United States
Environmental Protection
Agency
Office Of Air Quality
Planning And Standards
Research Triangle Park, NC 27711
EPA-454/R-96-003 V
February 19%
Air
PROCEDURES FOR VERIFICATION
OF
EMISSIONS INVENTORIES
-------
EPA-454/R-96-003
PROCEDURES FOR VERIFICATION
OF
EMISSIONS INVENTORIES
U S. Environmental Protection Agtnc»
Region 5. Library (PL-12J>
77 West Jackson Boulevard, iZtn tarn
Chicago.lt 60604-3590
Emission Factor And Inventory Group
Emissions, Monitoring, And Analysis Division
Office Of Air Quality Planning And Standards
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
February 1996
-------
This report has been reviewed by the Office Of Air Quality Planning And Standards, U. S. Environmental
Protection Agency, and has been approved for publication. Any mention of trade names or commercial products
is not intented to constitute endorsement or recommendation for use.
EPA-454/R-%-003
-------
ACKNOWLEDGMENTS
This report has been developed over a period of three years under the direction of Mr. J.
David Mobley, Leader of the Emission Factor And Inventory Group, Office Of Air Quality
Planning And Standards, U. S. Environmental Protection Agency (EPA). Many of the ideas and
approaches discussed in this report were provided by members of the United Nations Economic
Commission For Europe's Task Force On Emission Inventories, and others on the Task Force
provided review and critical comment on earlier versions of this report. The dedication and
professionalism of the Task Force has made this report possible. Specifically, the author
recognizes the contributions of Mr. Tinus Pulles of the TNO Institute in The Netherlands, who
co-chaired the Verification Expert Panel, and Mr. Gordon Mclnnes, the Chair of the Task Force.
The author wishes to acknowledge the assistance of Dr. Graham Johnson of the CSIRO,
Division Of Coal And Energy Technology, North Ryde, New South Wales, Australia; and Dr.
Harvey Jeffries of the University Of North Carolina At Chapel Hill, on the development of
concepts for application of the Integrated Empirical Rate (IER) model to emission inventory
validation. Dr. Johnson and Dr. Jeffries provided insights to the use, of the IER technique for
inventory validation and encouragement for the concept of these applications. Figures 3 and 4
herein come directly from work completed through a collaboration between Drs. Johnson and
Jeffries and are reproduced with their permission. The author also wishes to acknowledge Mr.
David Kirchgessner and Ms. Rima Dishakjian of EPA, the principal investigators of ongoing
projects in that Agency that involve the use of remote measurement techniques for estimating
emission rates from industrial area and volume sources. Mr. Lee Beck of EPA is also recognized
for his contributions to the development and application of the Data Attributes Rating System,
which estimates emission inventory reliability and accuracy.
In many cases, pertinent examples of a specific approach or the application of such have
been provided in the text, to facilitate the transfer of verification ideas to other programs. The
author was responsible for selection of those examples and recognizes that there are numerous
other examples that have been overlooked or omitted entirely. Apologies are offered for these
oversights, and the fact that some examples are not referenced in this report does not imply that
those approaches or ideas are not of value. The fact that other approaches are available is merely
an indication that interest in emission verification techniques is increasing and will continue to
grow in future programs.
111
-------
CONTENTS
PAGE
DISCLAIMER ii
ACKNOWLEDGMENTS iii
CONTENTS iv
TABLES vi
FIGURES viii
BACKGROUND AND OBJECTIVES 1
INTRODUCTION 1
HISTORICAL PERSPECTIVE 2
CURRENT PERSPECTIVE 4
EMISSIONS ESTIMATION AND THE SCALES OF
AIR QUALITY ISSUES 5
CONCEPTUAL ISSUES OF EMISSIONS VALIDATION 8
DEFINITION OF TERMS 9
OBJECTIVES 12
ORGANIZATION OF THE REPORT 13
PRINCIPAL FINDINGS 16
DOCUMENTATION OF DATA QUALITY 24
ESTABLISH DATA QUALITY OBJECTIVES 25
METHODS AND ASSUMPTIONS 27
QUALITY ASSURANCE AND QUALITY
CONTROL PROCEDURES 29
INDEPENDENT THIRD PARTY REVIEW 30
APPLICATION OF THE DATA 32
ASSESSMENT STUDIES 32
REGULATORY APPLICATIONS OF INVENTORIES 33
Air Quality Modeling Applications 33
Spatial Requirements for
Modeling Applications 34
Temporal Requirements for
Modeling Applications 37
Chemical Speciation Requirements for
Modeling Applications 39
Verification of Modeling Inventories 40
iv
-------
CONTENTS (continued)
PAGE
Control Strategy Analyses 43
Verification of Inventories for
Regulatory Applications 44
RELATIONSHIPS TO AMBIENT DATA 44
COMPARISON OF ALTERNATE ESTIMATES 46
OPPORTUNITIES FOR DATA COMPARISONS 46
APPLICATION OF COMPARISON DATA 49
UNCERTAINTY ESTIMATES 52
CAUSES OF EMISSIONS UNCERTAINTY 53
APPROACHES FOR ESTIMATING EMISSIONS
UNCERTAINTY 55
Analysis of Uncertainty Assuming
Normally-Distributed Data 56
Analysis of Uncertainty Using
Distribution-Free Data 60
SENSITIVITY ANALYSES 61
DATA ATTRIBUTES RATING SYSTEM 61
An Example of a Data Quality Rating System 64
SUMMARY 66
GROUND TRUTH VERIFICATION 67
STATISTICAL SURVEY ANALYSES 68
MONITORING ANALYSES 71
Direct Source Testing 71
Indirect Source Testing 73
Ambient Measurements 77
REFERENCES 86
APPENDIX A SUMMARY OF THE DATA ATTRIBUTES
RATING SYSTEM A-l
Description of Attributes A-2
Measurement Attribute A-2
Source Definition Attribute A-3
Pollutant Specific Attribute A-4
Spatial Scale Attribute A-4
Temporal Scale Attribute A-6
Application of Data Attribute Rating System A-7
An Example of Application of DARS A-9
-------
TABLES
No. PAGE
1 EMISSIONS INVENTORY VERIFICATION APPROACHES
BY SOURCE TYPE 18
2 INVENTORY PRIORITIES AND DATA QUALITY
OBJECTIVES FOR THE 1985 NAPAP INVENTORY 26
3 DISTRIBUTION OF EMISSIONS IN THE 1985
NAPAP EMISSIONS INVENTORY BY EMISSION
FACTOR QUALITY RATING 28
4 SUMMARY OF DATA COMPARISON OPPORTUNITIES
IN EMISSIONS INVENTORY APPLICATIONS 50
5 EXAMPLE OF AN UNCERTAINTY CALCULATION 58
6 UNCERTAINTY ESTIMATES FOR NOX EMISSIONS IN THE
1985 NAPAP EMISSIONS INVENTORY 59
7 EMISSION INVENTORY UNCERTAINTY RATINGS 65
8 MONITORING TYPES, EXAMPLES, AND USES
FOR EMISSIONS INVENTORIES 72
A-1 SUGGESTED SCORING SCALE FOR THE DARS
MEASUREMENT ATTRIBUTE A-3
A-2 SUGGESTED SCORING SCALE FOR THE DARS
SOURCE DEFINITION ATTRIBUTE A-5
A-3 SUGGESTED SCORING SCALE FOR THE DARS
POLLUTANT ATTRIBUTE A-6
A-4 SUGGESTED SCORING SCALE FOR THE DARS
SPATIAL SCALE ATTRIBUTE A-7
A-5 SUGGESTED SCORING SCALE FOR THE DARS
TEMPORAL SCALE ATTRIBUTE A-8
VI
-------
A-6 EXAMPLE OF DARS APPLICATION FOR VOC LOSSES FROM
REFUELING OF GASOLINE AUTOMOBILES IN ONE
NONATTAINMENT COUNTY A-10
vn
-------
FIGURES
No. PAGE
1 EXAMPLE OF COUNTY TO GRID CELL LAND
AREA ALLOCATION 36
2 COMPARISON OF HOURLY ALLOCATION FACTORS FOR
AN ACTUAL ELECTRIC UTILITY BOILER 42
3 OZONE FORMATION AS A FUNCTION OF TIME OF DAY ... 84
4 OXIDATION AS A FUNCTION OF CUMULATIVE
SUNLIGHT INTENSITY 84
Vlll
-------
PROCEDURES FOR VERIFICATION OF EMISSIONS INVENTORIES
BACKGROUND AND OBJECTIVES
INTRODUCTION
The Task Force on Emissions Inventories was commissioned by the United Nations
Economic Commission for Europe (UN/ECE) in the spring of 1992. The Task Force, comprising
environmental professionals from 24 countries and six international organizations, agreed to a
three-year work plan to develop information requirements to prepare emissions inventories for
European countries and to establish procedures to ensure compatibility and transparency among
the inventories developed for these nations. The Task Force established eight Expert Panels to
address the principal needs in specific areas of interest to the Task Force. These Expert Panels
are listed below:
• The Strategic Overview Expert Panel
• The VOC Expert Panel
• The Ammonia Expert Panel
• The Expert Panel on Heavy Metals and Persistent Organic Compounds
• The Power Plant and Industry Expert Panel
• The Mobile Source Expert Panel
• The Marine Expert Panel
• The Verification Expert Panel
Each Expert Panel was charged with preparation of a contribution to The Atmospheric
Emission Inventory Guidebook to identify the issues and to summarize approaches to be followed
in emissions inventory programs in Europe. This report is a compilation of the information and
techniques discussed by The Verification Expert Panel. The report discusses some of the
principles of emissions inventory preparation and the techniques and procedures that can be
applied during the planning and development of emissions inventories to increase the accuracy and
reliability of emissions estimates. Methodologies that have been applied in previous emissions
inventory projects and additional proposed activities to validate the emissions inventory data are
discussed.
Although the information discussed in this report was prepared for application by the
nations participating in the UN/ECE Task Force, the issues and approaches discussed are relevant
to any emissions inventory development project. The summary of approaches and techniques can
be used to document data quality and to establish confidence in emissions estimates in State,
national, and global applications.
-------
An emissions inventory is the foundation for essentially all air quality management
programs. Emissions inventories are used in two primary applications by air quality managers.
The first use of emissions inventories is in assessments that identify the largest air pollution
sources in a region. The second use of inventories is in regulatory and policy making applications
including as input data for air quality models, and in the development, implementation, and
tracking of control strategies. Each of these uses provides information that is needed by decision
makers to establish priorities and to make effective air quality management decisions. A number
of approaches can be used to compile emissions inventories, and the selection of an appropriate
technique is largely dependent on the needs of the specific program, the personnel and funding
resources available for the program, and the schedule constraints that are placed on the program.
Because emissions inventory data are usually estimates, it is always desirable to provide an
assessment of the validity of the estimated emissions magnitudes as a component of emissions
inventory development projects. The emissions validation or verification process assists decision
makers in choosing appropriate regulatory options.
Public awareness of the health and other environmental problems associated with
industrialization have made air quality management problems a high priority in some European
countries where these issues have not been a high priority in the past. Many of the important air
quality issues are regional in scope and may require emissions inventories that cross country
borders. Inventory verification is particularly important for programs involving multiple countries
with different levels of resources for inventory development.
The purpose of this document is to provide guidance related to the development of
emissions inventory data and suggestions for procedures and techniques that can be used to assess
the validity of the emissions data included in the inventories. Since most emissions data are
estimates, it is often difficult to derive statistically meaningful quantitative error bounds for
inventory data. Frequently, it is possible, however, to provide ranges that bound the likely
minimum and maximum for an emissions estimate or to develop a qualitative data quality
parameter to indicate the relative confidence that can be associated with various estimates.
This report begins with a background discussion of the approaches for emissions inventory
development and the role of emissions inventory data in air quality management activities.
Considerations applicable in planning phases of inventory development efforts that can improve
the reliability of inventory data are discussed. The remainder of the report discusses specific
analyses that can be performed to assist in validating emissions inventories.
HISTORICAL PERSPECTIVE
Over the past two decades, the industrialized countries in North America and Europe have
cooperated on air quality management programs. Air quality standards, that set ambient
-------
concentration limits for air pollutants, are a cornerstone of these programs. Air quality standards
are established to protect public health, and to limit other impacts of air pollutants on agriculture,
structures and other economic interests. Coordinated research to identify appropriate emissions
control options were developed to reduce the amount of pollutants that are emitted to the
atmosphere, as a control strategy to achieve the ambient concentration limits. The following
discussion briefly summarizes the history of these programs in the United States as an example
of similar programs that have been implemented in many of the other industrialized countries.
Originally, the United States established air quality standards for sulfur dioxide (SO2),
-nitrogen dioxide (NO2), carbon monoxide (CO), total suspended paniculate matter (TSP), and
photochemical oxidant expressed as ozone (O3). The U.S. EPA provided guidance for the control
of air quality to achieve these ambient standards. Most of the oxidized nitrogen is emitted in the
form of nitric oxide (NO). Atmospheric reactions can rapidly convert NO to NO2, and emissions
control programs for NO2 were based on limiting emissions of total oxides of nitrogen (NQ ).
Ozone and the other primary photochemical oxidants are not emitted directly from sources but are
formed through a complex reaction sequence involving nonmethane hydrocarbons (NMHC) and
NOX in the presence of solar ultraviolet light. In the United States an ozone control policy, based
on the control of emissions of NMHC, was adopted. Many other countries have adopted similar
ozone control policies.
Since the original definition of the criteria pollutants, the nomenclature for NMHC has
changed to volatile organic compounds (VOCs), the PM standard has been changed to an ambient
concentration standard for paniculate matter of aerodynamic diameter less than 10 microns (PM-
10), and a NAAQS standard for lead (Pb) has also been established.
Throughout the 1970s and 1980s, air quality management in the United States, and other
countries, was perceived as a local problem, and local jurisdications were charged with the
responsibility for the development and implementation of air quality control programs. National
programs were established in most countries, to provide guidance and to support the local
agencies. In the United States, EPA conducted research on source-specific control techniques and
mobile source control to assist the local agencies in developing their air quality management plans.
Emissions control research focused on the emissions processes of individual source types.
This research led to the compilation of recommended source-specific emission factors for each of
the important air pollution species. The recommended emission factors developed in the United
States, were published in the document entitled Compilation of Air Pollutant Emission Factors,
Volume I, Stationary Point and Area Sources, AP-42.1 This document has been revised through
the years and is currently undergoing another revision; it will soon be released in its fifth edition.
Emission factor compilations similar to the AP-42 have been used by many countries to develop
emissions inventories.
-------
The procedures recommended for emissions inventory development generally involve the
application of a representative emission factor that relates the magnitude of air emissions to some
other operating parameter associated with the activity. The emission factors are specified in units
of pounds of emissions per unit of that specified activity. Some common examples include
emission factors that are expressed in terms of the number of tons of coal burned, the number of
barrels of oil refined, tons of chemical produced, tons of solvent consumed, square feet of surface
coated, vehicle miles travelled, density of biomass, number of employees, the population of a
geopolitical unit, and acres of harvested land. In general, the activity data required for emissions
inventories for industrial sources are known at the annual level with a great deal of confidence.
The activity data for consumer- based source categories, such as the use of personal hygiene
products, lawn and garden care products, and household consumer products, however, are
generally not as well known and are often estimated. Since a large amount of inventory data is
based on estimates, there is commonly a large amount of uncertainty in the final emissions
inventories.
CURRENT PERSPECTIVE
Unfortunately, air quality issues rarely involve analyses of annual emissions of a particular
pollutant. The effects of poor air quality are more commonly associated with repeated, short term
occurrences of high concentrations of pollutants. The most intractable of these issues in industrial
countries also involve complex interactions of emitted pollutants with atmospheric systems that
result in the generation of secondary pollutants. Examples of problems related to secondary
pollutants are ambient ozone that is formed through complex chemical interactions involving
VOC, NOX and solar ultraviolet radiation, and the production of acidic deposition species through
interactions of SO2, NOX, ammonia, and other alkaline components. Coordinatedairpollution
research over the past 20 years has led to the development of analysis tools including air quality
models and statistical data analysis techniques. These analysis tools require emissions data that
are at a finer level of spatial and temporal resolution than the data of the original estimates.
Emissions data often need to be formatted in a regular spatial grid pattern that represents the
region for which the inventory was developed. The emissions data are also often required at daily
or hourly levels of resolution. Finally, since the individual components that make up the VOC
and PM-10 often have different effects on secondary pollutant formation, the emissions estimates
must be resolved (speciated) to represent their chemical composition. Each of these steps of
resolution requires assumptions and estimates of the distribution functions, which are seldom well
known, thereby adding uncertainty to the emissions estimates that are ultimately used to evaluate
control strategies.
-------
The largest sources of air pollutants were considered for control in the early efforts of air
quality management. Overall, control programs for the major sources have resulted in significant
reductions of emissions, and the existing air quality in the countries with aggressive control
programs is considerably improved relative to the conditions that would be present if no controls
had been implemented. The effects of increased population and expanded industrial development,
however, continue to produce emissions that cause air quality standards to be exceeded and public
health to be threatened. Inventory-based control plans have been successful in reducing the
number of urban areas that violate air quality standards, and in terms of the number and severity
of violations of air quality standards in many locations. This success has not been universal,
however, and additional pollution control is necessary to achieve complete success. As a result,
air pollution control programs must be expanded to include small sources and other activities that
have not been regulated in the past. These requirements place increased emphasis on the
preparation of complete and accurate inventories of the emissions sources in urban areas that are
in violation of air quality standards. Activities to promote pollution prevention, recycling, and
conservation are key components in programs to identify achievable emissions reductions from
the previously unregulated sources and source categories.
The needs related to allocating emissions data to the resolution required by air quality
models and the need for improved emissions estimates for a larger number of small sources have
increased the demand for reliable emissions estimation techniques and for appropriate verification
methodologies. Emissions inventory verification programs will not only provide information
useful to air quality decision makers, but will also help to prioritize future activities for inventory
improvements.
EMISSIONS ESTIMATION AND THE SCALES OF AIR QUALITY ISSUES
The concept of emission factors and the fact that almost all emissions inventories are based
on the application of emission factors has already been presented. There are, however, several
approaches for estimating emissions that rely on the use of emission factors. All of the approaches
can be categorized as being either top-down or bottom-up. The top-down process involves the
estimation of the inherent activity at a highly aggregated level of resolution such as at the country,
or provincial level. In the bottom-up approach attempts are made to estimate emissions for
individual sources and other specific human activities. These individual estimates are then
summed to represent the total emissions. The choice of the specific emissions estimation approach
to be used depends on the scope of the project, the data requirements, the resources and schedule
constraints, the amount of data available, and the level of detail represented in the available data.
-------
Obviously, it is preferable to produce a detailed bottom-up inventory for application to
modeling activities and area-wide control strategy development. The amount of uncertainty
associated with the modeling emissions inventories is decreased if source-specific information can
be obtained. Similarly, if specific source operating conditions are known, the benefits and costs
of emissions control can be evaluated in more detail than is possible if only a national average
emission estimate is available. The methods and techniques that can be applied to emissions
verification efforts are also dependent on the type of methodology used to develop the inventory.
Emissions from large point sources, mobile sources, dispersed area sources, and biogenic
and other natural sources are all important in the atmospheric processes that result in air quality
problems. The needs associated with emissions inventory development projects depend on the
type of air quality program to which they will be applied. National assessment of the major
sources would require only annual estimates of the total contribution of pollutants by major
industry types, whereas a study to evaluate the health impacts of a proposed industrial facility
would need hourly operating rate data and specific pollutant emissions anticipated from the
proposed industrial process. The degree of accuracy and precision, and the type of emissions
inventory verification program that is applicable to these problems is in turn related to the
inventory needs.
Historically, air quality problems have been addressed through pollutant specific regulatory
responses that are implemented and enforced independently. The general approach to addressing
these issues was to identify the specific health or environmental effect, evaluate the pollutant
emissions that contribute to that effect, and implement economically achievable controls to reduce
the emissions of the pollutants. These problems are now known to involve interactions among
primary source emissions, atmospheric processes, long-range transport and transformation, wet
and dry deposition, and ecological feedback systems. Air quality problems can be thought of as
occurring in the four regimes listed below.
• Local, source-specific problems occur over a period of hours to a
day, during which primary pollutants emitted from a source have
acute and cumulative toxicity effects on the population in the
immediate area. The pollutants of interest in this regime are not
influenced by chemical reaction.
• Urban area problems involve the collection of emissions sources in
an urban area and the interactions of those emissions with one
another. These interactions occur under the influence of local
meteorological conditions to produce secondary pollutants that have
-------
health-related implications to the population in the urban and
surrounding areas. The primary examples are violations of the
ambient ozone and CO standards. These issues involve
photochemistry and other complex atmospheric chemistry processes
that take place over a period of one to three days.
Regional air quality problems are those that occur on the scale of
major weather systems and generally involve transport conditions
where the effects can occur at distances of up to 1,000 kilometers
from the source areas. These types of problems occur over periods
of several days to more than a week and may or may not involve
atmospheric chemistry. The examples are elevated rural ozone
concentrations, air toxics deposition, and acid deposition problems.
Global air quality problems occur over periods of years. These
problems involve the collective emissions of species from all
countries. The compounds of interest in global problems are
subject to atmospheric removal processes that are slow relative to
the global emissions rate. Global climate change and stratospheric
ozone depletion are examples of these problems.
Often the individual problems included in each of these regimes are related to one another.
In particular, the oxidation chemistry associated with urban ozone formation processes can affect
and are influenced by several of these other processes. Obviously, regional rural ozone
concentrations are affected by the emissions of VOC, NOX, and CO that initially occur in and
around the urban centers. The same species that oxidize VOCs in the ozone formation process
also affect the oxidation of sulfur and nitrogen compounds in the formation of acid deposition
species. The photochemical processes that occur on urban scales can also affect the global
distribution of radicals and other intermediate species that can influence the concentrations and
removal rates of species active in the global problems.
All of the issues concerning emissions inventory preparation and the uses of emissions
inventories summarized above influence the decisions that affect the approach that will be followed
to validate the emissions inventory. An emissions verification program can range from simple
efforts that compare total national emissions per person to those of other similar countries, to
detailed emissions range estimations for specific sources. Each is useful to add credibility to the
estimates and to prevent inappropriate applications of the resulting data.
-------
CONCEPTUAL ISSUES OF EMISSIONS VERIFICATION
The preceding discussion emphasized the importance of appropriate emissions inventory
estimates to the success of air quality management and decision making. Of equal importance are
the procedures that will be included in the effort to ensure the accuracy, completeness, and
representativeness of the emissions data. These activities are collectively referred to as quality
assurance and are implemented through a quality assurance plan. Each air quality analysis effort
has specific emissions inventory needs that are related to the scale and objectives of the program.
In most applications, it is highly desirable to include in the program design methodologies and
procedures to ensure and document the quality of the emissions data. The effort that is
appropriate for these quality assurance and data verification components is dependent on many
factors. Primarily, considerations of the objectives of the program and the resources available for
the program influence the decisions on what type of quality assurance and verification effort are
reasonable.
An emission verification program is inherently related to quality assurance procedures and
objectives, but the concept of emissions verification extends beyond quality assurance activities.
Quality assurance procedures are defined and followed to ensure that the data that serve as the
foundation for the emissions estimates are the most representative, complete and meaningful data
available for the intended application. An emissions validation or emission inventory verification
program is performed to test how well the completed emission inventory supports correct decision
making in air management planning. As such, a verification program will provide data that can
be applied to demonstrate that the final emissions inventory is useful for its intended application,
rather than to provide data to test the accuracy of emissions or operating data for individual
sources.
This report presents the elements that should be considered in the development of an
emissions inventory and in the quality assurance aspects of inventory development efforts to
establish the validity of the emissions data. Whenever possible, examples of some quality
assurance activities that have been performed in emissions inventory programs in the United States
and Europe have been presented. Some of these examples are derived from the experience
obtained during the development of the National Acid Precipitation Assessment Program
(NAPAP) emissions inventory and in the preparation of emissions inventories used by the State
agencies and EPA in developing air quality control plans for criteria pollutants.2-3
This report is intended to represent a comprehensive discussion of verification concepts and
approaches that can be applied before, during, and after the development of emissions inventories
to improve the usefulness of emissions data. Ideally, some of the verification techniques discussed
here should be completed by the inventory developer and presented along with the inventory
databases to establish reliability and usefulness. If program resources and time schedules prohibit
8
-------
the completion of these tasks by the developer, users and/or third party reviewers can complete
these tasks with equal credibility if the procedures followed in the development of the inventory
are adequately documented.
