REPORT TO THE U. S. EPA
OF THE
SPECIALISTS' CONFERENCE
ON THE
EPA MODELING GUIDELINE
February 22-24, 1977
Chicago, Illinois
-------
REPORT TO U.S. EPA OF THE SPECIALISTS CONFERENCE
ON THE EPA MODELING GUIDELINE
February 22-24, 1977
Chicago, Illinois
Organized by: Energy and Environmental Systems Division
Argonne National Laboratory*
Conference Chairman:
Conference Organizers:
Conference Coordi nators:
John J. Roberts
Donald M. Rote
Albert E. Smith
Kenneth L. Brubaker
Miriam L. Holden, Director
Conference Planning & Management
Albert E. Smith
Sponsored by: Office of Air Quality Planning and Standards
United States Environmental Protection Agency
EPA Project Officer:
Herschel H. Slater
*Under Interagency Agreement No. EPA-IAG-D7-0013 between ERDA and EPA;
-------
TABLE OF CONTENTS
Contents Page
List of Participants 7
Conference Schedule 11
Preface 13
1. Policy Issues and General Technical Matters Addressed in
the Plenary Sessions 23
1.1 The Guideline Tends to Standardize Air Quality
Analysis 23
1.2 Specification of Computer Codes Versus Approved
Algorithms 25
1.3 Application of Standard Computer Codes for "Screening"
in the New Source Review Process 26
1.4 Evaluation and Calibration 27
1.5 The Draft Guideline Suggest a Heirarchy which Evolves
Toward Rollback as the Problems Become More Difficult .... 28
1.6 Estimating Short-Term Air Quality Impacts: Statistical
Versus Enumerative Approaches . . 29
1.7 To What Extent Should the User's Burden in
Acquisition of Data and Performance of Calculations
be Considered in Recommendations Set Forth in the
Guideline? 31
1.8 The EPA Office of Air Quality Management and
Standards Decided Not to Include Oxidant Modeling
in this Edition of the Guideline 32
1.9 Air Quality Data 32
1.10 Meteorological Data 32
1.11 Proprietary Computer Codes 33
1.12 Supplemental Control Systems 34
1.13 Structure of the Guideline 34
1.14 Periodic Review and Upgrading of the Guideline 35
2. Working Group Reports 37
2.1 Group 1-1: Multi-Source, Short-Term and Long-Term for
Set 1 Pollutants 39
2.1.1 Changes in Models Recommended in Draft
Guideline 40
2.1.2 Meteorological Data 44
2.1.3 Source Data 44
2.1.4 Supplementary Comments and Information 45
2.1.5 Further Recommendations 47
2.2 Group 1-2: Single-Source, Short-Term and Long-Term,
Set 1 Pollutants 49
2.2.1 Distinction Between Models and Computational
Procedures 50
2.2.2 State-of-the-Art Point Source Model . 50
-------
2.2.3 Availability of Regulatory Algorithms
(Programs, Codes) 52
2.2.4 Recommendation for "Screening Procedure 53
2.2.5 On the Question of Enumerative vs. Statistical
Use of the Estimates of Short-Term Concentrations . . 54
2.3 Group 1-3: Set 2 Pollutants (CO, NO™, Photochemical
Oxidant) 59
2.3.1 Overview and Philosophy 60
2.3.2 General Considerations 61
2.3.3 Carbon Monoxide Model Recommendations 63
2.3.4 Short-Term Photochemical Pollutant Model
Recommendations 64
2.3.5 Long-Term N02 Model Recommendations 67
2.4 Group II-l: Long-Range Transport and Loss Mechanisms ... 69
2.4.1 Summary of Discussion 70
2.4.2 Recommendations 71
2.5 Group II-2: Plume Dynamics Under Specail Conditions .... 75
2.5.1 Introductory Remarks 76
2.5.2 Aerodynaic Dovrawash 76
2.5.3 Sea Breeze and Other Anomalous Circulation ..... 77
2.5.4 Interactions of the Plume with an Elevated
Inversion Layer 78
2.6 Group II-3: Complex Terrain 81
2.6.1 Summary of Discussions ......... 82
2.6.2 Recommended Procedure for Complex Terrain 84
2.6.3 Assistance in Defining Screening Techniques in
Complex Terrain 85
2.6.4 Supplementary Comments and Information 95
2.7 Group II-4: Characterization of Turbulence 99
2.7.1 Introductory Remarks 100
2.7.2 Vertical Profiles of Wind Speed 100
2.7.3 Comparison of STAR, AT and OQ 101
2.7.4 Mixing Height Interpolation Scheme ......... 104
2.7.5 Vertical Dispersion Estimates ..... 104
2.7.6 Plume Dimension Stabilization Height 105
2.7.7 Horizontal Dispersion Estimates 105
2.7.8 Randomization of Wind Vector 106
2.7.9 Review and Comment on Pasquill's Recommendations
for Interim Changes to the Pasquill-Gifford
Curves 107
2.7.10 Use of Models in Urban vs. Rural Areas 107
2.8 Group II-5: Validation and Calibration . Ill
2.8.1 Principles 112
2.8.2 Considerations 113
2.8.3 Other Issues ........ 114
2,8.4 Proposed Mechanism for Meeting Requirements 114
2.8.5 Example Model Evaluation Information 115
2.8.6 Levels of Model Performance and Quality 116
3. Supplementary Materials 121
3.1 Group 1-1 122
3.1.1 Description of Texas Climatological Model 122
-------
3.1.2 Description of Texas Episodic Model 126
3.1.3 Minority Report on Application o£ Multi-Source,
Urban Model 130
3.1.4 Description of the ERTAQ Model 131
3.2 Group 1-2 135
3.2.1 Comments on Sec. 2.2.5 135
3.2.2 Comments on Short-Term Analysis and on M. Williams'
"Rationale for Elimination of the Maximum of the
Second Highest for Modeling Purposes" (3.2.3,
below) 137
3.2.3 Rationale for Elimination of the Maximum of the
Second Highest for Modeling Purposes 138
3.3 Group 1-3 139
3.3.1 Description of the SAI Reactive Plume Model 139
3.3.2 Description of the DEPICT Model
3.3.3 Description of the SAI Urban Air Pollution Simulation
Model
3.3.4 Description of the ARTSIM Model 149
3.3.5 Description of the TAPS Model 153
3.4 Group II-l 159
3.4.1 Description of the EGAMA Model 159
3.5 Group II-2 I63
(No supplementary materials) 163
3.6 Group II-3 165
3.6.1 Description of the TAPAS Model 165
3.6.2 Validation Data on the Valley Model 170
3.6.3 Comments by D. Henderson 176
3.6.4 Comments by M. Williams 178
3.7 Group II-4 179
3.7.1 Comments on the Group II-4 Discussion of the
Pasquill Gifford Oz Curves 179
3.7.2 Supplementary Comments on Discussion Topics of
Working Group II-4 184
3.7.3 Comments on Report of Working Group II-4 198
3.8 Group II-5 199
(No supplementary materials) 199
3.9 General Supplementary Materials 201
3.9.1 Use and Formulation of the Hanna-Gifford Model .... 201
3.9.2 Recommended Changes in Draft Guidelines 206
3.9.3 Comments on the RAM Urban Model 213
3.9.4 Comments Regarding EPA Models • 214
3.9.5 Descriptions of C.E.G.B. Air Quality Models 216
3.9.6 Description of the Air Quality Short-Term Model
(AQSTM) 221
3.9.7 Application of Air Quality Models Under the
Ontario Environmental Protection Act 225
3.9.8 Comments Regarding CRSTER 228
3.9.9 Comments on the Need for Model Accuracy 231
-------
3.10 Policy Issue Supplementary Materials 233
3.10.1 On the Use of Statistical Techniques for the
Prediction of Second High 24-Hour Concentra-
tions 233
3.10.2 Methods for Estimating Levels Not to be
Exceeded More than Once per Year 237
3.10.3 Position on Recently Proposed Amendments to
the Clean Air Act 238
3.10.4 Consistency and Standardization 241
3.10.5. Statistical Evaluation of Compliance 244
3.10.6 Statistical Evaluation of Compliance (Revised) . . . 245
3.10.7 Comment on Statistical Guide for Compliance .... 246
3.10.8 Validity of Hour-by-Hour Estimates of Air
Quality (The Use of the Nowcast in Pollution
Potential Forecasting) 247
3.10.9 A Comment Concerning the Use of Unverified
Models as "Relative Predictions" 248
3.10.10 Comments on Various Issues 249
3.10.11 Comments on the Use of Proprietary Models
and Some Examples 254
3.11 Description of Air Quality Models 259
4. Materials Distributed to Participants Prior to the Conference . . 315
-------
Conference Invited Participants
Dr. Sumner Barr
Section Leader, Atmospheric
Science Section
Los Alamos Scientific Laboratory
P.O. Box 1663, M.S. 490
Los Alamos, NM 37545
Dr. Norman Bowne
Certified Consulting Meteorologist
TRC Research Corporation of NE
125 Silas Deane Highway
Wethersfield, CT 06109
Mr. A1,E. Boyer
Senior Meteorologist
Ontario Hydro
700 University Avenue
Toronto, Ontario, Canada
Dr. Harrison E. Cramer
President, H.E. Cramer Co., Inc.
P.O Box 8049
Salt Lake City, UT 84108
Dr. Noel de Nevers
Professor, Department of
Chemical Engineering
University of Utah
Salt Lake City, UT 84112
Dr. Bruce A. Egan
Technical Director of
Environmental Services
Environmental Research and
Technology, Inc.
429 Marrett Road
Lexington, MA 02173
Dr. Douglas G. Fox
Meteorologist and Project Leader
Rocky Mountain Forest & Range
Experiment Station
Forest Service
U.S. Department of Agriculture
240 W. Prospect Street
Fort Collins, CO 80521
Dr. Steven R. Hanna
Meteorologist
Atmospheric Turbulence and Diffusion
Laboratory
National Oceanic and Atmospheric
Administration
Cheyenne Hall Building
P.O. Box E
Oak Ridge, TN 37830
Mr. Donald Henderson
Meteorologist
Region VIII, Environmental Protection
Agency
1860 Lincoln Street
Denver, CO 80203
Dr. Eugene Y. Leong
Chief, Air and Water Quality Division
Association of Bay Area Governments
Hotel Clairemont
Berkley, CA 94705
Mr. Gary L. Melvin
Manager of Air Resource Analysis Section
Division of Air Pollution Control
Illinois Environmental Protection Agency
2200 Churchill Road
Springfield, IL 62706
Dr. David J. Moore
Central Electricity Research Laboratory
Leatherhead, Surrey KT22 7SE
England
Mr. Lawrence E. Niemeyer
Director, Met. and Asses. Division
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Dr. William A. Perkins
Special Consultant
Member, EPA Science Advisory Board
Metronics
3174 Porter Drive
Palo Alto, CA 94304
-------
Mr. Richard A. Porter
Institute for Environmental Sciences
University of Texas, Dallas
P.O. Box 688
Richardson, TX 75080
Dr. Andrew J. Ranzieri
Senior Air Quality Engineer
California Air Resource Board,
Modeling Section
1709 llth Street
Sacramento, CA 95814
Dr. Phillip M. Roth
Director of Research
Systems Applications, Inc.
950 Northgate Drive
San Rafael, CA 94903
Dr. Jerry L. Shapiro
Chief Environmental Engineer
Bechtel Power Corporation
12400 E. Imperial Highway
Norwalk, CA 90650
Dr. Conrad Simon
Director, Environmental Programs
Division
Region II, Environmental Protection
Agency
Federal Officer Building
26 Federal Plaza
New York, NY 10007
Mr. Maynard E. Smith
President, Smith-Singer Meterologists,
Inc.
134 Broadway
Amityville, NY 11701
Dr. D. Bruce Turner
Chief, Environmental Applications
Branch, MD, ESRL
U.S. EPA
Environmental Sciences Research
Laboratory
Research Triangle Park, NC 27711
Dr. Isaac Van Der Hoven
Chief, Air Resource Environmental
Laboratory
National Oceanic and Atmospheric
Administration
Granax Building
8060 13th Street
Silver Spring, MD 20910
Mr. Robert I. Wevodau, Jr.
Engineering Services Division
E.I. DuPont de Nemours & Co., Inc.
13W38 Louviers Building
Wilmington, DE 19898
Dr. Michael D. Williams
Research Coordinator
John Muir Institute for Environmental
Studies, Inc.
Route #5, Box 299A
Santa Fe, NM 87501
-------
EPA Resource Staff
Dr. David K. Berry
Director, Policy Analysis Staff
Office of Air Quality Planning and Standards
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Dr. Peter Finkelstein
Regional Meteorologist
U.S. Environmental Protection Agency
Region III
6th and Walnut Street
Philadelphia, PA 19106
Mr. Michael Lazaro
Chief, Technical Analyst Section
Air and Hazardous Materials Division
U.S. Environmental Protection Agency
230 South Dearborn
Chicago, IL 60631
Mr. Herschel H. Slater
Physical Scientist
Monitoring and Data Analysis Division (MDAD)
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Mr. Joseph A. Tikvart
Chief, Source Receptor Analysis
Branch (MD-14), OAQPS
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
Energy and Environmental Systems Division/ANL Staff
Dr. John J. Roberts
Deputy Division Director
Dr. Donald M. Rote
Air Resources Section Director
Dr. Albert E. Smith
Dr. Kenneth L. Brubaker
-------
10
-------
11
Schedule - Specialists Conference on the EPA Modeling Guidelines
Tuesday
9:00
Registration
12:00
12:00
Lunch
1:00
1:00 _ . ,.
Opening address
1:45
1:45
Orientation
3:45
Wednesday
Breakfast
9:00
9:00 TT ..
Working groups:
1-1
1-2
1-3
(Coffee at 10:00)
12:00
12:00 „ ,.
Working group
lunch
1:00
1:00 _,
Plenary session -
Working group
reports
3:00
3:00 „ -_ , ,
Coffee break
o /c ! 3=30
3:45
Coffee break
' 4:15
4 : 15
Plenary session
on policy issues
6:00
6:00 _
Open
7:00
7:00 _.
Dinner
8:00
8:00 _-
Plenary session
(Cont'd)
10:30
3:30 „ .,
Working groups:
1-1
1-2
1-3
5:00
5:00 _
Open
6:30
6=30 nj
Dinner
7:30
7:30 TT
Writing sessions
for working groups
1-1, 1-2, 1-3
10:30
Thursday
Breakfast
9:00
9:00 TT . .
Working groups:
II-l
II-2
II-3
II-4
II-5
(Coffee at 10:00)
12:00
12:00 IT . .
Working group
lunch
1:00
Plenary session -
Working group
reports
3:00
3:00 _ .... , .
Coffee break
3:30
3:30 TT .
Working groups:
II-l II-4
II-2 II-5
II-3
5:00
5:00 _
Open
6:00
6:00
Dinner
7:00
7:00 T31
Plenary session -
Wrap-up of
issues
10:30
-------
12
-------
13
PREFACE
J. J. Roberts
The problem of air pollution control can be approached via two
alternative though not mutually exclusive policies: management of emissions
and management of air quality. In the former (which could include emission
tax policies), emphasis is placed on the technology for control of effluent
(including in-plant process changes); regulatory requirements based pri-
marily on cost/effectiveness considerations are expressed in terms of
emission standards reflecting reasonably available or best available control
technology.
Air quality management, on the other hand, entails the establishment
of air quality goals based upon stated criteria or public policies.
Emissions are then limited to the extent necessary to attain and maintain
such goals. For example, the Clean Air Act Amendments of 1970 require the
federal Environmental Protection Agency to establish National Ambient Air
Quality Standards (NAAQS) in terms of a national policy requiring the pro-
tection of public health and welfare. Air quality increments designed to
prevent significant deterioration (PSD) reflect an air quality rather than
an emission management policy.
In the Clean Air Act of 1967, the amendments of 1970, and in the
most recently proposed amendments of 1976, the U.S. Congress has con-
sistently endorsed a national policy of air quality management. Though
not exclusive of specific requirements for emission control (e.g., new
source performance standards, motor vehicles), the thrust of these legis-
lative acts is the achievement of air quality goals through a combination
of emission limitations and land-use-related measures.
The success of a policy of air quality management depends very
much upon the availability of calculational procedures to relate emissions
of air contaminants to resulting levels of pollution in the ambient air.
-------
14
Such "air quality models"* must be sensitive to quantities and geographical
location of emitted pollutants, meteorological conditions governing the dis-
persion of airborne contaminants, and, as appropriate, the physical and
chemical processes affecting transformation and removal.
A number of regulatory programs which implicitly call for the applica-
tion of air quality models are listed in Fig. 1. Proportional rollback and,
in some cases, multi-source urban dispersion models such as the Air Quality
Display Model (AQDM) were employed in the development of State Implementation
Plans (SIP) pursuant to clean air legislation of 1967 and 1971. Currently many
such SIPs are being revised to reflect regional growth, recent pressures to
permit substitution of coal for scarce petroleum resources, and other reasons
including failure of original analyses to properly predict future levels of
air quality and failure of emission sources to achieve required levels of
control. Recent court rulings and subsequent EPA regulations concerning pre-
vention of significant deterioration (PSD) call for use of air quality models
in two modes: (1) initial classification or reclassification of geographical
areas, where the decision requires estimates of the economic and energy impacts
of alternative PSD increments; and (2) determination of compliance of a pro-
posed new source with established PSD increments. The latter determination
is part of a federally mandated new source review (NSR) process which also
entails a determination that the source will not cause a violation of the NAAQS.
Where the source is a "major" one as defined by EPA directives and where the
source would cause or exacerbate a violation of the NAAQS, then the source can
be constructed if and only if a balancing or emission offset procedure is
followed.** In many situations, the emission offset evaluation will require
use of an air quality model in what is, in effect, a SIP revision.
Figures 2 and 3 present the NAAQS and PSD increments (current EPA
values and those recently proposed by Congress). They set forth the quanti-
tative targets and limits which govern the air quality management process.
