Working procedures for the control and containment of biotoxins.
(5) Biomedical Test Materials Program: Production Methods and Safety Manual -
Detailed instructions for BTM/flsh oil processing operations.
(6) Additional, selected publications on chemical safety and first aid.
19
-------
4 *
BASIC RULES AND PROCEDURES
1. Minimize all chemical exposures. Do not taste laboratory chemicals and avoid any contact
with your skin. Smell chemicals only briefly and sparingly if necessary to identify them.
2. Don't underestimate risks. Use general precautions for handling all lab chemicals and use
special precautions for chemicals that present special hazards.
3. Provide adequate ventilation. The best way to prevent exposure to airborne chemicals is to
use hoods and other ventilation devices.
4. Usevthe hood for operations which might result in the release of toxic chemical vapors or
dust. "* ' .
5. Use only those chemicals for which the quality of the available ventilation system is
appropriate. If hood operation is questionable, arrange for use of a more efficient hood for more
hazardous chemicals.
6. Vent apparatus which may discharge toxic chemicals (vacuum pumps, distillation columns,
etc.) into local exhaust devices (hoods or direct discharge from building).
7. Do not allow release of toxic substances in cold rooms, clean rooms, incubators, or any
rooms which have closed, recirculated atmospheres.
8. Smoking is not allowed in the laboratory building. Do no eat, drink, or apply cosmetics in
rooms where lab chemicals are present. Wash hands before conducting these activities.
9. Do not store food or beverages in refrigerators or other areas where chemicals are stored.
Glassware or utensils which are used for laboratory operations should not be used to store or
consume food.
10. Do not use mouth suction for pipeting or for starting a suction.
11. Do not engage in horseplay or other behavior which might startle or distract another worker
or cause an accident.
12. Keep long hair and loose clothing confined. Wear shoes at all times, but sandals or
perforated shoes should not be worn where chemicals are being used or mechanisms work is
being done.
13. Wear safety glasses or goggles when hazardous chemicals are being handled in the
laboratory.
20
-------
4
14. Wear appropriate gloves when the potential for contact with toxic materials exists.
15. Avoid the use of contact lenses in the laboratory. If they are necessary inform the
supervisor and takes precautions.
16. Wear lab coat when working with chemicals. Remove lab coat if it becomes significantly
contaminated and decontaminate it if it would be hazardous to laundry personnel.
17. Wash areas of exposed skin before leaving the laboratory.
18. Avoid working alone in the building. Do not work alone in the laboratory if procedures
could be hazardous.
wv
19. Seek information about hazards, and plan appropriate equipment layout and protective
procedures prior to performing work.
20. Be alert to unsafe conditions and see that thev are corrected. Warn others of unsafe
practices when they are detected. Those working regularly in an area are most aware of possibly
unsafe conditions. Formal reports of unsafe conditions may be made on Form CD-351, "Report
of Possible Safety/Health Hazard".
21. Pets and unsupervised children are not allowed in the Laboratory.
CHEMICAL PROCUREMENT, DISTRIBUTION AND STORAGE
1. Before a chemical is ordered, investigate the potential for special hazards. Request
Material Safety Data Sheet (MSDS) from the supply source. If special hazards exist, less
hazardous substitutes should be considered.
2. New solvents will not be ordered when an excess amount that is suitable for use is already
in storage.
3. Before a substance is received, information on proper handling, storage and disposal should
be determined by those involved.
4. No container may be accepted without an adequate label. Labels must identify the contents
and hazard class, and should contain first aid information.
5. All incoming orders will be received and noted by the purchasing agent, Sandra West
(Karen Bauersfeld, backup), prior to distribution.
21
-------
6. Material Safety Data Sheets (MSDS) are normally mailed by the supplier to the Area
Safety Representative (ASR, Malcolm Hale) for each new shipment of chemicals received. The
ASR will check to see if the sheets are needed for our central MSDS file or if special precautions
are needed. Copies of the MSDS will then be passed on by Sandra to the person who ordered the
chemical. If you are missing a MSDS, first check our MSDS file in the library, then the LabLink
computer data base or request one from the supplier of the chemical.