These approaches include some techniques that have been widely applied in past programs
and others that are experimental and have only been applied in selected applications. The concept
of a coordinated emissions verification program using a mix of routine and experimental
techniques together is itself relatively new. Such a program should strive to ensure that completed
inventories not only include high quality emissions estimates but also that those estimates are
applicable and reliable for the intended purpose of the programs they support.
As emissions verification programs are applied in additional emissions development
programs, the strengths and weaknesses of traditional technologies and approaches will become
well defined. Researchers should be aware of the weaknesses of common techniques and promote
the application of innovative techniques and approaches to overcome these weaknesses. New
computer systems and mathematical analyses should be considered in all coordinated verification
efforts. Examples of systems include Geographic Information Systems (GIS), statistical
applications, graphic presentations, Monte Carlo simulations, and relational databases. Examples
of innovative mathematical analyses that should be considered for emissions development and
verification include fuzzy logic, artificial intelligence systems, chaos theory and fractal geometry.
DEFINITION OF TERMS
It is useful to define some key terms relative to the inventory verification process that will
be used throughout the remainder of this report. Often individual inventory developers and/or
inventory users will define these terms in the context of their own applications or goals.
Therefore, the following definitions of key terms are provided to promote common usage in the
context of this report.
Accuracy Accuracy is a measure of the truth of a measurement or estimate. The term
accuracy is often used to describe data quality objectives for inventory data,
however, accuracy is hard to establish in inventory development efforts
since the truth for any specific emission rate or emissions magnitude is
rarely known.
Precision The term precision is used to express the repeatability of multiple
measurements of the same event. In experimental applications a
measurement or measurement technique could have high precision but low
accuracy. The term precision is also used to describe the exactness of a
measurement. The term precision is not well suited for use in emissions
inventory development.
-------
Confidence
Reliability
Quality Control
Quality Assurance
Uncertainty
The term confidence is used to represent trust in a measurement or
estimate. Many of the activities discussed in this report are designed to
increase the confidence that inventory developers and inventory users have
in the databases. Having confidence in inventory estimates does not make
those estimates accurate or precise, but will help to develop a consensus
that the data can be applied to problem solving.
Reliability is trustworthiness, authenticity or consistency. In the context of
emissions inventories reliability and confidence are closely linked. If the
approaches and data sources used in an inventory development project are
considered reliable, then users will have an acceptable degree of confidence
in the emissions data developed from those techniques.
Quality control activities are those procedures and tests that can be
performed during the planning and development of an inventory to ensure
that the data quality objectives are being met. These activities may include
criteria tests for data on operations, completeness criteria, or averaging
techniques for use in developing default parameters. Quality control
activities are generally applied by the developers.
Quality assurance describes the activities that are completed after the
development of a product, usually by an independent party to verify that
data quality objectives were met and that the product conforms to
specifications. In experimental programs, audits with standard instruments
and standard measures are used to establish the reliability of the
experimental procedures. In emissions inventory development, however,
few such standards exist. One effective activity discussed in this report is
the use of an independent review team of experts to monitor the
developments as progress is made on the inventory. The review team can
identify alternate approaches and further documentation to enhance the
credibility and reliability of the emissions estimates developed.
In the context of emissions inventory development, and in general use in
this report, quality assurance is used to represent the sum of activities that
are implemented to ensure the collection and presentation of high quality
data.
Uncertainty is a statistical term that is used to represent the degree of
accuracy and precision of data. It often expresses the range of possible
values of a parameter or a measurement around a mean or preferred value.
In some applications involving emissions inventory preparation, it is
possible to describe in statistical terms the relative accuracy of an estimate
10
-------
Validation
Verification
and ultimately provide a preferred estimate or central value and a percent
range that bounds the actual value. Such opportunities are frequently
limited to sources that have requirements for extensive monitoring, through
continuous emissions monitors, to verify emissions rates. More often,
however, the data that is available is insufficient to develop statistically
based quantitative measures of the data accuracy. In these cases, subjective
rating schemes are often used to describe the relative confidence that is
associated with specific estimates.
In the context of this emissions verification report, uncertainty is used to
represent any of several techniques or procedures that can be applied to
establish a ranking or numerical scale to compare the reliability of and
confidence in the emissions estimates. In their simplest forms, such
ranking procedures are subjective evaluations that reflect the accuracy or
reliability of estimates based on the opinion of the developer. In other
applications the evaluation is guided to a specific attribute of the data. For
example, the completeness, coverage, or specificity may be of special
importance and developers may be asked to rate the final emissions
estimates relative to one or more of these components that can affect the
quality of the estimates.
Validation is the establishment of sound approach and foundation. The
legal use of validation is to give an official confirmation or approval of an
act or product. Validation is an alternate term for the concept of
verification as used in this context.
The term verification is used to indicate truth or to confirm accuracy and
is used in this report to represent the ultimate reliability, and credibility of
the data reported.
In the context of this report verification refers to the collection of activities
and procedures that can be followed during the planning and development,
or after completion of an inventory that can help to establish the reliability
of the inventory for the intended applications of that inventory. In this
context, the representativeness of the final data for the intended applications
is of more importance than the absolute accuracy of the final emissions
estimates. The procedures identified as verification activities will be
applied to establish confidence that the data are sufficient in terms of
coverage, completeness and reliability to guide decision makers to effective
policy options.
These verification approaches can be used to understand the strengths and
weaknesses of completed inventories relative to the desired applications of
11
-------
the data. In this context, verification procedures should be useful in
directing research to improve the underlying data or procedures used to
develop emissions estimates in future programs.
Transparency In the context of this report, transparency is used to represent the condition
of being clear and free from pretense. The use of the term implies that data
collected and reported by different agencies will be similar and, therefore,
easily understood by other parties and comparable to the data presented by
the other parties.
i,
Compliance Compliance is the act of conforming or yielding to a specified norm or
protocol. In the inventory development process, compliance may indicate
conformity to development protocols or international agreements. In this
sense, the compliance issue can be thought of as verification that these
established and agreed norms are achieved. The concepts of verification
discussed in this report, however, are not intended to support the idea of
compliance to norms or international protocol.
OBJECTIVES
This report is intended to provide a comprehensive summary of procedures and techniques
that can be applied to emissions inventory development projects to help establish the accuracy and
reliability of the emissions data for the intended applications of the inventory. The techniques
discussed range from routine quality assurance activities that are a mandatory part of any technical
effort to specific experimental and field measurement studies that can be applied to specific parts
of inventories. Each of the techniques can be used independently or in combination to satisfy the
demands for establishing credibility and reliability of emissions data for essentially any purpose.
The selection of the specific procedures that could be useful in any given inventory project are
dictated by the intended applications of the project, the resources available for the development,
and the time constraints. The objectives of this report are summarized below:
• Present a comprehensive list of possible methods for application to emissions
inventory verification procedures.
• Provide examples of the application of these techniques from past projects or
proposed technical applications.
• Document the background information for these techniques to provide a roadmap
for users to obtain additional more specific information.
• Present a summary list of possible techniques associated with broad emissions
source categories and to provide some guideance on the priority of these various
techniques if resources and time constraints were not an issue.
12
-------
ORGANIZATION OF THE REPORT
The report begins with an overall summary of the findings of the project. This section is
organized in matrix format with broad categories of emissions verification approaches matched
to broadly defined source categories. The matrix entried identify those procedures that have been
applied or proposed for application for verification of emissions for sources in the indicated source
categories.
The remainder of the report is organized to follow the approach agreed to by the
Verification Expert Panel of the Task Force on Emissions Inventories commissioned by the
UN/ECE. The five primary program elements listed below were identified in the preliminary
meetings of the Verification Expert Panel.
• Documentation of Data Quality
• Application of the Data
• Comparison of Alternative Estimates
• Uncertainty Estimates
• Ground Truth Verification
Each of these program elements provides information that is useful to promote the
development of high quality and representative emissions estimates. The information developed
through these activities can be applied to establish confidence in the final inventory data. The
information can also be applied in programs to establish comparability between different databases
and to identify the strengths and weaknesses of the resulting inventory. The first two of these
elements involve planning and coordination, and the last three represent specific activities that can
be included in the overall inventory development and application programs. Specific activities that
can be considered and implemented as components of emissions inventory development programs
include comparisons of alternative estimates or estimation techniques, assessments of emissions
uncertainty and reliability, and ground truth verification studies that rely primarily on field and
laboratory measurement programs.
Documentation of data quality includes the development of a set of data quality objectives
and criteria that will be applied to ensure that those objectives have been met. These efforts begin
with the planning of the program and are directly related to the purpose of the study, the resources
available to complete the study, and the schedule constraints imposed on the study. These
activities require the definition of needs, criteria for accepting data sources, contingency plans for
the use of backup approaches, and the definition of a quality assurance and quality control
(QA/QC) protocol; activities may also include provisions for a third party review of the inventory
development program and resulting inventories.
13
-------
Considerations of the applications of the data and the needs of the inventory user
community are the primary factors that control the requirements of any inventory development
project. Often it is necessary to compromise based on the availability of data and other technical
resources. Detailed planning and program flexibility are keys to the success of many inventory
development programs. The applications of the inventory define parameters such as the
geographic scope; the spatial, temporal, and species resolution of the databases that are to be
generated in the program; and the approaches that can be implemented to achieve those needs.
The types of program applications that influence inventory needs include source permitting and
source-specific emissions increments, area-wide assessment studies to determine the relative
magnitude of source category groups, air quality modeling analyses, source-receptor modeling
analyses including relative VOC/NOX and CO/NQ ratios, and policy analyses to establish,
implement, and track urban or regional control strategies.
Often multiple approaches can be applied to develop an emissions inventory. While the
different approaches identified for any specific project may not always meet the statistical
requirements that represent independent methods, it can generally be assumed that convergence
of multiple approaches to a common emissions estimate contributes to the confidence that users
have in the resulting inventory. Such analyses are also very useful in establishing priorities for
any further efforts that are anticipated to improve second or third generation inventories. It is
always desirable to establish ranges to represent the likely maximum and minimum emissions
magnitude in addition to the mean or preferred estimate. These ranges can provide an assessment
of the uncertainty or reliability of the emissions estimates. These analyses can assist in
understanding the weaknesses of the resulting inventory and in interpreting the results of other
technical- or policy-related applications using the inventory data. Emissions data are almost
always based on estimates of at least one and often more than one critical parameter. Therefore,
standard types of propagation of error analyses are normally not applicable to emissions data. It
is possible in many cases, however, to provide reasoned bounds on the resulting inventory data.
These types of analyses are particularly important for sensitivity studies associated with modeling
applications. Sensitivity studies are used to determine the effect on control strategy conclusions
for the different extremes represented in the inventory. Examples of these types of analyses will
be discussed in more detail later.
The ideal emissions inventory would be based on direct measurement of all sources of
interest in a program. This approach is obviously impossible in all but the smallest scale and most
specific of cases. Therefore, verification of emissions estimates is sometimes related to
comparison of the estimates with a small set of measurements or through some other innovative
comparison technique. The Verification Expert Panel selected the term "ground truth verification"
to represent these types of analyses. There are many opportunities for such analyses. Some
14
-------
examples that have been applied previously in programs in the United States, and some techniques
that have been suggested for application to future programs will be discussed later in this report
in some detail.
15
-------
PRINCIPAL FINDINGS
This report is the culmination of an effort that has been ongoing for over two years. The
concepts and techniques for emissions inventory verification discussed in this report have been
developed through interactions with a large number of experts in air quality and emissions
inventory studies from the United States and European countries. The ideas expressed here,
therefore, cover a wide range of emissions verification opportunities including simple approaches
and some rather sophisticated techniques that rely on emerging technologies.
The selection of the most appropriate technique or combination of techniques for
application to any given inventory development project is dependent on several factors including
the nature of the projects that the inventories are expected to support, the visibility of those
projects, whether or not the projects involve cooperative efforts of several countries or other
geopolitical entities, budgets and other resources available for the project, and schedule
constraints. It is recognized that the more experimental and research oriented concepts discussed
in this report will not be used in the majority of inventory development projects. Increases in the
sophistication of modeling and data manipulation techniques, and the need for more detailed
control strategies that target less traditional source categories, however, do suggest that more
accurate and flexible emissions inventory data will be required in the future. Those future
inventory development programs could benefit from the application of many of the approaches
discussed in this report.
The advantages of high quality and accurate emissions data in air quality management
projects is obvious, but it is perhaps more important to mention the disadvantages of using poor
quality data with limited credibility. At the least, any inventory development project should
include a comprehensive list of data quality objectives and criteria for evaluating the resulting
inventory data against those objectives. It is also recommended to include as a minimum a simple
data quality rating approach and to document the assumptions used to derive the qualitative
ratings. Without such an analysis, the results of the inventory and the analyses that depend on the
inventory can be questioned. Regulatory initiatives and policy options can be adversely affected
if there is a general perception that the underlying data is incorrect or unreliable. Obviously,
more rigorous statistical uncertainty and measurement based verification approaches are preferred
and should be considered whenever they are possible. All inventory developers are also
encouraged to share their experiences with emissions inventory verification approaches so that the
wider emissions inventory community can benefit from those experiences.
16
-------
The procedures and techniques discussed in this report are intended to support inventory
development programs, and efforts that apply emissions inventory data to regulatory or policy
making activities. It is possible to apply similar emissions inventory verification activities to
demonstrate compliance with international protocol or agreements. Although this more universal
application is not addressed specifically in this report, many of the principles outlined here could
be used in such compliance demonstrations.
The most significant finding of this work is simply that there are many possible methods
and techniques that can be applied to establish the reliability of emissions inventories and to help
understand the weaknesses in inventories. This report presents many of these, although there are
undoubtedly some techniques and methods that have been overlooked. The report presents
examples of the application of some of these approaches to give inventory developers some ideas
of how these approaches can be applied and the benefit that can be derived from including such
procedures in their inventory development programs. To the extent possible work describing the
approaches discussed has been referenced to assist users in locating additional information on the
specific techniques and their application.
A summary of recommended emissions verification approaches is provided in Table 1.
This summary is prepared in a matrix format, linked to the second level of detail represented in
the Selected Nomenclature for Air Pollution (SNAP90) source category listings that have been
adopted by the CORINAIR 1990 project and selected as an organization approach for use in the
UN ECE guidebook. These broad categories should be familiar to any researchers interested in
the development of emissions inventories and the matrix format provides for direct transfer to any
particular distribution of source categories or sector descriptions associated with an emissions
database.
This matrix addresses the priorities for the adoption of specific verification techniques in
a general sense. It reflects those cases where specific inventory verification concepts have been
applied or are proposed for application in actual programs. Although some approaches are more
suited to particular source categories, nearly all of the approaches could be applied to any source
category.
17
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
su
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
PUBLIC POWER, COGENERATION AND DISTRICT HEATING PLANTS
Public Power and Cogeneration Plants
District Heating Plants
X
X
X
X
X
X
X
X
X
X
X
X
X
COMMERCIAL, INSTITUTIONAL AND RESIDENTIAL COMBUSTION PLANTS
Comm., Inst. and Res. Combustion
X
X
X
X
X
INDUSTRIAL COMBUSTION
Combustion in Boilers, Gas Turbines, and Stationary Engines
Process Furnaces Without Contact
Processes With Contact
X
X
X
X
X
X
X
X
X J
X
X
X
X
X
X
X
X
PRODUCTION PROCESSES
Processes in Petroleum Industries
Processes in Iron and Steel Industries and Collieries
Processes in Non-Ferrous Metal Industries
X
X
x J
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
oo
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE (continued)
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
su
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
PRODUCTION PROCESSES (continued)
Processes in Inorganic Chemical Industries
Processes in Organic Chemical Industries (bulk production)
Processes in Wood, Paper Pulp, Food and Drink Industries
and Other Industries
Cooling Plants
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
EXTRACTION AND DISTRIBUTION OF FOSSIL FUELS
Extraction and 1s' Treatment of Solid Fossil Fuels
Extraction, 1" Treatment and Loading of Liquid Fossil Fuels
Extraction, 1" Treatment and Loading of Gaseous Fossil Fuels
Liquid Fuel Distribution (except gasoline distribution)
Gasoline Distribution
Gas Distribution Networks
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE (continued)
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
su
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
SOLVENT USE
Paint Application
Degreasing and Dry Cleaning
Chemical Products Manufacturing or Processing
Other Use of Solvents and Related Activities
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
ROAD TRANSPORT
Passenger Cars
Light Duty Vehicles < 3.5 t
Heavy Duty Vehicles > 3.5 t and Buses
Mopeds and Motorcycles < 50 cubic cm
Motorcycles > 50 cubic cm
Gasoline Evaporation From Vehicles
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
to
o
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE (continued)
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
su
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
OTHER MOBILE SOURCES AND MACHINERY
Off Road Vehicles and Machines
Railways
Inland Waterways
Maritime Activities
Airports (LTO cycle and ground activities)
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
WASTE TREATMENT AND DISPOSAL
Wastewater Treatment
Waste Incineration
Sludge Spreading
Land Filling
Compost Production From Waste
Biogas Production
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE (continued)
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
SU
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
WASTE TREATMENT AND DISPOSAL (continued)
Open Burning of Agricultural Wastes
Latrines
X
X
X
X
X
X
X
AGRICULTURE
Cultures With Fertilizers (except animal manure)
Cultures Without Fertilizers
Stubble Burning
Animal Breeding (enteric fermentation)
Animal Breeding (excretion)
X
X
X
X
X
X
X
X
X
X
X
X
X
X
NATURE
Deciduous Forests
Coniferous Forests
Forest Fires
X
X
X
X
X
X
X
X
X
X
X
KJ
to
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
TABLE 1. EMISSIONS INVENTORY VERIFICATION APPROACHES BY SOURCE TYPE (continued)
SOURCE CATEGORY
VERIFICATION APPROACH
High Priority
DSS
SU
Second Priority
DQR
AE
Third Priority
SA
ISS
Low Priority
AM
AFA
NATURE (continued)
Natural Grassland
Humid Zones (marches-swamps)
Waters
Animals
Volcanoes
Near Surface Deposits
Humans
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
to
DSS Direct Source Sampling
SU Statistical Uncertainty Estimate
DQR Data Quality Ratings
AE Alternate Estimates
SA Survey Analyses
ISS Indirect Source Sampling
AM Ambient Measurements
AFA Allocation Factor Assessments
-------
DOCUMENTATION OF DATA QUALITY
The activities discussed in this section are related primarily to the planning stages of an
inventory development program to ensure that the resulting inventory and quality assurance
program applied to the inventory will be properly documented. These activities are completed to
ensure that the finished inventory will be useful for the intended projects that depend on the
inventory and to facilitate the application of the inventory in future programs. These records
«
establish the strengths and weaknesses of the inventory and can be assessed at later times to
determine whether the current inventory is suitable for other applications or whether major
revisions or modifications are needed to support those additional programs. The major
considerations discussed in this section include the following:
• Defining Data Quality Objectives
• Selecting Inventory Development Methods and Assumptions
• Defining Quality Assurance and Quality Control Procedures
• Defining Needs for Independent Review of Activities and Procedures
Activities associated with the documentation of data quality begin with the initial planning
phases of any inventory development project and extend through the final production and
documentation phase. The first step involved in these functions is to understand as completely as
possible the needs of the users of the final inventories and to develop a set of data requirements
and data quality objectives that are consistent with those needs. The remaining activities are
included under the general heading of quality assurance. The inventory developers must establish
the criteria, protocol, and data handling procedures that will be applied to the effort. Contingency
plans are needed, to specify backup approaches, in the event that the primary approaches are
found to be infeasible or ineffective. Quality assurance procedures should include a detailed plan
for the review and acceptance of databases that are used in the inventory effort. Procedures for
the correction or replacement of primary data inputs are required. AH data applied in the
inventory development should be documented with respect to their origin, date, coverage, and
sponsoring agency. Finally, some programs may employ a separate outside committee to review
the procedures and resulting databases. These additional reviews are very useful to establish
credibility of the emissions estimates, the inventory preparation methodologies, and the inventory
quality assurance activities. These review committees should be involved in the process from the
initial inventory planning activities through the completion and documentation of the inventory
development program.
24
-------
ESTABLISH DATA QUALITY OBJECTIVES
Data quality objectives (DQOs) are defined as part of the initial planning phase of any
inventory development project. The DQOs can be presented as a formal written component of
the inventory research plan or developed informally through consideration of the project
objectives. The DQOs specify the geographic scope, the spatial and temporal resolution, and the
pollutant and source coverage, and, in some cases, the accuracy criteria to be applied to inventory
components. The requirements for the accuracy of the inventory do not need to be specified in
rigid quantitative or statistical terms as is common in laboratory experiments. Often it is desirable
to define DQOs for emissions inventory programs in less stringent terms, to accommodate changes
in the user objectives, and to respond to data weaknesses and other problems encountered during
the inventory development project. Emissions data are estimates, and it is difficult to assign
quantitative error bounds on these estimates; therefore, it is sometimes inappropriate to define
quantitative error bounds in the development of DQOs.
The primary purpose of DQOs is to guide the inventory development team in the
completion of a final inventory database that will fill the needs of the intended user community.
This procedure requires a detailed understanding of the intended applications of the inventory.
The development of DQOs is also influenced by the resources and time constraints of the
inventory program. In some cases, some of the needs of the user community may not be
achievable. In these cases, coordination and cooperation between the inventory development team
and the intended users is required to agree on compromises and other procedures to ensure the best
possible inventory to support the application program.
The experience obtained through the development of the 1985 NAPAP Emissions
Inventory is used as an example of setting DQOs. The NAPAP inventory development effort was
the largest and most demanding international inventory program ever completed. The inventory
includes all emissions sources in the 48 contiguous United States and the ten Canadian provinces.
Emissions were estimated for over 66,000 stationary point sources and for approximately 100 area
source categories developed for each of the 3,057 counties in the United States and the 10
provinces of Canada. The inventory was resolved spatially to grid cells of 1/4 degree longitude
by 1/6 degree latitude. Emissions data were allocated to represent hourly emission rates for the
typical weekday, Saturday, and Sunday in each of the four seasons. Primary emissions species
were resolved to represent 49 species classes for use in a variety of modeling activities. The final
inventory was organized in over 100 files and consisted of well over one billion data elements.
Priorities were assigned to specific data elements as part of the DQOs developed for the
1985 NAPAP emissions inventory. These priorities are shown in Table 2.2 This type of planning
function serves as an example of how DQOs can be specified without using a strict quantitative
format. Such priority assignments serve as a guide to the inventory development
25
-------
TABLE 2. INVENTORY PRIORITIES AND DATA QUALITY OBJECTIVES
FOR THE 1985 NAPAP INVENTORY
Data Record Parameter
State Data Submittal
Emissions
Estimates
SO,
NO.
VOC
TSP
CO
Source Classification Code
Control Equipment/Efficiency
Operating Rate Data
Location Data
Stack Parameters
Temporal Operating Data
Other Key Data
Plant Confirmation of Data
Standard Industrial Classification Code
Category I Plants8
H
H
H
H
M
L
H
H
H
H
3
H
M
4
L
Category II Plants"
M
1
1
2
M
L
H
1
1
M
L
M
L
L
L
Ke\ for Table 2. "
Category' I Plants Emissions Magnitude >500 TPY VOC;
> 1000 TPY for Other Pollutants
b Category II Plants Emissions > 100 TPY for all Pollutants
1 High Priority for Combustion Sources; Medium for Other Sources
2 High Priority for Petroleum Refineries and Chemical Production;
Medium for Other Sources
3 High Priority for Stacks > 100 feet; Low for Stacks < 100 feet
4 High Priority for Plants with Emissions of SO2 and NOX > 2500 TPY
team and help to ensure that the appropriate level of effort is allocated to each part of the
inventory. The priority assignment also served as a reference for the users so that they knew what
information would be available in the completed inventory. Changing needs and expanded
applications of the inventory were identified during the NAPAP inventory development effort.
As a result, some analyses using the NAPAP inventory data were affected by data weaknesses,
26
-------
even though the priority data elements were defined in the initial planning with input from both
the inventory development and user teams. This condition is emphasized here to point out the
need to build as much flexibility as possible into the inventory development process. In high
visibility programs, such as the NAPAP effort, it is very likely that increased demands on the
inventory will be requested after the initial planning is complete.