*We shall distinguish such "air quality models" from individual algorithms
(e.g. estimation of plume rise) which may be combined in the model.
Further, for our purposes a Gaussian dispersion kernel may be considered
a sub-model employing algorithms to estimate the rate of dispersion as a
function of downwind distance or time. Finally, since complex models
are frequently implemented on a digital computer, they may be termed
herein computer models or computer codes. Often the name of the
computer code (e.g. AQDM) is used synonymously for the air quality model
from which it is derived.
**41 Fed. Reg. 55525 (Dec. 21, 1976)
-------
SIP
c
APPLICATIONS OF AIR QUALITY MODELS
IN AIR QUALITY MANAGEMENT
DEVELOPMENT
REVISION
PSD CLASSIFICATION
t
PSD
NAAQS
NON-
ATTAINMENT
INDIRECT SOURCE REVIEW
ENVIRONMENTAL IMPACT STATEMENTS
LITIGATION (ENFORCEMENT/ VARIANCE )
EMISSION
OFFSET"
FIG.
-------
16
NATIONAL AMBIENT AIR QUALITY STANDARDS (NAAQS)
POLLUTANTS
PARTICULATES
PRIMARY
SECONDARY
S02
PRIMARY
SECONDARY
CO
HYDROCARBONS
OXIDANTS
N02
CONCENTRATIONS IN /-ig/m3
1-hr
40°
160
3-hr
1300
160+
8-hr
10°
24- hr
260
150
. 365
1-YEAR
75 (G)
60 (G)*
80 IA)
100 (A)
0 CO CONCENTRATIONS MEASURED IN mg/m3
(A) ARITHMETIC AVERAGE
(6) GEOMETRIC AVERAGE
* AS A GUIDE TO BE USED IN ASSESSING IMPLEMENTATION PLANS TO ACHIEVfc
THE 24 HOUR STANDARD
+ AS A GUIDE TO BE USED IN DEVISING IMPLEMENTATION PLANS FOR
ACHIEVING OXIDANT STANDARDS
FIG. 2
-------
PREVENTION OF SIGNIFICANT DETERIORATION (PSD)
INCREMENTS
POLLUTANT
CLASS I AREAS CLASS II AREAS CLASS III AREAS
PARTICULATE
MATTER
ANNUAL
24-HOUR
5/5
10/10
10/19
30/37
UP TO NAAQS/37
UPTONAAQS/75
SULFUR
DIOXIDE
ANNUAL
24 HOUR
3 HOUR
2/2
5/5
25/25
15/29
100/19
700/512
UPTONAAQS/40
UP TO NAAQS/182
UP TO NAAQS/700
TOP NUMBER -EPA,
BOTTOM NUMBER-CONFERENCE REPORT ON S.32I9 TO
AMEND THE CLEAN AIR ACT, CONGRESSIONAL
RECORD, SEPTEMBER 30, 1976
FIG. 3
-------
18
Simultaneously, they establish implicitly a level of accuracy desirable in
applicable air quality models.
Scientists have been engaged for many years in the study of atmos-
pheric dispersion of windborne material from industrial and other pollutant
sources, including military activities. Much has been published on the
subject. In addition to definitive texts such as Pasquill (1961 and 1974)
and to summary reports such as Turner (1967) are numerous articles and
reports describing research on the formulation and evaluation of algorithms
and models characterizing the various aspects of the problem. Given these
research activities, coupled with the emergence of air pollution control
agencies at state and local levels and of consulting firms established
to aid both industry and government, one should not be surprised to find
extensive proliferation of algorithms, sub-models, models and related
computer codes, each designed to facilitate one or more aspects of air
quality analysis. Further, even though many models employ a Gaussian
kernel to characterize the dispersion phenomenon, results can vary widely
depending upon other calculational features such as manipulation of emission
and meteorological data, treatment of plume rise and interactions with
elevated inversions, evaluation of dispersion coefficients, averaging
techniques and statistics routines to estimate compliance with short- and
long-term standards. Different user-oriented input/output features may
distinguish two otherwise identical calculational procedures. And,
finally, the same computer code can be employed in different ways (e.g.
one versus five years of meteorological data) or results interpreted in
different ways (e.g., in determining compliance with short-term NAAQS).
Thus, with numerous models in the literature and numerous users in
the field, the potential exists for chaos, or at least for frequent, honest
disagreement, even among different units within the same organization.
Confusion of this sort has already emerged where an industry seeking a
construction permit in a multistate air quality control region is faced
with differing calculational procedures (and thus, in effect, differing
emission limits) among the various state and local agencies and, quite
possibly, between adjoining Regional Offices of the federal EPA.
It is evident that there is a need for some measure of standardiza-
tion in air quality analysis. As a minimum such standardization should be
-------
19
applied to calculatlonal procedures used throughout the federal EPA and by
state and local agencies in their official dealings with the EPA. The
issue then becomes one of determining the degree of specificity (e.g.,
specific computer codes), what in fact will be standardized (e.g., models
vs. algorithms), the rigidity of the requirements (or conversely, the
flexibility of the user to select or even develop alternative calculational
methods), the extent of coverage (e.g., specification of a particular model
vs. specification of the model along with procedures for input of data and
interpretation of results), and finally, and perhaps most importantly, the
performance criteria which a calculational procedure must meet to quality
for inclusion in any official statement of such standards.
The need within the national air pollution control program for a
guidance document on air quality analysis persists. Despite the difficulties
outlined above, the EPA/Office of Air Quality Planning and Standards (OAQPS)
has been given the responsibility for developing the requisite Guideline.
A first draft was prepared and circulated for initial in-house review during
December 1976. A second draft was published in early February. The target
date for a final version of the Guideline is July 1, 1977.
In December Argonne was apprised of this schedule and requested to
organize a three-day conference/workshop wherein specialists in air quality
analysis would critique the second draft of the Guideline. Recognizing that
the federal EPA and the states had to fulfill the important air quality
management responsibilities outlined in Fig. 1 employing the state-of-the-
art in analytical methods, the conferees were challenged to advise EPA on
the best available approaches to modeling air quality impacts and to concur
with or where possible recommend improvements to the many aspects of
the problem addressed in the Guideline.
The Specialists Conference on the EPA Modeling Guideline was held at
the Nordic Hills Inn, Itasca (Chicago), Illinois on February 22-24, 1977.
Members of the staff of the Energy and Environmental Systems Division of
Argonne National Laboratory served as conference organizers and reporters.
Several representatives of the EPA/OAQPS assisted as resource persons.
The twenty-four conferees were drawn from private industry, universities,
public interest organizations, and federal and state government. A list
of participants and affiliations appears at the front of this report.
-------
20
These participants and the conference, as a body, are considered essential
parts of the decision-making process for development of the Guideline. The
sum of the scientific group's experience in the private sector included many
years of service as staff members for large corporations and as consultants
to the entire spectrum of industrial activity of this Nation. The experience
and procedures followed in Canada and in Western Europe were brought to the
group by scientists from Canada and the United Kingdom. Those involved in
the public sector are active at the local, state, federal and even inter-
national levels. Several of the conferees are valued consultants to both
private and public activities as well as independent public interest associa-
tions promoting environmental issues. All are highly knowledgeable of the capa--
bilities and limitations of air quality modeling.
The conference was organized around a mix of plenary sessions and
concurrent workshops. Major policy issues and general technical matters
were reviewed by the conferees at-large. These general discussions provided
guidance to working groups responsible for examining specific aspects of the
Guideline (see Fig. 4). To assist everyone in preparing for the conference,
a reference notebook was prepared and distributed along with the draft
Guideline several weeks in advance. This notebook contains a brief charac-
terization of each model referenced in Tables 2 and 3 of the draft Guideline
along with key excerpts of material referenced in the Guideline. The
portion of the notebook containing the model descriptions is reproduced here
in Section 3.11.
I wish to thank my colleagues at Argonne for their assistance in
preparing the background material, planning the conference, and in general
seeing to the myriad of details essential to a successful meeting. We have
acted only as organizers of the conference and as recorders and reporters
of the positions articulated by the conferees. These conferees endured a
very intensive three-day marathon. Their exceptional efforts attest to
the seriousness with which they assumed their responsibilities.
The following quotation from Chief Justice David L. Bazelon of the
U.S. Court of Appeals for the District of Columbia seems appropriate to con-
clude these introductory comments. Speaking on the difficulties that the
courts face in resolving environmental debates, this jurist observes that
the courts are not the proper forum
-------
WORKING GROUPS
I-l MULTISOURCE, SHORT TERM AND LONG TERM, SET I POLLUTANTS
1-2 SINGLE SOURCE, SHORT TERM AND LONG TERM, SET I POLLUTANTS
1-3 SET 2 POLLUTANTS - N02 , OXIDANT.CO
LONG RANGE TRANSPORT AND LOSS MECHANISMS
H-2 PLUME DYNAMICS: WAKES, DOWNWASH, FUMIGATION,MIXING LAYER EFFECTS
E-3 COMPLEX TERRAIN
U.-4 CHARACTERIZATION OF TURBULENCE
E-5 VALIDATION AND CALIBRATION
FIG. 4
-------
22
...either to resolve the factual disputes or to make the
painful value choices. What the courts and judges can do,
and do well, when conscious of their role and limitations -
is scrutinize and monitor the decision-making process to
make sure that it is thorough, complete, and rational; that
all relevant information has been considered; and that
insofar as possible, those who will be affected by a
decision have had the opportunity to participate in it.
[Decision makers should openly disclose] where and why the
experts disagree as well as where they concur, and where the
information is sketchy as well as complete. When the issues
are controversial, any decision which is reached may be
unsatisfactory to large portions of the community. But
those who are dissatisfied with a particular decision will
be more likely to acquiesce in it if they perceive that their
views and interests were given fair hearing.
(Address to the Atomic Industrial Forum, January 10, 1977)
This Specialists' Conference and the report which now follows represents
an important element in such a decision-making process.
-------
-23-
1 POLICY ISSUES AND GENERAL TECHNICAL MATTERS
ADDRESSED IN THE PLENARY SESSIONS
Moderator/Reporter: J.J. Roberts
Policy issues and technical matters addressed in the plenary sessions
dealt primarily with the broader considerations surrounding the application
of models, purpose and structure of the guideline document, and criteria and
procedures for inclusion therein of "approved" models. These discussions
provided guidance to the working groups so that the individual position
papers are consistent with philosophical and technical perspectives of the
conferees at large. Fourteen specific policy issues and general technical
matters are summarized below.
1.1 THE GUIDELINE TENDS TO STANDARDIZE AIR QUALITY ANALYSIS
The principal issue for the conferees is the degree to which it is
appropriate for a Guideline to prescribe calculational methods and/or
specific computer codes to be used in the analysis of air quality problems.
The desirability of some guidance is generally recognized; the degree of
specificity of that guidance is the issue. Reference was made to the
Congressional Conference Report of September 30, 1976 on Clean Air Act
Amendments of 1976 where an amended Section 318 addressed "Standardized Air
Quality Modeling" and set forth a general prescription for a conference on
this subject which presumably would lead to a "guidance document," possibly
published in the Federal Register. Again, here, the degree of standardiza-
tion is left unclear in that the word "standardized" never appears within
the text of the article and, further, that the term "appropriate modeling"
appears somewhat in its stead.
A related aspect to the discussion on the extent of standardization
concerns the likelihood of approved models attaining some special status in
legal contests. It was generally recognized that such would, in fact, be
the case although wording could be incorporated to soften a rigid interpre-
tation by emphasizing the opportunity and/or desirability under certain
circumstances for the user to, at his discretion, select alternative
analytical procedures. The near certainty that the Guideline would attain
special status in regulatory matters argues for a cautious approach to
designating any particular model as "approved" and emphasizes the importance
-------
-24-
of proper evaluative procedures. .These legal implications also led to
recommendations of "conservative" modeling techniques for initial "screening"
in the new source review process.2
Two widely different perspectives on the issue of standardization
were expressed by the conferees. On the one hand, a "scientific" viewpoint
emphasized that air quality problems are generally unique and should be
treated on a case-by-case basis with appropriate professional counseling.
Reference in this case was made to a position recently drafted by the
American Meteorological Society (not yet approved by the AMS Council) which
responds principally to the above-referenced proposed amendments to the
Clean Air Act. The draft position statement expressed specific concern
with "the concept that there should be currently established one particular
model which will be capable of analyzing all conceivable situations," and
went on to point out that the wide range of meteorological, geographical,
and other aspects of the air quality modeling problem, especially regarding
short-averaging-time standards, precludes the uniform application of a
single model and, in some cases, is "not amenable to simple mathematical
treatment."
The other distinct perspective could be termed a "regulatory
perspective." It was most clearly articulated by spokesmen from industries
having to deal with the new source review process as applied to state
implementation plans and prevention of significant deterioration. Here,
the call was for clearly defined analytical procedures for determining air
quality impacts and, thus, compliance with applicable limitations. In
this sense, the Guideline was seen as requiring the use of reasonably
accepted, state-of-the-art mathematical models by which the applicant could
independently assure himself of a reasonable likelihood of compliance
before the investment of significant monies in the pre-construction phases
of a new development. The call was to identify those types of problems
which could be treated by approved models and the gray areas within which
special case-by-case analysis would be necessary. It was felt that
industry's need for clarity, for reducing uncertainties in the review
Section 2.4 and report of the Working Group II-5
2See Section 2.3
-------
-25-
process, was the principal motivation behind the above-referenced section
in the Congressional Conference Report. A related argument in favor of some
degree of standardization concerned the review of new pollution sources
impacting on more than one state and possibly more than one federal EPA
region. This situation is likely to occur under the prevention of signifi-
cant deterioration regulations. With different states and EPA regional
offices applying different air quality models to determine compliance of the
proposed source, chaotic regulatory situations can occur.
The conferees have reached a consensus on this issue in that it is
agreed that for certain standard types of situations, it would be desirable
for the Guideline to identify an "appropriate" or, possibly, "approved" set
of algorithms or particular computer code(s) to be applied if the particular
model(s) have been properly verified under the conditions for which the
standard application would be prescribed. The decision in this regard would
be made in accord with an approved set of procedures for testing, validation,
and possibly calibration. Secondly, significant flexibility should be
provided to the user whereby, with appropriate professional counsel, other
analytical methods could be employed where the standard situation was not
encountered.
1.2 SPECIFICATION OF COMPUTER CODES VERSUS APPROVED ALGORITHMS
The Guideline as currently proposed lists approved or recommended
computer codes to perform different air quality analyses. In fact, however,
with the exception of rollback, they all employ the same Gaussian dispersion
kernel and thus could be viewed as the same models. It is also recognized
that a computer code consists of many algorithms critical to the estimation
of air pollution levels; among these are emission inventory estimates
(seasonal average and temporally varying), statistical analysis of meteoro-
logical data including the initial acquisition and processing of such data,
prediction of plume rise, dispersion coefficients, and approximation of area
sources.
Thus, while there is a strong sentiment among the conferees to have
the Guideline focus as exclusively as possible on the identification of
approved algorithms which could then be used in creating an (implicitly)
approved computer code, it becomes readily apparent that the number
-------
-26-
of steps in performing the overall air quality analysis are so closely
interrelated that such an approach would be difficult to pursue, especially
at this time. Thus, the Guideline should, where appropriate, reference
specific computer codes, although evaluation studies should be conducted on
individual algorithms as well as on such complex air quality models. The
working groups, however, had many reservations about the specific models
listed in the Guideline, due in part to the lack of sufficient verification
experience, and in many cases could not give blanket approval to their use.
1.3 APPLICATION OF STANDARD COMPUTER CODES FOR "SCREENING" IN THE NEW
SOURCE REVIEW PROCESS
Whereas there is strong reluctance on the part of many conferees to
accept a particular code as a standard for a particular application, there
is general agreement that certain codes could be applied in a "screening"
process wherein a determination could be made that the source is in compli-
ance because the associated estimate of air quality impact is suitably far
below any regulatory threshold. The concept is one of applying the analyti-
cal procedure with sufficient conservatism to identify those situations
which are clearly within the allowable limitations. This approach would
distinguish a computer model as being appropriate for screening but not
necessarily sufficiently reliable to be endorsed as a standard for analysis
where the decision is a close one and where a more refined analysis and/or
interpretation of the results of such an analysis may be required. There
still remains the necessity of characterizing those geometrical, meteoro-
logical and other conditions under which the code is applicable in the
screening process.
The use of standard models as screening tools to eliminate
"non-problems" from more extensive new source review requirements has
several constructive ramifications:
1. If the conservative estimates demonstrate that air quality
goals could be readily achieved, a quick approval action is
possible.
2. If the conservative estimates indicate a potential problem,
on the basis of the results, the source can alter its
design or implement a program involving further measurements
and/or a refined modeling approach which might yield a more
realistic assessment of air quality impact. The fact that
-------
-27-
the standardized model results were conservative provides a
procedural recourse for sources to provide more information
on the expected impact.
3. From an advancement-of-technology point-of-view, pressure
is maintained on all parties to produce accurate modeling
techniques. If the standardized models are too conserva-
tive, too few "non-problems" are identified and EPA will
therefore strive for more realistic prediction methods
to reduce the conservatism. Similarly, because refined
approaches will be required for a reconsideration of a
potential denial, sources will be interested in engaging
in efforts which will tend to improve our understanding of
the technical issues.
1.4 EVALUATION AND CALIBRATION
The ability of a computational procedure or any of its algorithms to
properly characterize the physical processes that govern the transport and
aerochemistry of air pollutants and thus to predict pollution concentra-
tions with acceptable accuracy is clearly the most important criterion in
evaluating that model for inclusion in this Guideline. Currently, there is
no systematic procedure being employed by the federal EPA for the
evaluation of air quality models. Such a procedure would require criteria
for acceptability and for inclusion in a guideline, procedures for carrying
out a test program including specifications on data and on the range of
conditions under which the model must be evaluated. The evaluation process
would include an internal (EPA) review as well as, ultimately, an independent
peer group review. This evaluation process certainly would be an extensive
one involving significant effort on the part of persons developing and
promoting particular theories and models. A more extensive discussion of
the requirements for such a systematic procedure are described in the report
of Working Group II-5.