7. Chemicals which generate acute hazardous waste (e.g. carbon disulfide) or unstable
chemicals (e.g. diethyl ether) will be identified and monitored from purchase to disposal. The
ASR will keep a log on future purchases of such chemicals. Wastes will be stored in appropriate,
labeled containers and not mixed with incompatible wastes. Smaller containers shall be
purchased, unless there is a short-term need for larger volumes.
8. Chemicals such as diethyl ether or dioxane, which may form dangerous peroxides, will be
protected from light and heat and kept for no more than 2 years. Small quantities of ether that
are being used in the lab will.be stored in an explosion-proof refrigerator.
9. Diethyl ether bottles will be dated when opened and disposed of within 6 months of
opening. The waste or outdated ether will be added to solvent wastes in a steel drum for
approved commercial disposal.
10. All samples of marine biological toxins shall be stored in the padlocked, explosion-proof
refrigerator in Room 216.
11. Environmental samples which appear edible will be marked as unsafe for human
consumption during refrigerated storage.
ENVIRONMENTAL MONITORING
Regular instrumental monitoring of airborne concentrations is not usually justified or
practical in laboratories but may be appropriate when testing or redesigning hoods or other
ventilation devices, or when a highly toxic volatile substance is stored or used regularly (e.g. 3
times/week).
Monitoring procedures and requirements for employees working with radioactive materials
are described in the Laboratory Radiation Safety Manual, available on the library safety shelf, or
see John Bemiss, the Radiation Safety Officer.
-------
HOUSEKEEPING, MAINTENANCE AND INSPECTIONS
1. Laboratory inspections for safety, housekeeping and chemical controls will be performed at
least once each six months. Informal inspections of specific areas may be carried out at any time.
2. Areas should be kept orderly and floors cleaned regularly.
3. Eyewash fountains will be inspected every three months. They should be operated at least
weekly to prevent rust buildup in the water.
4. Safety showers will be inspected and tested every six months.
5. Access to exits, emergency equipment and utility controls shall not be blocked.
* v
6. Hoods will be monitored, and modified if inadequate. If you suspect that a hood is
malfunctioning, notify your supervisor or the ASR.
7. The Area Safety Representative is primarily responsible for scheduled inspections.
Informal inspections and safety surveillance are primarily the responsibility of supervisors.
8. The results of scheduled inspections will be given to the appropriate supervisors along with
suggestions for correcting any and all violations.
MEDICAL PROGRAM
1. For medical emergencies dial 911 and take appropriate first aid measures.
2. For less critical injuries that require medical attention the employee will be transported to
James Island Medical Care (Dr. Costa) at 430 Folly Road. Form CA-16, authorization for
medical treatment, should be obtained from the Administrative Officer (Karen Bauersfeld) and
signed by the supervisor prior to leaving if possible.
3. When an employee develops symptoms associated with a hazardous chemical exposure, the
employee shall have the right to an appropriate medical examination free of charge.
4. In the event of an accident in the laboratory resulting in the likelihood of hazardous
exposure, an affected employee has the right to a medical consultation to determine the need for
a medical examination. The employee must first notify his/her supervisor and complete a Form
CA-16.
5. All medical examinations and consultations shall be performed by a licensed physician
without cost to the employee, without loss of pay and at a reasonable time and place. The
23
-------
employee may choose his/her personal physician, but the form (CA-16) authorizing medical
procedures must be completed first.
6. Any medical conditions resulting from hazardous chemical exposure which requires
treatment will be documented with details of the exposure and detailed statements from the
attending physician.
PERSONAL PROTECTIVE APPAREL AND EQUIPMENT
1. Protective apparel (gloves, safety goggles, aprons, etc.) shall be provided from operating
funds for the degree of protection required for substances being handled.
*
2. Ari*easily accessible drench-type safety shower shall be available in areas of higher risk.
3. An eyewash fountain shall be provided in high risk areas.
4. Fire extinguishers shall be readily accessible to all work areas.
5. Respiratory protection, a fire alarm and a telephone for emergency use shall be available
nearby to all work areas.
RECORDS
1. Accident records, including recommendations to prevent a recurrence, will be retained in
the Administrative Office (KB) and by the Area Safety Representative (MH). Employees
incurring an accident or illness believed to be work related will complete and file a Form CD-137
(Report of Accident/Illness) within 24 hours.