METHODS AND ASSUMPTIONS
Another important part of the inventory development planning process is to specify the
methods to be applied to data collection and data processing, which in turn requires definition of
the criteria to be applied to assess the various data sources that are available. These criteria are
then used to make acceptance and rejection decisions for the various data suggested for application
to the inventory. The criteria should cover the completeness of the data, the geographic scope of
the data, the year of record for the data, the sponsoring agency that was responsible for developing
the database, and the accuracy of the data, if known.
To the extent possible, consistency of data sources is desirable. The use of consistent
databases for all similar sources in a region will facilitate inventory updates, projections to future
years, and evaluation of alternative control strategies. It is sometimes possible to mix data sources
in the development of a regional scale inventory. As an example, detailed traffic flow data may
be available for an urban area through an urban transportation planning program, but traffic count
or traffic flow data for the outlying suburban and rural areas may not be available. In many urban
areas, the bulk of the mobile source emissions that affect the understanding of the problem and
the design of control strategies would be included in the urban transportation planning area.
Under these circumstances, it may be acceptable to use the detailed urban area data in combination
with other simple estimates of vehicle traffic for the outlying area.
The issue of consistency is particularly important in programs that require inventories that
cross country borders. Data consistency between countries is, however, sometimes very difficult
to achieve. This may be particularly true in Europe, where recent political events have changed
the communication and accessibility between countries, and in some cases, have changed the
countries themselves. When significantly different emissions estimation techniques are applied
in bordering countries, large discontinuities can result at the border. Such discontinuities can not
only interfere with control strategy development and implementation, they can also
affect the performance of air quality models. When planning an emissions inventory for a
regional program that includes multiple countries, the consistency issue should be considered and
common estimation methodologies should be applied whenever possible.
One useful approach to document inventory strengths and weaknesses is to implement a
data quality rating scheme for emission factors and other supporting data that are to be used in the
27
-------
inventories. This approach has been advocated and implemented in inventory development
activities in the United States. The U.S. EPA includes data quality rating factors to indicate the
confidence in the emission factors that are listed in AP-42.
The data quality rating factors used in AP-42 are specified as A, B, C, D, and E with A
representing the highest quality and E representing the lowest quality. Although these quality
ratings are rather subjective, they do provide an assessment of the confidence that can be
associated with those factors. For example, these ratings were applied to the NAPAP inventory
to establish the distribution of emissions by emission factor quality rating. These results are
• summarized in Table 3. In this example, it is clear that the confidence in the emissions for SO2
and NOX is much higher than for the other pollutants.
It is also possible to develop more complex data quality rating systems for application to
emissions inventories. Such a system would consider factors such as geographic and temporal
resolution of emission factors and activity data, as well as the quality and quantity of
measurements made in actual systems to derive a weighted numerical score that would be assigned
to emissions estimates. The numerical score would then provide a semi-quantitative quality rating
scale that could be used to compare the reliability of two or more independent estimates for the
same pollutant and source category combination. One such rating scheme is being considered for
application to an international database of greenhouse gas emissions. Although the specifics of
the rating system have not yet been applied to any published emissions data, this type of system
could be applied at both global and country specific scales for application to global inventory
programs and to source specific processes in regional and urban inventory programs.4
TABLE 3. DISTRIBUTION OF EMISSIONS IN THE 1985 NAPAP EMISSIONS
INVENTORY BY EMISSION FACTOR DATA QUALITY RATING
POLLUTANT
SO,
NOY
VOC
TSP
CO
EMISSIONS
TOTAL, TPY2
10,593,049
5,321,622
444,454
608,977
1,945,290
PERCENT OF EMISSIONS BY DATA QUALITY FACTOR
A
95.3
67.9
7.0
18.3
17.5
B
1.6
14.5
9.2
20.3
10.5
C
0.3
10.5
10.6
8.3
18.8
D
0.1
1.7
5.2
6.4
0.4
E
0.9
2.1
13.1
5,4
16.0
OTHER"
1.8
3.3
54.9
40.8
36.8
'' Total emissions estimated with emission factor files
h Unrated emission factor or use of an emission factor for a source category for which it is unrated
28
-------
It is also important to prepare contingency plans and backup approaches that can be applied
in the event that desired approaches do not provide adequate data, are not available within time
constraints, or are incomplete or otherwise unacceptable. Contingency plans may involve the use
of default values, surrogate factors, or other parameterizations to represent specific activities or
source categories. It is desirable to define these contingency plans in the inventory planning phase
and to involve the inventory users in the definition of the contingency plans.
QUALITY ASSURANCE AND QUALITY CONTROL PROCEDURES
A quality assurance plan is another important component of any emissions inventory
development project. The QA plan provides the blueprint for all activities that are included in the
program to ensure data quality and to achieve the particular objectives of the program. The QA
plan specifies all of the activities discussed previously and the specific data checking and data
correction steps that will be implemented during the preparation of the inventories. Many of the
components of inventory verification that will be discussed in more detail in the remainder of this
document are also activities that should be documented in the QA plan.
The QA plan specifies the types of data that will be collected, the procedures that will be
applied to assess the applicability of those data to the program, the steps that will be taken to
correct or modify questionable or incorrect data, the procedures for documenting data corrections
or modifications, and the procedures that will be applied to process the data into formats that are
consistent with the inventory applications. The QA plan must be consistent with the DQOs and
the resources that are available.
The QA plan should include a well-defined corrective action plan and any specific
procedures that are necessary to document what was done to substitute or modify the original
databases. In most cases, it is useful to maintain a separate QA file. The file should be organized
to allow simple and timely access to specific information concerning quality assurance activities
performed on the data. One approach is an electronic audit trail. An audit trail file can be
structured to facilitate review of information on the date of receipt of original data, the magnitude
of the reported data, the type of QA problem associated with the data, the correction to that data,
the date of the correction, and the person responsible for that correction. This type of file can
allow a quick and comprehensive review of all data manipulations and corrections and can provide
a format that allows the entire process to be archived so that analyses of these corrections can be
completed long after the inventory development project is over. These electronic documentation
techniques can be particularly useful if the inventory data are used in additional programs in the
future.
29
-------
Some specific QA activities that were applied to the 1985 NAPAP Emissions Inventory
are the following:
• the definition of DQOs and priority data elements (see Table 2)
• completeness checks against previous inventories and follow-up to verify
facility closures and start-ups
• checks that all high-priority data elements were included in data records
• checks that the data represented a consistent base year 1985
• checks of aggregate total emissions and operating data against alternative
data sources (e.g., State and national fuel use data against data reported to
the Department of Energy)
• internal consistency checks for specific sources to ensure that operating and
emissions data were compatible
• an emissions confirmation by the facility for sources over 2,500 TPY of
SO2 and NOX
• detailed computer and manual reviews of the 1,000 largest sources of SO2,
NOX, and VOC
• maintenance of detailed and complete QA files for each State
• two separate mailings to the reporting agency with a request for correction
or verification of questionable, missing, or incorrect data
• maintenance of a project docket containing references for all data sources
and processing techniques used in the project
• maintenance of a computerized audit trail
• development and documentation of default values
• separate computer program elements to check the results after each phase
of data processing
• computerized checks to ensure that all spatial, temporal, and species
allocation files accounted for 100 percent of the input data
• identification of data records with inaccurate locations (data records outside
of specified boundaries or over water bodies, etc.)
• complete and detailed series of reports documenting the inventory effort
INDEPENDENT THIRD PARTY REVIEW
If time and resources permit, an independent review panel or third party review can be
useful for large, complex, and high visibility programs. It is useful for these review functions to
be operative through the entire project, from planning phases through completion. The specific
responsibilities of the review teams are to ensure that the proposed project will satisfy the
objectives of the users, that resources are adequately applied to ensure the best product within time
constraints, that QA objectives are reasonable and applicable, that activities conducted are
consistent with the needs and DQOs of the program, that corrective actions are logical and
30
-------
appropriate, and that the documentation is reliable and clear. Written reports from the review
team should be included in the overall QA files for the project. These written reports can be in
the form of meeting minutes of regularly scheduled and ad hoc meetings or as a formal written
report. After the final inventory is prepared, the review team should include an assessment of the
overall effort that points out any deficiencies in the final product, but that also recognizes the
strengths of the final product, in terms of whether the project adhered to plans, responded to the
DQOs, and substantially fulfilled the needs of the project. During the development of the 1985
NAPAP Emissions Inventory, both an independent review panel and a third party review of the
final inventory were employed. Both of these functions provided recommendations that resulted
in improvements to the inventory.
31
-------
APPLICATION OF THE DATA
The intended application of the completed inventory is the principal consideration when
preparing and implementing an inventory verification exercise. The major uses of emissions
inventory data and the specific needs for inventory verification programs related to each of these
uses are discussed in this section. In some cases, it is possible to use data and analyses developed
while conducting these activities for emissions verification exercises. Some examples of such
opportunities are discussed in this section. The two primary uses of emissions data and emissions
inventories are listed below:
• assessments of the specific air quality problems in an area and identification of the
most important sources and source categories that influence those air quality
problems
• input for regulatory activities including air quality modeling analyses and the
design, implementation, and tracking of the effects of air quality control strategies
ASSESSMENT STUDIES
The requirements for specificity and accuracy of inventory data are less stringent for
assessment studies than for the other applications of inventories. Top-down inventory
development methodologies are usually suitable for application to assessment studies and annual
emissions estimates for an entire industrial sector are often adequate for these purposes.
Assessment studies, are generally intended to provide the background understanding of the
primary causes of the air quality problems being evaluated. One example of an assessment type
inventory is the preparation of annual trends inventories. In these applications, it is not necessary
to develop estimates for specific source categories; it is usually more suitable to develop estimates
for large industrial sectors such as electric power production and chemical manufacturing
operations. Another example is an assessment study to define the ten largest source categories of
VOC and NO, emissions in an urban area. For both of these applications, total emissions of NOX
resulting from fuel combustion can be estimated from total energy demand estimates. In this type
of study it would not be necessary to represent the emissions by boiler design or even fuel type.
Similarly, total emissions of VOC, CO, and NOX from mobile sources could be estimated from
vehicle registration records, total regional fuel sales data, and assumptions about the fleet average
fuel economy. Details about road type and speed classifications would not be required.
32
-------
Alternative approaches for these highly aggregated estimates can sometimes be applied to add
validity to the estimates if the emissions data developed through alternative approaches are
comparable. In these cases, the detail required in the emissions estimates would not justify the
use of a complex and detailed approach to emissions verification.
REGULATORY APPLICATIONS OF INVENTORIES
Regulatory activities are performed by air quality control agencies to define programs and
policy options to reduce the negative impacts caused by air pollutants. The development of an air
quality control policy usually uses an emissions inventory to estimate the potential for mitigating
those problems and the costs that are associated with the control options. Some of the issues
associated with regulatory activities that depend on emissions inventories are listed below:
• understanding the importance of local emissions relative to the impact of air
contaminants that are transported from other regions
• understanding the importance of biogenic or other naturally occurring
sources relative to the anthropogenic sources in an area
• establishing the relative importance of the various anthropogenic source
categories to the overall controllable emissions burden
Air Quality Modeling Applications
The requirements for emissions inventories in air quality modeling applications, and the
methodologies and activities required to validate these inventories are significantly more
demanding than for assessment inventories. For application to air quality modeling programs,
representative inventories of the appropriate species are needed at spatial and temporal scales
consistent with the model formulation. It is necessary to develop the baseline inventory data at
source-specific detail to represent the species, spatial, and temporal variability associated with the
emissions. In the most complex modeling programs, the needs of the model community often
change during the development of the inventory. It is important to build in flexibility to allow
changes in program requirements to accommodate changing user needs.
An air quality model is a computer program that represents atmospheric transport,
chemical reactions, and pollutant deposition phenomena as a collection of mathematical
expressions. These models require meteorological and emissions input data and are used to
simulate actual atmospheric conditions that result in air quality problems. Air quality problems
such as urban ozone formation are influenced by local meteorological factors including wind
speed, wind direction, temperature, and sunlight intensity. Since these factors are variable on
33
-------
hourly time scales, emissions data are also required at hourly time scales. The models used to
simulate urban ozone formation are usually based on a mass balance approach in which pollutants
are transported in response to the hourly behavior of the wind and temperature conditions. The
models are structured to represent physical distance scales that can simulate the hourly changes.
This spatial resolution is commonly represented by a regular grid pattern overlaid on the modeling
region. Emissions data, therefore, must also be represented at these spatial and temporal scales
to be compatible with the model formulation. Since emissions data are commonly estimated at
Jhe annual-level, and at country-, province-, or county-levels, techniques must be applied to
convert the emissions estimates into the appropriate spatial and temporal resolution. VOC, NO,,
and particulate emissions data must also be resolved to represent the source specific chemical
compounds included in the VOC, NOX or particulate matter estimate, to adequately track the
complex chemistry that occurs throughout an urban area. The following discussion provides
information on the requirements of spatial, temporal, and species resolution of inventory data for
modeling applications and provides some approaches that can be used to evaluate the quality of
resolved inventory data.
Spatial Requirements for Modeling Applications
Emissions inventories used in urban and regional modeling studies require the most
demanding specifications on any emissions inventories, because of the scope, range and level of
detail needed. The verification needs for modeling inventories are dependent on the physical
domain of the modeling exercise, the chemistry simulated by the model, and the specific
regulatory applications that the model results will support. Specifically, the verification needs
must address the data sources and methodologies applied to resolve the emissions data.
Spatial allocation refers to the development of spatially resolved emissions estimates
consistent with the appropriate spatial scales of the model application. These requirements include
both the physical location of the plant or source and the elevation at which emissions are released
to the atmosphere through smoke stacks. Plant-specific location data are often available for large
stationary point sources and, if such data are available, they can be used directly to locate those
sources. Verification is sometimes accomplished by checking the physical location data or through
mapping the data to identify sources that are inaccurately located. Misrepresented location data
can place emissions points outside of the geographic domain of the study or over water bodies.
Once inaccurate location data are discovered, they can be corrected by contacting representatives
of the facility.
It is also necessary to represent the appropriate height of the emissions release for large
point sources from elevated smoke stacks. The effective release height is defined as the stack
height plus the plume rise, which is the vertical distance travelled by the emissions plume before
34
-------
the plume levels out and begins to disperse in response to the wind at that height. The plume rise
is affected by the speed at which the emissions are released from the stack and by the buoyancy
of the hot stack gas. Plume rise is a function of the stack height, stack mouth diameter, stack gas
temperature, and volume flow rate of the stack gas. In some cases, these parameters are known
and can be included in the data collection protocol. When these data are not known, it is
necessary to apply default parameters to estimate the effective release height. Stack parameters
are frequently checked in inventory applications by determining whether the parameters are
outside of a typical range of values. For example, stack heights for large coal burning utility
plants should not have stack heights lower than 200 feet or higher than 1250 feet. Similarly, stack
diameters should be less than 20 percent of the reported stack height.
Commonly, mobile and other dispersed area sources are estimated at some aggregated level
of spatial resolution (e.g., in the United States, these estimates are often made at the county level).
The spatial processing of these data involves formatting the inventory data to represent a regular
grid pattern. The appropriate scale is dependent on the particular modeling application. For
urban air quality modeling, the grid pattern is usually on the order of one kilometer to five
kilometer squares. For regional ozone, visibility, or acid deposition modeling, the scales can be
expressed in kilometers or degrees of latitude and longitude. The patterns represented in these
models can range from 10-kilometer to 150-kilometer squares or equivalent resolution expressed
in units of latitude and longitude. Global models are usually represented by degrees of latitude
and longitude and typically range from 0.5- degree to 10-degree cells.
The methodology used to spatially allocate emissions is to apply spatial allocation factors.
The spatial allocation factors represent the fraction of the activity that is included in each grid that
overlaps the area covered by the emissions estimate. Figure 1 illustrates this concept for an
example case involving a single grid cell that overlaps three emissions regions, represented here
as counties.5 In this example the allocation is based on land area. The area units represented in
this illustration are arbitrary but could represent any standard area measurement units. In the grid
cell shown in Figure 1, emissions would be allocated in the following manner for all source
categories that were linked to the example spatial factor:
Total emissions in grid cell X = 0.119 (6.37/54.1) * emissions for County A
+ 0.032 (1.58/48.8) * emissions for County B
-1- 0.053 (2.81/52.5) * emissions for County C
Similar functions that describe the proportion of overlap of the selected surrogate data are
developed for all occurrences of county and grid cell overlap, so that all of the emissions in each
35
-------
ON
Overlap of County A
in the Grid Cell
area = 6.37
allocation factor =
6.37/54.1 = 0.119
Overlap of County B
in the Grid Cell
area = 1.58
allocation factor
1.58/48.8 = 0.032
Overlap of County C
in the Grid Cell
area = 2.80
allocation factor =
2.8/52.5 - 0.053
NOTE: Area Units Are
Arbitrary
Figure 1. Example of County to Grid Cell Land Area Allocation
-------
county are ultimately assigned to a grid cell. The development of these spatial factors requires
the preparation of a separate spatially resolved data file for each spatial allocation factor and can
be a costly component of the inventory development process. Some allocation factors can be
developed from detailed census data, while others may require analysis of satellite imagery data.
Therefore, the effort is usually limited to the preparation of a small set of factors that can be used
to estimate the distribution of a large number of sources. For example, population and
employment statistics for specific economic sectors are appropriate surrogates for many activities.
Other useful factors are urban land area, agricultural land area, forested land area, and the area
of water bodies.
The types of sources and source categories that require spatial processing depend on the
scale of the modeling application. On urban scales, mobile sources, residential fuel combustion,
pesticide application, road paving operations, and dry cleaning operations are examples of the
activities that would be allocated using spatial factor files. Inventories prepared for extended
regional modeling and for global modeling applications would likely use spatial allocation
techniques for a larger number of sources than for urban modeling scales. The needs for source-
specific location data become less important in larger scale modeling applications because the
spatial averaging inherent in those larger scale models does not support the use specific location
data. In the global applications, all sources can be estimated at national levels and spatially
aggregated to the model grid resolution through the use of spatial allocation factors.
Geographic Information Systems (GIS) provide efficient spatial processing capabilities for
application to emissions inventories. GIS programs are computerized data processing tools that
are structured to facilitate mapping and distributing data in a spatial format. These programs
allow the application of layers of data that can automatically be interrelated. It is necessary to
construct spatially resolved surrogate data similar to those used in non-GIS spatial allocation
techniques. The GIS programs, however, can overlay the spatial distribution data onto the
emissions data and automatically calculate the allocation fractions, to conform to user-defined
resolution. These approaches allow the user to produce several gridded emissions files at different
scales using a single file of spatially resolved surrogate data. The use of GIS systems can,
therefore, reduce the cost associated with spatial factor development if applications at different
scales are anticipated. Any future emissions inventory development projects should consider the
use of one of these GIS systems to simplify the application of emissions data in different modeling
projects.
Temporal Requirements for Modeling Applications
Temporal allocation refers to the processing of inventory data prepared at annual resolution
to represent seasonal, monthly, daily, or hourly fractions that are required by the models. Most
37
-------
manufacturing, utility, and service industries keep records of the level of activity conducted at
individual facilities on at least an annual level. In many cases, detailed operating schedule data
are available for the large stationary point sources. These data can then be processed through
computer programs to represent data at the appropriate level of temporal resolution. In some
cases, assumptions about continuous operation or a regular operating cycle provide fairly specific
estimates about operating rates. These assumptions, however, do not always adequately represent
variations in the efficiency of control equipment, process upsets, or accidental releases, and many
times do not represent shutdowns for scheduled or unscheduled maintenance and repair activities.
Therefore, assumptions about normal operating schedules can often add considerable uncertainty
in temporally resolved emissions estimates, especially for application to an episodic simulation of
an actual observed air quality event.
Typical temporal factors represent conditions of continuous operation: an eight-hour work
day, five days per week, and a ten-hour work day, 6 days per week. While these assumptions
about spatial and temporal allocation factors can be assumed to represent a large number of
sources, the surrogate factors are not directly applicable to the distribution of any individual
source category.
In the NAPAP application, annual emissions were resolved to represent hourly emissions,
for the typical weekday, Saturday, and Sunday in each of the four seasons. This was
accomplished by specifying seasonal fractions, days of the week in operation, and hours of the day
in operation. The seasonal splits represented simply the fraction of the annual emissions that
occur in each season. The days of the week operation data were processed to develop fractions
for the weekday, Saturday, and Sunday. For example, a five-day work week schedule caused
emissions to be equally distributed through all five weekdays and a six-day operating schedule
allocated emissions equally to weekdays and Saturday. Hourly distributions were represented as
typical work days. Typical assumptions included continuous operation with emissions distributed
equally to each hour of the day and eight-hour operation with emissions equally distributed from
7 a.m. to 5 p.m. local time. Other more complex operating schedules were developed for
application to highway vehicles, electric generating sources, and other source types that typically
operate on irregular schedules.6
An example of using the application of the data as a verification tool is offered through the
the experience of NAPAP. An analysis of daily emissions of CO was performed over the
modeling region. It was discovered that emissions of CO in the grid cells surrounding New York
City were higher on Saturday than on the weekdays. This was unexpected and further
investigation revealed that emissions for pleasure craft were allocated spatially following the
population distribution surrogate. Since it is expected that pleasure craft activity would be
concentrated on weekends the emissions for this category were concentrated in high population
38
-------
density areas on Saturday leading to the higher CO emissions for New York City on Saturday.
These types of analyses can be useful to understand the weaknesses of inventory data and to help
with better interpretation of modeling and other analytical results.
Chemical Speciation Requirements for Modeling Applications
Species allocation involves representing the components of aggregated primary emissions
species for modeling or risk determination applications. In the United States, emission factors
represent the primary emission pollutants, namely SO2, NOX, VOC, PM-10, CO, and lead (Pb).
In urban and regional ozone modeling, the components of the NOX and VOC emissions are
ultimately more important than the total NO, or VOC estimate. Oxides of nitrogen include both
NO and nitrogen dioxide (NO2). Volatile organic compounds include literally hundreds of
individual hydrocarbon compounds ranging from those that are only marginally active in urban
photochemistry to those that are highly reactive. All air quality models used to simulate ambient
ozone distributions represent the complex mix of VOC species as a group of representative
hydrocarbon classes. The hydrocarbon classification involves either grouping compounds with
similar reaction characteristics into a representative psuedo-species, or grouping similar carbon
bond types that can be simulated by common reaction mechanisms. The degree of VOC grouping
varies in various modeling analyses. The VOC species grouping approach and the number of
representative VOC classes included in any modeling exercise is dependent on the specifics of the
program and ultimately on the objectives of the analyses that use the modeling results. Estimation
techniques based on source-specific profiles are often used to allocate the VOC source emissions
estimate to the hydrocarbon classes that are required by the models. Similarly, source-specific
allocation profiles are applied to represent the split between NO and NO2. It is sometimes
necessary to estimate size ranges for paniculate emissions for application to visibility studies.
Both VOC and paniculate matter include individual compounds or elements that are considered
hazardous air pollutants or air toxics. Modeling activities for risk assessments or other analyses
involving the distribution of specific toxic compounds require inventories that have been speciated
to represent those compounds of interest.
The speciation profiles represent the weight percent of the component hydrocarbons and
elements included in the VOC and particulate emissions, respectively. The fractions were
determined from source-specific emissions testing and subsequent laboratory analysis of the
samples. There were a total of 313 specific hydrocarbon speciation profiles and 131 paniculate
speciation profiles applied in the NAPAP inventory. These profiles were applied to speciate
emissions of VOC and paniculate for well over 1,000 individual source categories of each
pollutant.7 Obviously, speciation profiles were applied to sources for which test results were not
available. The assumptions about speciation characteristics can have significant influences on the
39
-------
chemistry represented in the air quality models, and the use of these surrogate speciation profiles
adds large uncertainty to the modeling inventories. Since operating conditions vary even between
sources that fit the same source category description, the use of a single speciation profile for all
similar sources also adds unknown error to the emissions estimates.
Verification of Modeling Inventories
Air quality modeling teams are always interested in having a quantified error bound placed
on the resulting hourly, gridded, and speciated emissions estimates. Each emissions allocation
step adds uncertainty to the resulting emissions estimates, and it is almost impossible to quantify
the error associated with the resolved estimates applied in modeling exercises. There are
significant uncertainties in the model formulations themselves which are nearly impossible to
evaluate without a clear understanding of the uncertainties in the model inputs. Specific activities
that can be applied to modeling inventories to assist in estimating the validity of the emissions data
will be discussed in later sections of this report.