Most of the models referenced in Table 3 of the proposed Guideline
and recommended by the EPA have undergone some evaluation. However, in
general, it was agreed that the existing models would not meet the criteria
likely to be imposed via the above-mentioned systematic evaluation procedure.
EPA itself is well aware of these limitations and will welcome specific
recommendations for a comprehensive evaluation procedure.
-------
-25-
Th e multi-source, long-term air quality models in Table 3 of the
draft Guideline generally employ a calibration step to fit, usually by linear
regression techniques, the model predictions to a set of observed long-term
averages. The resulting regression coefficients are then employed in all
further use of that model for predictions of future levels of air pollution
under varying source and meteorological conditions. Positions for and
against such calibration were expressed. That calibration has been commonly
used with apparently acceptable results would argue for its endorsement as
an acceptable procedure; that the use of calibration implies a failure of
the model to properly represent the physical processes would suggest that
more attention be paid to those failings that to a statistical procedure to
conceal their details. The decision on this matter was left up to the
working group on multi-source models, wherein the use of calibration
techniques was endorsed for evaluation of long-term average concentrations
in urban, multi-source configurations.
1.5 THE DRAFT GUIDELINE SUGGESTS A HIERARCHY WHICH EVOLVES TOWARD ROLLBACK
AS THE PROBLEMS BECOME MORE DIFFICULT
The conferees could not understand the rationale for this seemingly
contradictory proposition. Rollback techniques, as the EPA admitted, are
rarely an acceptable option in the eyes of the EPA regional offices. While
perhaps appropriate for a first-cut estimate of the degree of required
emission control and possibly where the geography and aerochemistry of the
problem would permit an assumption of homogeneity, it makes no sense to use
a super-simple model in complex situations. In general, it was recommended
that rollback not be included as an acceptable option for S02, particulates,
and CO.* Its application to photochemical oxidants and N02 is discussed in
more detail under the report from Working Group 1-3.
Another aspect of concern for the use of rollback is under condi-
tions where there are an insufficient number of monitoring sites to assure
that the maximum pollution levels are used in determining the constants of
*Working Group 1-1 did, however, accept the proposition that "if the
meteorological or topographic complexities of the region are such that the
use of any available air quality model is precluded, then the model used
for strategy evaluation may be limited to a Rollback Model."
-------
-29-
proportionality. Thus, one might significantly underestimate the degree of
required controls in areas not monitored. Conversely, rollback can over-
estimate the degree of control required by focusing on "hot spots" which are
not characteristic of the area.
1.6 ESTIMATING SHORT-TERM AIR QUALITY IMPACTS: STATISTICAL VERSUS
ENUMEEATIVE APPROACHES
Two very controversial issues emerge under this general topic.
Firstly, the PSD increments are stated as increments never to be exceeded, a
condition which has somewhat frightening implications when viewed in terms
of a statistical treatment of meteorology and air pollution and the
unpredictability of rare events. The EPA explains that this dilemma would
be resolved by applying the models to a fixed and limited meteorological
history (up to five years) wherein, if the model did not indicate a violation
of the PSD increments, approval would be given for construction of the source.
Further, it is suggested that the PSD increments would be enforced only
during the review process and thus constitute a procedural type of regulation
rather than one which would be enforced after the fact if monitored
pollution levels showed the increments to be violated. This position seems
somewhat contradictory to the basic concept of SIP development and revision.
The second issue which comes under this general heading concerns the
procedure for the calculation of short-term impacts, whether for determining
compliance with short-term (three and twenty-four hour) national ambient air
quality standards or with PSD increments. The most common way of employing a
model such as CRSTER is to calculate air pollution levels at each receptor
site for each hour in a given period of recent history, usually one to five
years. The results of this analysis are then applied in an enumerative sense
whereby as long as no more than one violation per year at any site is found
to exist, the NAAQS short-term values are assumed met, and where no value
exceeds the appropriate PSD increment, that limitation is assumed satisfied.
An alternative, and to many persons a more satisfying approach, would
be to generate or in some way employ a statistical representation of the air
quality at any given receptor site and apply the results of such an analysis
to the test of compliance. This approach would require some, possibly minor,
redefinition of the NAAQS and PSD limitations. Consistent with this
-------
-30-
statistical approach, a model such as CRSTER would be run over an
appropriate period of time with the resulting computed values fitted to a
suitable cumulative frequency distribution function. It was observed that
similar problems occur in interpreting the short-term data from air quality
monitoring sites which may be subject to occasional and undetermined
malfunctions producing erroneously high values.
It is common engineering practice to use statistically computed
criteria for design to meet extreme meteorological conditions. Drainage
for flood control and wind loads on structures are two examples. Appro-
priate meteorological parameters are treated by special statistical
procedures to estimate conditions for specified periods ranging from ten to
one hundred years. These estimates often exceed the period of record by a
factor of two or more. Although meteorological conditions associated with
atmospheric transport models are very different from those associated with
the above examples, the concept of using a statistical approach to determine
compliance is analogous to these established practices.
Although the consensus is that a statistical approach is theoret-
ically more valid, it is recognized that a detailed investigation of
the nature of the statistics should be completed prior to implementation..
In particular, concern is expressed regarding the appropriate form of the
cumulative distribution function to be used in any curve-fitting procedure.
To the extent that enumerative (i.e., deterministic) methods are
used to determine compliance with short-term standards, two approaches
suggest themselves: (1) the above-described enumerative approach using
CRSTER, where the major question is the period of record over which calcu-
lations are to be made; and (2) a "worst case" approach where the challenge
is to identify the conditions under which maximum ground level concentra-
tions might occur.
In the first situation, it was argued by one conferee that the period
of record be equivalent to the plant life (i.e., as long as twenty to thirty
years). It was also recognized that even a five-year set of calculations
could be unreasonably costly for certain smaller sources.
The worst case approach via models such as CRSTER makes sense to the
conferees as a screening mechanism whereby compliance can be assumed if the
-------
-32-
worst case estimate is conservatively below the allowable increment.
However, the use of an idealized plume rise algorithm (as in CRSTER) might
overlook special high pollution conditions associated, for example, with
aerodynamic downwash. Other specialized situations such as unusual
circulations (e.g., sea breezes) and fumigation should be considered as
potentially creating the worst case situation.
Finally, in applying a "worst case" analysis, some recognition must
be taken of the potential for occurrence of such a situation in order to
determine compliance with short-term NAAQS.
1.7 TO WHAT EXTENT SHOULD THE USER'S BURDEN IN ACQUISITION OF DATA AND
PERFORMANCE OF CALCULATIONS BE CONSIDERED IN RECOMMENDATIONS SET FORTH
IN THE GUIDELINE?
It is generally recognized that for some types of calculations and
some computer codes the burden of acquisition of emission and meteorological
data and the computer costs can be excessive for small pollution control
agencies and for many industrial applicants. Thus, another argument can be
made in favor of an inexpensive screening process to limit the number of
situations in which more extensive air quality analysis would be required.
However, in those situations where the standards are in fact threatened and
a more refined analysis is required, the best available methods should be
employed. This burden would most likely fall upon the federal EPA and upon
state agencies.
To the extent the Guideline recommends models for screening purposes,
such models may reflect a desire for simplicity and ease of use. To the
extent the Guideline recommends models for general application in air
quality assessments, they should be characterized in terms of the potential
burden upon the user but the decision as to the appropriateness of that
burden should be left to the user.
In a somewhat related aspect, the Guideline would permit the
Regional Administrator to select a model other than that recommended "if the
data bases required" are "unavailable or inadequate." The degree to which
it is incumbent upon the EPA or upon the applicant in a new source review
procedure to delay the project until adequate data are assembled should be
clarified.
-------
-32-
1.8 THE EPA OFFICE OF AIR QUALITY PLANNING AND STANDARDS DECIDED NOT TO
INCLUDE OXIDANT MODELING IN THIS EDITION OF THE GUIDELINE
EPA representatives indicate that they have recently organized a task
force to develop methods for relating levels of photochemical oxidants to
the required degree of hydrocarbon controls. Thus, they chose to leave
oxidant modeling out of this edition of the Guideline. The appropriateness
of currently available models for the description of chemically reactive
pollutants is considered briefly in the report of Working Group 1-3.
1.9 AIR QUALITY DATA
The importance of air quality data in evaluating the performance of
an air quality model is evident. The conferees are concerned that, in
general, insufficient consideration has been given to criteria for proper
siting of monitors and acquisition and processing of air quality data for
validation and for calibration purposes. The EPA has a task force currently
studying the monitoring situation in the USA, but the major thrust of this
effort would appear to be a cost/effectiveness approach to minimizing the
number of monitors necessary to assure compliance with ambient air quality
standards. The establishment of appropriate criteria for monitoring data
to be employed in model validation studies would be a part of the evalua-
tion procedure developed pursuant to policy issue 1.4, above. Finally, in
addition to the need for better air quality data, there is a need for
modelers to be more aware of the data which is available.
1.10 METEOROLOGICAL DATA
In a matter analogous to the concern for the quality of monitoring
data, concern is expressed for the inadequacy of meteorological data
currently available via the National Weather Service. Further, hourly
meteorological observations are not available on magnetic tape from the
National Weather Records Center for many sites in the USA. Finally, the
NWS program of twice per day urban soundings has been discontinued.
It is generally agreed that little can be done in the near future to
change the method of observation and to reinstate the sounding program. The
National Climatic Center is preparing hourly records for selected locations at
the request of EPA under an interagency reimbursement agreement. Several other
parties from time to time obtain climatic data of this type under similar arrange-
ments.
-------
-33-
Concern is also expressed for a lack of adequate meteorological data
at elevations approximating the effective stack height of major elevated
sources, as well as for how the available meteorological data are used in
evaluation and in various algorithms.
1.11 PROPRIETARY COMPUTER CODES
The appropriateness of referencing proprietary computer codes in the
Guideline is a controversial issue. The large majority of conferees,
including several private consultants, oppose the inclusion of such models
in this government publication. EPA representatives indicate that they do
not intend to reference such models in the Guideline.
Arguments in favor of including references to proprietary models are
consistent with a structure for the Guideline which does not recommend
specific computer codes but rather approved, generic methods and algorithms
for air quality analysis. It is felt by some that this approach would
allow a listing of computer models, public and private, which satisfied the
generic criteria. This position is also consistent with a concern for
excessive standardization of calculational methods in contrast to more
emphasis on case-by-case analysis by qualified experts. Regarding the
image of secrecy associated with a proprietary model, it is argued that the
regulatory agency could be provided with documentation and possibly
listings of the proprietary code as long as the agency did not release this
information to the public. Finally, proprietary models can be tested
against sample problems in order to demonstrate their consistency with
other, approved, public domain models.
This rationale for including proprietary models and the associated
procedures for their evaluation as well as their safeguard are countered by
a number of arguments. From the viewpoint of a government agency, it would
appear to be impractical, if not directly contrary to law, for proprietary
models to be used in new source review and, quite probably, in the design
of regulations and SIP revisions. Regarding new source review, spokesmen
from industry express great concern that the agency could be using a model
to which they as applicants had no access. The agency would at the same
time face requirements for compliance with various public disclosure laws,
state and federal, and in all likelihood could not honor an agreement to
-------
-34-
safeguard the details of the proprietary model. For example, the general
counsel for EPA Region V forbade the use of a proprietary model in
regulation development for these very reasons.
As to the use of standardized test problems for establishing some
degree of equivalence between models approved via the Guideline and
alternate, proprietary models, it is recognized that such may be desirable
whenever a private party chooses to use a proprietary model in a permit
application or in litigation. The challenge then becomes one of developing
a set of test problems which explore the full range of application of the
models in order to determine with a fair degree of confidence the question
of equivalence .
1. 12 SUPPLEMENTAL CONTROL SYSTEMS (SCS)
The Guideline makes no reference to SCS. Although many of the
computer codes referenced in the draft Guideline are applicable to SCS
determinations, some codes more suitable to transient analysis were not
considered. Further, an important part of an SCS strategy is the on-line
analysis of air quality and meteorological data and the related, near-
term forecasting techniques.
Since the Guideline does not address SCS, the conferees did not
consider this subject in significant detail. It is recommended that the
Guideline be amended to indicate, probably by footnote, that SCS is not
considered within the Guideline and then to provide one or more references
for the user who is interested in pursuing the federal requirements and
available methods for the implementation of SCS.
1. 12 STRUCTURE OF THE GUIDELINE
The utility of the draft Guideline would be greatly improved if
certain minor structural modifications were incorporated. Chapter
subheadings and greater reliance on tabular presentations along with a
table of contents would make the document more readable. As mentioned
earlier, SCS must be referenced in order to clarify the intended scope of
the Guideline. Similarly, the section on growth does not provide adequate
guidance for handling this complex subject. It would be preferable for the
-------
-35-
Guideline to acknowledge its limitation in this regard and provide
references wherein the user can assess the state-of-the-art of growth
projections for air quality management. Finally, the section on model
application should be expanded to provide more guidance on use of appropriate
emission data.
1.14 PERIODIC REVIEW AND UPGRADING OF THE GUIDELINE
By recognizing the generally unsatisfactory status of past efforts
in validation of air quality models and the need for a more extensive,
formalized procedure (see policy issue 1.4), the conferees wish to under-
score the need for periodic review and upgrading of the Guideline. Further,
the absence of coverage of topics such as SCS and photochemical oxidants
and the limited treatment of oxides of nitrogen call for a significant
expansion in the scope of future editions.
The conferees would be pleased to examine subsequent drafts of this
first edition and to participate in the review of future editions. However,
it is generally agreed that a more extended and less intensive review
schedule, perhaps involving several standing committees, would be preferable
to the highly compressed schedule imposed by a two- or three-day intensive
conference/workshop.
-------
-36-
-------
-37-
2 WORKING GROUP REPORTS
The working group reports are based on discussions held within the
working group meetings. These reports incorporate corrections and editorial
changes recommended by working group members. Extensive revisions or addi-
tional material submitted subsequent to the group meetings by group members
or other conference participants have been referenced where appropriate and
collected together in Section 3.
-------
-38-
-------
-39-
2.1 GROUP 1-1: MULT I-SOURCE, SHORT-TERM AND LONG-TERM FOR SET 1 POLLUTANTS
PARTICIPANTS
N. Bowne R. A. Porter
A. E. Boyer C. Simon
H. E. Cramer J. A. Tikvart
S. R. Hanna D. B. Turner
W. A. Perkins M. Lazaro
Moderator/Reporter: A. E. Smith
CONTENTS
2.1.1 Changes in Models Eecommended in Draft Guideline
2.2.2 Meteorological Data
2.1.3 Source Data
2.1.4 Supplementary Comments and Information
2.1.5 Further Recommendations
-------
-40-
2.1.1 Changes in Models Recommended in Draft Guidelines
The consensus of the group was that Gaussian plume models are the best
available at this time for multi-source situations. The situation most fre-
quently referred to throughout the discussions was the urban area with a large
number of sources. Concerns were expressed regarding the need to specify the
constraints on the use of a model and on various algorithms implementing a
model. In this regard, the group recognized the need for well written user's
guides which detail the limits of an algorithm's applicability. It was sug-
gested that a list of features which would make algorithms more desirable might
be developed. This was considered a reasonable suggestion but no attempt was
made to produce such a list.
After accepting the Gaussian plume model as the best currently available,
the specific algorithms in Table 3 of the draft guideline document were con-
sidered. It was concluded that knowledge of the state of validation of these
models was lacking and that a need existed for validation studies and, more
generally, packaged test cases to test new algorithms. The group considered
whether to retain the hierachical structure of the draft Table 3 or to simply
list acceptable algorithms with the choice of a prrticular algorithm being re-
lated to the level of detail required by the task at hand. The consensus
retained the hierachy.
The use of the Larsen procedure for estimating short-term averages in
AQDM and CDM was discussed. The group concluded that hour-by-hour calculations
as done by RAM are a preferable method but did not feel that the statistical
conversion of averaging times should be precluded in urban areas where large
point sources are not a major influence. The possibility of giving CRSTER the
capability of handling multiple sources was discussed and some validation data
on CRSTER were presented. It was noted that short-term predictions tend to be
accurate to within about a factor of two and that the suggested long-term algor-
ithms tend to overpredict. The point was also made that AQDM, CMD, and the
Hanna-Gifford model (properly applied) give equivalent correlations with mea-
sured data. The importance of representative meteorological data and a good
emissions inventory was stressed by several members. Possible problems with
the existing parameterizations of a in the guideline models were noted.
z
The Texas Climatological Model (TCM) and Texas Episodic Model (TEM)
were suggested as additional models for Table 3. These models were repre-
sented as being essentially computerized algorithms applying the Hanna-Gifford
-------
-41-
model with provisions for including the effects of individual point sources.
Although there was some reluctance to including models not available from
EPA or models in which a member of the group had a particular interest in
Table 3, the Texas models were recommended for inclusion given the fact that
they have been widely used. So that other models might be included in future
revisions of Table 3, development of a formal procedure for approving equiva-
lent models was suggested.
Throughout its deliberations, the group was aware that the data to
make the best technical, scientific decision are frequently unavailable but
that a need does exist for guidance in making the administrative decisions
required by law.
The final concensus of the group for specific changes and additions to
the February 1977 draft guideline follow. Where doubt might exist as to where
changes were made, additions or changes have been indicated by a bar in the
left hand margin. Recommendations of the group for areas needing attention
or study follow the suggested changes in the guideline.