2. Chemical Hygiene Plan records will document that facilities and precautions are
compatible with current knowledge and regulations.
3. An inventory of Room 266, volatile solvents storage, will be recorded each year and
purchase restrictions applied to specific items if needed.
4. Records of hazardous chemicals stored in individual work areas will be posted inside the
entrance of each room and periodically updated. Once a year workers will be instructed by
Division Chiefs to update their lab inventories. Any chemical with one or more NFPA
(diamond) Codes rated at 2 or above may be considered hazardous.
5. Records of high-risk substances (e.g., toxins or HCN) will be maintained in the laboratories
by Project Leaders, and will include amounts on hand, amounts used and names of workers
24
-------
involved.
6. Medical records will be retained by the Laboratory in accordance with the requirements of
state and federal regulations.
SIGNS AND LABELS
1. All containers shall be labeled as to contents and date made or purchased. This includes
reagent bottles, chemical waste containers and receptacles. The labels on purchased chemical
containers, including hazard and first-aid information, shall not be removed.
2. Emergency telephone numbers will be posted for supervisors, emergency personnel and
workers tesponsible for equipment and facilities.
3. There will be prominent signs for locating fire extinguishers, exits, safety showers,
eyewash fountains and other safety and first aid equipment.
4. There will be warnings at areas or equipment where special or unusual hazards exist.
SPILLS AND ACCIDENTS
1. Procedures in case of fire, disaster or incidents that require evacuation are described in the
Charleston Laboratory Emergency Preparedness Plan, of which each employee should have a
copy. A reference copy is on the Safety/MSDS shelf in the Charleston Laboratory Library.
2. General rules for accidents and spills involving hazardous chemicals:
Eye Contact: Promptly flush eyes with water (15 minutes are recommended) and get
medical attention.
Ingestion: Check container label and MSDS for first aid recommendations. Get medical
attention as soon as possible.
Inhalation: Remove to fresh air and get medical attention.
Skin Contact: Flush thoroughly with water and immediately remove contaminated clothing.
3. In the event of a spill of hazardous material, take steps to contain and absorb it, if personal
safety permits. Inform others in the area who might be affected and/or could assist.
4. In the event of a large spill of a hazardous material such that cleanup requirements exceed
25
-------
* ->
the in-house capability, sound the fire alarm and have the building evacuated. Fire department
personnel having self contained breathing apparatus (SCB A) and protective equipment should be
informed of spill conditions upon their arrival.
5. In the event of a serious accident, do not move a worker with a possible serious injury
unless the immediate area is not safe. Call 911 for medical assistance, alert others in the
laboratory and provide first aid.
6. Use or transfer of hazardous chemical should take place in a properly operating hood if
possible. Have absorbent or neutralizing materials available in case of spills.
7. Contaminated absorbent materials will be placed in a plastic bag or other suitable container
and disposed of according to waste regulations.
8. All accidents or near accidents will be investigated and analyzed and the results distributed
to all who might benefit. The Administrative Officer (KB) and the Area Safety Representative
(MH) are required to maintain records of any accidents that involve injury or property damage.
TRAINING AND INFORMATION
1. Aim: To assure that all individuals at risk are adequately informed about the work in the
laboratory, its risks, and what to do if an accident occurs.
2. Supervisors will provide employees with information and training on hazardous chemicals
in their work area at the time of their initial assignment, and whenever a new hazard is
introduced.
3. Employees will be informed of the location and availability of the written hazard
communication program for the laboratory.
4. The Hazard Communication Program, including the required inventory list of hazardous
chemicals, will be available in the Library in the section with the Material Safety Data Sheets.
5. Employees shall receive initial Hazard Communication training and annual refresher
training in accordance with the requirements of the Hazard Communication Program (HCP).
Initial training shall include at least:
(i) Method and observations that may be used to detect the presence or release of a
hazardous chemical in the work area.
(ii) The physical and health hazards of the chemicals used in the work area.
26
-------
4
(iii) The measures employees can take to protect themselves from these hazards, including
work practices, emergency procedures and personal protective equipment.
(iv) Details of the Hazard Communication Program, including an explanation of labeling
systems and material safety data sheets.