If time and resources permit, it is sometimes possible to implement activities that can help
to establish the reliability of the techniques used to resolve the inventories. In all cases, the data
sources and the methodologies used to process these data into allocation factors should be
documented and referenced. Such documentation can be helpful in evaluating the inherent validity
of the raw data and the assumptions applied to process those data.
For the case of spatial allocation, estimates developed through the application of one
allocation profile can be compared to the distribution that would result from another independent
but related activity. For example, if mobile sources are allocated following a population
surrogate, it is possible to compare that distribution to one based on traffic density estimates or
a file of the distribution of roadways in an area. These types of checks generally do not provide
quantitative error estimates, but they can often be useful to estimate the relative confidence that
can be associated with a particular source category.
Checking temporal allocation profiles is not as straightforward as checking spatial
allocation. If specific operating schedule data are available for some of the largest sources, it is
possible to compare those data to the surrogate temporal profiles. These types of comparisons
cannot provide quantitative assessments of uncertainty for other specific facilities, but they can
often be used to provide a qualitative assessment of the reliability of those allocation factors. For
example, in the NAPAP program, average operating schedule data were applied to many of the
large coal burning utility sources as a default. Independent operating data were made available
for the coal burning sources included in the Tennessee Valley Authority (TVA). The NAPAP
operating schedules and the plant-specific operating schedules were compared, and it was found
that the default operating schedule data were similar to the actual TVA operating data, in most
40
-------
cases. An example of one comparison of the various assumptions about hourly temporal allocation
is shown in Figure 2. This result suggested that overall, the application of the temporal profiles
representing the default operating schedules did not seriously affect the validity of the inventory.
Not all comparisons were as good as that shown in Figure 2, and there was significant error for
some particular sources. Figure 2 also shows the error that can result when continuous operation
assumptions are used. In particular, routine and unscheduled shutdowns for maintenance and
repairs are not represented by average or default operating schedule data. The effects of
neglecting shutdowns of major coal burning utility sources were tested in sensitivity analyses using
the Regional Acid Deposition Model (RADM) in the NAPAP program. These tests revealed that
these types of shutdowns could seriously affect model results. Since the effects of facility
shutdowns were so serious, and the temporal allocation methodologies did not accurately represent
such occurrences, actual emissions data were represented for the largest 100 sources of SO2 in the
RADM modeling domain. A program was established to collect hourly emissions data from those
sources that operated continuous emissions monitors (CEMs). Facility load and other operating
data were collected for those sources that did not have CEMs, and hourly emissions were
calculated. The detailed data were substituted for the average emissions data for modeling
applications of episodic events in the NAPAP program.
Checking of speciation allocation methodologies is difficult. The only effective way to
check the assumptions inherent in the speciation profiles is to conduct source tests and compare
the results of the test to the speciation profile. Obviously, if all of the important sources are
tested, there is no need for the application of the surrogate profiles. Even if selected sources can
be tested, the tests themselves are limited by the selection of the individual compounds that are
measured. One technique applied in programs in the United States is the evaluation of the
speciation profiles quality and the assignment of a quality rating factor to those profiles as they
are applied to individual source categories. It is then possible to assess the distribution of the
resulting speciated data following the rating scheme. If a large fraction of the important
compounds is associated with low quality ratings, the validity of the data can be considered
suspect. Other techniques that involve comparisons to ambient measurement data are possible and
will be discussed later in this report.
In all cases when allocation profiles are applied, the allocation profiles should be checked
to ensure that they sum to unity. This will at least verify that the application of the profiles will
not result in the loss or addition of emissions during the processing. It is also useful to check the
emissions sums after each step in the processing to verify that the computer programs used in the
allocation steps function correctly and do not result in changing the emissions estimates relative
to the totals developed at the more aggregated level.
41
-------
-U
K)
0.05
0.045
w
I— I
Kj
PH
0.04
>
I
i
0.035
0.03
/ \
\ ''
Continuous Operation Assumption
NAPAP DEFAULT OPERATING SCHEDULE
ACTUAL PLANT OPERATING SCHEDULE
i i i
i i i i i i
j i i
1 3 5 7 9 11 13 15 17 19 21 23
2 4 6 8 10 12 14 16 18 20 22 24
TIME OF DAY, (hours)
Figure 2. Comparison of Hourly Allocation Factors For An Actual Electric Utility Boiler
-------
Control Strategy Analyses
Regulatory analyses are applied to all pollutants and at all scales. The analyses usually
involve the identification and rank ordering of the sources or source categories as they are believed
to affect the air quality problem under study. For example, a regulatory assessment of an urban
ozone issue would seek to estimate the contribution of NOX and VOC emissions from mobile
sources relative to the magnitude of emissions from other fuel combustion, manufacturing, and
service related activities. An additional component would be estimating the contribution of forests
and other naturally occurring biomass to the total VOC emissions.
These types of activities are needed to estimate the total amount of emissions control that
can be achieved and to provide preliminary control cost estimates. If an upwind location
contributes to the local air quality problem, the regulatory agency may need to initiate discussions
with their colleagues in the upwind area to quantify improvements that can be expected from their
control programs. Similarly, if the assessment indicates that more than 50 percent of the ozone
precursor emissions affecting an urban area ozone problem result from local mobile sources, the
air quality decision makers should seek alternatives that could be adopted to improve public
transportation or otherwise reduce the amount of personal automobile traffic.
Since these types of studies are used to better understand the principal issues and to develop
policy planning approaches, the emissions data do not need to be highly resolved. Emissions data
for these applications can often be developed using top-down approaches based on gross statistics
such as fuel sales or even on population and economic statistics in some cases. Obviously, the
level of detail required and the confidence associated with the absolute emissions estimates are not
as rigorous as for other scientific and engineering studies. Often verification exercises can be
implemented in which comparisons of per capita or other aggregated emissions estimates are
compared to similar regions or even to other countries that have similar population density and
economic development status.
Following the formulation of an initial policy based on the assessment, more detailed
emissions estimates for the baseline case are required. These more detailed emissions estimates
should be developed for specific types of sources that are targeted for control. One approach
widely used in the United States is to characterize an industry as a series of model plants. These
model plants can be differentiated by size and/or operating characteristics. The estimates of
emissions rates for these model plants are developed through a detailed engineering study of the
processes employed in the industry and a sound understanding of the causes of air emissions.
Once these issues are defined, it is possible to evaluate the potential for add-on emissions control
devices, changes in the operations, or other modifications, such as pollution prevention activities
or recycling opportunities that could reduce air emissions. The costs associated with each of these
options can then be evaluated, and cost benefit analyses can be developed. Decision makers can
43
-------
then assess the costs to society of the various options and the expected improvements in air quality
that would result from the implementation of those activities.
Often the effects of a control strategy based on a collection of emissions control options
are evaluated through the application of air quality models. Therefore, the needs for verification
of emissions inventory data and the techniques that can be applied are similar to those discussed
previously. Of primary concern in evaluating the impacts of control strategies is whether the
range of emissions magnitudes associated with the uncertainty in emissions estimates leads to a
similar decision about a proposed control strategy. Inventories that are bounded by ranges large
enough to result in different policy decisions are difficult to apply in the regulatory framework.
For example, if there are large ranges in emissions estimates associated with NOX and VOC in an
urban area, it is possible that the models would predict control strategies based on NOX control
at one extreme and VOC control at the other extreme.
Once the primary sources for control are understood and a cost-effective control strategy
is proposed, the policy analyses must consider the implementation and tracking of the control
program. The first consideration of emissions data used in this way is the need to develop
projection inventories to represent the conditions likely to be encountered in the future. All
control programs take time to implement, and changing populations, economic development, and
new industrial activities all affect the future emissions scenarios. The development of verification
techniques should consider not only the application to the baseline inventory but to the historical
and the assumed future conditions as well. The success of control programs can then be evaluated
against the expected range of uncontrolled and controlled emissions represented in the future year
emissions projections.
Verification of Inventories for Regulatory Applications
The level of detail required in inventories used in regulatory activities is similar to that
needed for modeling activities. Therefore, the activities suitable for verification are also similar.
One approach introduced in the United States, with passage of the Clean Air Act Amendments of
1990, is to require each facility that emits more than 25 tons of a pollutant per year to report its
emissions to the regulatory agency and ultimately to the U.S. EPA. The methodologies and data
used by the facilities to estimate emissions must be documented to the satisfaction of the regulatory
agency. These data will become a primary source of information to verify inventory estimation
techniques and to improve existing inventories.
RELATIONSHIPS TO AMBIENT DATA
One of the primary indications of the potential inadequacies of ozone precursor emissions
estimation methodologies is the apparent lack of agreement between VOC and NOX concentrations,
44
-------
the VOC to NOX ratios (VOC/NOJ, and CO to NOX ratios (CO/NOJ represented in emissions
inventories and those measured in urban atmospheres. Ambient air measurement programs have
long been a tool for assessing the air quality conditions in a location. Early morning VOC and
NOX concentrations are often measured in and near the urban centers as part of ambient
measurements programs. These early morning measurements in urban centers are designed to
provide emissions data during the period of high density morning traffic (6 to 9 a.m. local time),
before significant chemical reaction has a chance to change the hydrocarbon and NOX
concentrations. The ambient measurements are then used to initialize air quality models in terms
of VOC and NOX concentration and the associated VOC/NOX ratio.
The predictions of complex air quality models that simulate the interactions between all
sources in the urban area over multiple days often do not simulate the measured VOC/NOX ratios
in the urban center. Much of the cause for the failure of the models to simulate these ratios has
been attributed to inadequacies of the emissions input data. A recent study conducted by the
California Air Resources Board (CARB) has presented detailed analyses of the application of
comparisons of measured VOC/NOX and CO/NQ to evaluate the weaknesses in emissions
inventories in the Los Angeles area.8 The study design attempted to relate the measured
abundance of VOC and CO to the amount of VOC and CO that would be expected from the
highway mobile source categories in the inventory. The results found that the emissions models
used to estimate mobile source emissions underestimate CO emissions by a factor of about 2.7 and
underestimate VOC emissions by a factor of about 3.8. These and other measurements studies9'10
suggest that much of the discrepancy may result from the treatment of a small percentage of high
emitters as if they had emissions characteristics similar to the average test vehicle. The results of
this program will be discussed in more detail in the section of this report covering ground truth
verification methods.
The results of these and other studies suggest the current methodologies for inventory
preparation do not consider all of the factors that are important in episodic analyses. While the
current methodologies do seem to do a good job of estimating emissions at national levels and for
annual time periods, there is a need to improve existing and develop new methodologies for
temporally resolved, area-specific emissions estimates. Further application of data from programs
comparing ambient measurement data to emissions inventory data will help to identify those
sources and source categories that need improved emissions estimation methodologies and those
that are missing altogether from the inventory preparation process. In the United States,
additional programs are ongoing to identify and develop estimation methodologies for source
categories that have not been included in previous inventory development efforts.11
45
-------
COMPARISON OF ALTERNATE ESTIMATES
It has already been mentioned that alternate approaches exist for estimating emissions
magnitudes from selected source categories. These alternate approaches can be used to derive
independent estimates of emissions. The validity of the data resulting from two or more
independent emissions estimates can be inferred from the degree of agreement among the
^estimates. Some examples of such data comparisons are discussed in this section. While some
opportunities for data comparisons require extensive efforts in deriving alternate estimates, others
require only a limited additional resources. This discussion will focus on the simpler kinds of data
comparisons, which will be most useful in the widest range of applications.
OPPORTUNITIES FOR DATA COMPARISONS
Opportunities for data comparisons exist on many levels. It is possible to compare
alternate emission factors, alternate operating rate data, and the final calculated emissions based
on independent methodologies. Statistical comparisons of aggregate emissions totals may be
applied between countries or regions of countries that have similar population and economic
status. In each of these cases, the convergence of estimates derived through alternate estimation
strategies adds credibility and validity to the final reported estimates.
One of the first data comparison opportunities in any inventory development effort is a
completeness check. A completeness check involves a comparison of the facilities included in the
current inventory to those included in previous inventories, if a recent emissions database is
available. Completeness checks ensure that all of the important source categories are considered
in the inventory, and that all of the facilities within a source category are included. If individual
facilities are represented in the base line inventory, all of the significant sources of the pollutants
of interest can be compared to the sources included in the previous inventory. In an area with a
dynamic economy, some sources may no longer be in operation or new sources may have been
added since the earlier inventory was developed. Records of the parent company or records
maintained by trade associations can be checked to determine whether sources were closed or
started up in the area. Where a discrepancy between the sources in two inventories is discovered,
the inventory development team should seek to verify whether a source was actually closed or a
new source was started.
In some cases, the inventory developers may encourage the application of source-specific
emission factors, in lieu of the default or otherwise recommended factors, if sufficient
documentation of the source-specific emission factors can be provided. The agency or authority
46
-------
responsible for the development of the inventory should review all source-specific emission factors
to assess their applicability in relation to their quality and consistency with the rest of the
inventory. In general, source-specific emission factors, for well-characterized sources, should
agree to within about 20 percent of default or otherwise recommended factors. If the discrepancy
is larger, the data upon which the specific emission factor is based should be reviewed. If the
documentation of the development of the factors is unclear or otherwise inadequate, application
of standard factors should be considered. In some cases, the revised factor could be based on
actual emissions, and the default based on uncontrolled emissions. When uncontrolled emission
factors are used, the predicted emissions magnitude must be corrected for the effect of the control
equipment. In these cases, the comparison of final calculated emissions would be more
appropriate than the comparison of the emission factors themselves.
Emissions estimates for major source categories at the country or regional level through
two alternate measures of the inherent activity can often be developed. For example, emissions
estimates for some manufacturing industries can be based on raw material feed to the process or
on the total amount of product produced. Similarly, estimates of total gasoline fuel sales can be
compared to the predicted vehicle miles travelled adjusted for the average vehicle fleet fuel
economy to provide an assessment of the activity associated with motor vehicle use. Many of
these types of comparisons can be facilitated through the use of graphical or regression analyses.
An emissions inventory project should include as many of these simple comparisons as possible.
An agreement among these comparisons can add confidence in the final emissions calculation
without a large investment of resources or time.
In many cases, comparisons of alternate methodologies involve the comparison of a top-
down methodology to a bottom-up methodology. The concept of top-down and bottom-up
approaches has been introduced earlier in this report. Basically, a top-down methodology uses
a country-level or some other aggregate estimate of an inherent activity and an average emission
factor representative of the aggregate activity. Conversely, bottom-up approaches are based on
the additive emissions resulting from specific conditions and operating characteristics of the
population of specific facilities included in the source category. When the preferred inventory
development methodology is based on a bottom-up approach, it is often possible to develop an
estimate using a top-down approach quickly and efficiently. In such cases, the two emissions
estimates can be compared, again without the expenditure of significant resources or time. For
example, the national-level sales of a particular solvent could be compared to the source-specific
estimates of the emissions of that solvent corrected for the amount of the solvent that is recycled
or otherwise reclaimed.
Another type of comparison is based on the average emissions rate or emissions density,
which can then be compared to the sum of source-specific emissions estimates for the population
47
-------
of facilities in a given source category. These comparisons are most valuable when applied to
suorce categories that are comprised of a large number of small sources. An example of this type
of comparison is an estimate of solvent emissions from dry cleaning operations expressed as a per
capita rate, which can be compared to the emissions magnitude derived from records of the
amount of dry cleaning fluids purchased by all of the dry cleaners in an area. Other opportunities
for these types of comparisons include aggregate emissions expressed on a per-employee basis or
on a per-square kilometer basis. These types of comparisons can be particularly useful in a
country with little experience in inventory development. It is often possible to compare an
aggregate average per-capita, per-employee, or per-area emissions density to similarly expressed
emissions for a nearby country of similar economic status. Comparisons of aggregate emissions
densities among countries can be particularly useful in Europe. In general, newly emerging
nations and existing nations that have only recently moved toward cooperation in international
environmental planning do not have well- established environmental programs. Resource and time
constraints may make it difficult for such nations to develop comparison databases internally.
Opportunities for comparison of emissions inventory data with other nearby countries with more
experience in environmental programs and emissions inventory development may represent the
only credible way of assessing the accuracy of these countries' emissions estimates.
Data comparisons based on control totals refer to comparisons of aggregate emissions
and/or activity data to some other independent assessment of the aggregate emissions. The most
common of these types of comparisons involves the comparison of the total emissions developed
through a detailed emissions inventory program to aggregate estimates of emissions completed for
a trend analysis. A trend analysis commonly relies on economic or industrial statistics that are
published by government agencies. These statistics are generally developed for economic analyses
and are often prepared at levels of aggregation that are not consistent with source categories used
to develop emissions inventories. They are sometimes useful to provide a general assessment of
annual emissions with the expenditure of modest resources. These approaches can be applied to
historical records and to projections of the various economic parameters to estimate historical and
projection inventories on the same basis as that for the current year. These analyses are useful in
assessing the impact of control programs and the expected impact of future control programs that
are under consideration by the regulatory authorities. While comparisons of detailed source-
specific emissions estimates cannot be expected to match exactly with the aggregate trends types
of analyses, reasonable agreement should be expected. Comparisons based on these types of
analyses can be useful to point out large uncertainties or inadequacies in the detailed emissions
inventories.
In applications involving the development of emissions control strategies, data on the
extent and effect of existing control measures are also important. In the United States emissions
48
-------
data are often represented as uncontrolled emissions and additional information on control
practices and control efficiencies are included as part of facility records. These control efficiency
estimates can then be applied to the uncontrolled emissions to represent actual emissions. It is
important to check the internal consistency of the individual source records to ensure that the
assumptions about control devices and typical operating control efficiencies are reasonable. The
concept of internal record consistency checks has been discussed previously. These checks are
performed to verify that operating rate data are represented in the proper units, that control
equipment has been identified correctly, and that stated control efficiencies are in the range
expected for that control equipment type. These data can also be checked against similar operating
and control device information for neighboring or nearby countries with a similar history of air
pollution control. A summary of the types of comparisons involving emissions estimates and
supporting data used in the development of emissions inventories is presented in Table 4.
APPLICATION OF COMPARISON DATA
While comparisons based on alternate data sources and alternate emissions estimation
methodologies provide valuable tests of the validity of emissions data, they do not in themselves
prove the accuracy of emissions inventories. The only true test of the accuracy of emissions
estimates is a detailed comparison of the emissions
magnitudes developed through the estimation procedures with simultaneous emissions
measurements or detailed materials balance procedures. Emissions measurements involve actual
source testing methods. Detailed materials balances involve an analysis in which all of the
material input to a system or process is accounted for as being either bound in the product;
released to air, or water or in solid form; or recovered for off-site disposal or for recycle. Each
of these procedures is expensive and time-consuming and can only be accomplished on a limited
scale. Naturally, if source test data are available they provide an excellent opportunity to check
the emissions estimation procedures for those sources that have measurement data.
Emissions comparisons that provide results that are in general agreement do offer
credibility to the final reported emissions estimates, and these types of checks should be included
in emissions inventory development projects whenever possible. In addition to providing data that
can be used to establish the validity of emissions data in an existing inventory, these types of
checks can be used to identify weaknesses and uncertainties in the databases or methodologies used
to generate the emissions estimates. The data or methodology weaknesses identified in these
comparison analyses can direct inventory researchers in the development of future research
activities to improve upon those weaknesses.
The examples included in this discussion suggest some of the types of comparisons that can
be completed. Many opportunities exist for these types of comparisons, and it should be noted
49
-------
TABLE 4. SUMMARY OF DATA COMPARISON OPPORTUNITIES
IN EMISSIONS INVENTORY APPLICATIONS
DATA COMPARISON TYPE
EXAMPLES
Alternate estimation methods
• Emissions magnitudes based on raw material feed
versus product
• Emissions magnitudes based on alternate measures
of the inherent activity
Top-down versus bottom-up methodologies
• National- or regional-level estimates versus source-
specific totals within source categories
Emission density comparisons
• Aggregate estimates for per-capita, per-employee
or per-area compared to total emissions from all
facilities in a source category
• Aggregate emissions densities compared to similar
estimates from other countries or regions
Emission factor comparisons
• Source-specific emission factors compared to
default or average factors
• Uncontrolled emission factors with average level of
control to controlled emission factors
• Emission factors based on alternate measures of the
inherent activity
Control total comparisons
• National totals compared to sum of all source
categories or facilities within source categories
• Summed emissions totals from detailed inventories
compared to national totals in trends analyses
• National totals compared to national totals of
nearby countries corrected for population and
economic status
Completeness checks
• Comparison to earlier inventories to check that all
significant sources are considered
• Checks that all important source categories are
considered
• Checks that all important data elements are
included for facility records
Consistency checks
• Internal consistency for facility data records
• Consistency of methodology for source categories
• Consistency of methodologies between countries in
multiple country inventory development
50
-------
that comparisons at any level of detail are valuable for inventory development programs. The
opportunities available include comparisons of emission factors, activity data bases, and emissions
magnitudes developed through alternate techniques. Care must be taken, however, to ensure that
comparisons are done for sources and source categories that are indeed comparable. Factors
related to climate, economic development status, and other external influences can affect the
degree of comparability that is expected.
51
-------
UNCERTAINTY ESTIMATES
Experimental measurement data are commonly reported as average or preferred values with
an associated error bound expressed in either absolute or relative units. The standard techniques
for estimating experimental uncertainty depend on the known accuracy and precision of the
measurement methods employed in the experiments. Since the accuracy and precision
specifications of the data elements associated with emissions estimates are rarely known, these
standard approaches for developing uncertainty are, in general, not applicable. There is a need,
however, for reporting quantitative error bounds associated with emissions estimates. Several
studies have been completed over the past decade to explore approaches for arriving at reasonable
estimates of the uncertainty associated with emissions estimates. Perhaps the best example of
attempts to establish emissions credibility is represented by the experience with the 1985 NAPAP
Emissions Inventory. The discussion presented in this section includes the procedures and
analyses of emissions uncertainty presented in the State of Science and Technology (SOS/T)
Report No. 1 on Emissions Involved in Acidic Deposition Processes, prepared as part of the
documentation of the NAPAP program.12 Several other approaches have also been proposed,
some of which take issue with the methods used in the SOS/T Report. This report references a
number of recent publications in which the uncertainty issues are discussed in more detail, and
interested parties are encouraged to refer to these documents for further reading on the subject.
Uncertainty estimates for emissions data are important for assessing both the inherent
uncertainty of the emissions estimates for individual facilities and the range of emissions
magnitude represented by all sources in a study area. The uncertainty estimates for individual
facilities are useful to understand the likely impacts of source- specific control options.
Uncertainty associated with the collection of facilities in a source category description and in the
complete inventory are useful to assess the overall quality of the emissions data and the relative
quality of the aggregate estimates of specific pollutants relative to the other pollutants active in
the air quality issues of concern. The issue of emissions data uncertainty continues to present a
significant challenge in air quality management programs. While the statistical techniques
discussed in this report are useful for application to well- characterized sources, they are often not
applicable to many sources of air pollutant emissions. The understanding of the effects of the
estimation assumptions on individual facilities and the effects of assumptions inherent in the
emissions allocation procedures has not yet advanced to the point that allows routine statistical
uncertainty analyses of completed inventories. Efforts to improve this situation are continuing.
52
-------
Since statistically rigorous uncertainty analyses are often difficult to complete, a variety
of data quality rating approaches have been used. These approaches all rely to some degree on
subjective analyses by the developers. Most of these approaches use a letter grade to indicate the
confidence that is associated with the estimates, similar to the A through E rating scheme used in
AP-42. One additional semi-quantitative approach is being considered by the U.S. EPA. This
technique, known as the Data Attribute Rating System (DARS), is being developed jointly by the
Office of Research and Development of the U.S. EPA.13 The goal of this development effort is
to provide a consistent rating system that results in quantitative scores that can be applied in
comparative analyses. The technique is intended to be useful in a wide range of inventory
applications. Conceptually, DARS evaluates the attributes associated with emission factors and
activity data and the reliability of those parameters for the intended applications. The details
associated with DARS are described following the discussion of more traditional statistical
uncertainty analyses.
CAUSES OF EMISSIONS UNCERTAINTY
The fundamental approach for estimating emissions data is based on the application of the
algorithm or model represented in Equation 1:
E = (AF)*(EF)*(1-C eff)*(All Fac) (1)
where: E = Emissions estimate for the source
AF = Activity factor for the source
EF = Emission factor for the source
C eff = Fraction of emissions removed by a control device in
use at the source
All Fac = Factors for spatial, temporal, or species allocation
The uncertainty in the emissions estimate results from the combined uncertainty in each
of the factors on the right hand side of Equation 1. With the exception of the activity factor, each
of these parameters is an estimate that is based on a small number of measurements and is then
universally applied to all sources within a given source category. Even the activity data is often
an estimate, although the inherent uncertainty of the activity data is frequently less than that for
the other parameters.