On page 15, change Table 3;
Table 3. Multi-Source Models Applicable to Specific
Pollutants and Averaging Timesa
S02 and TSP
Annual Average
AQDM/CDM26'2?/TCMb
SO and TSP
24-Hour Average
RAM31'32
TEMb
AQDM*/CDM*26'27
so2
3-Hour Average
RAM31'32
TEMb
* *?fi 77
AQDM /COM Zb'
Statistical conversion of averaging times required.
Numerical references in the table cite references in the draft Guideline.
Appropriate reference to user's manual. Descriptions of TCM and TEM are
contained in Sees. 3.1.1 and 3.1.2.
On page 16:
9
Similar models have been summarized and discussed by Lamb et.al.,
Moses , Stern , and others. They are available from private consultants
and other governmental agencies. However, to meet the need for consistency
-------
-42-
identifled in the Introduction to this guideline, selected models have
been specified. Based on a determination by the Regional Administrator
that another air quality model already in use by a state agency provides
equivalent or more reliable concentration estimates, that model may be
12
used. Guidelines which provide a procedure for comparing air quality
models are in preparation.
There are limitations to the use of all atmospheric transport and
dispersion models. Users must be aware of these limitations and not apply
the models outside their limitations. Models listed in Table 3 are readily
available; there are many other published multi-source models.
On page 17:
Multi-Source Models Required for Sulfur Dioxide and Total Suspended
Partioulates (Annual Average)
j /-
The Air Quality Display Model CA-QDM) or the Climatological Dispersion
27
Model (CDM) , or TCM may be used to evaluate multi-source complexes.
On page 18;
If a more detailed or more suitable model is available, especially in
a Region with major meteorological or topographic complexities, that model
may be used.
Also, if the meteorological or topographic complexities of the region
are such that the use of any available air quality model is precluded, then
the model used for strategy evaluation may be limited to a Rollback Model.
Multi-Source Models Required for Sulfur Dioxide and Total Suspended
Particulates (Short-Tern Averages)
31 32
The Real-Time Air-Quality-Simulation Model (RAM) ' may be used to
evaluate multi-source complexes.
If the data bases required to apply RAM are unavailable or inadequate,
the TEM may be used.
If the resources required to operate RAM or TEM are not available,
AQDM or CDM may be used to estimate short-term concentrations of SCL and
particulate matter. These models must be used with procedures for the
-------
-43-
25
statistical conversion of averaging times as discussed by Larsen to
convert annual average concentration estimates to 3-hour and 24-hour average
concentrations. This technique is valid only in urban, multi-source situations
and should not be used in situations dominated by single point sources.
Other similar techniques for making this conversion may also be used.
On page 19:
If a more detailed or more suitable model is available, especially
in a Region which has major meteorological or topographic complexities,
that model may be used.
If the meteorological or topographic complexities of the Region are
such that the use of any available air quality model is precluded, then the
model used for control strategy evaluation may be limited to a Rollback
Model.30
-------
-44-
2.1.2 Meteorological Data
In discussing the section on meteorological data it was pointed out
that the data listed was that required by the models listed and that more
detailed data was likely to be needed as models improve. Several comments
indicated that representative meteorological data is not always readily
available despite an implication to the contrary in the draft guideline. The
group felt that the requirement of a five year data base was reasonable if
five years of data were available. This was strengthened by one
member who reported on a situation where consistent results began to be ob-
tained at about 5 years when a total dosage model was used with increasingly
long meteorological data bases in the 1-10 year range.
The consensus of the group is expressed in the following recommendations
for specific changes in the draft guideline.
On page 30:
Specific meteorological data required to describe transport and dis-
persion in the atmosphere are wind direction, wind speed, wind shear, atmos-
pheric stability and mixing height appropriate to the site. These parameters
may be derived from routine measurements by National Weather Service (NWS)
stations and the data may be available both as individual observations and in
summarized form from the National Climatic Center, Asheville, N.C. If other
sets of data which encompass wind direction, wind speed, atmospheric stability,
mixing height or other indicators of atmospheric turbulence and mixing are
available, they may be used. Local universities, industrial companies, pol-
lution control agencies and consultants may be sources of such data. A five
year data base is desirable.
On page 32 add the paragraph:
It is to be noted that future availability of meso and micro meteor-
ological data collections will make practical more detailed meteorological
analysis and subsequent improvement of model estimates.
2.1.S Source Data
There were no major problems with the draft guideline statements about
source data for multi-source urban situations. The consensus held, however,
that other than area-wide diurnal variations could be important and that
-------
-45-
emission models should be used where available. The group suggested the
following change in wording to express its consensus.
On page 35:
For multi-source urban situations, detailed source data are generally
impossible to obtain. In these cases, source data shall be based on annual
average conditions. Area source information required are types and amounts
of pollutant emissions, the physical size of the area over which emissions
are prorated, representative stack height for the area, the location of the
centroid or the southwest corner of the source in appropriate coordinates.
Where emission models are available, output from such models should be used.
Short-term models may be modified to accept such data.
2.1.4 Supplementary Comments and Information
Supplementary comments and information on multi-source modeling were
submitted by several conference participants. Both members of the working
group itself and others contributed. This material can be found in Sec. 3
and is referenced below.
R. Porter submitted additional information on the Texas models. Sections
3.1.1 and 3.1.2 describe the Texas Climatological Model (TCM) and the Texas
Episodic Model (TEM), respectively, in the format used in the conference note-
book.
G. Melvin presented a description of Illinois' Air Quality Short Term
Model (AQSTM) in the format of the notebook in Sec. 3.9.6. The group did not
discuss this model.
A. Boyer presented a minority report (Sec. 3.1.3) suggesting limitations
on who should be required to apply multi-source urban models. The group as
a whole did not take a position on this subject. He has also described the
use of urban and single source models by the Ontario EPA in Sec. 3.9.7.
S. Hanna has provided an extensive description of the development,
application, and suggested uses of the Hanna-Gifford model in Sec. 3.9.1.
Included is a description of the model in the notebook format including a
formulation treating chemical reactions.
-------
-46-
M. Williams has suggested adding monitoring requirements to the condi-
tions under which rollback may be used and has commented on the choice of
receptor sites. He recommends specific wording changes for pp. 19 and 29
of the draft guideline in Sec. 3.9.2.
M. Smith submitted criticisms of RAM in Sees. 3.9.3 and 3.9.4. Section
3.9.4 is a letter from Dr. Howard M. Ellis of Enviroplan, Inc. The criticisms
relate to the lack of validation, problems with a values, and the full load
z
operation assumption. Section 3.9.4 also suggests that some measure of
actual operating rates be used if five, rather than one, years of meteorology
data are employed.
-------
-47-
2.1.5 Further Recommendations
1. Packaged test cases should be developed for comparing and testing models.
These cases should include the required emissions and meteorological data
for input and the appropriate measured air quality data against which
model predictions can be checked.
2. User's Guides should delineate clearly the limits of applicability of
model algorithms in order to avoid misuse. There is also a frequent lack
of clarity in the specification of the inputs needed by particular
algorithms.
3. EPA should encourage continuing effort, either in-house or under contract,
for validating the models they are recommending. Such information is
necessary for an informed scientific endorsement of particular models and
such information has not been presented at the conference.
4. A formal procedure should be developed for approving other models and
placing them in the guideline.
-------
-48-
-------
-49-
2.2 GROUP 1-2: SINGLE SOURCE, SHORT-TERM AND LONG-TERM,, SET 1 POLLUTANTS
PARTICIPANTS
S. Barr J. L. Shapiro
N. H. de Nevers M. E. Smith
D. Henderson I. Van der Hoven
D. J. Moore R- !• Wevodau, Jr.
H. H. Slater M. D. Williams
Moderator/Reporter: D. M. Rote
CONTENTS
2.2.1 Distinction Between Models and Computational Procedures
2.2.2 State-of-the-Art Point Source Model
2.2.3 Availability of Regulatory Algorithms (Programs, Codes)
2.2.4 Recommendation for "Screening Procedure"
2.2.5 On the Question of Enumerative vs. Statistical Use of the
Estimates of Short-Term Concentrations
-------
-50-
2.2.1 Distinction Between Models and Computation Procedures
For the purposes of these deliberations a clear distinction should
be made between models and computational algorithms. The latter are
sometimes called computer programs or computer codes. A model is a set
of mathematical relationships, based on scientific principles, often using
adjustable parameters. Examples of atmospheric concentration models are:
rollback, fixed box models, Gaussian models, and moving box models.
A computational algorithm is a set of detailed instructions for
implementing a model- For example, AQDM, CDM, CRSTER, PTMAX, PTDIS, and
PTMTP are all computational algorithms for implementing the same basic
Gaussian model. For identical input data they must give the same result,
if they faithfully represent the agreed underlying model.
This distinction was made because the group agreed that the Gaussian
model is the state-of-the-art model for operational estimates of TSP and
SO . The choice of particular algorithms to be used depends on the amount
and quality of input data available, the detail required in the answer,
the available budget, and other factors.
2.2.2 State-of-the-Art Point Source Model
The working group recognized that no model available today, nor
probably in the foreseeable future, is free of serious limitations, but
it was unanimous in recommending the Gaussian statistical model as the
1
state-of-the-art model for all point source evaluations.
It should be mentioned that the use of the term "Gaussian" is
understood to refer to the distribution of the pollutant about a plume
centerline. Consequently, the use of a particular algorithm to compute
the position of the plume centerline does not alter the fact that the
model is still basically a Gaussian model.
Each of the computational algorithms listed on page 14 (Table 2)
of the draft Guideline contains the Gaussian assumptions for the repre-
sentation of plume dispersion as a key feature. Therefore, the group
agreed that each is acceptable in principle.
-------
-51-
Several concerns were expressed regarding specific assumptions or
the management of data in several of these algorithms. However, none were
sufficiently basic to warrant discarding the Gaussian model in favor of
2 3
another approach. The major concerns are listed below:
1. There is evidence that the az curve representing
the A stability category in the Pasquill-Gifford
formulation may result in serious overestimates of short-term
maximum values from tall-stack sources.
2. The system of using a random adjustment to
winds recorded in 10° intervals may not be
adequate to represent full horizontal plume spread. This
system is used in CRSTER (see results of Group II-4 dis-
cussions) .
3. From analysis of limited power plant data it appears that
the 24-hour predictions of CRSTER may be too low for relatively
flat terrain. It is worth determining whether this is a typical
result of the model.
4. There is controversy over the technique used for adjusting
the algorithm to terrain variations. More generally, how
should the Gaussian model be modified when a plume approaches
elevated terrain?
5. There is evidence that stability estimates, and the resultant
az and possibly ay values, should be functions of source
height, and not independent of it as shown in the Pasquill-
Giff ord curves. This may influence the applicability of
the Gaussian model to tall stack calculations.
6. Effluent released during one segment of time may have an
effect on the concentration at a receptor at some later
time. This could influence multi-hour concentration estimates.
7. Plume rise is not considered as a final, settled issue. As
with the diffusion algorithms these estimates should be reviewed.
It should be noted that some of these concerns are specific to the
"CRSTER" algorithm, namely points 2 and 3, while others apply to the
Gaussian model in general. It should be emphasized that in all cases,
the participants were concerned with the problem of calculating short-term
ground-level concentrations resulting from point source and especially
tall-stack emissions. Each of the items listed were only briefly discussed
by this working group since the subsequent working groups on specific
issues would presumably deal with these concerns.
-------
-52-
This working group also expressed concern about applying this set
of algorithms beyond 50 kilometers from the source. The concerns ranged
from vague uneasiness because of the lack of data on dispersion and model
validation at greater distances to the recognition that physio-chemical
processes, such as deposition at the air-land or air-sea interface would
alter the Gaussian distribution of the pollutant. Conversion and loss
processes are not treated by the Guideline models. Neglect of these
latter factors should result in an overestimate of ground level concentra-
tions. However, extrapolation of the Pasquill-Gifford dispersion curves
to large distances could result in an underestimate of ground level con-
centrations.
Although not specifically agreed upon as a recommendation, it
was clear that all members of the group favored expanded efforts to refine
and improve the algorithms listed in the proposed Guidelines by careful
comparison with data.
2.2.3 Availability of Regulatory Algorithms (Programs, Codes)
It was unanimously agreed that regulatory agencies (EPA, state
and local agencies) should be required to state clearly which models and
algorithms (programs, codes) they will use to review construction permit
applications.
If published, widely available algorithms (e.g., those on the
UNAMAP tape, available from NTIS) are used, then a listing of algorithms
together with a brief description of how optional features are selected
and which algorithms apply to particular cases should be adequate. If,
on the other hand, they use models which are not readily available, they
should be required to make copies of computer tapes and complete program
documentation, and make these available to any interested party at nominal
cost.
The purpose of this requirement is to enable those who apply for
permits to know exactly how their request will be evaluated.
Wherever practical, regulatory agencies should use nationally-
available models, rather than ones peculiar to their agency. This
recommendation minimizes the number of different regulatory situations
-------
-53-
a multi-state source must face. If there is some significant advantage
(i.e., special algorithms for accurate treatment of land/sea breeze,
fumigation, etc.) in using a non-standard model, then this recommendation
should be ignored. In such cases the requirement for adequate documentation
must be re-emphasized.
There should be some criteria for deciding whether alternative
algorithms or models are acceptable. It was unanimously agreed that such
criteria could include the use of reference or bench-mark type test problems
designed by the EPA.
2.2.4 Recommendation for "Screening Procedure"
The computational effort and resource expenditure required to carry
out single source calculations with the CRSTER algorithm as outlined in the
Guideline were judged likely to be excessive for small sources. This
judgment took into account the fact that both the CRSTER and alternative
computation algorithms (some of which are much easier to implement) were
based on the same Gaussian model. Instead of the indiscriminate use of
CRSTER for all sources, a screening procedure was recommended. Only
sources failing the screening test or failing to qualify for the screening
test should be subjected to an analysis with the more elaborate CRSTER
algorithm.
The question of precisely what criteria should be used in qualifying
a source for the screening procedure and in deciding compliance with the
controlling standard was discussed at some length. It was generally agreed
that sources having stack heights greater than 100 meters would automatically
be subjected to the more complete analysis by the "CRSTER" model. Applica-
tion of the "CRSTER" model must of course be limited to cases where the
physical stack height is substantially greater than the surrounding terrain
variations. Reservations were expressed because applicants might use this
cirteria as a basis for stack design. Hence, it may be necessary to consider
additional criteria which take into account effluent or heat emission rate
as well as stack height.
To pass the screening test the predicted concentration using the
screening procedure must be less than or equal to 50% of the controlling
standard concentration.
-------
-54-
The recommended screening procedure is as follows:
a) A hypothetical joint frequency distribution of wind speed
and stability is developed containing all possible reasonable
combinations of those 2 meteorological variables. This dis-
tribution is to be used in the selection of appropriate short-
term meteorological conditions.
b) Computations are carried out using one of the UNAMAP point
source models to determine the 1-hr, maximum concentrations.
c) Longer-term concentrations estimates are developed using
simple ratios to the hourly maximum concentration estimate.
It is tentatively suggested that the ratio of 1-hr, maximum
to 24-hr, maximum be taken as 3 for stack heights less than 100m.
d) The estimated maximum concentration is compared with the
controlling standard concentration. If the estimated
value is <_ 50% of that standard,the test is passed and no
further analysis using the CRSTER model may be required.
It should be pointed out that the screening procedure may have to
take into consideration special effects such as fumigation, downwash,
or terrain complexities using appropriate algorithms. (See the reports
of the Working Groups on Complex Terrain and Plume Dynamics.)
This recommended screening procedure is quite similar to that
outlined in "Draft-Guidelines for Air Quality Maintenance Planning and
Analysis, Volume 10: Reviewing New Stationary Sources" SRAB, MDAD, OAQPS,
EPA, February 1977. The latter guideline suggests the use of an
approach similar to that used by the "VALLEY" model for complex terrain
cases.
See footnote 4 for further comments.
2.2.5 On the Question of Enumerative vs. Statistical Use of the Estimates
of Short-Term Concentrations
In regard to this question there was general, although not unanimous,
agreement that a statistical approach should be used. However there were
differences of opinion regarding the precise nature of the statistical
approach, the length of meteorological records required, and which con-
centration value (highest, second highest) should be applied as the control.
It was generally felt that the question of the precise nature of the
approach is sufficiently important to warrant a detailed investigation
subsequent to the conference.
-------
-55-
The following statement reflects the consensus of the majority of
participants of this working group:
In evaluating the modeling results to predict compliance with short-
term NAAQS (not to be exceeded more than once a year) or PSD (not to be
exceeded), the test shall consist of computing the predicted concentrations
at each of the several worst receptor points, for a specified period of time ,
using the state-of-the-art modeling technique (e.g., CRSTER or its equivalent)
These worst receptor points are defined by the highest concentrations for the
averaging time of concern (3 hours or 24 hours). The resulting set of
computed values for each of the several points chosen shall be fitted by a
suitable cumulative frequency distribution function .
In order to be consistent with both the inherent uncertainties
in the frequency of occurrence of rare events and the use of the second
highest concentration as required by the NAAQS, the following demonstration
of compliance seems appropriate: The distributions shall be considered
to demonstrate compliance if the cumulative distributions functions indicate
that the short-term NAAQS will not be exceeded twice in a calendar year,
more than one year in each jc years at any given point, and that the PSD
increments will not be exceeded more than once per £ years.
We understand that the use of the second highest value was chosen
to correspond to the realities of ambient air quality monitoring. The
realities of modeling are different. Thus - for modeled compliance -
a preferable demonstration of compliance would be: The distributions shall
be considered to demonstrate compliance if the cumulative distribution
functions indicate that the short-term NAAQS will not be exceeded, on the
average, more than once per calendar year, and that the PSD increments will
not be exceeded more than once per y years. In the latter statement the
term "on the average" is understood to be based upon a limited but specified
length of record, say one to five years.
The x and y in the above paragraphs have been purposely left undefined.
It was believed that a thorough study of the consequences of possible choices
of those values should be made before values can be assigned .
-------
-5(5-
FOOTNOTES FOR SECTION 2.2.