6. The objectives of the HCP training are to enable employees to recognize and respond to
chemical hazards, become familiar with material safety data sheets and personal protective
equipment, and safely handle, use and store chemicals.
WASTE DISPOSAL
W*-
1. Aim: To assure that minimal harm to people, other organisms, and the environment will
result from the disposal of laboratory chemicals.
2. Room 266 is designated for organic solvent storage, and not for acids, alkalies or waste
chemicals. Moderate amounts (1 to 4 gallons) of waste organic solvents may be stored
temporarily (on floor to the right inside door) if they are properly labeled as to contents and if the
ASR is notified.
3. Waste solvents will be transferred from temporary storage or directly from labs to storage
drums in Container C behind the Laboratory building until pickup by a commercial waste
disposal company can be arranged. Transfers will be made by the ASR or by Laboratory
personnel in consultation with the ASR.
4. The ASR, with assistance from project personnel if needed, will assure that waste will only
be added to drums containing other wastes that are chemically compatible. The volumes and
compositions of wastes added will be recorded as the drums are filled. Each staff member is
responsible for keeping records of contents of temporary waste containers used in each
laboratory.
5. Wastes will be removed from the premises only by EPA certified waste disposal
companies.
6. Most waste acids or bases can be disposed of by neutralization in a sink and flushing down
the drain with excess water. Special care and sink within a fume hood should be used.
7. Radioactive waste materials will be handled in accordance with the Laboratory Radiation
Safety Manual and will be stored in the restricted area in Room 410 until they can be disposed
of.
8. All supplies used in toxin bioassays and extracted cell debris are collected of in approved
27
-------
biohazard disposal bags. All animal cages are to be treated with a strong sodium hypochlorite
solution and steam cleaned following toxin bioassays. When toxic wastes are neutralized by
autoclaving or bleach treatment, they will be placed inside a plain plastic bag without the
biohazard warning before they are placed in the dumpster for disposal.
9. Environmental samples that appear edible will be made unpalatable and marked "unsafe for
human consumption" prior to disposal.
28
-------
4 *
f ^ \ UNITED STATES DEPARTMENT OF COMMERCE
1t National Oceanic and Atmaapheric Administration
\./ NATIONAL MARINE FISHERIES SERVICE
Southeast Fisheries Science Center
Charleston Laboratory
PO Box 12607
Charleston, SC 29422-2607
June 9, 1995
MEMORANDUM FOR:
Staff
FROM:
Malcoim Hale
SUBJECTS
(1) HVAC recirculation
(2) Sharps Disposal
Two subjects are included in the interest of conserving paper:
(1) HVAC - Our new HVAC system is more energy efficient than the old one due to the fact that,
most of the laboratory air is recirculated and mixed with a smaller volume of fresh make-up air.
Because of this and as a normal safety precaution, please use a fame hood for anv volatile
chemicals. Do not transfer or work with volatile chemicals on open lab benches, because fumes
will be distributed to ail other work areas that are on the same air handler system.
(2) Sharps Disposal - Disposable hypodermic needles, used blades and other "sharps" should be
accumulated in plastic "sharps" containers. These secure plastic containers can be added to the
biohazardous waste boxes in freezer C for later transportation and incineration. Even if your
sharps are not true biohazards, they should be disposed of in this way to avoid any
misunderstandings if they are found in normal trash containers.
29
-------
APPENDIX E
CHARLESTON LABORATORY
EMERGENCY PREPAREDNESS PLAN
OCCUPANT EMERGENCY
PROCEDURES AND RESPONSIBILITIES
April 1992
30
-------
4
EVACUATION INFORMATION
Persons authorized to order evacuation:
Laboratory Director - Sylvia Galloway
Administrative Office - Karen Bauersfeld
Bldg.. Maint. Supervisor - Robbie Meyer
Notification for Evacuation
i
Fire - Fire Alarm Bells
Explosion or Gas Leak - Public Address announcement
Suspicious Obiect - Public Address announcement
Reporting Site
All employees should report to the parking lot in front of the building near the flagpole.
Building Re-entrv
Re-entry to the building will be announced by the Laboratory Director or designated
official.