The recommended values for emission factors and control efficiencies are often based on
the mean of a limited number of measurements for a small sample of the sources included in a
source category. Differences in operating characteristics, maintenance and repair procedures, and
53
-------
in some cases climate and local weather conditions can affect the actual emission factor and
control efficiency as applied to individual sources. In general, the variability of these parameters
is determined by assuming that the individual measurements used to establish the average emission
factor or control efficiency are normally distributed. The assumed uncertainty in these parameters
is then related to the standard deviation of the individual measurements about the mean value.
The uncertainty of activity or operating data for point sources that maintain records of the
raw material feed or amount of product produced is generally low, since these facilities normally
maintain accurate records of these parameters. The estimate of the mean and variation about that
mean for some area and mobile sources are often established through an engineering assessment
or through the use of a model, and the resulting variability of the activity data for these sources
can be quite high. In general, rigorous estimates of the uncertainties in these parameters have not
been developed.
The uncertainty of allocation factors is also often based on engineering assessments. In
select cases, for large point sources estimates of the variability in the application of typical
temporal allocation profiles can be assessed through analyses of actual operating rate data (e.g.,
Figure 2). Even when such analyses are available the application of those estimates to all other
sources in a source category can result in additional unknown error. Spatial allocation of point
sources is generally known with a great deal of accuracy from plant-specific location data. The
spatial allocation of area and mobile sources, however, is normally accomplished by the
application of spatial allocation surrogates that do not reflect the variability in those activities
resulting from personal lifestyles or other external influences. Speciation allocation factors are
the largest source of uncertainty in these applications, and only limited information is available
to assess these uncertainties. In general, uncertainty estimates based on the techniques discussed
later in this section have not been developed for highly resolved speciated inventories, because the
information to complete these analyses is simply not available.
The variabilities in the parameters used in the emissions estimation algorithm (equation 1)
are assumed to result from random errors in the application of mean values for individual sources.
There is also the potential for systematic error or bias in the emissions estimates. Bias results
when one of the parameters used in the emissions estimation algorithm is based on
unrepresentative data or does not consider some essential component of the emissions process.
For example, if an emission factor is developed from source test data that were collected under
load conditions or at operating capacities that are not representative of the normal operating
conditions, that emission factor may include a bias when applied to the normal operating
conditions. The uncertainty in an emission factor, estimated by considering of the standard
deviation of emissions rate measurements performed at 100 percent capacity, will not reflect the
54
-------
systematic under- or over-prediction that would result from a bias in the application of that
emission factor at normal operating conditions at a lower capacity.
For mobile sources, biogenic sources, and other source categories that use an emissions
model to predict emissions, the models often consider only the most important variables that
contribute to the variability in emissions rates. For example, biogenic VOC emissions are often
predicted in the United States with an algorithm that depends on air temperature and sunlight
intensity. Those two parameters explain only about 50 percent of the variability in the measured
emissions rates.14 Neglecting the effect of the variables that contribute to the remaining variability
may introduce an unknown bias into the emission factors applied to biomass sources. Bias in
aggregate emissions estimates can also be caused by the failure to consider all of the sources or
source categories that contribute emissions in an area. Systematic error in emissions estimates is
difficult to predict and the effects of the bias introduced in emissions estimates as a result of
systematic error can have significant effects on air quality analyses that rely on emissions
inventories.
APPROACHES FOR ESTIMATING EMISSIONS UNCERTAINTY
A number of approaches can be taken in ascertaining the level of uncertainty in emissions
estimates for a particular inventory. The approach chosen can be based upon knowledge of the
data distribution characteristics and should include any assumptions used in performing the
emission estimates. The methods used for collection of the data also need to be considered so that
the sample variance can be computed correctly before determining an uncertainty value.
One common assumption made in analyzing uncertainty in emission inventories is that the
data are normally distributed and that the factors used in estimating emissions are independent.
These assumptions allow the use of standard statistical analyses to determine the mean, variance,
standard deviation, and confidence intervals. This approach has been applied in the NAPAP
analyses, and is discussed in detail in the NAPAP SOS/T Report No. 1. A summary of the
technique is presented later in this section.
It has been argued that emission inventory data are not normally distributed and that the
use of the same emission factor for many sources in a given source category negates the
assumption of independence. These factors may limit the ability to calculate the uncertainty in
emission inventory data. These issues have been discussed in some detail by Benkovitz and
Oden15. The methods for calculating uncertainty for an emission source category using a sum of
individual, site-specific emission estimates or using averaged emission factors for all sources were
reviewed. The authors state that the use of emission estimates for individual source sites reduces
uncertainty in almost all circumstances. The authors also acknowledge, however, that the use of
such techniques are not appropriate for source categories where a limited amount of empirical data
55
-------
is available on which to calculate some of the statistics needed to perform these estimates of
uncertainty (e.g., many area sources). The utility of such methods is limited to sources that are
quite well characterized such as power plants. Assuming that the original data used to develop
the averaged emission factor is available, and that the parameters being used in the emission
estimate are independent and do not vary with time, equations developed in this paper could be
used to estimate the mean square error (MSB) of the sum of emissions and the variances needed
to estimate the uncertainty in the emission estimate.
In a second paper, Oden and Benkovitz investigate some of the assumptions commonly
associated with uncertainty analyses.16 Specifically, they explored the assumption that parameters
used in estimating emissions are independent and that their values do not vary with tune.
Although these investigations have not verified the assumptions of independence or that these
variables change significantly over time, the results imply that the factors do not contribute
significantly to the bias of the estimation. The general conclusion was that the emission estimates
based on these assumptions will be unbiased for a wide range of time series. They caution,
however, that proper accounting of the emission estimation parameters is necessary to ensure that
the base assumptions used in determining uncertainty are valid.
A second approach to determining uncertainty involves the use of statistical techniques that
do not require an assumption of data that are normally distributed. Non-parametric analyses, also
known as distribution-free methods can be performed without consideration of the underlying
distribution. A third alternative involves completing a logarithmic transformation of the data to
approximate a normal distribution and then applying standard statistical analyses to the
transformed data. Both of these approaches are discussed in detail by Ferreiro and Cristobal17.
Examples of the application of these procedures are presented in the following discussion.
Analysis of Uncertainty Assuming Normally-Distributed Data
One approach to estimating emissions uncertainty, based on the technique discussed in the
NAPAP SOS/T Report No. 1, is outlined in the following paragraphs. Readers are encouraged
to refer to the SOS/T report for more detail concerning this methodology.'2 The approach relies
on simple statistics including the standard deviation, the coefficient of variation, and the 90
percent relative confidence interval. The standard deviation (S.D.) is a commonly used statistic
that describes, quantitatively, the spread of data points in a population of measurement data. The
coefficient of variation (C. V.) is a measure of the standard deviation relative to the mean value
(i.e., C.V. = S.D./mean). The 90 percent relative confidence interval is used to define the limits
that include 90 percent of all possible measurements in a population assuming that the distribution
of the measurements is a normal distribution. In a normal distribution, 90 percent of the possible
measurement values lie within a range bounded by ± 1.64 tunes the standard deviation.
56
-------
In this approach, the mean and standard deviation are determined for each element of the
emission estimation equation. For the example presented here, we assume that the emissions
estimate is the simple product of the AF and the EF, or E = (AF)*(EF). The C.V.s associated
with the activity factor and the emission factor are then calculated. The C.V.s are combined to
determine the C.V. of the final emissions estimate by applying Equation 2.
(1 + (C.V.)2emiss) = (1 + (C.V.)2AF) * (1 + (C.V.)2EF) (2)
The 90 percent relative confidence interval is then calculated for the final emissions estimate by
Equation 3.
90 percent RCI = C.V.emiss * 1.64 (3)
An example, derived from the NAPAP SOS/T Report No. 1, is summarized in Table 5.12
The 90 percent confidence interval can also be estimated for the sum of these two sources
following similar procedures. The mean emissions for the sum of the two sources is 900. The
standard deviation of that mean can be estimated from Equation 4.
(S.D. [combined emissions])2 = (S.D. [plant #1])2 + (S.D. [plant #2])2 (4)
The standard deviation estimated for the sum of the emissions from the two plants is then 79.16.
This value gives a C.V. of 79.16/900, or 0.09. The 90 percent relative confidence interval is then
0.15, which yields a 90 percent confidence interval of 767 to 1033. The result gives a range for
the 90 percent confidence interval that is less than the sum of the 90 percent confidence interval
for the two individual sources. This result demonstrates the concept that the overall uncertainty
decreases as the number of sources increases.
Similar analyses can be performed for emissions estimates from a collection of source
categories. An analysis demonstrating this concept based on emissions estimates from the 1985
NAPAP Emissions Inventory is presented in Table 6. The C.V.s for the emissions categories
listed in Table 5 were derived from previous work completed by Chun.18 Two analyses
representing a lower and upper limit of the variabilities of NOX emissions from the utility,
industrial combustion, and transportation sectors are represented. These C.V.s are based on the
available data and the most reasonable assumptions and as such represent the best estimates at this
time. These analyses have been applied to establish that the overall uncertainty in the NOX
emissions estimates for the 1985 NAPAP emissions inventory is between 6 and 11 percent.
Similar statistical relationships were also provided for the SO2 estimates in the NAPAP inventory,
and the results indicate that the overall uncertainty in the annual SO2 emissions estimates is
between 4 and 9 percent. Unfortunately, similar analyses were not completed for the annual VOC
57
-------
TABLE 5. EXAMPLE OF AN UNCERTAINTY CALCULATION
PLANT # 1
PLANT # 2
ACTIVITY DATA
MEAN
STANDARD DEVIATION
C.V.
100.00
2.00
0.02
50.00
1.00
0.02
EMISSION FACTOR
MEAN
STANDARD DEVIATION
C.V.
5.00
0.5
0.10
8.00
1.2
0.15
EMISSIONS ESTIMATE
MEAN
STANDARD DEVIATION
C.V.
90 % RCI
90% CONFIDENCE
INTERVAL
500.00
51.00
0.10
0.16
416 to 584
400.00
60.54
0.15
0.25
301 to 499
emissions estimates, because credible assumptions about the variability of the VOC emissions for
the major emissions sectors have not been developed. The overall variability in VOC estimates
is much larger that those determined for NOX and SO2. The procedure presented in Table 5 could
be applied to VOC if valid C.V.s for the emissions estimates for the large VOC sources are
developed. The development of valid C.V.s will require a significantly improved understanding
of the errors associated with the application of surrogate allocation data, emission factors and
control efficiencies, and the relationships between these factors that affect the independence
assumption.
58
-------
TABLE 6. UNCERTAINTY ESTIMATES FOR NOX EMISSIONS IN THE 1985 NAPAP EMISSIONS INVENTORY
SOURCE CATEGORY
EMISSIONS
Tg/year
C.V.
S.D.
(S.D.)2
90%
RCI
90% CONFIDENCE
INTERVAL
Low Variability Case3
Electric Utilities
Industrial Combustion
Industrial Process
Residential/Commercial
Transportation
All Other Sources
6.09
2.25
0.84
0.62
803
0.81
0.07
0.15
0.25
0.30
0.02
0.40
0.43
0.34
0.21
0.19
0.16
0.32
Total 18.64 0.04 0.71
0.182
0.114
0.044
0.035
0.026
0.105
0.505
0.11
0.25
0.41
0.49
0.03
0.66
0.06
5. 42 to 6.76
1.69 to 2.81
0.50 to 1.18
0.32 to 0.92
7.79 to 8.27
0.28 to 1.34
17.52 to 19.76
High Variability Case"
Electric Utilities
Industrial Combustion
Industrial Process
Residential/Commercial
Transportation
All Other Sources
6.09
2.25
0.84
0.62
8 03
0.81
0.09
0.20
0.25
0.30
0.10
0.40
Total 18.64 0.06
0.55
0.45
0.21
0.19
0.80
0.32
1.15
0.300
0.203
0.044
0.035
0.645
0.105
1.331
0.15
0.33
0.41
0.49
0.16
0.66
0.10
5. 18 to 7.00
1.51 to 2.99
0.50 to 1.18
0.32 to 0.92
6.75 to 9.31
0.28 to 1.34
16. 78 to 20.50
The low and high variability cases reflect different assumptions about the variability of emissions from utility sources with low NO, burners and from the industrial combustion
and transportation sectors from Chun.
For the totals, the overall S.D. is calculated as the square root of the sum of the squares of the S.D.s of the individual entries. The overall C.V is calculated from the overall
S.D., which is then applied to determine the 90% confidence interval.
C.V. = coefficient of variation; S D. = standard deviation; RCI = relative confidence interval
-------
Another example of the application of uncertainty analyses has been presented by
researchers at the TNO Institute of Environmental Sciences in The Netherlands.19 This publication
provides an excellent discussion of the sources of emissions uncertainty and the concept of
propagation of errors in emissions inventory development applications. This report presents
quantitative uncertainty estimates for NOX emissions from on-road vehicle emissions and stationary
combustion sources. The report recognizes the difficulties in developing quantitative uncertainty
estimates for other sources of NO, and nearly all sources of VOCs.
The TNO report presents an analysis of uncertainties or expected variabilities for the major
factors (e.g., the miles travelled by specific vehicle categories, the emission factor for specific
vehicle categories, combustion and fuel characteristics in stationary combustion sources,
measurement instrument error, sampling errors, and compounded errors in scaling spot
measurements to annual aggregation) that affect NOX emissions from these two important
categories. The results of the study indicate that the overall uncertainties associated with motor
vehicle and stationary combustion sources of NOX are 4 percent and 2.6 percent, respectively.
These analyses suggest that the emissions inventory techniques applied to The Netherlands national
inventory underestimate true NOX emissions slightly and that those estimates contain an overall
uncertainty of approximately 3.5 percent.
Analysis of Uncertainty Using Distribution-Free Methods
Ferreiro and Cristobal have reviewed the methods for performing analyses using non-
parametric or distribution-free techniques in an overall discussion of statistical handling of
emission inventory sampling and analyses.17 The methods that use summary descriptive statistics
to characterize the distribution of emissions related data are presented along with methods for
identifying the median, lower and upper fourths, outlier cutoffs, and extreme values (outliers) are
discussed in some detail. The techniques to calculate confidence limits for the sample estimates
and the description of the sample structure are also given in this paper. In this paper, the authors
concentrated on the statistical treatment of these parameters and did not present any applications
of these techniques to existing inventory data. The interested reader should review the paper for
details and formulations.
An example of the application of non-parametric statistics to actual data has been presented
by Khalil.20 The uncertainties associated with global budgets of several atmospheric trace gases
were estimated using a method that relies only on the ranges of emission rates. A simple range
estimate based on the preferred value and expected range for each major category of the inventory
was compared with a 90 and 95 percent confidence limit on the ranges constructed statistically
from the range of each component. The resulting calculation of a 90% confidence interval
produces a narrower range of emissions than that derived through an additive analysis. The base
60
-------
assumption for this technique is that the probability that any value in a sample range is equally
likely (uniform distribution) instead of being clustered around some sample mean. These analyses
can produce more meaningful range estimates for assessment and sensitivity studies.
SENSITIVITY ANALYSES
The types of uncertainty analyses presented above for annual emissions estimates are useful
to assess the credibility and validity of the emissions data and the emissions inventory development
process. The application of the emissions data at the annual level are, however, limited. Many
of the analyses that use the emissions data require the chemically speciated inventories resolved
to appropriate temporal and spatial formats. Estimates of the 90 percent confidence intervals for
the resolved emissions estimates are used by modelers to perform sensitivity studies. Sensitivity
studies are used to test the model predictions over the likely range of emissions to determine if
different results are obtained at the extremes of the emissions range. Ideally, the sensitivity
of model predictions to emissions inputs would address the individual components that contribute
to the variability of the emissions. These results would assist both modeling and inventory
development researchers to target the most critical issues for further research and improvements.
Unfortunately, the current state-of-the-science on emissions inventory development cannot support
these types of analyses. For example, consider the case of an industrial process that uses a
solvent-based metal degreasing operation. If fugitive emissions of the solvent are not included in
the point source inventory, the spatial allocation methodology may incorrectly add all or part of
the emissions to an inappropriate grid cell. If the degreasing process is conducted intermittently
in batch-mode, the use of a standard eight-hour work day would result in emissions allocation to
hours of the day when no emissions are present. If outdated data are used to specify the chemical
formulation of the degreasing materials, the speciation profiles will generate specific chemicals
that may not even be used at the plant. In these cases, the idea of means and C.V.s to describe
the expected variability of the hourly emissions rate of a particular compound would be
meaningless, in the context of the analyses presented above. It could be stated that the uncertainty
of the emissions data is 100 percent of the specified value; however, these types of estimates are
not very useful for application to model sensitivity studies.
DATA QUALITY RATING SYSTEMS
Emissions inventory estimates are often calculated as functions of process rates,
manufacturing units, control technologies, and factors for spatial, temporal, and species allocation.
Estimates of each of these parameters are often based on a small number of measurements and the
estimate is then universally applied to all sources within a given source category. Differences in
operating characteristics, maintenance and repair procedures, and in some cases climate and local
61
-------
weather conditions can affect the actual emission factor and control efficiency as applied to
individual sources.
Spatial allocation of point sources is generally known with a great deal of accuracy from
plant-specific location data, but the spatial allocation of area and mobile sources usually requires
the application of spatial allocation surrogates that often do not reflect the variability in those
activities resulting from personal lifestyles or other external influences. Similarly, surrogates of
temporal operating characteristics are often applied to allocate emissions to seasonal, daily and
hourly levels when specific operating data are not available. Species allocation factors are the
largest source of uncertainty in these applications, and only limited information is available to
assess these uncertainties. In general, uncertainty estimates based on the techniques discussed here
have not been developed for highly resolved species inventories, because the information to
complete these analyses is simply not available.
Therefore, a meaningful measure of the overall reliability of an emissions inventory can
sometimes be developed by the application of a data rating scheme. Rating schemes can have
different formats, but each sets up some arbitrary scale that is applied to score individual
emissions estimates at the appropriate level of aggregation. Several rating schemes have been
discussed in the context of the UN ECE Task Force. Each of the schemes is briefly summarized
below.
The U.S. EPA has long used a rating system for its preferred emission factor listings
included in its AP-42 document.1 This technique uses a letter rating system of A through E to
represent the confidence in emission factors from best to worst. In this system A factors are based
on several measurements of a large number of sources, and E factors are based on engineering or
expert judgement. The U.S. EPA has recently expanded this approach to include a letter based
rating of the emissions estimate as well as for the emission factor. While there are some
guidelines for the assignment of the letter score, this approach is largely subjective.
A similar method has been used in Great Britain for assessing the overall quality of
emissions estimates.21 In this approach letter ratings are assigned to both emission factors and the
activity data used in the emissions estimates. The combined ratings are reduced to a single overall
score following an established schedule. The emission factor criteria for the letter scores are
similar to those applied in the U.S. EPA's approach and scores for the activity data are based
largely on the origin of the data. Published data either by a government agency or through an
industry trade association are assigned C ratings and extrapolated data based on a surrogate would
receive an E rating.
The IPCC has included a rating scheme in its guidelines for reporting of greenhouse gas
emissions through international conventions.22 This scheme uses a different approach. For each
pollutant associated with major source categories a code is specified to indicate the coverage of
62
-------
the data included in the estimate. The codes indicate if the estimate includes full coverage of all
sources or partial coverage due to incomplete data or other causes. Additional codes can be
specified to indicate if the estimate was not performed, included in some other category, not
occurring or not applicable. An additional rating is then applied to each pollutant for each source
category to indicate the quality assessment of the estimate as either high, medium, or low quality.
Two additional ratings are requested that apply to the source categories without reference to
specific pollutants. These ratings cover the quality of the documentation supporting the estimates,
rated as either high, medium, or low; and a rating to indicate the level of aggregation represented
in the estimate. The possible choices are 1 for total emissions estimated, 2 for sectoral split and
3 for a sub-sectoral split. This rating scheme has more detail but retains a simplicity that allows
the analyst to quickly review the quality ratings and to compare the quality ratings to other
estimates.
Another rating approach has been developed and is being used by researchers in the
Netherlands.23 This approach recognizes the difficulties in getting agreement from several
organizations in international efforts on the specific needs of emissions data quality and on
definitions of data acceptability criteria. In this approach two specific issues are addressed
concurrently in the rating scheme. The first is an assessment of the accuracy or uncertainty in the
emissions estimate, and the second is an assessment of whether decision makers have confidence
in the application of the estimates for regulatory and policy activities.
In this approach two scaling indicators are applied to represent these two concerns. The
first is a letter grade from A through E that indicates the inventory developers assessment of the
overall quality of the estimate. A ratings imply the highest quality and accuracy and E ratings
imply that the estimate is an educated guess. The second rating scale applies a letter code to
indicate the purpose for which the estimate was prepared and offers the policy maker a quick
assessment of the reliability of the estimate for a given application. These rating categories and
their associated applicability are listed below.
Applicability Rating Description
N National Level
R Regional Level
L Local Level
I Industry Level
P Plant Level
63
-------
This indicator is meant to provide information to the user to enable judgment of the level
of aggregation put into the estimate. For example, when an emission factor is based on national
averaged numbers and therefore aimed at estimating the national total emissions, it is assigned a
rating of "N", and the user would be cautioned against the application of this factor for any
specific plant, or for only one section of a country where conditions may be different. Likewise
estimates based on plant level data, with a rating of "P", would not be used with high confidence
to estimate the regional total emissions for an emission sector.
One more quantitative approach to address this need is being developed by the Office of
Research and Development of the U.S. Environmental Protection Agency.13 In this approach a
numerical value is associated with the quality of the various components or attributes of an
emissions inventory. This technique, called the Data Attribute Rating System (DARS), seeks to
establish a list of attributes that can affect the quality or reliability of the emission factor and
activity data associated with the emissions estimate for any given source category. A numerical
scale is used to rank these attributes in a relative priority against a set of criteria selected to
represent the reliability of each attribute estimate. This procedure will allow a comparative
assessment of the overall quality of the alternate emissions estimates for a specific category or for
a group of high priority source categories in an urban or regional inventory. Some additional
details of this proposed approach are provided in Appendix A and further information will be
available during 1995.
An Example of a Data Quality Rating Approach
Table 7 represents an application of the concepts of qualitative data rating schemes for the
most important pollutants of concern in air quality analyses organized by major source category
groupings. It is important to note that any such qualitative summary is subjective and that
individual researchers may not agree with every entry listed in the table. While the subjective
nature of this approach is recognized, the data ratings summarized in Table 7 represent a general
consensus among the members of the UN/ECE Expert Panel on Verification, given the current
understanding of emissions inventory estimation methods. The letter grade ratings summarized
in Table 7 are primarily applicable to the estimation approaches for emissions inventory
preparation that rely on emission factors and estimates of activity indicators. In all cases, the
application of more direct approaches based on measurement would receive higher quality ratings.
The application of these subjective ratings for the aggregated source category groupings
represented, can be misleading in some cases. For example, the rating specified for heavy
metals/persistent organic pollutants for road transport is listed as E to apply in general to the
understanding of the contribution of these pollutants from mobile sources. In fact for the specific
case of lead from mobile sources the emission factors and emissions estimates are known with
64
-------
TABLE 7. EMISSION INVENTORY UNCERTAINTY RATINGS
SOURCE CATEGORY GROUPING
1 public power, cogeneration and district heating
2 commercial, institutional & residential combustion
3 industrial combustion
4 industrial processes
5 extraction & distribution of fossil fuels
6 solvent use
7 road transport
8 other mobile sources and machinery
9 waste treatment and incineration
disposal activities other
10 agriculture activities
11 nature
SO2
A
B
A
B
C
C
C
B
C
D3
NO,
B
C
B
C
C
C
D
B
C
D
D
voc
C
C
C
C
C
B
C
D
B
C
D
D
CO
B
C
B
C
C
C
D
C
C
D
E
NH3
E
E
E
D
E
HM/POP
D
E
D
E
E
E1
E2
E
D
E
E
E3
CO2
A
B
A
B
D
B
C
B
C
D
D
CH4
C
C
C
D
D
C
D
C
D
D
E
N2O
E
E
E
D
E
D
E
E
E
E
1 In some cases, solvents may be toxic compounds
2 Rating representative of typical pollutant source category combination; some specific cases may have higher ratings
Natural sources could be contributed from volcanoes and other geothermal events
-------
significantly more confidence. In such an analysis at that level of disaggregation, lead from
mobile sources would receive a B rating. Also at this level of aggregation several source category
pollutant combinations are irrelevant in that emissions of the pollutant from that source category
are zero or so minimal as to be of little or no importance.