1. This recommendation was made for cases where there is no appreciable
depletion of plume material and where the topography is relatively simple.
2. In a post-conference contribution (see Section 3.9.2) M. Williams
expressed some additional concerns regarding CRSTER and RAM and pointed
out a documented case in which plume rise was suppressed by an elevated
inversion layer. Such phenomenon are not properly treated by these
algorithms.
3. At the conference, M. Smith passed out copies of comments prepared by
H. Ellis regarding the "RAM" and other guideline recommended algorithms.
See Sections 3.9.3 and 3.9.4.
4. It is worthwhile calling attention to a post-conference submittal in
which A. Boyer (see Section 3.9.7) describes a regulatory approach taken
in the Ontario Environmental Protection Act. This approach, which is
a variant of the screening procedure concept, requires the applicant to
make estimates of impact based on a simple point source model applied
under certain meteorological conditions. It is then the responsibility
of the control agency to determine whether the combined effects of mul-
tiple sources, each complying with single source standards, when operating
under a wide spectrum of conditions, is acceptable.
5. The length of meteorological record to be used for the calculations was
not firmly agreed upon. Some were happy with a minimum of one year and
a maximum of five years if the data were available while others preferred
a minimum of two years. One participant suggested that the projected
life time of the plant might be a more appropriate record length to insure
compliance with the standard. The majority, however, were reluctant to
recommend a specific record length without first examining the whole
question in much greater detail.
6. At least one working group member expressed formal concern over the
difficulty of fitting air quality data with a suitable functional form.
See comments by Wevodau, Section 3.2.1. This concern plus others lead
this member to conclude that the enumerative approach, as recommended
in the draft guideline, was more acceptable at this time. Also see a
comment submitted by a non-working group member, R. Porter, Section 3.2.2,
-------
-57-
in which the suggestion was made to include with the fitting procedure
a formal test of "goodness of fit".
In addition to the statement above, Mike Williams expressed reserva-
tions about the use of modeled second highest concentration estimates.
His statement was submitted after the Group 1-2 meeting was adjourned
and was therefore placed in Section 3.2.3 under the title "Rationale for
Elimination of the Maximum of Second Highest for Modeling Purposes".
An additional comment made outside the group meeting which addressed
William's comment was submitted by R. Porter and was placed in Section
3.2.2.
-------
-58-
-------
-59-
2.3 GROUP 1-3: SET 2 POLLUTANTS (CO, M? , PHOTOCHEMICAL OXIDANT)
PARTICIPANTS
B. A. Egan
P. Finkelstein
D. G. Fox
E. Y. Leong
Moderator/Reporter: K. L. Brubaker
G. L. Melvin
L. E. Niemeyer
A. J. Ranzieri
P. M. Roth
CONTENTS
2.3,1 Overview and Philosophy
2.3.2 General Considerations
2.3.3 Carbon Monoxide Model Recommendations
2.3.4 Short-Term Photochemical Pollutant Model Recommendations
2.3,5 Long-Term NO Model Recommendations
&
-------
-60-
2.S.I Overview and Philosophy
In order to recommend a model for general application, it is necessary
that the model meet appropriately specified performance criteria. In our
opinion, while a number of models are now available for use, it is not
possible as of this writing to appraise the performance of these models for
one or more of the following reasons:
- performance criteria do not now exist
- evidence of verification has not been satisfactorily reviewed
- adequate evidence of verification is not available
- acceptable evaluation exercises have not been undertaken
As a consequence, we recommend the identification of only generic classes
of models for use, avoiding the recommendation of any specific model.
While this position is in our view a wise and appropriate one to
adopt, we realize that it falls short of meeting the short-term needs of
the user community. In order to partially satisfy this need, we propose
a two level hierarchy of model usage —
- screening (general estimation and potential problem identification)
- refined exercise (attempt at "best" prediction)
*
In the case of CO prediction models, we tentatively recommend specific
models for screening use, but at this time we can only recommend several
generic types of model as being potentially suitable for refined calculations
in practical applications. The specific models tentatively recommended
for screening purposes may be deemed suitable for certain types of refined
calculation at some time in the future upon completion and/or presentation
of adequate verification studies.
In the case of the photochemical pollutants, we again recommend
specific methods for screening purposes but only generic types of approaches
for refined calculation. In this case however, the screening procedures
are not potentially suitable for refined predictions.
In any case, it is incumbent upon the user to assure himself that a
model is truly appropriate for a particular application, be it for screening
purposes or refined calculation.
Subject to evidence of verification.
-------
-61-
Since we are unable to recommend specific models for refined use
at this time and recognizing the eventual need for such recommendation, we
would strongly encourage that the following steps be undertaken as quickly
as possible and with sufficient talent and resources:
- Establish relevant and unambiguous performance criteria. We
would encourage the convening of a panel of experts to achieve
this end.
- Establish an independent group having a charter to perform a
continuing evaluative function — that is, determining if
candidate models satisfy the specified performance criteria.
- Incorporate new information and knowledge into these recommenda-
tions annually and introduce specific model endorsements as rapidly
as feasible (perhaps bimonthly or quarterly).
2.3.2 General Considerations
Applications requiring the use of air quality models for set 2
pollutants fall into a wide variety of categories, of which the following
are examples:
- State Implementation Plan Revisions
- Air Quality Maintenance Plans
- Regional Transportation Plans, including Transporation
Control Plans, Transportation System Management, Trans-
portation Improvement Program Requirements
- EIS/EIR's
- NSR
- Indirect Source Review
- PSD
- Others
A need for both regional and local models is apparent from a consideration
of these applications and both needs can rarely be satisfied by a single
model. By our definition, local models deal with distance scales of the
order of 1 km or less and regional models with distance scales greater
than 1 km.
Several generic classes of models can be identified, each having
advantages and disadvantages. The principal classes and some of their
limitations are:
-------
-62-
- Gaussian
a) considerable simplification of real phenomena
b) cannot treat low wind speed cases
c) cannot properly treat cases involving complex terrain
d) cannot properly treat cases involving complex source geometries
(mainly depressed and elevated roadway sections)
e) cannot treat time varying behavior
- Numerical
a) specification of the appropriate values of the diffusion
parameters can be difficult
b) generally requires more meteorological data than routinely
available
c) the effects of numerical errors (e.g., pseudo-diffusion) must
be evaluated
- Statistical/Empirical
a) low spatial and/or temporal resolution
b) often site or region-specific
- Physical
a) specification of the appropriate values of the diffusion
parameters difficult
b) site-specific
c) difficulty in achieving dynamic and geometric similarity
A general assignment of model classes to local and regional scale
applications may be made as follows:
Local/Site-Specific Regional
Gaussian Numerical
Numerical Statistical/Empirical
Physical Gaussian
Care needs to be exercised in complex circumstances in which these general
assignments may be inappropriate and in which a sophisticated analysis may
be required. We emphasize that the burden is on the user to consider his
specific application and needs carefully, taking the limitations of different
models or model classes into account.
The main criterion that should be used in selecting a model for a
specific application, is the past performance of the model in situations
which approximate the applications at hand as closely as possible. This
verification evidence should include work by an independent group other than
the model developer if possible, and should cover a wide range of conditions
-------
-63-
in order to clearly define the limitations of the model. If past verification
is inadequate, a verification study should be required as part of the
analysis, including the specification of evaluation procedures and per-
formance guidelines. Verification is important because it provides the
best mechanism for identifying model limitations in practice and for
determining the suitability of different models for specific applications
in an appropriate manner.
Cost/benefit considerations should play a role in determining the
level of sophistication with which the air quality analysis is to be carried
out. Some of the relevant factors are:
- the economic consequences or cost of making a specific
decision compared to the cost of the analysis necessary
to justify the decision technically,
- the extent to which an increased level of effort in doing
an analysis is justified by an increased level of confidence
in the results,
- the available resources (money, skilled manpower, computer)
- the data requirements (meteorological, emissions, air
quality).
We suggest that the information in the Guideline on the following topics
be updated periodically:
- the codification or identification of typical applications
- model verification information, classified particularly with
respect to typical applications
2.2.3 Carbon Monoxide Model Recommendations
Consistent with the philosophy outlined in Sec. 2.3.1, we can
tentatively recommend certain models for use as "screening" tools, subject
to evidence of verification. The term "screening" tools applies to
methodologies which provide approximate and conservative estimates of
the air quality impact of a specific source type so as to identify those
situations which require no further air quality impact evaluation:
-------
-64-
Model
HIWAY
Hanna - Gifford - HIWAY
(Point and Area Source
Models Recommended in
Sections 2.1 and 2.2)
Source Category
Individual Project Reviews
Transportation Network Review
Point Sources
These models must, of course, be used in a manner consistent with the
scientific principles and assumptions inherent in their development. Further,
the user should apply these models only to situations similar to those in
which they have been verified. Also, it is recommended that U.S. EPA
immediately subject these models to further verification procedures.
2.2.4 Short-Term Photochemical Pollutant Model Recommendations
In this section we divide our discussion into two parts, one dealing
with the use of models for screening purposes and one with the use of models
for more refined predictions. We also give recognition to the problems
of point source and regional modeling of photochemically reactive pollutants.
2.3.4.1 Screening Purposes
The purpose of a screening exercise is to eliminate from further
consideration those sources which are very unlikely to cause or contribute
to ambient concentrations in excess of the air quality standards. A suggested
technique for screening NO sources with respect to NO standards is to use
X ^
a simple gaussian diffusion model in conjunction with the appropriate
Briggs formula to compute plume rise. It should be assumed that all NO
is emitted in the form of NO and that NO is a conservative pollutant.
Future verification work may allow the assumption of 100% conversion of
NO to NO to be relaxed, thereby increasing the utility of this technique
A £
as a screening procedure. The Gaussian model is used to estimate the
highest 1-hr, maximum ground level concentration taking into account the
existing background concentration, potential for downwash of the plume,
Subject to evidence of verification. A specific concern is the validity
of this model for "worst-case" low wind speed conditions.
-------
-65-
terrain, other sources and the full range of possible meteorological
conditions. The approach should be conservative in nature.
As a first approximation, several techniques may be used to estimate
the regional impact on oxidant levels of the emission of oxidant precursors.
The technique under development by EPA (EPA, January 1977), commonly known
as the Dimitriades - Dodge Isopleth technique, should be viewed at present
as the preferred screening procedure. The technique, if adapted for the
user's specific area and circumstances, provides a quick estimate of the
degree of control required. (A code for such use is now being developed
under contract to EPA). If the necessary data are unavailable for develop-
ing a site-specific system of curves, the more general curves should be
used. If it is not possible for either set of curves to be used, the only
available alternative is Rollback; however, we have serious reservations
concerning the use of this procedure.
2.3.4. 2 Eefined Predictions
Once it has been determined through a screening exercise that
exceedances of the air quality standard may occur, it may be appropriate
to pursue a more refined modeling approach. Such approaches can be
divided into point source and regional models.
In order to calculate the temporal and spatial distributions of
NO and 0 that occur in the atmosphere, it is essential that a numerical
£* J
modeling approach be used. Numerical models have the ability to treat
through realistic mathematical representation of the transport, diffusion,
and chemical reaction processes that occur in the atmosphere. Two classes
(grid and trajectory) of regional numerical models have been developed.
These models are currently undergoing testing and evaluation. Until these
validation studies are completed, these models must be exercised with caution.
2.3.4.3 Models for Reactive Plumes from Point Sources
Several models which are designed for this specific type of application
have recently been developed or are currently under development. Simulation of
the transport, dispersion and chemical evolution of plumes of reactive material
from point sources requires treatments of the simultaneous spread of the plume,
entrailment of ambient air with pollutant concentrations and chemical composition
-------
-66-
different from that within the plume, the effects of these phenomena on the
chemical dynamics within the plume, and advection of the plume, all on a
time dependent basis. Clearly, models designed specifically for this type
of application should be used.
The data requirements for reactive plume models are more demanding
than for Gaussian models. In many cases considerable care must be exercised
in selecting the most appropriate and representative inputs, in particular
the meteorological and ambient air quality variables. This may require the
expertise of experienced modelers and meteorologists, depending on the type
of problem. Of course, this does not preclude the applicant from supple-
menting existing data bases, designing an aerometic monitoring system, and
collecting field data to more adequately establish inputs. A model's
performance can be evaluated only through appropriate and well considered
verification studies based on sound inputs and comparative data.
2.3.4.4 Regional Models
The assessment of the temporal and spatial impacts of a new
emission source of NO and RHC on regional air quality requires the use
of regional airshed models. Airshed models require substantial amounts
of data, including a gridded emission inventory for both mobile and sta-
tionary sources of NO (NO + NO) and RHC. In addition, depending on the
} x 2
chemical mechanism, it may be required to further classify the RHC according
to subgroups (such as olefins, aldehydes, paraffins and aromatics). This
gridded emission inventory generally requires a resolution of 1 to 5 km
depending on the application. In addition detailed meteorological data
are required to describe the wind flow, diffusion, and inversion fields.
Grid models are most appropriate for assessing air quality impacts on a
systems basis while reactive plume models are most useful for assessing the
impact of isolated sources. These models are complex and require experienced
modelers who can properly interpret the results and assess the impacts. It
is of importance to note that because of complexities of the models and their
substantial data requirements, their use will most likely be restricted
to applications in those regions where sufficient data are available.
Even in such regions, a careful editing of the data bases must be undertaken
to ensure proper interpretation of model results.
-------
-67-
Again, as with the reactive plume models, the reliability of model predictions
can only be assessed through carefully planned and designed validation studies,
Where input data are available, it is strongly recommended that validation
studies be considered an integral part of the modeling endeavor.
2.3.5 Long-Term N0y Model Recommendations
ILl
An appropriate screening procedure for determining impacts of new
sources on long-term (annual) NO levels would be to make use of an appro-
priate model now in use for conservative pollutants (such as COM in urban
areas, for example) together with the assumptions: 1) that all NO emissions
X
are really NO emissions and 2) that NO is a conservative pollutant. After
sufficient experience with this procedure, it may be possible to assume less
than 100% conversion of NO to NO and thereby to improve the usefulness of
X ^L
the technique as a screening tool.
In order to obtain a refined estimate it will be necessary to make
use of more sophisticated techniques applied on a case-by-case basis.
Supplementary Material
Descriptions of two reactive plume models may be found in Sec. 3.3.1
and 3.3.2.
-------
-68-
-------
-69-
2.4 GROUP II-I: LONG-RANGE TRANSPORT AND LOSS MECHANISMS
PARTICIPANTS
A. E. Boyer A. J. Ranzieri
E. Y. Leong J. A. Tikvart
D. J. Moore M. D. Williams
Moderator/Reporter: K. L. Brubaker
CONTENTS
2.4,1 Swmayy of Disaussi-on
2.4.2 Recommendations
-------
-70-
2.4.1 Summary of Discussion
The section in the EPA draft Guideline that is directly related to long-
range transport considerations is the first paragraph on page 21, which
reads in part "It is recommended that estimates should not be made for
distances greater than 100 kilometers and that air quality impact beyond
this distance be assumed to be negligible." The point is made further on
in the paragraph that impacts at this distance and beyond are "likely to
be small and for all except very large facilities," and the Regional
Administrator is directed to consider specific large individual sources
on a case-by-case basis if a threat to air quality standards or prevention
of significant deterioration (PSD) increments exists at these distances.
In the general discussion which arose from these remarks, the follow-
ing points were made and agreed upon by the participants:
• The guideline should refrain from specifying a single distance
cutoff. It was considered preferable to give a range of distances
to indicate some degree of uncertainty regarding the maximum distance
at which relatively simple models could be used for estimating pollu-
tant concentrations, and a range of 50-100 km was deemed appropriate
although a lower value of 25 km was also considered.
• Observational evidence indicating that Class I PSD increments
have been or could be exceeded at distances of the order
of 100 km from a single or a small number of sources was presented
and discussed. It was felt that the assumption of negligible
impact beyond 100 km would be invalid in a sufficiently large
number of cases that it should be deleted from the guideline.
(Specifically, data from Ref. 2 were presented which indicated
that S02 concentrations approaching 50 yg/m^ were observed
approximately 90 km downwind of a group of four power plants in
eastern England. Several participants reported knowledge of other
studies which also provided evidence for potential PSD increment
exceedances. These were provided after the conference and are
given as Refs. 1, 3-5.)
• Regarding the different averaging times specified in the PSD
regulations for S02, it was felt that if the short-term (3 and
24-hour) increments were not exceeded, then neither would the
annual increment. Attention should therefore be focused on the
possibility of a short-term exceedance in evaluating the potential
impact of a new source.
• The need for treatment of individual situations on a case-by-case
basis should be emphasized.
• Plume depletion and chemical conversion in plumes from elevated
sources can be significant at distances of the order of 25 km,
-------
-71-
and are normally significant at distances of 50-100 km and beyond
under conditions of appreciable vertical mixing. Plume depletion is
normally significant at considerably shorter distances for near-
ground level sources.
Attention was then focused on questions relating to the modeling of transport,
dispersion and depletion over distances beyond 50 km and the following
points were agreed upon:
• Techniques for long range transport modeling are available.
(Refs. 6-16, provided by participants after the conference.)
• Only limited comparison of predictions of these models with
observational data is available and the models cannot be
considered validated.
• A treatment of vertical dispersion using an approach based
on K-theory is generally more appropriate than one using
a Gaussian approach whenever plume depletion by dry deposition
is significant,due to the ability of the former to incorporate a
more realistic boundary condition at the earth's surface.
• Trajectory analysis of the type usually done in the available
models requires wind, pressure and temperature data both at
ground level and aloft at representative sites within the
area of interest.
Based upon this discussion, several changes in the EPA guideline
were recommended.