Drill Schedule
Before a quarterly fire drill, the Laboratory Director and the Fire Department must be
notified - other staff will not be notified.
31
-------
EMERGENCY PERSONNEL
Designated Official/ Laboratory Director - Sylvia Galloway
Occupant Emergency Coordinator/ Chief. Office of Administration
Karen Bauersfeld
Team Coordinator/ Area Safety Representative
Malcolm Hale
Area Monitors
Primary/Alternate
Debbie Braddock/Karen Bauersfeld
Malcolm Hale/Jon Ahlquist
Tod Leighfleld/Bemie Lanoue
Al Fortner/Scott Sivertsen
Greg Mitchum/
Laura Webster/John Bemiss
Carl Kinerd/Tom Edwards
Robert Roberts/Aaron Dias
Handicapped Monitor
Karen Bauersfeld
Damage Control Team
Robbie Meyer
After Hours/Weekend/HoHdav Team
Paul Bauersfeld (Coordinator)
John Babinchak
Alan Fortner
Robert Roberts
Malcolm Hale
Area
Administration (Rm 100-118)
Open Office Area (Rm 200-214)
Biotoxin/Toxicol. (Rm 215-224)
Back Hall, West (Rm 225-230)
Back Hall, East (Rm 231-247)
Microbiology (Rm 248-257)
ADP/Office Area (Rm 258-260)
Processing et al. (Rm 400-426)
Home 795-0586
795-8738
795-7716
795-8553
795-0267
32
-------
i
FIRE EVACUATION PROCEDURES
1 Upon detection of a fire, regardless of the size, do the following immediately and in
sequence, unless there are others available to permit simultaneous execution.
Activate fire alarm - located in rooms or hallways as identified by signs
On hearing the fire alarm the receptionist will not attempt to silence the alarm.
She/he will ascertain from the annunciator panel in which one of the five zones
the alarm is coming from and transmit this information over the paging system.
Attempt to extinguish fire immediately, if possible - Using fire extinguishers
located throughout the building, immediately attempt to extinguish the fire if it is
safe to do so. If this attempt fails, or is not possible due to the size or nature of
the fire, leave the building.
2. Upon hearing the fire alarm, the personnel will:
Stop work immediately. Terminate all telephone calls, conferences, meetings,
and if time permits, close all doors, turn off lights, and other equipment except
those marked "Do Not Turn Off'. Do not attempt to remove personal belongings
or records.
Evacuate the building. When area/zone in which the fire is located is known,
the receptionist will make the following announcement over PA system:
"ATTENTION: - There is an alarm in zone (give #). Please leave
the building via the nearest exit, avoiding zone (give #). Repeat.
There is an alarm in zone (give #). Please leave the building via
the nearest exit, avoiding zone (give #)."
Area monitors should take their stations and assist in evacuating personnel. Check to see
that lights are off and doors are closed. If alarm sounds when you are far from your area,
leave the building by the nearest exit.
During evacuation, do noj run, push, shout or congregate, and limit talking to a
minimum. Move in a swift, orderly manner to the nearest exits (marked by red
lighted exit signs) avoiding if possible the zone where the fire is located. Room
236 has a fire exit door and Rooms 243 and 246 have fire escape windows. Large
or heavy objects (e.g., chair) may be used to break through any window, in the
event the hallways are impassable, however, the main hallway is to be used as the
first choice for evacuation.
Upon exiting the building, move to the parking lot in front of the building near the
flag pole, report to your supervisor and remain there until advised that it is safe to
return to the building.
33
-------
All driveways and roadways must be kept clear for the use of emergency
apparatus and fire equipment. Under HQ circumstances will personnel be
permitted to move any vehicles, unless otherwise directed.
3. Return to the building only after an official 'all clear' is given.
VISITORS AND HANDICAPPED EVACUATION INSTRUCTIONS
Visitors
Visitors pose a different problem which must be dealt with on the spot, according to the
emergency conditions. In this respect, all employees are expected to serve. In the event the
evacuation alarm is sounded, the employee with whom the visitor is visiting, or the employee
closest torthe visitor will immediately inform the visitor that the alarm requires the evacuation of
the building and that they will be assisted in this evacuation. If the visitor cannot leave the
building on his/her own power, the employee will instruct the visitor to remain in place while
they (the employee) go for assistance. The employee will go immediately to the Area Monitor,
who will be located in the vicinity of the exit of that area, and report the exact nature of the
problem. The Area Monitor will obtain the necessary assistance needed while the reporting
employee returns to the handicapped person.