The value of a rating scheme such as that summarized in Table 7 is enhanced when applied
in conjunction with a table of total emissions from each pollutant organized in the same matrix
format. The researcher can then compare relative quality ratings in consideration of the overall
contribution of that category to the total loadings of emissions of the specific pollutant species.
The appearance of sources of significant amounts of pollutants with corresponding low quality
ratings can serve to caution researchers on the applications of the inventory and direct efficient
research efforts in future programs to improve the quality of the overall inventories.
SUMMARY
In summary, procedures for estimating emissions uncertainty have only recently been made
available for specific, well-understood emissions sources and the pollutants that are unmistakably
associated with those sources. While these techniques are useful to assess the relative accuracy
and validity of the aggregate emissions estimates, they have not yet evolved to the level where
they can be rigorously applied to sensitivity studies of photochemical air quality models and other
more detailed analyses associated with emissions control options and
control strategy development. Further efforts are ongoing and opportunities are being explored
for analyses that can be used to evaluate the validity of an emissions inventory in terms of its
influence on air quality analyses. Some of these opportunities will be discussed in the next section
of this report.
The deficiencies in inventory development methodologies and the lack of rigorous
estimates of the error in existing inventories have serious impacts on air quality analyses and air
quality management programs. These concerns have been expressed concisely in a recent book
published in the United States by the National Research Council:
The development and refinement of methods for estimating emissions
is a fundamental element in the implementation of the regulatory
approach to air quality management. Demonstrating the accuracy
of emissions estimates has been the Achilles' heel in evaluating the
effectiveness of air quality management control strategies and is a
fundamental flaw in the regulatory process.14
66
-------
GROUND TRUTH VERIFICATION
The implementation of the procedures and activities discussed in the preceding sections of
this report will provide the basis for the development of a well-documented emissions inventory
database. The application of techniques to establish a likely range for the final emissions
estimates, and the methodologies for the estimation of uncertainty can assist in a qualitative
assessment of the representativeness and relative accuracy of the inventory data. Ground truth
verification involves techniques that make direct comparisons between emissions estimates and
some other known quantity that is related either directly to the emissions source or indirectly to
the underlying process that results in emissions. While ground truth verification procedures can
be resource intensive they will often provide the most powerful and quantitative method for data
verification and should be incorporated into emissions inventory development programs whenever
possible.
The purpose of this section is to discuss some of the techniques that can be applied to
ground truth verification. Many of the techniques discussed in this section have been applied in
the United States and elsewhere, and have provided useful insights into the major weaknesses of
existing inventory estimation methodologies and resulted in additional research programs to
improve these methodologies. Other promising approaches based on innovative applications of
modeling and mathematical concepts are also discussed. One approach is based on the work
reported by Johnson, et al,24'25'26 that describes a unique way to analyze the urban ozone formation
process. Other approaches based on artificial intelligence and the mathematical formulations
associated with artificial intelligence techniques are also mentioned. The conceptual basis for
these approaches is presented here, although the details of the analysis technique are necessarily
limited, since practical applications of these techniques are not available.
The increase in verification studies within the global scientific community has been
discussed in previous sections of this report. A valuable example of this interest is the work on
verification that is being conducted in support of the IKARUS project (Instruments for Greenhouse
Gas Reduction Strategies) which is discussed in a recent publication of the Research Center Julich
(KFA) sponsored by the Federal Republic of Germany.27 This research addresses many
verification issues in the context of coordinated international climate change programs. One
aspect of this work involves verification of emissions inventory data and other data used in support
of the development of emissions inventories. This work is of particular importance to European
emissions inventory development projects, because it addresses specifically the needs related to
transparency of emissions and other analysis efforts that involve more than one country. This
67
-------
report discusses many of the specific ground truth analyses that are discussed in the remainder of
this section of this report. Interested readers are encouraged to explore the IKARUS project and
to consider the adoption of interesting results that will be developed through this program.
This discussion of ground truth verification techniques is organized into the three major
groupings listed below. In some cases, the actual implementation of specific activities might
involve elements from more than one of these groups. Some results of actual field measurement
programs based on some of these techniques are discussed, for applications where such results are
available. As mentioned earlier in this report, innovative mathematical approaches such as
artificial intelligence, fuzzy logic, chaos theory and fractal geometry should be explored for
application in emissions verification programs. Since these approaches have not yet been adapted
to this application, no examples are available and they have not been discussed in this section of
the report. The major groupings included here are:
• statistical survey analyses
• monitoring analyses
• modeling analyses
STATISTICAL SURVEY ANALYSES
Some common methodologies for estimating emissions from area source emission
categories rely on a per- capita, per-employee, per-land area or some other surrogate emission
factor. While these approaches may be adequate for estimating national or regional emissions.
they may introduce bias when applied to specific locations or during specific time periods. One
method for verification of emissions estimates based on a surrogate or aggregate emission factor
is to perform a survey on a selected sample of the target population of individual sources included
in the aggregate source category. Ferreirro and Cristobal have discussed several approaches that
can be followed to select an appropriate sample for survey analyses.17
Although the focus of this work is related to the verification of an existing emission
inventory, the approaches discussed could also be used in the development of emission factors or
other parameters used in calculating emissions for a source category (activity measures, control
efficiency, etc.). The discussion stresses the need to characterize the target and sampling
populations and to specify the sample design as an initial step. Characterization of the source
populations includes descriptions of the sectoral, spatial and temporal dimensions of the inventory;
an extremely important consideration when the emission inventories developed by a number of
organizations are expected to be compared, each with differing categorical schemes.
Haphazard, judgement, and probability principles are discussed as sample selection alternatives.
The haphazard method is only appropriate when the target population is homogenous. Selection
68
-------
of the most easily accessible sources is favored using this technique, which may introduce biases
and is not amenable to assessing the accuracy of the estimates derived. The judgement method
relies solely on the subjective knowledge of experts to select representative sample units.
Although the use of experts is a prerequisite for developing a sound sampling design, it is difficult
to verify if it is used as the sole method for choosing samples. Probability sampling includes
simple random sampling, stratified random sampling, two-staged cluster, systematic, and double
sampling. The first two methods are discussed in detail by Ferreirro and Cristobal. Simple
random sampling assigns an equal chance of any one unit in a population being chosen. The
simple random sampling method is a valid technique only if there is low variability among subsets
within the target population.
Stratified random sampling benefits from simplicity of design and that reduction in the
overall variance of the sampling errors can be achieved in an efficient and cost effective manner.
The stratified random sampling approach is recommended whenever possible. The target
population is first divided into a number of subsets which are non-overlapping and together cover
the whole population. These subsets or strata should be homogenous within themselves and show
diversity between themselves. Strata can be partitioned based on size, technologies, or any other
attributes of the population units.
Data collected prior to a verification (i.e., in the original emission inventory) can also be
used in the verification process. This is true for both the simple random sampling approach or
in stratified random sampling. Care must be taken, however, to place previously collected data
into the proper strata for stratified random sampling, and to assure that the samples are randomly
selected before using previously collected data (i.e., sample selection should be performed before
determining what data are available).
Methods for presenting the results of these analyses are also discussed, with the
recommendation of the use of the box plot and K-number representations. These representations
of the data depend on the median and other categorical factors to determine the sample
distribution. Determination of the confidence limits is possible for these categorical factors by
using either normal distribution statistics or non-parametric procedures. Another alternative is to
perform a logarithmic transformation of the data to approximate a symmetrical distribution and
thereby "normalize" the data before using in standard statistical techniques. Methods for testing
the data for conformity to a standard distribution, identification of outliers, and testing the general
consistency of the empirical data are also discussed.
An example of such an approach is the application of a per capita emission factor for
estimating VOC emissions from dry cleaning operations. On average in a particular country, a
per capita emission factor can provide an accurate estimate of annual emissions on the national
scale. It is likely, however, that highly populated urban areas with a large concentration of
69
-------
service-related or government employment would have higher per-capita emissions than rural,
agricultural areas. The development of total emissions using the per-capita emission factor and
the distribution of those emissions based on the population spatial surrogate factors might result
in a skewed distribution of the emissions rate in specific areas. It is also possible that the use of
dry cleaning services could be related to seasonal factors that could affect the temporal distribution
of emissions.
Statistical sampling techniques, such as those discussed by Ferrierro and Cristobal17, could
be implemented to identify the population of dry cleaning establishments that need to be sampled
in detail to provide a statistically rigorous definition of the regional and temporal distribution of
the activity associated with the dry cleaning industry. The results of a statistical sampling based
on these principles could be applied to develop emission or allocation factors that depend on
population density, economic demographics, or the distribution of employment by major industrial
and commercial sectors.
These approaches may be used to develop more relevant emissions estimation
methodologies or to assess the potential regional or temporal bias introduced by application of the
simple population-based approach. These approaches could be applied to other source categories
in addition to the dry cleaning industry. Some of these categories include autobody repair,
residential wood-fuel use, and heavy-construction equipment. The application of these techniques
requires an understanding of the emission sector under consideration to ensure that a representative
and random sample of facilities is included in the survey.
Another possible survey approach could be implemented to evaluate the representativeness
of resolved emissions estimates applied in urban modeling analyses. In this approach, some of
the grid cells that are thought to contribute significant emissions magnitudes could be surveyed
in a methodical way. This would involve a detailed manual survey of all stationary activities in
a given grid cell location that can result in emissions. Emissions estimates for the collection of
facilities in the grid cell would be developed based on material balance or other sound engineering
assessment. Details of the specific chemicals used in these industries, the diurnal nature of the
processes resulting in emissions, and other factors related to the emission sources could be
obtained through this survey. The results of these compilations could then be compared to the
emissions magnitude and temporal and species assumptions applied in the allocation approach used
to develop the modeling inventory, to assess the weaknesses of those assumptions.
In survey applications, the initial sampling and manual survey could be accomplished
simply to identify the facilities to be included in a follow-up survey, using questionnaires that
could be sent to the facilities to acquire the required data. The use of a follow-up questionnaire
would reduce the personnel resources required for the study, but would also result in a burden to
those facilities selected for the survey. The use of questionnaires is also affected by the
70
-------
willingness of the selected facilities to respond and by any bias in the estimates of the activities
associated with those facilities that would be introduced by the person or group of people who
would be responsible for completing the questionnaire.
MONITORING ANALYSES
Monitoring analyses include three principal types of measurement activities: direct source
testing, indirect source testing, and ambient measurements. All monitoring programs are
expensive to implement and should be well planned and executed to maximize the data recovery
and to ensure the collection of high-quality measurement data. It is possible, in some cases, to
apply measurement data that are routinely collected as part of a government-sponsored air quality
management program, and data that are routinely collected by individual facilities to monitor
process operation and efficiency to an emissions verification exercise. Whenever a monitoring
program is considered, a thorough review of all existing measurement data should be completed
and the program should be designed to make use of these data whenever possible. Table 8
presents a summary of the potential uses of monitoring data for emissions inventory verification.
Direct Source Testing
Direct measurement of stack gas emissions, using CEMs, is sometimes required to
establish compliance with environmental regulations. Obviously, if such measurements are
available to the agency responsible for the development of an emissions inventory, those data can
be applied directly to the inventory. In these cases, there is no need for the application of an
emissions estimation technique. More commonly, however, such compliance data are only
available for limited periods of time, or for only a subset of the population of sources in a given
area. However, compliance data collected for some specific facilities or over a limited time
period, along with similar data collected specifically for application to an emissions verification
program, can be used to evaluate emission factors and emissions estimation techniques.
The application of this type of direct source testing data simply involves the comparison
of the emission factor derived from the measured data to the emission factor used in the estimation
technique or the measured emissions to the emissions calculated through the application of the
estimation technique. When using compliance data, it is important to consider the operating
conditions in effect during the test. In many applications, compliance tests are performed during
periods of maximum load. If normal operating conditions are at a slightly reduced load, the
results of the compliance test under maximum load may result in a bias when compared to the
estimation technique.
71
-------
TABLE 8. MONITORING TYPES, EXAMPLES, AND USES FOR
EMISSIONS INVENTORIES
Monitoring
Class
Examples of Monitoring
Programs
Uses of the Data for Emissions
Inventories
Direct
Measurements
• Inprocess emissions
measurements
• Process operating parameters
• Random sampling of process
units or potential leak tests
• Comparison to estimated values
• Identification of ranges of
application estimates (operating
parameters, emissions factors)
• Specification of fugitive emissions
or process leaks
Indirect
Measurements
• Remote measurement
systems: FTIR, UV, Gas Filter
Correlation
• Ambient VOC/NOX ratio
studies
• Comparison of estimated emission
rates with near source concentrations
• Estimation of emission factors for
sources that do not have stacks or
vents
Ambient
Studies
• Tunnel Studies
• Aircraft Studies
• Upwind-downwind
difference studies
• Receptor Modeling
• Identification of obvious
weaknesses in procedures or
underestimation of emissions
• Checking of ambient impacts of
sources or mixtures of sources
• Identification of principal emissions
sources in a region
The U.S. EPA and Environmental Agencies in other industrialized countries have
published a series of standard methods for application to emissions testing of stationary sources.
These standard methods have been developed to promote consistency and comparability between
independent measurement programs.
The U.S. EPA has also established methods for testing highway gasoline-powered motor
vehicles. Many States and local agencies have adopted the Federal motor vehicle test procedures
or variations of those procedures, known as motor vehicle inspection and maintenance (I/M)
programs, as part of their air quality management plans. Although the I/M programs are designed
to identify in-use vehicles that do not comply with Federal emissions standards, they are in fact
a direct source emissions test. Applications of these tests and the experience gained through their
routine operation in many urban areas in the United States has led to the development of
72
-------
assumptions about the distribution of emission rates in the overall vehicle fleet. These tests
monitor directly the exhaust gas generated during idle and/or specific load conditions.
Indirect Source Testing
Direct source testing methods are primarily applied to large stationary sources where
emissions are vented through a clearly identifiable stack or vent. Indirect source testing methods
are used to estimate emissions from dispersed sources. These types of sources are either too
numerous to consider individually , like residential space heating, or arise from unexpected
sources, like leaks in chemical plants or petroleum refineries. Some examples of indirect source
testing are described below.
Measurement of Operating Parameters. It is not always necessary to measure the direct
emissions from a source to quantify the actual emissions. For sources that have relatively high-
quality emission factors or emission estimation algorithms that have been demonstrated to predict
emissions with a high degree of accuracy over the typical range of operating conditions, emissions
can be monitored by collecting and processing these operating rate data. For example, SO2
emission factors for coal- and oil-fired boilers are relatively well known. These emission factors
are expressed as a function of the sulfur content of the fuel. Therefore, accurate monitoring of
the fuel use rate and the sulfur content of that fuel can be applied to make highly accurate
estimates of the SO2 emission rate as a function of time.
Random Sampling of Leak Sites. Developing emissions estimates for fugitive losses of
VOCs from equipment and process leaks in large, complex chemical plants, pharmaceutical
manufacturing facilities, and petrochemical refineries is particularly difficult. Typically, emission
factors based on an average measured emission rate for pumps, flanges, valves, storage equipment
and process equipment, are applied to all similar plants, even though emissions measurements for
those specific plants are not available. Any generic approach based on this general methodology
will introduce bias when applied to specific operating facilities. Differences in operating
conditions, chemicals used, maintenance procedures, and products produced in individual facilities
contribute to the potential for bias when applying these techniques. In addition, the emission rate
measurements frequently focus on those devices that are characterized as being the most significant
leaking sources. One more refined approach is to routinely monitor a random sample of these
emission points in an individual facility to apply a statistically representative distribution of those
data to all of the potential leak sites. In this approach a more representative emission estimate can
be developed for any individual facility without the need for direct measurement of all of the
specific leak sites in operation at the facility.
Remote Measurement Techniques. The sampling procedures used in direct emissions
source tests are based primarily on the collection of a captive sample, which is then directed to
73
-------
an analysis cell or enclosed in an airtight container that is delivered to a controlled laboratory
setting for analysis. These types of measurements are frequently referred to as in-situ sampling
techniques. Remote measurement techniques are measurement methods that do not rely on the
collection of the captive sample. Instead they make a determination of concentration or some
other physical property in an undisturbed air parcel. A common remote measurement method that
is familiar to everyone is radio detection and ranging, or radar. Radar systems can pinpoint the
location of a mass that reflects radio waves emitted from a centrally located transmitter. The
vector and time delay of the radio reflection identify the location of the mass that is measured.
Similar principles can be applied to measure the concentration of gases in air. Some examples of
remote measurement technologies applicable to measure air quality parameters include the
following:
• light detection and ranging (LIDAR)
• differential absorption LIDAR (DIAL LIDAR)
• Fourier transform infrared spectroscopy (FTIR)
• ultraviolet (UV) spectroscopy
• gas-filter radiometer
Each of these technologies can be used to measure the concentration of pollutants along
a line of sight through the ambient atmosphere or at the mouth of an emissions source. The
measurement principle of all of these techniques is based on the physical properties of molecules
that serve to alter the wavelength or intensity of light waves as those light waves pass through an
air sample. All molecules have properties that result in the absorption of characteristic
wavelengths of electromagnetic energy. The energy that is absorbed by the molecules alters either
the vibrational, rotational, or electron orbit energy state of the molecule. Simply stated, the
measurement is performed by monitoring the change in the energy signal of the light between the
light source and the light receiving equipment. The LIDAR techniques involve a narrow
waveband laser as a light source. The infrared, ultraviolet, and gas-filter correlation spectroscopy
techniques use a lower power light source that emits energy over a wide range of wavelengths.
All of these measurement techniques are experimental and only FTIR and Gas Filter
Radiometry have been used in field tests to explore opportunities in emissions measurements
programs. These techniques may have applications in emissions verification programs in the
future. The following sections present a brief summary of some of the experimental applications
of these technologies.
Example of an FTIR Emissions Test. The U.S. EPA Office of Research and
Development is supporting research to demonstrate the application of remote measurement systems
based on FTIR technology for estimating emission rates from large dispersed area sources.28
74
-------
These particular tests were performed to evaluate the technique in measuring emissions from a
surface coal mine to develop an emission factor. However, the method could just as easily be
applied to check or verify the results of an emission estimation methodology for the source. The
example discussed here is designed to measure methane (CH4) emissions from surface coal mining
operations, but the basic approach could be applied to other similar sources including but not
limited to; landfills, waste water treatment systems, emissions from agricultural pesticide
application, controlled field burning, and wildfires. Adaptations to the basic approach are also
being explored by U.S. EPA to measure emission rates from other industrial activities such as
fugitive emissions from process leaks, and fugitive emissions from chemical storage facilities.29
The approach involves the measurement of a path averaged concentration along a line of
sight that is immediately downwind of the area source. In the application described here, the line
of sight measurement was obtained at approximately one meter above the surface along the lip
edge on the downwind side of the coal pit. The application of the method requires simultaneous
measurements of the wind speed and direction, temperature, and atmospheric pressure. These
parameters are used to estimate the shape and extent of the emissions plume using simple
dispersion modeling techniques. The path average measurements are used along with the results
of the simple dispersion modeling techniques to calculate the emissions source strength required
to produce the calculated plume.
The results of these tests provided reasonable emissions estimates; however, experimental
and equipment problems were encountered during the exercise. Following these preliminary tests,
EPA initiated additional research to evaluate the technique and to further develop the approach.
These tests were carried out in a research field in the Research Triangle Park area of North
Carolina in the summer of 1992.
The results of the tests were encouraging. A large percentage of the experiments yielded
calculated emission rates within 25 percent of the known emission rate from a simulated area
source. Meteorological factors do affect the results of the modeling approach. Research efforts
to refine these methodologies for use in emissions factor development and leak detection from
volume sources are currently being supported by the U.S. EPA Office of Air Quality Planning
and Standards.29
Another application of FTIR measurements of emissions from wildfires in Australia has
been reported.30 In these experiments, emissions of CO2, CO, CH4, CH2O, CH3OH, HCOOH,
CH3COOH, NO, NO2, NH3, N2O, and HCN were obtained simultaneously for several grass fires
and fires of piled forest debris. The emissions of all compounds were expressed as a ratio relative
to the emissions of CO2. The tests conducted in these experiments provide good examples of the
use of the FTIR open path measurements as an emissions verification technique. The results of
the study indicate higher than expected emission rates for reduced and partially reduced species
75
-------
including NH3, CH,, and CH,O. In fact, the results of these tests revealed that NjH was the
dominant form of nitrogen emissions from these fires. Conventional thought would not have
predicted that the emissions of NH3 would be higher than the oxidized forms of nitrogen.
Example of a Gas-Filter Radiometer Emission Test. Researchers at the University of
Denver have developed a gas-filter radiometer that is used to remotely measure emissions of CO
from in-use highway motor vehicles.31'32 The system uses an infrared light source that is mounted
on one side of a single lane road segment. In these tests measurements were made on an on-ramp
to a restricted access highway. The detector unit is operated on the other side of the roadway.
The line of sight is set at approximately 10 inches above the road surface to be in line with the
exhaust pipe. The instrument uses three detectors, one for CO, one for CO2, and one as a
reference channel. As each car passes the detector, there is a drop in the reference signal caused
by the car interfering with the light beam. This drop in the reference initiates a measurement.
As the car exits the beam, a one-second voltage versus time trace is obtained and stored
electronically in the device. The system uses a rotating filter wheel, with one cell filled with CO
and H2 and the other cell filled with N2.
A freeze-frame video system is used to make a permanent record of each vehicle tested.
This system can be operated unattended and more than 1,000 vehicles can be tested per day. After
the test period, the system is retrieved and taken to the laboratory where information is
downloaded into computer-based files for further processing. This instrument has been operated
through several measurement periods and has collected CO emissions data for more than 117,000
individual cars. An instrument based on a similar method but without the gas-filter wheel has
been developed by General Motors Corporation.33 Both of these instruments have been operated
in side-by-side tests to verify that they give comparable results.
These systems are primary examples of the use of indirect source measurements for
inventory verification. The results of the measurements were compared to emissions estimates
obtained by use of the standard MOBILE4 motor vehicle emission factor model and the EMFAC7
motor vehicle emission factor model developed by the State of California. In all such tests, a
significant discrepancy was found between the CO emissions estimates predicted with the models
and the resulting CO emission rates measured with the instruments. The results of these studies
indicate that 50 percent of the CO emissions from all vehicles result from less than 10 percent of
the cars. The standard motor vehicle emissions models do not adequately treat the high emitting
vehicles and, consequently, underestimate total exhaust emissions. These indirect measurement
methods of in-use vehicles are more representative of actual fleet emissions than controlled tests
in I/M programs because vehicle owners can plan for and make adjustments to the onboard
emissions control equipment prior to the I/M tests, yielding tests that are not representative of the
actual conditions under which the vehicles are operated in normal use. In-use measurement
76
-------
studies have identified the use of the standard motor vehicle emission factor models as a major
source of error in urban VOC and CO emissions inventories. The results of these studies provide
excellent examples of measurements-based approaches to emissions inventory verification projects
that can help focus future research and method development that will result in improvements in
emission inventory estimates.
Ambient Measurements
Several techniques have been developed and applied that seek to relate ambient
measurement data to emissions source strengths. The studies using these techniques have been
conducted to assess the reliability of general overall emissions estimation methods for use in
regulatory applications. These studies can be categorized in one of three major groupings:
ambient ratio studies (YOG/NO,, and CO/NOX), tunnel studies, and receptor modeling studies.
The concepts of using these types of studies for emissions verification and some examples of each
are discussed in the following paragraphs.
Ambient Ratio Studies. In the United States, ambient measurement programs are routinely
operated in urban areas that are classified as nonattainment for the ambient ozone standard.
Typically, these measurement programs include a rural measurement site in a location that is in
the typical upwind sector, two or more sites in the downtown area near the urban core, and two
or more sites in the downwind sector at locations that are thought to represent the location of the
ozone maxima events. Both grid-based and trajectory modeling approaches are used to simulate
the urban area and model predictions are compared to the observed concentrations of ozone and
ozone precursors. The use of grid-based models allows the investigator to track the temporal
distribution of ozone precursors in the urban center in addition to the ozone maxima. Frequently,
these models are reasonably successful in predicting the ozone maximum in the downwind
locations, but are less successful in tracking the concentrations of precursors in the downtown
area. These types of results have led researchers to question whether the underlying emissions
inventories are adequately representing the actual emissions fields, and to question if control
strategies based on these modeling predictions are valid.