2.4.2 Peeomnendati-ons
It is recommended that the paragraph in the draft EPA Guideline on
Air Quality Models and Associated Data Bases which deals with long range
transport, specifically the first paragraph on page 21 of that document,
be altered to read as follows:
"The administration of the national prevention of significant
deterioration policy may require that the air quality impact of a
source be estimated for great distances downwind. It is uncertain,
however, what the impact of sources at such great distances is. There
are several reasons for this. Our knowledge of the dispersion coefficients
-------
-72-
*
for air quality models becomes increasingly tenuous with downwind distance.
Distances beyond 50-100 kilometers require substantial travel time at the
most frequently experienced wind speeds. As travel time increases, the
daily weather cycle is more likely to alter plume trajectories and dis-
persion characteristics. The impact at distances greater than 50-100
kilometers is likely to be small, however the Impacts are not necessarily
negligible for large sources. Techniques are available to examine
these impacts, but only limited experience in their use is currently
available. If it appears that a large source (for example, a 2000-MW
coal-fired power plant meeting new source performance standards) may
constitute a threat to ambient air quality standards or prevention of
significant deterioration increments at large distances, that source
should be considered on a case-by-case basis with available techniques."
Supplementary Material
A description of models used by the Central Electricity Generating
Board, England, for medium and long range predictions may be found in
Sec. 3.9.5.
Vertical dispersion in these situations is more appropriately treated
with models which are based on K-theory. There also are inherent
difficulties with Gaussian models in cases where plume depletion is
significant. Plume depletion would normally be significant at dis-
tances of the order of 50-100 kilometers and beyond under conditions
of appreciable vertical mixing, and at considerably shorter distances
for near-ground level sources.
-------
-73-
REFERENCES FOR SECTION 2.4.
Observations of Effects from Single Sources
at Long Range (50-100 km and beyond)
1. Drivas, P. and F. H. Shair, A Tracer Study of Pollutant Transport and
Dispersion in the Los Angeles Area, Atmos. Env. 8^, 1155 (1974).
2. Fisher, B. E. A. and P. R. Maul, The Mathematical Modeling of the
Transport of Sulphur Dioxide Across Country, presented at the symposium
"Systems and Models in Air and Water Pollution," organized by the
Institute of Measurement and Control, London, England, Sept. 22-24 (1976)
3. Heimbach, J. A., A. B. Super and J. T. McPartland, Dispersion from an
Elevated Source Over Colstrip, Montana, presented as paper no. 75-26.6
at the 68th Annual Air Pollution Control Association Meeting, Boston,
Massachusetts (1976).
4. Lamb, B. K., A. Lorenzen, F. H. Shair, Tracer Study of Power Plant
Emission Transport and Dispersion from the Oxnard/Venture Plain,
final report prepared for the California State Air Resources Board
under contract no. ARB-5-306, February (1977).
5. Williams, M. D. and R. Cudney, Predictions and Measurements of Power
Plant Plume Visibility Reductions and Terrain Interactions, pp. 415-
420, preprints of the Third Symposium on Atmospheric Turbulence,
Diffusion and Air Quality, sponsored by the American Meteorological
Society, October 19-22 (1976).
Long Range Transport Models
6. Bolin, B. and C. Persson, Regional Dispersion and Deposition of Atmos-
pheric Pollutants with Particular Application to Sulfur Pollution over
Western Europe, Tellus _24_, 281-310 (1975) .
7. Egan, B. A., K. S. Rao and A. Bass, A Three-Dimensional Adveotive-
Diffusive Model for Long Range Sulfate Transport and Transformation,
presented at the 7th International Technical Meeting on Air Pollution
Modeling and Its Application, Airlie, VA, Sept. 7-10 (1976).
8. Eliassen, A. and J. Saltbones, Decay and Transformation Rates of SO*,
as Estimated from Emission Data, Trajectories and Measured Air Con-
centrations, Atmos. Env. 9., 425-429 (1975).
9. Fisher, B. E. A. and P. R. Maul, The Mathematical Modelling of the
Transport of Sulphur Dioxide Across Country, Presented at the symposium
"Systems and Models in Air and Water Pollution," organized by the
Institute of Measurement and Control, London, England, Sept. 22-24 (1976)
10. Heffter, J. L. and G. A. Ferber, A Regional Continental Scale Transport,
Diffusion and Deposition Model, NOAA Technical Memorandum ERL-ARL-50,
June (1975).
11. Knox, J., Numerical Modeling of the Transport, Diffusion and Deposition
of Pollutants for Regions and Extended Scales, J.APCA. 24, 660-664 (1974)
-------
-74-
12. Sao, K. S., J. S. Lague and B. A. Egan, An Air Trajectory Model for
Regional Transport of Atmospheric Sulfates, pp. 325-331, preprints
of the Third Symposium on Atmospheric Turbulence, Diffusion and Air
Quality, sponsored by The American Meteorological Society, October
19-22 (1976).
13. Scriven, R. A. and B. E. A. Fisher, The Long Range Transport of Airborne
Material and Its Removal by Deposition and Washout - I. General Consi-
derations, Atmos. Env. 9_» 49-58 (1975).
14. Scriven, R. A. and B. E. A. Fisher, The Long Range Transport of Airborne
Material and Its Removal by Deposition and Washout - II, The Effect of
Turbulent Diffusion, Atmos. Env. 9^ 59-68 (1975).
15. Sheih, C. M., Application of a Lagrangian Statistical Trajectory Model
to the Simulation of Sulfur Pollution over Northeastern United States,
pp. 311-317, preprints of the Third Symposium on Atmospheric Turbulence,
Diffusion and Air Quality, sponsored by The American Meteorological
Society, October 19-22, (1976).
16. Wendell, L. L. D. C. Powell and R. L. Drake, A Regional Scale Model for
Computing Deposition and Ground Level Air Concentrations of 30% and
Sulfates from Elevated and Ground Sources, pp. 318-324, preprints of
the Third Symposium on Atmospheric Turbulence, Diffusion and Air Quality,
sponsored by The American Meteorological Society, October 19-22 (1976).
-------
2.5 GROUP II-2: PLUME DYNAMICS UNDER SPECIAL CONDITIONS
PARTICIPANTS
L. E. Niemeyer C. Simon
R. A. Porter R. I. Wevodau
P. Finkelstein
Moderator/Reporter: J. J. Roberts
CONTENTS
2,5.1 Introductory Remarks
2.5.2 Aerodynamic Dotinuash
2.5.3 Sea Breeze and Other Anomalous Circulation Patterns
2.5.4 Interactions of the Plume Dith an Elevated Inversion Layer
-------
-76-
2.5.1 Introductory Remarks
Despite the wide range of problems implied by this title, the working
group focused on only a few situations not routinely treated by the standard
models but which may lead to violations of short-term NAAQS or PSD incre-
ments, especially in Class I areas. Three situations were considered:
1. Aerodynamic downwash;
2. Sea breeze and other anomalous circulation patterns;
3. Interactions of the plume with an elevated inversion layer:
a. Fumigation
b. Trapping
The members were in unanimous agreement on the proper way to approach
these issues in a guideline document. Firstly, the guideline should state
clearly that it is incumbent upon the various parties involved in the
decision making process to determine whether or not any of these phenomena
are likely to occur, and, if so, whether in a manner likely to cause viola-
tions of short-term limits. Secondly, this screening process, as well as
any subsequent analyses, must be carried out under the guidance of persons
knowledgeable in the physical processes involved. Thirdly, since the
situations for downwash and unusual circulations require a case-by-case
analysis, no standard models or rules-of-thumb should be formally recommended
in the guideline. And, finally, to assist the user of the guideline in
selecting the best available and most appropriate analytical methods for
estimating such impacts, one or more pertinent references should be cited.
2.5.2 Aerodynamic Downuash
Included in this problem area are all situations wherein the dynamics
of the effluent in the immediate vicinity of the source are influenced in a
significant manner by nearby structures and terrain. Two closely related
approaches have, in general, been taken to simulate these conditions within
the structure of the standard Gaussian plume model:
1. An equivalent volumetric source is assumed, with a cross-
sectional area approximating that of the obstacles -
usually the building upon which the stack or vent is
situated;
-------
-77-
2. An enhanced DZ is employed, usually with an assumption of
zero plume rise.
These approaches are documented in references 1, 2, and 3 in sufficient
detail to permit the user to handle the most common problem of a single
building and its stack.
In a closely related matter, the members were agreed that the rule-
of-thumb, "a proper height for a stack is 2 1/2 times that of any nearby
structure," should not be arbitrarily applied to avoid downwash
conditions since (a) in many instances the rule is unreasonably conserva-
tive, causing unnecessary expense and unsightly high stacks, and (b) in
other instances it may be inadequate especially if localized terrain effects
influence the air flow in the vicinity of the source. (See Ref. 4 for a
further discussion on these topics.)
The members recommended that the guideline include a listing of
situations in which downwash effects might occur. These would serve as
warnings to the user that special attention should be given to the potential
for downwash. They would not represent explicit criteria. Included in this
list should be:
1. Relative heights of stack and nearby structures or land
features (both upwind and downwind);
2. Dimensions of the building and stack, in particular the
aspect ratio relative to the wind direction;
3. Emission characteristics - thermal flux and effluent
velocity at part load operation as well as at design
values; and
4. Multiple stack configurations.
2.5.5 Sea Breeze and Other Anomalous Circulation Patterns
The phenomenon commonly known as a"sea breeze" can have a substantial
influence on the transport and diffusion of pollutants. The EPA modeling
guideline should recommend that is be accounted for in major modeling studies
in areas where the sea breeze can be important. Three areas in which the
sea breeze can be important are in long—term modeling, short—term modeling
on a regional scale, and short—term modeling of individual sources. For
individual sources the sea breeze fumigation, documented by Lyons and
-------
-78-
Hales , can lead to high short-term concentrations of pollutants. For
short-term studies of a large area, the sea breeze can strongly influence
local circulation,changing wind and stability patterns and, again,
possibly causing high short-term levels in areas where they wouldn't be other-
wise expected. This has been documented by Gatz . For long-term models
the sea breeze effect can have a marked effect on the stability climatology
of the shore area, affecting wind patterns, stability, cloud cover, and
precipitation patterns. Great care must be taken to ensure that the data
Q
used in such a model is representative of the area. (See Estoque ).
A meteorological problem which is similar physically - but dissimilar
topographically - is that of the mountain/valley winds. This, too, is a
local, thermally-induced wind field which must be taken into account in any
type of diffusion modeling in an area prone to these winds, or serious
errors can and will occur. The need for correct, representative data is
never more crucially felt than while modeling in both these problem areas.
While the committee feels compelled to point out this deficiency in the
present draft guideline, we do not feel competent to draft an alternative
and wish to suggest that an experienced, knowledgeable scientist in this
area be retained to draft this portion of the Guideline.
2.5.4 Interactions of the Plume with an Elevated Inversion Layer
The group considered two topics under this heading:
1. Short-term downwash of a plume during transition from stable
to unstable conditions (fumigation)
2. Plume interaction with mixing height
Fumigation downwash may be a significant factor in evaluating large
sources relative to Class I 3-hr. PSD increments. The guideline user should
be aware of this and seek expert advice if fumigation is reasonably
anticipated to be a significant factor. The group recognizes that there are
many cases in which fumigation will not significantly affect 3-hr, (or
longer) concentrations, but we cannot offer specific procedures to determine
if fumigation is a significant factor. Fumigation is discussed in Turner's
Workbook^, "Meteorology and Atomic Energy"10 and TVA Studies11. In
evaluating the effect of fumigation on 3-hr, concentration, the user must
-------
-79-
take into consideration the short duration of fumigation at a given receptor
site.
Current models (e.g., CRSTER) assume no concentration at ground level
if effective stack height is greater than the mixing height.* The group
generally agreed this assumption is not uniformly valid.** Trapping
probably does occur sometimes even when the calculated effective stack height
is greater than the height of the mixed layer, and thus the idealized
assumption in models such as CRSTER may lead to underestimation of short-term
maxima. However, we are unable to recommend a change in current models at
this time, in part because of the uncertainty in the manner currently used to
estimate hourly mixing heights. The group believes this topic needs addi-
tional field evaluation in order to identify those physical conditions under
which plume trapping can be assumed to occur.
* See Sec. 3.9.6 for a summary of a model called "AQSTM" which was
contributed by G. Melvin. This model treats atmospheric trapping and
other phenomenon discussed in this section.
**See Sec. 3.9.2 for a. discussion of a case in point by M. Williams which
was submitted after the conference.
-------
-80-
REFERENCES FOR SECTION 2.5.
1. Briggs, G. A., Diffusion Estimation for1 Small Emissions, Environmental
Research Laboratories, Air Resources Atmosphere Turbulence and Diffusion
Laboratory 1973 Annual Report, USAEC Report ATDL-106, National Oceanic
and Atmospheric Administration, December 1974.
2. Huber, A. H. and Snyder, W. H., Stack Placement in the Lee of a Mountain
Ridge - A Wind Tunnel Study, Environmental Protection Agency, EPA-600/4-
76-047, 1976.
3. Riley, J. J., et al, A Numerical and Experimental Study of Stably
Stratified Flow Around Complex Terrain, Environmental Protection Agency,
EPA-600/4-76-021 and 022, Vols. 1 and 2, 1976.
4. Snyder, W. H. and Lawson, R. E., Jr., Determination of a Necessary
Height for a Stack Close to a Building - A Wind Study, Atmospheric
Environment, Vol. 10, pp. 683-691, 1976.
5. Lyons, W. A., et al, Detailed Field Measurements and Numerical Models of
SC>2 from Power Plants in the Lake Michigan Shoreline Environment, Report
to Wisconsin Electrical Power Company, p. 218, 1974.
Lyons, W. A. and Rubin, E. M., Aircraft Measurements of the Chicago Urban
Plume at 100 KM Downwind, Proceedings of the 3rd Symposium on Atmospheric
Turbulence, Diffusion and Air Quality, Raleigh, N.C., pp. 358-365, 1976.
6. Hales, J. Vern, Draft EIS for Indian River Power Plant, Delmarva Power
and Light Company, 1974.
7. Gatz, D., Chicago Air Pollution System Analysis Program," Final Report,
ANL/ES-CC-099, J. J. Roberts, Editor, Sec. 4.3 and 4.4, Feb. 1971.
8. Estoque, M. G., A Numerical Model of the Atmospheric Boundary Layer,
Journal of Geophysical Research, 68:4, pp. 1103-1113, 1963.
9. Turner, D. B., Workbook of Atmospheric Dispersion Estimates, Office at
Air Programs, Environmental Protection Agency, Pub. No. AP-26, 1970.
10. Slade, D. H., Meteorology and Atomic Energy - 1968, U.S. Atomic Energy
Commission, July 1968.
11. Full Scale Study of Inversion Breakup at Large Power Plants, Tennessee
Valley Authority report, p. 177, 1970.
-------
-81-
2.6 GROUP II-3: COMPLEX TERRAIN
PARTICIPANTS
S. Barr D. Henderson
B. A. Egan G. L. Melvin
D. G. Fox
Moderator: H. H. Slater
Reporter: A. E. Smith
CONTENTS
2.6.1 Summary of Discussions
2.6.2 Recommended Procedure for Complex Terrain
2.6.3 Assistance in Defining Screening Techniques in Complex Terrain
2.6.4 Supplementary Comments and Information
-------
-82-
2.6.1 Summary of Discussions
The group was in unanimous agreement that it is not possible to specify
a general model for providing a definitive statement concerning the air qual-
ity impact of a source or group of sources locating in complex terrain. While
a number of generic types and specific algorithms are available, it is not
possible to appraise the performance of these models or algorithms during
this workshop for one or more of the following reasons:
Lack of performance criteria,
Lack of evidence for adequate simulation of physical
processes for all possible situations, and
Lack of acceptable evaluation exercises.
The discussion was limited to topographic complexities (e.g., lake or
sea breezes were not discussed) and single sources. Considerable discussion
was devoted to the use of the Valley algorithm. Some validation data for
Valley were presented and explained(see Sec. 3.6.2). it was clear that Val-
ley was not uniquely defined for various members of the group. One opinion
held that specific meteorological conditions, 2.5 m/sec winds and F stability
for a total of 6 out of 24 hours, were intended to be used in the algorithm.
(This is the version described in the conference notebook.) Other users re-
ported that they relax this restriction in practice and use site-specific
meteorology. The point was also made that conditions conducive to highest
concentrations depend on the specific situation and may frequently occur with
moderate wind speeds, not always under stable conditions. It was also noted
that even as a screen, Valley is not applicable in all situations because it
might not treat critical situations like circular flows and inversions in
valleys. Doubts were also expressed concerning the use of Valley as a conser-
vative screening technique. Concern was also expressed that specifying the
meteorology would allow some group to seize on these conditions and say they
should always be used to estimate maximum concentrations even when such an
assumption would be incorrect. One Valley user expressed confidence in using
the algorithm as a screening procedure in mountainous terrain with the specific
meteorological input. Another noted that in less severe terrain results ob-
tained from Valley with site-specific meteorology were only negligibly less
well correlated with observed concentrations than results from a Gaussian
model using a plume half-height correction and that in general it appears to
-------
-83-
give satisfactory results, at least from viewpoint of a regulatory agency. It
was decided that two positions concerning Valley had developed and should be
noted:
Some members of the group felt comfortable with Valley, at
least as a reasonable screening procedure in complex ter-
rain, and
Most members of the group, however, felt that Valley could
not be recommended for general use in complex terrain or
even as a screening tool in all situations although steps
could be taken to render its use less objectionable as a
screen in certain circumstances.
Additional discussion centered on other available algorithms for and
experience with treating complex terrain. Results using a half-height correc-
tion to a Gaussian plume model in three river valleys in Illinois were pre-
sented. This experience also indicated that the worst agreement occurred
under calm conditions or with circulating flows. It was noted that poor
agreement between predicted and measured values is usually attributable to
poor source or meteorological data. In general, the group felt that good
meteorological data is almost always a problem, particularly in complex ter-
rain.
A method of correcting the Gaussian formula based on theoretical con-
siderations of flow around a hemispherical obstacle was also presented. The
method appears to work better than the full-lift assumption.