Handicapped Employees
Area Monitors will be made aware of handicapped employees in their respective areas
and the nature of their handicap. A method of evacuation will be planned, by the Area Monitor,
consistent with the handicap, prior to any emergency. Handicapped Monitors will be specifically
assigned to each handicapped employee. Each handicapped employee will be fully advised of
their individual evacuation plan by the Handicapped Monitor.
34
-------
SOUTHEAST FISHERIES SCIENCE CENTER
tQAEWQBOCS
2?ssซfฎsliii;
Reproduced from
bast available copy
-------
BOMB THREAT OR OTHER SUBVERSIVE ACT
If a bomb threat against SEFSC Charleston Laboratory is received by an employee, it
shall be reported immediately to the Laboratory Director or his designee for appropriate action.
The Laboratory Director will direct the evacuation of the building. The following personnel
have been designated to act for NOAA/NMFS in determining course of action to be taken for
employees located in SEFSC; this listing is in descending order of priority:
Laboratory Director - 8525
Administrative Officer - 8561 (will contact the Acting Director, if LD is unavailable)
When the decision is made to evacuate the building, the Laboratory Director or his representative
will: **ฆ
Order immediate evacuation of the building by notifying the Administrative Officer.
(NOTE: The Fire Alarm system will not be used for evacuation during a bomb threat.^
Make the following announcement over the P.A. system:
"Attention Staff. The Laboratory has received a bomb threat. Please turn off all
unnecessary equipment including gas, water, and electricity, close doors and leave the
building by the nearest exit. Go to and remain in the front of the guard shack, just inside
the main gate, until notified otherwise. This is not a test. Do not stay in the building."
Repeat.
The Laboratory Director or his designee will call the Charleston County Police and Fire
Department (dial 911).
No evacuation order shall be given without approval of the Laboratory Director.
However, when there is an immediate danger to persons such as an actual fire or
explosion, the premises shall be evacuated at once, without consultation, by sounding the
fire alarm system in the building and notifying Charleston County Fire Department (dial
911), the following normal evacuation procedures.
Special Instructions to all Employees
If you receive a call from a person stating there is a bomb in the building:
Keep the caller on the line as long as possible. Ask the caller to repeat the message.
Record (write) every word spoken by the person making the call.
If the caller does not indicate the location of the bomb or the time of possible detonation,
the person receiving the call should ask the caller to provide the information.
36
-------
Pay particular attention to any strange or peculiar telephone noises such as motors
running, background music, and any other noises which might give even a remote clue as
to the place from which the call is being made. Listen closely to the voice (male -
female), voice quality, accents or speech impediments. It may be advisable (if time
permits) to inform the caller that the building is occupied and the detonation of a bomb
could result in death or serious injury to many people.
Immediately after the caller hangs up, the person receiving the call should report this
information to the Laboratory Director.
In the event you see a suspicious-looking object during or before an evacuation, observe
the following:
- Do not disturb the object
-Notify the Laboratory Director (8525) or the Administrative Officer (8561) and inform
him/her of its location.
INTRUDER ON PREMISES
In the event an intruder(s) is/are observed on the premises during the working hours, call
the reception desk (8511) stating nature of intrusion, i.e., demonstration, vandalism, theft, bomb,
arson, etc. State whether the intruder has a weapon. The receptionist will relay your call to the
County Police and/or MRRI Security. During off hours notify the MRRI Security Office
(762-5044) or MRRI Law Enforcement (762-5018 from 7:00a - 9:00p and 1-800-922-5431 after
hours and holidays).
In no case attempt to apprehend an intruder. If possible record appearance of
individual(s), sex, race, clothes, hair, behavior, etc., and means of escape, on foot, car, truck,
bike, etc. Record license plate number if possible.
37
-------
U S DEPARTMENT OF COMMERCE
Technology Administration
National Technical Information Service
Springfield, VA 22161 (703) 487-4650
------- |