A detailed research study, sponsored by the California Air Resources Board (ARB), has
been completed to assess the ability of standard emissions inventory estimation methodologies to
adequately represent the concentrations of nonmethane organic gases (NMOC), CO, and NOX, and
the ratios of NMOC/NOX, and CO/NO, observed in selected regions of the South Coast Air Basin
surrounding Los Angeles.8 In this study, ambient air measurement data were evaluated to estimate
the NMOC/NOX and CO/NQ under various temporal scenarios in regions that are primarily
influenced by highway motor vehicle emissions. Similar ratios were determined from the gridded,
hourly, and speciated emissions inventory under a variety of spatial averaging assumptions.
77
-------
Statistical analyses were performed to determine the most appropriate temporal scales for these
comparisons. Ratios derived from early morning measurements (7 to 8 a.m. local time) were the
most appropriate for comparisons of the NMOC/NO, and CO/NOX in the summer months, and
similar ratios derived from overnight measurements were found to be most appropriate in the fall
months. Spatial averaging was evaluated for the single grid cell covering the measurement site,
for multiple grid cells surrounding the measurement site, and for the basin average emissions
estimate. In the summer months, comparisons between the measured data and the highway motor
vehicle inventory were analyzed, and in the fall months some contribution of stationary source
NOX emissions was included. The results of the study suggest that the measurements-based ratios
for NMOC/NOX and CO/NOX are about 2 to 2.5 and 1.5 times higher, respectively, than the
corresponding ratios derived from the emissions inventory. If it is assumed that the NOX
emissions inventory is representative of the motor vehicle NOX emissions, these results suggest that
current mobile source emissions methodologies underestimate NMOC and CO emissions by those
factors. Other measurements-based ambient data collected from tunnel studies (discussed below)
suggest that the overall NOX emissions from motor vehicle sources is reasonably estimated from
existing emissions methodologies. Further modeling exercises using grid-based photochemical
models yield similar discrepancies between the measured and estimated emission rates.
Additional studies, designed to evaluate the comparisons of ambient NMOC/NOX ratios
with those derived from emissions inventories, have been conducted by the U.S. EPA. In the first
study, similar comparisons of NMOC/NOX ratios derived from ambient measurements and
emissions inventory estimates were developed for 15 cities in the United States.9 The emissions
inventory ratios were based on county wide emissions estimates taken from the 1985 NAPAP
emissions inventory database. The baseline comparisons in this study showed that, on average,
the emissions inventory ratios were 24 percent lower than the ratios derived from the ambient
measurements. The baseline emissions inventory was modified to include enhancements to
account for biogenic VOC emissions, to update the mobile source emissions to be representative
of MOBILE4, and to account for rule effectiveness assumptions. The comparisons based on the
modified inventory data were, on average, 2 percent lower than ratios based on the ambient
measurement data. Comparisons for individual cities ranged from a value where the inventory-
derived NMOC/NOX ratio was 173 percent higher than the measurements-based ratio to a value
where the inventory ratio was 61 percent lower than the measurement ratio.
Further analyses on four cities including (Houston, Texas; Philadelphia, Pennsylvania (two
sites); St. Louis, Missouri; and Washington, DC) were explored in a follow up study.10 In these
analyses, gridded emissions data based on the NAPAP inventory were compared to measurements-
derived ratios for measurement sites contained within the grid cell. The same enhancements were
made to the emissions inventory data, although in these tests the effect of the biogenic component
78
-------
was minimal since the comparisons were made during early morning hours and the sites were
urban areas with little vegetative cover.
The results of these studies suggest that on average the NMOC concentrations derived from
the inventory data were underestimated by 50 to 80 percent relative to the ambient measurements.
At one site, the emissions inventory overestimated NMOC concentrations by about 20 percent.
The emissions inventory data for Houston, however, underestimated NMOC concentrations by
330 percent. This result may include bias due to the neglect of potentially significant point source
NMOC emissions resulting from the large population of petrochemical industry sources in that
area, which are outside of the selected grid cell. With the exception of the data for Houston, the
discrepancy between emissions derived NMOC and ambient NMOC was not strongly dependent
on ambient temperature, implying that the sources that are underestimated in the emissions
inventory are not temperature- dependent. The Houston data did show a significant relationship
to temperature, suggesting that sources such as organic liquid storage may be the cause of the
emissions that are not adequately reflected in the inventory.
There are several other examples of similar or related studies to assess the relationships
between emissions inventory-derived NMOC/NOX and CO/NOX ratios and those derived from
ambient measurements. In all cases, the trend is toward underestimation of VOC emissions in the
emissions inventories. Each of these studies is based on simplifying assumptions that can affect
the comparisons and, therefore, should be viewed as preliminary. The consistent results indicating
moderate to significant underestimation of VOC emissions, however, leads to the conclusion that
most emissions inventories underestimate VOC and CO emissions in the aggregate, and the
development of emissions control strategies based on these inventories may be inadequate to
achieve the desired reduction in ambient ozone and CO concentrations.
The U.S. EPA Office of Mobile Sources recently completed a study to review 5 specific
studies of comparisons between ambient VOC/NO, and CO/NOX ratios and similar ratios derived
from emissions inventories.34 This study concludes that these ratios in ambient monitoring data
are higher than those calculated from emissions inventory data, suggesting that the VOC and CO
inventories underestimate actual emissions or that the NOX inventories overestimate actual
emissions. There are several problems related to source distributions and mixing assumptions that
influence these results. Additional efforts are required to establish the true meaning of these types
of comparisons, but the studies completed so far have provided useful first steps to a potentially
valuable emissions verification tool.
Tunnel Studies. Highway tunnels offer an excellent location to sample the contribution
of emissions from in-use highway vehicles. Tunnel studies were first used in the United States
to estimate in-use emissions rates from highway vehicles in 1981.35 More recently, a study was
completed in 1987 in a tunnel under the Van Nuys airport runways in the Los Angeles area.36 In
79
-------
the Van Nuys study, concentrations of VOC, CO, NOX, and paniculate matter (PM) were
measured at both the upwind and downwind portals of the tunnel and the emissions rate was
measured by difference. Air flow was determined by simultaneously monitoring the exit
concentration of an SF6 tracer, which was introduced at the upwind portal at a known release rate.
Video images were recorded to accurately assess the distribution of vehicles in the tunnel during
the measurement periods. Average vehicle speed was measured in each lane of the tunnel during
each experiment. Altogether 22 measurement periods were monitored during October and
December of 1987. The measured concentration data were used to estimate the mass emissions
rate for the sampling periods.
The temperature, vehicle type distribution, and average vehicle speed data were applied
in the California mobile source emission factor model (EMFAC7) to calculate the mass emissions
rate for each monitoring period. The emissions rates estimated from the EMFAC7 model were
then compared with those derived from the measurements. In all 22 measurement periods, the
emissions rates of VOC and CO, based on the measured concentration data, exceeded those
predicted by the emissions model by factors of 1.4 to 6.9, and from 1.1 to 3.6 for VOC and CO,
respectively. The comparisons for NOX, however, were very consistent with those predicted by
the model. Ratios of measured to modeled emission rates for NOX varied from 0.6 to 1.4, with
a mean ratio of 1.0. These results suggest that the EMFAC7 model is adequate for estimating
NOX emission factors but consistently underestimates emission factors for both VOC and CO. If
only VOC emissions were underestimated it might be concluded that the EMFAC7 model did not
adequately treat running loss evaporative emissions. Since both CO and VOC are consistently
underestimated, however, it is more likely that the EMFAC7 model fails to account for the
percentage of vehicles that are gross emitters. It is assumed that the vehicles using the tunnel over
this time period represent a random sample of the overall vehicle fleet in the Los Angeles area.
The use of EMFAC7, therefore, could underestimate basin-wide motor vehicle emissions by a
factor of 2 to 5. The consistent underestimation of VOC and CO emissions from motor vehicle
sources has serious implications on the development of area-wide control strategies based on
emissions inventories that are developed through the use of the EMFAC7 or MOBILE4 mobile
source emission factor models.
The earlier studies completed in the Allegheny Tunnel35 also imply a consistent
underprediction of emissions by mobile source emission factor models. The earlier studies suggest
a consistent underprediction of VOC, CO, and NOX by the models. It is difficult to compare the
studies, however, because earlier studies used earlier versions of the mobile source emission factor
models that have since been updated to include additional evaporative loss emissions.
Comparisons between the results of these two tunnel studies are further complicated by differences
80
-------
in the mix of vehicle model years, the mix of vehicle types, and the speeds associated with traffic
in the two tunnels.
Aircraft Monitoring Studies. Results from an interesting measurement program have been
reported by researchers at the TNO Institute of Environmental Sciences.37 This paper presents
results of study comparing emissions inventory results to ambient monitoring results using an
instrumented aircraft platform. Aircraft measurements are made in both the upwind and
downwind direction from a cluster of sources of specific VOC emissions. The measurements are
obtained at various altitudes to define a flux of the selected pollutants in both locations and the
difference is attributed to the combined emissions strength of all sources in the area within the
measurement planes. In these preliminary studies the analyses focused on the major known
sources of a list of five specific VOC species representative of the source mix being studied. The
results of the study indicated a significant potential of this methodology to verify emissions
inventory procedures. Although the estimates of the emission strengths determined from the
aircraft measurements were consistently higher than the source strength estimated from the
inventory, the ranges of emissions estimates determined by the two estimation approaches overlap
in general. These results are encouraging and indicate that further exploration of this and other
related techniques should be explored in future studies.
Receptor Modeling. Receptor modeling is the use of ambient measurement data and the
chemical characteristics of specific source types to estimate the contribution of different sources
to the observed concentration of pollutants in a sample. The most common use of receptor
modeling techniques is in chemical mass balance (CMB) receptor modeling analyses. CMB
studies rely on the known distribution of VOC species in the major sources or source categories
in an area. The specific compounds selected for analyses are known as the fitting species. The
criteria for the selection of fitting species is that they cannot be emitted in large amounts from
other sources, and they cannot be subject to photochemical reaction during transport between the
emission point and the receptor monitoring location. For instance, acetylene is a primary
component of motor vehicle exhaust and is not emitted in large quantities from other types of
sources. Acetylene is also relatively unreactive in photochemical systems and, therefore, the
acetylene concentration measured in an ambient air sample can be assumed to result from motor
vehicle exhaust. The concentration of the fitting species in a given air sample relative to the total
VOC concentrations can be used to estimate the relative contribution of that source to the
measured VOC concentration. In this way, the relative contribution of the major sources can be
assessed.
In one example, the U.S. EPA has performed a CMB receptor modeling analysis of
ambient air samples collected in Atlanta, Georgia, during the summer of 1990.38 This study was
a preliminary analysis to demonstrate the technique and only the contribution of mobile sources
81
-------
was estimated. The Atlanta area is a good location for use of this analysis technique to estimate
the contribution of mobile sources because high-quality mobile source speciation data were
available and mobile sources dominate the VOC emissions in the Atlanta area. The results of the
study suggest that 77 percent of the VOC emissions result from mobile sources. Other studies that
provide the split between mobile sources and stationary sources have also been suggested for the
Atlanta study area. Similar studies can be implemented in any urban area that has a relatively
complete measurement program that provides speciated VOC measurement data. These studies
can be used to determine whether the emissions inventory database adequately represents the
relative contributions of general source categories. Receptor modeling approaches can also be
used to estimate the relative contributions of individual facilities in a group of four or five adjacent
industrial facilities. In this application of receptor modeling, ambient concentration excursions
over small time scales are analyzed relative to short-term variations in winds in the immediate
study area. These data can be interpreted with dispersion models to locate the specific source
regions of the fitting species. These analyses can be used to construct the distribution of emissions
magnitudes for the individual sources in the group of industrial facilities.
An approach for a more complex application of receptor modeling analyses has recently
been presented.39 In this approach, functions that describe the transformation of pollutants due
to photochemical reactions can be applied to account for the mix of sources that contribute to the
observed ozone precursors. The model proposed for these analyses is a hybrid receptor model,
called the source identification through empirical orthogonal functions (SITEOF). The model
relates gradients of empirical orthogonal functions and winds to the sum of sources and sinks of
a pollutant. The model was applied in a test case, and provided qualitatively reasonable estimates
of the formation rates for ozone in both winter and summer. These results could be applied along
with trajectory analyses to analyze the distribution of emissions represented in the emissions
inventory to those contributing to the formation of ozone derived through the receptor modeling
approach. The details of application of these techniques are complex and further reading is
suggested to any interested reader.39'40
A receptor modeling approach based on filter samples and combined x-ray fluorescence
(XRF) and instrumental neutron activation analysis (INAA) has been applied to identify the
sources of paniculate matter collected in ambient urban environments.40 The technique uses the
chemical mass balance approach and has been used to estimate the relative contribution of various
sources to the ambient paniculate concentrations in Philadelphia, Pennsylvania. The results of
these studies indicate that the vehicle exhaust portion of the PM-10 measurement was between 4
and 6 percent, stationary sources contributed 5 percent, and the sulfate component of PM-10 was
between 49 and 54 percent. Model trajectory analyses indicated that approximately 80 percent
of the sulfate paniculate resulted from local sources. Further applications of these techniques have
82
-------
been combined with computer-controlled scanning electron microscopy (CCSEM) and transition
electron microscopy (TEM) to enhance the data collection and resolution capabilities of these
techniques.41 The combined use of the electron microscope data, which provides particle
morphology information in addition to the chemical composition information obtained through
XRF and INAA, allows greater resolution of the source characteristics and enhances the resolution
at which source contributions to the ambient particle loaded can be determined.
Integrated Empirical Rate Model. A unique methodology appropriate for the study of
urban ozone formation processes has been developed by Johnson, et al, in Australia.24 The
approach, known as the integrated empirical rate (IER) model, simplifies the representation of
ozone formation processes relative to the gridded, emissions-based modeling approaches that have
been common in the United States and elsewhere for the past 15 years. The IER model was
developed to simulate a series of smog chamber experiments that were performed under conditions
that are actually observed in Sydney, Australia. Sydney is a large city with a mix of mobile and
stationary sources of ozone precursors that is representative of most industrial urban areas. Smog
chamber experiments performed in the United States are operated under conditions of VOC and
NOX concentrations and VOC/NOX ratios that are atypical of actual urban conditions. This is
because, in the United States, ambient air is used to flush the chambers, and it is necessary to
charge the chambers with elevated concentrations of the ozone precursors to overcome the effects
of precursor concentrations in the ambient air. In the Australian experiments, a high-efficiency
clean air generator was employed, which allowed the researchers to obtain a highly purified
background condition. The ability to establish a clean background allows simulations of VOC and
NOX conditions that represent ambient urban conditions.
The basis of the IER approach relies on the representation of the ozone formation process
on a temporal scale as a function of cumulative sunlight rather than as a function of time of day.
In more traditional analyses, the ozone concentration is plotted, whereas in the IER approach the
net photochemical oxidation is plotted. The net photochemical oxidation is defined as the sum of
NO to NO2 transformations plus the ozone production. Figures 3 and 4 represent actual smog
chamber data collected by the Australian researchers, as a function of time of day and as a
function of cumulative sunlight, respectively.
The striking observation of these plots is that, in the IER representation (Figure 4), oxidant
formation is a linear process as a function of cumulative sunlight, with a maximum that is defined
by the initial amount of NOX available in the system. Johnson labelled the portion of the ozone
formation process characterized by the linear increase the light-limited regime, and the maximum
possible ozone formation the NOx-limited regime. The slope of the line in the light-limited regime
is then a direct measure of the hydrocarbon reactivity of the sample. A less obvious but more
important observation of these plots is that the inherent reactivity of the hydrocarbon mix available
83
-------
oo
.u
•
O
CL
o
o
I
O
C
I
re
I
eg.
era
+ [NO] - [NO], ppm
p
b
p
b
p
01
3 m
P m
< T3
8
O
O
n
o
B3
O
O
•*9
H
n
O
Concentration, ppm
p
to
p
in
'83UejpeJ.il /\D
•£.
J5-*
vj
-------
in the system is conserved with time during the light-limited regime. Ambient measurements of
ozone, NO, NO2, and VOC can be analyzed to locate the air sample on the line representing the
light-limited regime. It is then a simple process to translate the cumulative sunlight intensity into
elapsed time, and to estimate the origin of the emissions sources by performing a back trajectory
analysis.
This approach can provide a significant opportunity for emissions verification to assess the
usefulness of emissions inventories prepared for the modeling-based regulatory applications. One
application of the approach would use the assessment of reactivity and oxidant formation on an
iterative basis to account for the contribution of emissions sources in each grid cell during each
hour. The results of these analyses would provide a high-resolution estimate of the total emissions
of reactive VOC and NOX. Comparisons of these estimates to those developed through the
standard inventory preparation and allocation methodologies would identify those critical regions
that contain sources that are inadequately treated by current inventory methodologies. The results
of the analyses could be used to identify those particular source types that have the greatest impact
on ozone formation processes and ultimately on ozone control strategies.
A Measurement Instrument Based on 1ER. One result of the research conducted by
Johnson, et al, is the development of an air quality measurement instrument that is designed
specifically to measure the chemical properties associated with the IER.42 The system This
automated measurement system, called AIRTRAK monitors the NO, NOX, ozone, photochemical
oxidation, and reactivity of samples. The system can be automated to collect canister samples for
laboratory analysis of speciated hydrocarbons. The advantage of this device is that it can be
programmed to collect the canister samples during regularly scheduled periods, such as 6 a.m. to
9 a.m.; during periods when a significant reactivity event is occurring: or when there is a change
in the oxidant formation rate. This would concentrate measurement resources on significant
events rather than on collecting routine measurements, many of which are of limited interest. The
use of a network of AIRTRAK instruments would provide base information that could be applied
to the model-based inventory verification approach discussed above, in addition to providing all
of the information routinely collected in standard monitoring programs.
The Lake Michigan Ozone Study used four AIRTRAK measurement systems during the
summer 1991 sampling period.43 The data were transferred to Australia for analysis by the
instrument developer. Results of these and additional measurement applications of AIRTRAK are
expected soon, to establish the capabilities of the instrument for air quality and emissions
estimation applications.
85
-------
11. S.L. Kersteter, et al, Identification and Characterization of Unaccounted For Area Source
Categories. (EPA-600-R-92-006) U.S. Environmental Protection Agency, Research
Triangle Park, North Carolina, January 1992.
12. National Acid Precipitation Assessment Program, State of Science and Technology Report
No. 1: Emissions Involved in Acidic Deposition Processes. December 1990.
13. L. Beck, L.A. Bravo, B.L. Peer, M. Saeger, and Y. Yan, A Data Attributes Rating
System. Presented at the International Conference on the Emission Inventory:
Applications and Improvement, Raleigh, NC, November 1-3, 1994.
14. National Research Council Committee on Tropospheric Ozone Formation and
Measurement, Rethinking the Ozone Problem in Urban and Regional Air Pollution,
Chapter 9, Emissions Inventories. National Academy Press, Washington, DC, 1991.
15. C.M. Benkovitz and N.L. Oden, Individual Versus Averaged Estimates of Parameters
Used In Large Scale Emissions Inventories. Atmospheric Environment, Vol. 23, No. 5,
pp. 903-909, 1989.
16. N.L. Oden and C. M. Benkovitz, Statistical Implications of the Dependence Between the
Parameters Used For Calculations of Large Scale Emissions Inventories. Atmospheric
Environment, Vol. 24A, No. 3, pp. 449-456, 1990.
17. A. Ferreiro and A Cristobal, Some Statistical Tools for Inventory Verification. Prepared
for the UNECE Task Force on Emission Inventories: Second Meeting and CORINAIR
Expert Meeting, Delft, The Netherlands, June 1993.
18. K.C. Chun, Uncertainty Data Base for Emissions Estimation Parameters: Interim Report.
ANL/TM-328, Argonne National Laboratory, Argonne, Illinois, March 1987.
19. T. Pulles and Hans-Peter Baars, Uncertainty of Emissions Inventories - Methodology and
Preliminary Results for the Dutch Emission Inventory, Presented at the EPA-AWMA
Conference on Emission Inventory Issues in the 1990's, September 10-12, 1991, Durham,
N.C.
20. M.A.K. Khalil, A Statistical Method For Estimating Uncertainties In the Total Global
Budgets of Atmospheric Trace Gases. Journal of Environmental Science and Health, Vol.
A27(3), pp. 755-770, 1992.
21. Summary of Verification Quality Ratings presented by Stephen Richardson AEA
Technology, United Kingdom, at the Third Meeting of the U.N. ECE Task Force on
Emissions Inventories, Regensburg, Germany, May 1994.
87
-------
REFERENCES
1. Compilation of Air Pollutant Emission Factors. Volume I: Stationary Point and Area
Sources. AP-42. Fifth Edition (GPO No. 055-000-00251-7), U.S. Environmental
Protection Agency, Research Triangle Park, North Carolina, January, 1995.
2. M. Saeger, et al, The 1985 NAPAP Emissions Inventory (Version 2): Development of the
Annual Data and Modelers' Tapes. (EPA-600/7-89-012a) U.S. Environmental Protection
Agency, Research Triangle Park, North Carolina, November 1989.
3. Quality Review Guidelines for 1990 Base Year Emissions Inventories. (EPA-450/4-91-
022) U.S. Environmental Protection Agency, Research Triangle Park, North Carolina,
September 1991.
4. Personal Communication with Lee L. Beck, U.S. Environmental Protection Agency,
Global Warming Control Branch, Office of Research and Development, Research Triangle
Park, NC 27711, February 23, 1993.
5. L.G. Modica and D.R. Dulleba, The 1985 NAPAP Emissions Inventory: Development of
Spatial Allocation Factors. (EPA-600/7-89-010b) U.S. Environmental Protection Agency,
Research Triangle Park, North Carolina, April 1990.
6. D.B. Fratt, et al, The 1985 NAPAP Emissions Inventory: Development of Temporal
Allocation Factors. (EPA-600/7-89-010d) U.S. Environmental Protection Agency,
Research Triangle Park, North Carolina, April 1990.
7. R.A. Walters and M.L. Saeger, The 1985 NAPAP Emissions Inventory: Development of
Species Allocation Factors. (EPA-600/7-89-010f) U.S. Environmental Protection Agency,
Research Triangle Park, April 1990.
8. E.M. Fujita, et al, Comparison of Emission Inventory and Ambient Concentration Ratios
of CO, NMOG, and NOX in California's South Coast Air Basin. Journal of The Air and
Waste Management Association. 42: No. 3, pp. 264-276. March 1992.
9. K. Baugues, A Review of NMOC, NOX and NMOC/NO, Ratios Measured in 1984 and
1985. (EPA-450/4-86-015) U.S. Environmental Protection Agency, Research Triangle
Park, North Carolina, September 1986.
10. K. Baugues, Further Comparisons of Ambient and Emissions Inventory NMOC/NOX Ratios.
Presented at the 85th Annual Meeting and Exhibition of The Air and Waste Management
Association, Kansas City, Missouri, June 21-26, 1992.
86
-------
22. Greenhouse Gas Inventory Reporting Instructions; Final Draft Volume I.
Intergovernmental Panel on Climate Change, PAris France, 1994.
23. Summary of Verification Concept presented by Tinus Pulles, TNO Institute of
Environmental Studies, Delft, The Netherlands at the Third Meeting of the U.N. ECE
Task Force on Emissions Inventories, Regensburg, Germany, May 1994.
24. G.M. Johnson, A Simple Model for Predicting the Ozone Concentration of Ambient Air.
Proceedings, Eighth International Clean Air Conference, Vol. 2, pp. 715-731. Eds. H.F.
Hartman, J.N. O'Heare, J. Chidodo, and R. Gillis. Clean Air Society of Australia and
New Zealand, 1984.
25. D.S. Wratt, et al, Predicting the Impact of a Proposed Gas-Fired Power Station on
Photochemical Pollution Levels Around Auckland. Proceedings, Tenth International
Conference of the Clean Air Society of Australia and New Zealand, March, 1990.
26. G.M. Johnson, et al, Management of Photochemical Smog Using the AIRTRAK Approach.
Tenth International Conference of the Clean Air Society of Australia and New Zealand,
March 1990.
27. G. Stein, IKARUS Instruments for Greenhouse Gas Reduction Strategies, Subproject 9:
Verification, Second Interim Report, October 1991 - November 1992. Contributors: W.