One proposal considered the use of a combination of meteorological and
air quality modeling to develop the concept of air shed zoning. 3 (See also
Sec. 3.6.1.) This approach was presented as being particularly applicable
to the designation of land areas for the purpose of the prevention of signi-
ficant deterioration.
A computerized flow correction model for use in mountainous terrain
was also described but the group did not take any position on such work.
The group made several attempts to develop some guidance material.
One suggestion was to enumerate the limitations on the use of various types
of models (Gaussian, K-theory, statistical, etc.) in complex terrain. This
approach was abandoned in favor of suggesting an administrative procedure
for use in complex terrain situations. The procedure is described in
Section 2.6.2. It includes a screening technique which is
-------
-84-
detailed in Section 2.6.3. Pertinent references are cited including some
analytical routines. Given the dearth of applicable, validated techniques,
the group's consensus called for requiring that a thorough and comprehensive
effort be made to evaluate sources in complex terrain and for assigning con-
siderable priority to developing and validating models for complex terrain.
2.6.2 Recommended Procedure for Complex Terrain
Given the agreement that no single model can adequately treat all com-
plex terrain situations, the group suggested that a specific procedure be estab-
lished that will lead to a rational and adequate assessment of a source locating
in complex terrain and that will result in a two level hierarchy of a model
usage consisting of conservative screening procedures followed by detailed
analytical studies when the screens are not passed and in the development of
data which will lead to further model refinement or input data acquisition.
The following process is recommended for use in evaluating sources desir-
ing to locate in complex terrain:
1) Review available data (air quality, meteorological, source
data, etc.) and the physical situation (action by source
alone).
2) Develop a rationale with regard to a particular analysis
technique (by analogy if necessary) (action by source
alone).
3) As a screening procedure, the source reviews initial re-
sults with Agency. (See recommendations in subsection 3.)
4) If no agreement is reached based on the screen or if the
source does not pass the screen, then the source has an
option to develop detailed site-specific data and perform
detailed analyses.
5) Agency evaluates detailed analyses.
6) Source constructs (if approval granted); however, air
quality monitoring may be required.
7) Air quality data reviewed as necessary.
This approach requires control agencies to have access to the services
of qualified personnel who are highly knowledgeable in dispersion climatology
and the characteristics of atmospheric flow in complex terrain. In addition,
the EPA must serve as a clearinghouse for data concerning the application of
various models in complex terrain situations.
-------
-55-
It is further recommended that specific procedures can be applied for use
as screening tools for specific physical situations. Within the time limits
of the session, such a procedure could not be fully developed. However, an
outline of phenomena which should be considered in various complex terrain
situations and suggestions for modifications to flat-terrain models are pre-
sented in the following section of this report along with the references to
some of the appropriate literature.
For those sources which do not pass the screening procedures, detailed
studies may be conducted by the sources to develop the meteorological data
necessary to select and verify an analysis procedure which will result in an
air quality impact assessment sufficiently refined to make the necessary sit-
ing decisions. At this point the assistance of a meteorologist knowledgeable
in complex terrain problems and familiar with the area of interest is required.
2.6.3 Assistance in Defining Screening Techniques in Complex Terrain
This section provides a general discussion of phenomena which should
be considered in addressing compliance with air quality goals in regions of
rough terrain. The purpose is to provide some guidance on appropriate model-
ing techniques as a function of various source configurations with respect to
topography and meteorological situations which might be expected. Modifica-
tions are suggested which could be incorporated in a dispersion model origi-
nally developed for flat terrain analyses and which would be one of a number
of generic types (e.g., Gaussian, "K-theory"). The modifications are proposed
in the context of developing "conservative" models which would tend to over-
predict air quality impact and which would be appropriately used for "screening"
purposes. Two broad categories of meteorological flow situations have been
defined: 1) plume interactions with terrain under organized flow conditions,
and 2) effects of local meteorological phenomena on dispersion in complex
terrain.
2.6.5.1 Plume Interactions with Terrain under Organized Flow Conditions
This section is concerned with how a plume affects ground-level concen-
trations in flow situations which might be considered to be organized in
the sense that the wind field would have high spatial correlations from one
location to another and would not be significantly affected by local circula-
-------
-86-
tions. These are flows which lend themselves to some simplified types of
theories (e.g., potential flow, modified potential flow, etc.). The approaches
discussed below are categorized by different geometry and atmospheric stabil-
ity classification considerations.
(A) Alterations to Plume Trajectories Due to Terrain Effects
Plume Height Greater than the Height of Nearby Terrain. Under these
conditions, models which are appropriate for flat terrain can generally be
utilized with some modest modifications. If the plume height is much greater
than the terrain, flat terrain assumptions may be valid if some consideration
is given to the possible enhancement of turbulence due to "roughness."1 For
plume heights not much greater than terrain heights, some modifications to
the plume centerline trajectory seem appropriate. If the plume is embedded
in a stable layer above the terrain, it is a reasonable approximation to
assume that the plume trajectory would be horizontal for purposes of computing
ground surface concentrations. (Note, however, the comment about the possible
effects of "lee wave" phenomena. )
For a plume approaching a terrain object during neutral atmospheric
stability conditions, modifications which are suggested from potential flow
theory appear appropriate. A commonly used modification to Gaussian models
involves adding an increment to the plume height over the terrain equal to
approximately one-half the terrain height (distances estimated from the valley
floor). A more conservative estimating technique would be to allow a smaller
lift of the plume centerline above the terrain. This is, however, a poten-
tially poor assumption when the plume height is nearly equal to the terrain
height. Under non-stable conditions, allowing full lift of the plume, that
is, a terrain following trajectory, would appear to be generally nonconserva-
tive, although it appears from a theoretical point of view to be reasonable
for flow normal to two-dimensional type ridge orientations.2
Plume Height Lower than the Height of Nearby Terrain. If the plume
height is initially lower than the height of nearby terrain, the approach to
defining the flow pattern and, hence, plume centerline trajectories, depends
critically on the local topography, atmospheric stability, and airflow. Poten-
tial flow theory provides some insight on the approach for neutral atmospheric
stability conditions. If the flow is normal to a ridge, a plume embedded in
-------
-87-
the flow will accelerate and lift over the ridge with an actual center line-
standoff distance equal to approximately one-half the plume height above the
upwind valley flow. Because of the flow acceleration, however, the stream-
line spacing decreases and, through distortion effects, the vertical dimen-
sions of the plume decrease. The net effect suggest no alternation of the
o
effective standoff distance in the exponential term in a Gaussian model.
A numerical simulation appropriate for this flow situation should account for
these kinematic effects automatically. If the flow is normal to a more iso-
lated three-dimensional terrain object (as opposed to a two-dimensional
ridge), the effective standoff distance (considering both trajectory alter-
nations and plume distortion effects) will be less than that for flow over a
ridge shape. A value of about one-half the effective plume height in the
absence of terrain is an approximation suggested for terrain objects with
roughly equal horizontal and vertical dimensions. Under stable atmospheric
conditions, for flow normal to a two-dimensional ridge, the flow may pass
over the ridge if the stability is weak and wind speed is not too low. If
the temperature inversion is strong and the wind weak, the flow may "block"
and effluents would tend to stagnate upwind of the terrain. In either case,
careful consideration would be given to how the principle of conservation of
mass will act to alter plume trajectories and constrain the types of flow
situations possible. On-site meteorological measurements, as well as any
information on air quality concentration patterns, are extremely useful in
such situations. For stable flow normal to more three-dimensionally shaped
objects, the possibility of "blocking" is greatly reduced because the flow
will tend to pass around the sides of the terrain. The possibilities for
relatively high ground level concentrations exists under these conditions,
however, if the plume flows directly toward the terrain. The frequency of
occurrence of possible meteorological conditions which would cause this flow
situation should be carefully assessed.
Under these postulated situations, insight regarding the expected plume
behavior can be greatly enhanced with field observations and/or scaled-down
physical modeling studies (stratified wind tunnel or water-towing tank experi-
ments). Eowever, the frequency of occurrence and persistance of meteorological
conditions cannot be determined by these techniques.
-------
-88-
(B) Alterations to Turbulent Diffusion fates in Complex Terrain Modeling
Studies
General Considerations. A number of field measurement studies have
indicated that the association utilized in flat terrain of stability classi-
fication indices with turbulent diffusion rates are not necessarily valid in
complex terrain. In particular, a number of phenomena tend to decrease the
possibilities for very low diffusion rates associated with stable classifica-
tions. 3»6»7 A shift of stable classifications toward neutral appears
often to be appropriate for purposes of estimating atmospheric turbulence
levels. Under stable conditions, the presence of terrain may also greatly
enhance crosswind "meandering" of a plume which in effect reduces time-
averaged concentration values.
Considerations in the Near-Field of Sources. For purposes of examining
the impact of emission sources on nearby high terrain (within 10 km) careful
consideration should be given to the effects of buoyancy-induced entrainment
on plume dilution during the rising phase of plume growth.1*'5 Inclusion of
buoyancy-induced turbulence can result in markedly different estimations of
plume centerline and ground-level concentration values.
Far-Field Air Quality Impact. Additional considerations must be given
to estimating the impact of plumes which travel large distances (30 km or
more) before encountering high terrain:
1) If at the location of the encounter, the plume dimensions
are the same order or larger than the terrain dimensions
and the plume centerline is below the terrain height, then
near centerline concentrations can be expected to occur at
ground level.
2) The transport wind field may be highly variable both spa-
tially and temporally. Therefore, methods for estimating
dispersion rates under these conditions need to be given
special attention.
3) Removal of gases and particulates by chemical reactions,
scavenging, and deposition processes may be important to
include in the estimation techniques to provide realistic
estimates.
-------
-89-
(C) Other Factors Affecting the Estimation of Air Quality Levels under
Organized Flow Conditions
Lee-Side FloDs. Observational data and physical model experiments
show that under stable atmospheric conditions, the flow may accelerate on the
lee side of mountain ranges and cause streamlines to closely approach the lee
surface. It is possible that this type of flow could cause a down draft of a
plume toward the surface and would increase ground-level concentrations.
This type of flow is often characterized by the production of lee-waves down-
wind of the mountains. Flow separation effects on effluent dispersion of
sources both upwind and downwind of an obstruction are discussed in a subse-
quent section.
Channeling Effects. Topographic features will alter larger scale
meteorological flows to tend to follow terrain contours. This results in
"channeling" of winds into preferred directions and therefore an increased
persistence of winds in directions along valleys. These effects are of spe-
cial importance in the analysis and application of wind data at a site for
purposes of estimating 24-hour and annual average concentrations where, for
point sources, persistence of wind direction is a major concern.
Fumigation and Limited Mixing Depth Effect. Special attention should
be given to possible high ground-level concentrations on high terrain during
meteorological flow conditions characterized by fumigation of elevated plumes
to the surface by vigorous mixing. Other factors being equal, larger ground-
level concentrations can be expected under these conditions in rough terrain
than in flat terrain. Another concern is limitations to dilution rates
associated with the "lidding" effects of an elevated inversion above high
terrain. This phenomena needs further understanding; an approach similar to
that adopted for fumigation simulations may be reasonable. Numerical simula-
tion models which can account for the effects of spatially and time varying
meteorological parameters would be especially appropriate for applications of
this type.
-------
-90-
2.6.S.2 Effects of Local Meteorological Phenomena on Dispersion in
Complex Terrain
(A) Surface generated flow systems
(i) Upslope-downslope
(ii) Surface inversion dynamics
(iii) Mechanical forcing and turbulence
(B) Interaction of surface generated flows with
larger scale meteorology
(i) Separation and decoupling
(ii) Channeling
(iii) Lee waves and blocking
(C) Worst case estimates
(A) Surface Generated Flow Systems
A major complication in dealing with dispersion over irregular topog-
raphy is the fact that a simple atmosphere-surface interface does not e^ist
as it does over a flat surface. The amount of solar energy received by
irregular terrain surfaces varies widely with elevation and orientation of
the surface toward the sun (aspect). This differential heating gives rise
to locally forced wind systems and atmospheric stability variations which
have to be considered. The irregular terrain, simply by its presence, also
acts to alter an otherwise straight wind.
(i) Upslope-Downs lope
The general character of surface winds in mountainous terrain is
readily identified. Winds blow upslope (toward higher elevation terrain) on
clear sunny days and downslope (toward lower elevation terrain) at other
times.12'13'11**19 This description is deceptively simple because complex
valley-mountain geometry often disrupts this flow.11*'19 Modeling the disper-
sion from ground or low level sources must take into account these persistent
and regular upslope-downslope winds. Since such winds are rather shallow,
often not exceeding 100-200 meters depth, 10,13,19 tall sources may be sub-
jected to quite different flows underlining the need for site specific data
taken at both the source emission and effective plume height levels. One
must be certain in areas of complex terrain to ensure that the winds used for
dispersion analysis in fact represent the immediate environment of the site.
-------
-92-
(ii) Surface Inversion Dynamics
Surface based inversions often form in mountain valleys.13'19 Ground
based emissions, low level sources, as well as tall sources, can be trapped
by such conditions. Maximum surface concentrations may occur as a result of
the limited vertical mixing, especially during times of minimum inversion
height. Gaussian models are of limited value for calculating such concentra-
tions. Rather, a model which divides the total emission by the volume flow
rate of air in the valley below the inversion, a so-called box model, is more
appropriate.15'22 This technique may be suitable as a conservative screening
method.
(iii) Mechanical Forcing and Turbulence
Turbulence can be forced, enhancing mixing and dispersion, by the
presence of topographic relief.16 Such a condition is difficult to analyze
although terrain types can be identified which give rise to mechanical turbu-
lence. Abrupt changes in elevation, isolated topographic features and very
irregular surfaces all can contribute to mechanical forcing.11' Physical
modeling techniques (wind tunnel, water channels) can be used most success-
fully to qualify and quantify such mechanical turbulence generation. Physi-
cal modeling techniques do not, however, deal with the frequency of occurrence
and persistance of the meteorological conditions modeled.
(B) Interactions of Surface Generated Flows with Larger Scale
Meteorology
A major complication of mountain meteorology is the fact that normal
passage of large-scale meteorological systems is greatly altered by the
topography. The expected surface induced flows, for example, may be strongly
altered or not occur at all as a result of larger scale patterns. These
local effects of larger scale systems can have a major effect on dispersion
modeling.
(i) Separation and Decoupling
Particularly when wind blows perpendicular to a ridge the flow may
separate from the downwind (leeward) side. Separation causes a closed circu-
lation which can bring a plume down to the surface as well as a wake effect
often far downwind of the obstacle.9 It is not uncommon for various types
-------
-92-
of flow systems to exist within a mountain valley reasonably independent of,
or decoupled from, the aloft meteorology. 9»1'*'19 The modeling of a plume
under such situations can be very complex, especially if the plume travels
between each flow regime. Phenomena based flow fields can be introduced into
Gaussian models to be used as screening techniques. However, a preferred
modeling technology is again the use of wind tunnels and/or water channels,
but the problem of persistence would need further evaluation.
(i,i) Channel-ing
Wind components blowing parallel with a valley often descend well into
the valley and channel a strong flow along the valley axis. 9 »12 »Jl* »19 The
conventional Gaussian model is appropriate so long as the wind field is
properly specified.
(Hi) Lee Waves and Blocking
Lee waves generally occur with stable flows and set up a standing pat-
tern of waves downwind of major terrain features.9'16 Depending upon how
close to the surface these waves and their associated flows occur, either
separation or a high velocity surface flow may result.9'19 Blocking, another
result of stable flow, occurs upwind (windward) of terrain and basically
represents a region of stagnation.9'18 Upwind slopes with a reasonable flow
component perpendicular to the ridge line may be subject to such blocking
under stable atmospheric conditions. A box model such as described above
could be calculated using the ridge top as mixing height and a surface area
appropriately defined. Such a process could be used as a screening technique.
An alternative is the use of combined meteorological and dispersion modeling.
(C) Worst Case Estimates
Worst case estimates of ground level concentrations under terrain-
disturbed flow conditions may be addressed with simple modeling techniques
if care is taken in selecting model input variables. Perhaps the most impor-
tant effect of topographic flow features is the severe constraint on repre-
sentativeness of input data including mean wind, stability, and turbulence,
since these flows exhibit strong space-time variability, particularly in the
vertical. Some determination must be made that the plume is involved in the
same flow regime as the meteorological input data throughout its path to the
potential receptor site.
2 3
-------
-93-
Many drainage flows are relatively shallow (< 200 m in depth) and wind
measurements near ground level may not apply to an elevated plume if it is
above the drainage regime. Also the lateral radius of representativeness of
wind measurements becomes very small in broken terrain.
If the net transport is properly estimated and if alterations to dilu-
tion rate due to local effects are accounted for, Gaussian plume or box model-
ing concepts may be reasonably applied to obtain bounds on ground level
concentrations of pollutants. In establishing procedures for estimating the
impact of a proposed facility the following steps should be considered :
1. Examination of topographic charts on several scales with
varying degrees of smoothing will aid in determining the
basic setting with regard to basin structure, channeling,
downwash, drainage,and stagnation potential.
2. A quantitative critical review of existing meteorological
records from sites within the same general topographic
domain will aid in establishing the existence and occur-
rence frequency of topographic effects on transport.
These data are likely to be sparse and not properly
located for a site analysis, so site meteorology must
be considered next.
3. Based on the insights gained from the first two steps,
perform a series of observations to establish the radius
of representativeness. This should include elevated
winds by means of pibals and concurrent observations from
combinations of topography such as different slope faces,
valley-mesa, hillside-plain, and source-receptor points.
In documenting local flows, consideration should be given
to their recurrence frequency, depth, lateral extent,
diurnal pattern, net wind speed,and some measure of tur-
bulence intensity.