Fischer, W. Katscher, U. Kotte, W. Lauppe, J.C. di Primio, S. Comes, U. Fritsche, F.
Matthes, J.F. Lauchbery, and J. Salt.
28. D.A. Kirchgessner, et al, Estimation of Methane Emissions From a Surface Coal Mine
Using Open-Path FTIR Spectroscopy and Modeling Techniques. Presented at the NATO
Advanced Research Workshop on Atmospheric Methane Cycles: Sources, Sinks
Distribution and Role in Global Change, October 6-11, 1991. Submitted for publication
in Chemosphere.
29. J. Rucker and M. Saeger, Final Test Plan for Emission Measurement Method
Development Using FTIR Open-Path Monitoring, Report Submitted to Emissions
Measurements Branch, U.S. Environmental Protection Agency, Rima Dishakjian Principal
Investigator, May 1994.
30. D. W. T. Griffith, et al, FTIR Remote Sensing ofBiomass Burning Emissions of CO2, CO,
CH4, CH20, NO, NQ , NI/, and ft O. Presented at the NATO Advanced Research
Workshop on Atmospheric Methane Cycles: Sources, Sinks Distribution and Role in
Global Change, October 6-11, 1991. Submitted for publication in Chemosphere.
31. D.H. Stedman, Automobile Carbon Monoxide Emission. Environmental Science and
Technology, Vol. 23, No.2, pp. 147-149, 1989.
-------
32. G.A. Bishop and D.H. Stedman, On-Road Carbon Monoxide Emission Measurement
Comparisons for the 1988-1989 Colorado Oxy-fuels Program. Environmental Science and
Technology, Vol. 24, No. 6, pp. 843-847, 1990.
33. R.D. Stephens and S.H. Cadle, Remote Sensing Measurements of Carbon Monoxide
Emissions From On-Road Vehicles. Journal of The Air and Waste Management
Association, Vol. 41, No. 1, pp. 39-46, January 1991.
34. G. Yarwood, et al, Evaluation of Ambient Species Profiles, Ambient Versus Modeled
NMHC:NOX and CO:NOX Ratios, And Source-Receptor Analyses, Office of Mobile Sources
U.S. Environmental Protection Agency, Final Report No. SYSAPP94-94/081, September
1994.
35. R.A. Gorse, Jr., On-Road Emission Rates of Carbon Monoxide, Nitrogen Oxides, and
Gaseous Hydrocarbons. Environmental Science and Technology, Vol. 18, No. 7, p. 500,
1984.
36. M.N. Ingalls, On-Road Vehicle Emission Factors From Measurements in a Los Angeles
Area Tunnel. Presented at the 82nd Annual Meeting and Exhibition of The Air and Waste
Management Association, Anaheim, California, June 25-30, 1989.
37. M.P.J. Pulles, J.J.M. Berdowski, and M.G.M. Roemer, Industrial Emissions of
Hydrocarbons: Inventoried Emissions and Measured Concentrations in the Atmosphere,
Presented at the First Meeting of the UNECE Task Force on Emissions Inventories,
London, England, May 4-6, 1992.
38. C.W. Lewis and T.L. Conner, Source Reconciliation of Ambient Volatile Organic
Compounds Measured in the Atlanta 1990 Summer Study, The Mobile Source Component.
Presented at the AWMA/EPA Conference on Emission Inventory Issues in the 1990's,
September 10-12, 1991, Durham N.C.
39. R.C. Henry, A Receptor Modeling Approach to Ozone Regulation. Presented at the 84th
Annual Meeting and Exhibition of The Air and Waste Management Association,
Vancouver, British Columbia, June 16-21, 1991.
40. T.G. Dzubay, et al, A Composite Receptor Method Applied to Philadelphia Aerosol.
Environmental Science and Technology, Vol. 22, No. 1, pp. 46-52, January 1988.
41. B.C. Henderson, et al, Rapid Acquisition/Storage of Electron Microscope Images. Journal
of The Air Pollution Control Association, Vol. 33, No. 10, October 1983.
89
-------
42. G.M. Johnson and S.M. Quigley, A Universal Monitor for Photochemical Smog.
Presented at the 82nd Annual Meeting and Exhibition of The Air and Waste Management
Association, Anaheim, California, June 25-30, 1989.
43. N.E. Bowne, The Lake Michigan Ozone Study: The Data. Presented at the 85th Annual
Meeting and Exhibition of The Air and Waste Management Association, Kansas City,
Missouri, June 21-26, 1992.
90
-------
APPENDIX A
SUMMARY OF THE PROPOSED DATA ATTRIBUTES RATING SYSTEM
One of the recommendations of this report is to complete statistical uncertainty analyses
or direct source measurements for specific components of any emissions inventory whenever it
is possible. These forms of verification are superior to all others because they provide a
quantitative assessment of the accuracy and reliability of the emissions inventory estimates.
Since direct measurement programs are prohibitively expensive in almost all inventory
development programs and the data required to perform statistical uncertainty analyses are often
unavailable, it is desirable to have an alternate approach that will provide some measure of a semi-
quantitative assessment of the quality of emissions inventory data. The report suggests that as a
minimum some form of data quality rating approach be implemented. These data quality rating
schemes are subjective and at best offer a qualitative assessment of the overall accuracy and
reliability of the emissions estimates. One option to provide a more quantitative quality rating
method is being developed by the Office of Research and Development of the U.S. Environmental
Protection Agency. The method, known as the Data Attribute Rating System (DARS), is still in
a conceptual stage. The principle behind this approach is to characterize the emission factor and
activity data applied in any source category in terms of the underlying attributes that affect those
parameters.
Currently, the attribute lists for both the emission factor and activity parameters include
a measurement attribute, a source definition attribute, a spatial scale attribute, and a temporal scale
attribute. The emission factor may also include a pollutant specific attribute. Each of the
attributes is then assigned a score based on a set of criteria. The score for all attributes are then
added and the sum is divided by the maximum possible score to define a dimensionless scale value
from 0.0 to 1.0, that can be used to compare alternate estimates for a given parameter or to
compare the reliability of different parameters in the same inventories. Once the parameter scale
values are derived the quality of the overall emissions estimate can be assessed by multiplying the
parameter scores for an overall score for the emissions estimate.
Although this approach is still considered a subjective approach, a reasoned selection of
attributes and tha acceptance of a defined set of criteria for scoring the individual attributes can
provide a more consisetent numerical value to replace the letter scale grading systems used in most
other accepted rating systems. This approach offers the potential for a significant breakthrough
in emissions inventory verification by providing two important elements. First, the approach can
be universally applied to inventories on many scales (source, urban, regional and global), and it
A-l
-------
is based on a common terminology for the rating of inventories, thereby enhancing the
transparency of the rating system and the rated inventories. The selection of this single approach
for a detailed discussion in this appendix is not meant to imply that other rating systems are not
useful. Rather, this approach is discussed separately in an attempt to begin a debate within the
emissions inventory community in the United States and Europe that will ultimately result in the
development of a universal approach that will satisfy the verification needs of a large number of
inventory types. The principals of the DARS are outlined in the following discussion.
Description of the Attributes
A discussion of the attributes that affect emissions data quality is presented in the following
paragraphs. A suggested scoring scale and a brief discussion of the criteria for selecting a
particular score is provided with the discussion of each attribute. It will become apparent to the
reader that the application of the suggested criteria and the ultimate scoring of the quality of each
attribute retains a subjective nature and that different individuals could easily arrive at different
absolute scores for the same attributes and subsequently for the same emissions inventory data.
The criteria and attributes are believed to be specific enough, however, to ensure that two
individuals would arrive at consistent relative rankings when scoring more than one inventory or
more than one approach for developing an emissions estimate for a specific source category.
Measurement Attribute
The measurement attribute is related to the type, quantity and coverage of the
measurements that were used to develop the value of the parameter. For example, an emission
factor that is based on repeated measurements of a large number of sources covering the range of
typical operating conditions is of higher reliability than an emission factor that is based on a single
measurement of one source when applied universally to all sources in a given emissions category.
Similarly, an emission factor based on a mass balance analysis would typically be considered less
reliable than one based on measurement data, and an emission factor based on an engineering
assessment would normally be considered even less reliable.
A similar analyses can be completed for the activity data used to develop an emissions
estimate. Activity data, such as raw material feed, or fuel consumed, for a specific source that
is based on continuous measurement would be of higher quality than an activity estimate based
on a related surrogate. For example, an measurement of the amount of fuel consumed in a boiler
over a given period of time would be scored higher than a fuel consumption estimate based on a
prediction of load demand for a certain process. Another example would be an estimate of mobile
source activity based on a measurement of vehicle miles travelled on specific road segments which
A-2
-------
would be of higher reliability than an estimate based on fuel sales and average miles per gallon
or kilometers per liter in the fleet.
One example for the scoring scale for the measurement data attribute that has been
suggested is summarized in Table A-l.
TABLE A-l. SUGGESTED SCORING SCALE FOR THE DARS
MEASUREMENT ATTRIBUTE
Score
Emission Factor Criteria
Activity Data Criteria
10
Continuous or near continuous mea-
surement of emission rates from all
relevant sites; data capture greater
than 90%
Direct continuous measurement of
raw material feed rate, production
rate or consumption rate; data cap-
ture greater than 90%
Measurement of emission rates from
a representative sample of sources
over a range of size, and load
Direct intermittent measurement of
raw material feed rate, production
rate or consumption rate covering a
representative sample of sources
Emission factor derived from labora-
tory, bench scale or pilot studies;
multiple samples representative of
actual process
Activity rate derived from a mea-
sured surrogate that has a demon-
strated statistical association with the
activity causing emissions (e.g.,
correlation); data covers a represen-
tative sample
Emission factor based on mass bal-
ance, known physical principles or
other first principles
Activity rate derived from engineer-
ing or physical principles (e.g.,
design specifications, nominal load
demands)
Emission factor based on expert
judgement
Activity estimate based on expert
judgement
Source Definition Attribute
The source definition attribute represents the degree of specificity associated with the
application of an emission factor or activity parameter to an individual source category. For
example, there are well documented emission factors to represent external combustion boilers that
A-3
-------
differ slightly for different types of boiler design (e.g., wall fired, overfeed stoker, and traveling
grate). These emission factors would receive a high source definition score when applied in an
inventory that segregates boilers at that level of specificity. In contrast, an average emission
factor developed from an assumed population of boilers expressed in terms of electricity demand
would not be scored as high. Another example is represented for the case of a well documented
emission factor for CH4 emissions resulting from enteric fermentation for dairy cattle. When
applied to dairy cattle the source specificity for the emission factor would score 10, the same
factor applied to all cattle would score 9 or 8, and the same factor applied to all domestic
ruminants might score 5 or 6.
Table A-2 presents a suggested scoring schedule for the source definition attribute.
Pollutant Specificity Attribute
The pollutant specific attribute is only applied to the emission factor component of the
emission estimate. The methodologies and data sources used to develop activity data are truly
unrelated to the choice of pollutants that are represented in inventories. While it is true that in
many cases it is desirable to develop emission factors for air toxics or specific components of PM-
10 or VOC relative to activity data units that are routinely used in more general emission
inventories the origin and development of the emission factors themselves have little relation to
the development of the activity data. This attribute is applied to distinguish between emission
factors that are developed for specific application to a particular pollutant and an emission factor
derived from a surrogate speciation profile applied to a source category or any other indirect
application to specific pollutants. Emission factors developed for the specific pollutant will score
high, those based on a measured ratio to a common pollutant will be lower and those based on a
speciation profile applied to an entire source category will be even lower. A suggested scoring
schedule for the pollutant specific attribute is presented in Table A-3.
Spatial Scale Attribute
The spatial scale attribute is used to score the data relative to its reliability for activities
and processes in specific regional or urban applications. Spatial variability can affect the quality
of emission factors and activity data. In many applications emission factors and activity data are
known or estimated at a spatial scale that differs from the spatial scale of the intended inventory.
For instance, fuel consumption associated with a particular activity may be measured and well
documented at a country-level, or at a State or similar geopolitical unit level. When it is
necessary to estimate emissions at a finer level of spatial resolution these aggregate data are often
A-4
-------
TABLE A-2. SUGGESTED SCORING SCALE FOR THE DARS
SOURCE DEFINITION ATTRIBUTE
Score
Emission Factor Criteria
Activity Data Criteria
10
Emission factor developed specifi-
cally for the intended source cate-
gory or process
Activity data is developed directly for
the source of process activity and is
expressed in units consistent with the
activity and the applied emission
factor
Emission factor developed for a
subset or superset of the intended
source or process units and the
variability of the factor for the
application is low
Activity data representative of a subset
or superset of the category with the
activity data expressed in similar units
Emission factor developed for a
similar category with limited vari-
ability or a high degree of correla-
tion to intended category or process
Activity representative of a similar
category that is highly correlated to
the intended category or process
Emission factor developed for a
subset, superset or similar source
category with moderate to high
expected variability from the in-
tended source or process
Activity data derived from a subset,
superset or similar category with a
high variability or low correlation to
the intended category
Emission factor developed for a
surrogate activity with limited
information to establish or estimate
its degree of correlation to the
intended category or process
Activity data representative of a
surrogate category with limited infor-
mation to establish or estimate the
degree of correlation to the intended
category or process
Emission factor developed for a
surrogate category and applied to
the intended category through
engineering or expert judgement
Activity data representative of a
surrogate category and applied to the
intended category through engineering
or expert judgement
A-5
-------
TABLE A-3. SUGGESTED SCORING SCALE FOR THE DARS
POLLUTANT ATTRIBUTE
Score
10
8
6
3
1
Emission Factor Criteria
Emission factor developed specifically for the intended pollutant.
Emission factor based on a measured ratio to a more commonly
measured pollutant that is strongly related to the intended pollutant in
that source category.
Emission factor based on a representative speciation profile relative to a
more commonly measured aggregate pollutant.
Emission factor based on a speciation profile for a similar source
activity.
Emission factor based on a expert judgement or assumption of process
conditions and common emission factors.
allocated to the smaller spatial scale by applying some surrogate distribution factor (e.g.,
population, or employment). Similarly, emission factors are sometimes available for application
to one country or region, and later applied to other countries or regions with a different economic
base or climate.
Table A-4 presents a suggested scoring schedule for criteria applied to the spatial scale
attribute.
Temporal Scale Attribute
This attribute is used to rate the application of the emission factor or activity based on the
time scale of the inventory. For example, an emission factor for a certain process that goes
through frequent start up and shut down procedures may have an emissions rate that varies
considerably through those cycles relative to the emissions rate at steady state production. If an
emission factor is based on the annual operating conditions it would be unreliable in those start
up and shut down phases. Emission factors and activity data can be affected by seasonal
influences, changes between day and 11% hi. Activity data can also be dependent on operations for
the typical weekday and weekend day. The specific conditions that can affect the temporal
variability of emissions data include but are not necessarily limited to temperature, wind speed,
solar radiation, humidity, snow and ice ground cover and soil moisture content. Biogenic sources
A-6
-------
TABLE A-4. SUGGESTED SCORING SCALE FOR THE DARS
SPATIAL SCALE ATTRIBUTE
Score
Emission Factor Criteria
Activity Data Criteria
10
Emission factor is developed for
and specific to the activity at the
given spatial scale
Activity data are developed for and
specific to the geographic region of
the inventory
Emission factor developed for a
spatial scale either smaller of larger
than that of the current inventory
effort for applications where spatial
scale variability is expected to be
low
Activity data scaled from a region of
either smaller or larger scale, scaling
factors are both correlated to the
actual activity and representative of
the specific geographic area
Emission factor developed for a
spatial scale either smaller or larger
scale and spatial variability large as
a result of different relief, climate,
economic base or other factor
Activity data representative of either
smaller or larger scale and the scaling
factors are not correlated well with the
activity
Emission factor based on unknown
spatial scale or is applied to a
category with unknown spatial
variability
Activity data spatial variability un-
known
are also dependent on the stage of the growth cycle. Any factors or activity data that are
developed for an average condition or a very specific condition will introduce uncertainty when
applied to problems covering different time scales. Table A-5 presents a suggested scoring
schedule for the temporal scale attribute.
Application of Date Attribute Rating System
Each attribute is assigned its score based on a set of criteria as suggested in Tables A-l
through A-5. The total scores for the attributes are summed and divided by the maximum score
for all attributes to define a numerical score between 0.1 and 1.0. The scores for the emission
factor and activity data can then be multiplied to derive an overall inventory ranking. The data
can be used to compare different emission factors or activity data and overall inventories.
A-7
-------
TABLE A-5. SUGGESTED SCORING SCALE FOR THE DARS
TEMPORAL SCALE ATTRIBUTE
Score
Emission Factor Criteria
Activity Data Criteria
10
Emission factor developed for and
applicable to the same temporal
scale as the inventory
Activity data is specific for the tem-
poral period represented in the inven-
tory
Emission factor is developed from
an average of repeated measurement
periods for the same temporal scale
(e.g., for several years covering the
same month)
Activity data is representative of the
same temporal period but is based on
an average over several repeated
periods (e.g., activity for the spring
but is average of the most recent 3
spring periods)
Emission factor derived for a longer
or shorter time period, or for a
different year or season, but the
temporal variability is expected to
be low
Activity data representative of a
longer or shorter time period or a
different year or season but temporal
variability is low
Emission factor derived for a longer
or shorter time period or a different
year and the temporal variability is
expected to be high
Activity data representative of a
longer or shorter time period or a
different year, and the temporal
variability is expected to be high
Emission factor basis difficult to
assess in temporal basis or infor-
mation is lacking to establish tem-
poral variability
Activity data representative of a
longer or shorter time period and the
temporal variability is difficult to
assess
Similarly, each attribute is scored individually so that the components that contribute
significantly to the uncertainty can be prioritized. This approach can offer a powerful capability
to understand the relative merits of alternate inventories and alternate approaches to developing
inventories but can also indicate the specific areas where improvements in understanding of
emission factors and/or activity data would have the greatest benefits.
It is important to remember that each attribute must be considered independently to extract
the most detail and meaning from the final ranking. This issue could be particularly confusing
A-8
-------
in relation to the interpretation of the measurement, source definition and pollutant specific
attributes. The measurement attribute is intended to assign a quality or reliability estimate solely
on the basis of the methods or extent of measurement data used to compile either the emission
factor or activity data. The source specific attribute, however, is intended solely to rank the
estimate as it applies to the specific source category or group of source categories represented in
the inventory. As such, the ranking is unaffected by the absolute accuracy of the emission factor
or activity data, or the perceived reliability of the data value. It is only used to assess whether the
value used is specific for the particular source category. Similarly, the pollutant specific attribute
does not reflect an assessment of the accuracy of the emission factor, but is only related to
specificity of that emission factor to the pollutant represented in the inventory.
Therefore, it is entirely possible that an emission factor could receive a high score for the
measurement attribute and pollutant specific attribute, but a low score for the source definition
attribute. This situation would arise in the case where an emission factor was based on many high
quality measurements of the emission rate of the specific pollutant, but those
measurements were made on sources that are similar but different from the source being assessed.
The appropriate use of this technique, therefore, will require practice and discipline on the part
of the researcher.
It is also important to realize that there is a significant amount of subjectivity associated
with the application of the technique. When two or more researchers apply this technique to a
single inventory or a component of an inventory it is almost certain that the final scores will
differ. If the various scorers have a common understanding of the underlying data and the
meanings of the attributes and their ranking criteria the differences in final scores should be small.
Therefore, analyses based on the application of this technique should not make distinctions relative
to the overall reliability of emissions estimates based on scores that differ marginally from one
another. When scores differ by 15 percent or more, there should be reason to question the results
and review techniques and the inventory development methodologies used in the different
estimates. In general, when there are ranges of scores, the inventory with the highest overall
score would be favored in terms of its reliability. A strength of this approach, however, allows
the review of the sources of the high and low scores and the attributes that contribute to the high
and low scores. Therefore, a knowledgeable researcher can make assessments of the importance
of each attribute for the particular inventory application and reach conclusions from the proper
perspective.
An example of application of DARS
As an example of the application of DARS consider the development of an inventory for
VOC emissions resulting from vehicle refueling of gasoline automobiles for one day with an ozone
A-9
-------
standard exceedence in one county in one State in the United States. The emission factor is
estimated using EPA's MOBILES.0 emission factor model. The source category is specific for
refueling activities and the pollutant is specific for VOC from automobiles. The spatial scale
attribute was ranked relatively low because the data used are state specific and temperature can
vary for counties within a state. Similarly, the temporal scale attribute was ranked intermediate
because average summer temperatures are applied to represent the emissions for a specific day.
The fuel consumption activity data is based on gasoline taxes, and although this is a
surrogate it is highly correlated with fuel consumption. The tax is collected at the state-level and
apportioned to county which decreases the confidence in the county-level estimate, and it is
assumed that gasoline sales per day in the summer months are relatively constant. These
assumptions are used to generate the analysis summarized in Table A-6.
TABLE A-6. EXAMPLE OF DARS APPLICATION FOR VOC LOSSES
FROM REFUELING OF GASOLINE AUTOMOBILES
IN ONE NONATTAINMENT COUNTY
Attribute
Measurement
Source Specific
Pollutant
Spatial Scale
Temporal Scale
Composite
Attribute Criteria Ratings
Emission Factor
8
10
10
7
8
0.86
Activity Data
7
8
_
7
9
0.78
Overall Estimate
0.56
0.80
1.00
0.49
0.72
0.67
This analysis shows a high degree of confidence in the overall emission estimate with a
composite score of 0.67. The major factors that contribute to the high confidence are the
pollutant, source definition and temporal scale attributes and the variables for the measurement
attribute for the activity data and the spatial scale attribute represent the areas where improvements
could be made.
A-10
-------
Essentially any emissions estimate could be assessed in this way. At times the specific
techniques used in some types of estimates would not coincide directly with the attributes and the
criteria suggested in this discussion. One alternative that is under discussion is a system to allow
weighting factors to be applied to each attribute. The advantage of adding this feature is that it
would allow the researcher to compensate for attributes that are either not applicable or not of an
equal importance in specific applications. The disadvantage is that it would add complexity and
increase the subjectivity hi the analysis. Further work is being completed to refine this approach
to develop a more consistent list of attributes and criteria that can be widely applied in a variety
of inventory development efforts.
A-ll
-------
TECHNICAL REPORT DATA
1. REPORT NO
EPA-454/R-96-003
3. RECIPIENTS ACCESSION NO
4 TITLE AND SUBTITLE
Procedures For Verification Of Emission Inventories
5. REPORT DATE
February 1996
6. PERFORMING ORGANIZATION CODE
7 AUTHOR(S1
Mark Saeeer
8 PERFORMING ORGANIZATION REPORT NO
9 PERFORMING ORGANIZATION NAME AND ADDRESS
Science Applications International Corp.
3100 Tower Blvd.
Durham, NC 27701
10. PROGRAM ELEMENT NO
11 CONTRACT/GRANT NO
12 SPONSORING AGENCY NAME AND ADDRESS
Emission Factor And Inventory Group (MD 14)
U. S. Environmental Protection Agency
Research Triangle Park, NC 27711
13 TYPE OF REPORT AND PERIOD COVERED
14. SPONSORING AGENCY CODE
15 SUPPLEMENTARY NOTES
16 ABSTRACT
This document reports on a study of the concepts and techniques involved in emission inventory development and
verification. The study is the result of interaction among the Verification Expert Panel of the Task Force On Emission
Inventories, which was sanctioned under the United Nation Economic Commission For Europe. The verification
opportunities discussed range from simple approaches to sophisticated modeling and data manipulation techniques that
require technolo^\ that is yet emerging. The five primary elements of a verification program, as identified by the Panel,
are:
Documentation of Data Quality Uncertainty Estimates
Application of the Data Ground Truth Verification
Comparison of Alternative Estimates
This report discusses geopolitical, budgetary, and time constraints that will affect the quality of emission inventories, and
it guides users in selecting appropriate methods for various situations and conditions. The future will require more accurate
and flexible emission inventory data and technology, and future efforts will benefit from the application of many of the
inventory approaches discussed in this report.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
b. IDENTIFIERS/OPEN ENDED TERMS
c COSAT1 Field/Group
Air Emissions
Emission Inventory Verification
Inventory Techniques
Inventory Technique Selection
18. DISTRIBUTION STATEMENT
19 SECURITY CLASS (Repon)
Unclassified
21 NO OF PAGES
110
Unlimited
20 SECURITY CLASS (Page)
Unclassified
22 PRICE
EPA Form 2220-1 (Rev. 4-77)
PREVIOUS EDITION IS OBSOLETE
------- |