4. The preliminary analysis outlined above gives a basis for
estimating bounds on ground level concentrations for screen-
ing purposes and aids in the design of subsequent monitoring
-------
-94-
networks. Using the morphology of local flows and the
design of the proposed facility, an assessment can be made
of the interaction of the plume with these wind domains,
including the likely pollutant trajectories.
Examples of simple model applications given the insights from steps 1-
4 above are:
a. Plume Involved in Drainage Wind. These flows maintain a
downslope component and are estimable from smoothed topo-
graphic charts. The turbulence in these winds is often
greater than would be implied by a simple AT predictor
and should be measured in order to not underestimate
dilution. Flows may vary from 1 to 10 m/sec but are
often remarkably steady and amenable to Gaussian plume
modeling along the determined (possibly curved) plume
axis. The daytime counterpart to the drainage wind is
generally weak, brief and sporadic and is dominated by
the large scale flow regime of the day.
b. Plume Elevated above Drainage Flow. If it is determined
that the plume and receptor are in the more uniform wind
field above the local drainage wind ,traditional modeling
techniques [for organized flows] may be applicable to
worst case determinations.
c. Stagnation. A plume imbedded in a stagnant flow created
by a basin structure or blocking by a ridge may be treated
using modified Gaussian plume techniques.21'22 (The Valley
model is an example of one such technique.) Plumes above
a stagnant region in a well defined wind field may be more
amenable to the traditional Gaussian plume model applica-
tions .
d. Separated Flows. These are local in character and depend
on the existence of some general meteorological conditions
of stability and wind speed. The geometry of the source
and the obstacle producing the separation is very important.
With an upstream source estimated to be involved in the
-------
-PS-
separated flow region, a well mixed zone on the lee side of
the obstacle ("box" model) may be adopted. With the source
in a lee side downwash, maximum ground level concentrations
would be due to a nearly direct path of pollutant to the
ground in a fumigation mode.
2.6.4 Supplementary Comments and Information
Supplementary comments and information were submitted by both members
of the working group and others. This material can be found in Sec. 3.6 and
is referenced below.
D. Fox submitted a description of the TAPAS model in the format of the
conference notebook (Sec. 3.6.1). This model can be applied in complex terrain.
In Sec. 3.6.2, H. Slater discusses the Valley model and some compari-
sons of observed and estimated concentrations including a description of the
conditions under which the data were obtained.
M. Williams (Sec. 3.9.2) has suggested a specific change in the wording
for p. 30 of the draft Guideline. The suggestion concerns the choice of wind
speed when an elevated source impacts high terrain. He presents data on the
unacceptability of ground level inferences of stability, wind speed, and direc-
tion during stable conditions. He also suggests some changes in the Valley
algorithm. In Sec. 3.6.4,he also comments on the use of plume half-height
corrections, the uniformity of the wind field in the lower atmosphere under
stable conditions, and the influence of terrain elements on stable plumes.
The letter from Howard M. Ellis of Enviroplan submitted by M. Smith
(Sec. 3.9.4) comments on the use of a half-height rather than a full-height
(CRSTER treatment) reduction in effective stack height in complex terrain and
on turbulence and dispersion enhancement.
G. Melvin (Sec. 3.9.6) has presented a description, in the notebook
format, of the Illinois' Air Quality Short Term Model (AQSTM) which can be
used in complex terrain.
In Sec. 3.6.3,D. Henderson has commented on the use of local meteoro-
logical conditions in Valley and on the conditions under which the model is
most applicable. He has also noted factors which determine the degree of con-
servatism of a box model.
-------
-96-
REFERENCES FOR SECTION 2.6.
1. Hovind, E. L., T. C. Spangler, and A. J. Anderson, 1974: The influence
of rough mountainous terrain upon plume dispersion from an elevated
source. Presented at the Symposium on Atmospheric Diffusion and Air
Pollution - American Meteorological Society. Santa Barbara, California,
September 1974. pp. 214-217.
2. Hunt, J. C. R., and P. J. Mulhearn, 1973: Turbulent dispersion from
sources near tuo-dimensional obstacles. J. Fluid Mech., 61. pp. 245-274.
3. Egan, B. A., 1975: Turbulent diffusion in complex terrain. Workshop on
Air Pollution Meteorology and Environmental Assessment, American Meteor-
ological Society, September 30-October 3, 1975, Boston, Massachusetts.
4. Pasquill, F., 1976: Atmospheric Parameters In Gaussian Plume Modeling
Part II, EPA-600/4-76-030b.
5. Egan, B. A., and A. Bass, 1976: Air quality modeling on effluent plumes
in rough terrain. Third Symposium on Atmospheric Turbulence Diffusion
and Air Quality. October 26-29, 1976, Raleigh, North Carolina.
6. Start, G. E., C. R. Dickson, and L. L. Wendell, 1975: Diffusion in a
canyon within rough mountainous terrain. J. Appl. Meteor., 24, 3, pp.
333-346.
7. Hinds, W. T., 1970: Diffusion over coastal mountains of southern Cali-
fornia. Atmospheric Environment, 2. pp. 541-558.
8. VanderHoven, I., 1976: A survey of field measurements of atmospheric
diffusion under low wind speed inversion conditions. Nuclear Safety,
Vol. 17, No. 2, March-April 1976.
9. Alaka, M. A. ed., 1968: The airflow over mountains . World Meterol.
Organ. Tech. Note 34, WMO No. 98 R.P. 43.
10. Bergen, J. D., 1969: J. Appl. Meteorol. 8. pp. 884-895.
11. Davidson, B. and J. Halitsky, 1961: Studies of the field of turbulence
in the lee of mountain ridges and tree lines. New York University.
12. DeFant, F., 1951: Local wind. In "Compendium of Meteorology," ed. T. F.
Malone. American Meteorol. Society, Boston, MA.
13. Geiger, R., 1959: The climate near the ground. Howard Univ. Press,
Ch. 18-20.
14. Panofsky, H. and B. Prasad, 1967: J. of Applied Meteorol. 6(3):493-499.
15. Thyer, N. and K. J. K. Buettner, 1962: On valley and mountain winds.
Dept. of Atmos. Sci., Univ. of Wash. AFCRL-62-1082.
-------
-97-
REFERENCES (Cont'dJ
16. Sato, Y. Onda and T. Saito, 1974: Turbulent diffusion in environmental
pollution, eds. Frankiel and Munn, Academic Press, New York, p. 241-251.
17. Snyder, W. H., 1972: Boundary Layer Meteorology 33 1.
18. Yih, C. S. 1965: Dynamics of Nonhomogeneous Fluids. Macmillan Co.
New York.
19. Yoshino, M. M. , 1975: Climate in a Small Area, University of Tokyo
Press.
20. Sagendorf and Dickson, 1974: Diffusion under low wind speed inversion
conditions. Proc. Symposium on Atmospheric Diffusion and Air Pollution,
American Meteorological Society, Santa Barbara, Calif. September, 1974.
21. Leahey, D. M. and J. Halitsky, 1973: Low wind turbulence statistics and
related diffusion estimates from a site located in the Hudson River
Valley. Atmos Env. 7, 49.
22. Fox, D. G. 1975: Man, leisure, and wildlands: A complex interaction.
Proc. First Eisenhower Consortium Res. Symp., Sept. 14-19, 1975, Vail,
Coro. Eisenhower Consortium Bull. 1. pp. 259-273.
23. Fox, D. G., M. A. Fosberg, W. E. Marlatt, and W. Reeser, 1976: Analysis
of Mountain Air Quality. Third Symposium on Atmospheric Turbulence Dif-
fusion and Air Quality. October 26-29, 1976. Raleigh, N.C.
-------
-98-
-------
-99-
2.7 GROUP II-4: CHARACTERIZATION OF TURBULENCE
PARTICIPANTS
H. E. Cramer
S. R. Hanna
M. E. Smith
Guest Participant: D. J. Moore
Moderator/Reporter: D. M. Rote
D. Bruce Turner
I. Van der Hoven
M. Lazaro
CONTENTS
2.7.1 Introductory Remarks
2.7.2 Vertical Profiles of Wind Speed
2.7.3 Comparison of STAR, A27 and aQ
2.7.4 Mixing Height Interpolation Scheme
2.7.5 Vertical Dispersion Estimates
2.7.6 Plume Dimension Stabilization Height
2.7.7 Horizontal Dispersion Estimates
2.7.8 Randomization of Wind Vector
2.7.9 Review and Comment on Pasquill's Recommendations for
Interim Chnages to the Pasquill-Gifford Curves
2.7.10 Use of Models in Urban vs. Rural Areas
-------
-100-
2.7.1 Introductory Remarks
This working group dealt primarily with detailed aspects of the
Gaussian plume model in general and with some issues specific to some of
the algorithms recommended in the proposed guideline. The problems asso-
ciated with special modeling situations such as complex terrain were not
discussed.
The general consensus of the group was that while some serious
concerns were raised and some important questions needed to be answered,
none were so serious as to prevent interim use of the Guideline recommended
simple terrain algorithms such as "RAM" and "CRSTER".
The most important issue raised by the group was the validity of the sub-
jectively extrapolated portion of the a curve for the Pasquill-Gifford stability
z
class A and its applicability to the computation of ground level concentra-
tions from tall stack emissions. Similar questions were also raised re-
garding the upward turning stability B curve and the downward turning
stability D, E, and F curves for a . The group supported the method of
z
multiple images for treating dispersion in a limited mixing layer and
stressed that for elevated releases the surface based stability parameter
must be accompanied by information about the layer containing the plume.
Of special importance is the position of the plume relative to a ground
based or elevated inversion layer.
There was significant difference of opinion about the validity of
the prediction of neutral hours by the "STAR" computer program and the
implications for both ground based and elevated releases.
The group unanimously agreed that these questions deserved serious
attention and recommended that a thorough and systematic study be con-
ducted of both the theoretical arguments and available data.
These issues are more fully discussed in the following text.
Discussions on several additional issues, including some specific to the
CRSTER algorithm are also presented.
2.7.2 Vertical Profiles of Wind Speed
The working group unanimously agreed that the representation of
the increase of wind speed with height now employed in the EPA modeling
-------
-101-
systems is satisfactory. This system takes the form of a power law with the
exponent varying according to stability categories.
2.7.S Comparison of STAR, A27, and a0
In order to make any dispersion evaluation, the available meteorological
data must be stratified according to categories representing the diffusive
capacity of the atmosphere. A variety of simplified approximations have been
used to represent this variable.
The group notes that the various systems, when applied to the same
data, give widely different distributions of stability categories. Many
authors have pointed out these differences, but a pair of examples show the
extent of the problem.
Table 1 compares the frequency of the Pasquill-Gifford stability
categories estimated from tower temperature difference (AT) measurements with
a set derived from standard deviation of the wind fluctuations (aA) measure-
0
ments. The discrepancies are enormous, with differences as great as a factor
of 3 or 4 among the major unstable, neutral and stable groups. Table 2 com-
pares a "STAR" calculation with the AT and a estimates. Again the variations
are huge.
A question was raised regarding the prediction of neutral stability
categories by the "STAR" program. While the group did not unanimously agree,
some participants expressed the opinion that the "STAR" program tended to pre-
dict unrealistically high numbers of neutral hours especially for low level
2
sources.
It is the consensus that the technique of estimating stability now used
in the EPA computer evaluations (STAR) can be used in the interim but that it
be reviewed to resolve this question and to develop an alternate system which
will eliminate the unrealistically large percentages of neutral hours if nec-
cessary.
We doubt that AT, at least over modest height intervals, will serve as
an alternative because of instrumental accuracy problems, and because it is
based on only one of the several key factors responsible for turbulence. A
possible alternative might be based on wind fluctuations for estimates of hori-
zontal dispersion and either AT or net radiaion measurements for vertical dis-
persion.
-------
Table 1
JOINT FREQUENCY OF OQ (30') and AT_nnl ., DATA* (%)
W ZOO - J
MARCH 1971 - FEBRUARY 1972
Category
A
B
C
D
E
F
G
Sum
A
0.20
0.57
1.41
0.82
0.01
0.00
3.20
I
B
0.10
0,33
0.84
0.45
0.03
0.00
1.84
13.13
Unstable
C
0.51
0.72
3.61
1.87
0.13
0.00
8.09
D
1.24
0.89
1.88
2.78
0.19
0.00
12.87
12.87
Neutral
E
5.58
4.02
9.97
29.21
0.26
0.00
60.29
F
2.03
1.33
1.97
3.35
1.50
0.00
12.29
76.23
Stable
G Sum
1.75 11.4(Tj
0.66 7.91 )36 28 Unstable
1.00 16.97J
1.59 45.90 45.90 Neutral
0.61 11.28^ ,
I ***
0.04 0.77 )20.05 Stable <»
I
5.65
102.23
From: Environmental Report, Trojan Nuclear Power Plant, Oregon,
-------
Table 2
**
STAR
AT
ae
STAR
AT
ae
ABC
1.08 7.68 7.93
3.20 1.84 8.09
11.40 7.91 16.97
Unstable
16.69
13.13
36.28
D E
57.17 8.74
12.87 60.29
45.90 11.28
Neutral
57.17
12.87
45.90
F
17.39
12.29
0.77
Stable
26.13
76.23
20.05
G
5.65
0.00
*
**
Data based on an Environmental Report on the Troj.an Nuclear Power Plant, Oregon as
reported in a letter dated November 22, 1976 from James Carson to Maynard Smith.
(See reference materials at end of report)
-------
-104-
For further discussion on problems associated with surface based meteoro-
logical data, see Section 3.9.2 for some remarks by M. Williams regarding the
LAPPES Program data.
2.7.4 Mixing Height Interpolation Scheme Used in CRSTER
There are reservations about the interpolation scheme used in CBSTER to
obtain hourly mixing heights. This system should be tested against available
data and the results should be documented. The scheme should be examined and
could be improved using hourly surface temperatures to interpolate the mixing
heights between the standard hours of radiosonde observations.
2.7.5 Vertical Dispersion Estimates
3
There is evidence that in their present form, the Pasquill-
Gifford a curves are unsuitable for calculation of the ground level concen-
z
tration due to emissions from tall stacks. In particular, the stability A
and, to some extent, the stability B curves may result in large overestimates
of the short-term maximum ground-level concentration (1-hr, to 3-hr, averages
especially) and in underestimates of its distance when compared with observa-
tions. Extrapolation of the A curves for a and
-------
-105-
elimination of the A curve and the use of the B curve for both A and B
stability categories. Alternatively, the 0 curves for the A, D, E, and
z
F categories may be reformulated to show an approximate linear dependence
on distance (on a log-log plot), starting at a distance 100 meters from
the source. If this procedure is adopted, the effects of thermal stratifi-
cation in limiting the vertical growth of plumes can be accounted for by
the use of multiple reflection terms in combination with a specified mixing
depth. These adjustments (i.e., inclusion of multiple reflections) apply
only to plumes that stabilize within the mixing layer and thus in practice
are most applicable to the A and D categories. It should be noted that the
algorithms RAM and CRSTER treat limited mixing in this manner for stabilities
A through D.
Additional comments on this subject were submitted after the conference
by M. Williams (see Sec. 3.7.1) and D. B. Turner and L. E. Niemeyer (see Sec. 3.7.3)
Finally, in order to account for variations in surface roughness
and urban/rural differences, it was suggested that some latitude be given
to exercise professional judgement in making minor adjustments in stability
class assignments. See related comments by Moore below.
2.7.6 Initial Plume Dimension
Plume size at the plume stabilization distance (at downwind
distances of approximately ten stack heights) should be determined and
incorporated into a and o determinations by adding variances or use of
virtual distances to account for initial size of buoyant plumes. A
2
possible algorithm for doing this has been discussed by G. Briggs .
2.7.7 Horizontal Dispersion Estimates
D. Moore3 of the Central Electricity Generating Board (CEGB)
presented some recent results of the determination of ground level a
values due to elevated sources. The results abtained from surface sampling
(with a sampling time of roughly one hour) to distances of 14 km, indi-
cated linear variation with distance for all cases studied. The constant
of proportionality varied from 0.04 under strong winds to 0.2 under light
winds. These results appeared to be very similar to the Pasquill-Gifford
-------
-106-
curves which apply to the plume distribution relative to its centerline and
which are characteristic of 3-minute sampling time. Currently, the CEGB
is using a values which are a continuous rather than discrete function of
y s
the stability variable .
The group concurred that the present PG curves for a should be used
for rural cases and for 1-hour sampling periods for elevated sources. This
position takes into consideration the original basis for these curves (ground-
level releases 3 to 10 minute sampling times) and the influence of wind shear
for elevated sources which effectively enhances the horizontal spread near
ground level.
The position was somewhat less clear for near ground releases. The
general opinion was that for ground level sources and 1-hour sampling times
the present PG curves for a should be adjusted for sampling time using the
1/5 power law.
To account for increased site roughness or heat flux, a values
should be shifted one stability category toward unstable (except a shift
from B to A).
The incorporation of the Briggs1 urban a curves in urban models
warrants further consideration.
Extreme caution is advised in the use of so-called "stability G"
category. Contrary to the trend for decreased dispersion with increasing
stability, the stable, light wind situation which characterizes the
"stability G" case has been shown (see for example Van der Hoven ) to
occur with rather large horizontal plume meander. Thus, the use of Pasquill G
results in serious overestimates of short-term concentrations from low-
level sources.
2.7.8 Wind Direction Randomization
Most wind data available for air pollution studies are recorded to
10° direction intervals, whereas the time hourly-mean values would have
-------
-107-
shown variation within these intervals. To assume that the winds are actually
fixed on the 10° radials will result in unrealistic maximization of concentra-
tions in these directions. To overcome this tendency, a randomizer has been
introduced in the CRSTER computer program to force deviation of the wind
from one hour to the next.
Occasionally this randomizer will repeat the same random number for
several hours in a row. It may therefore create an unrealistically large
3-hourly calculation. The repetition effect becomes very unlikely over
longer periods, and presents no problem.
It is recommended that this problem be reviewed, since three consecu-
tive identical wind directions is exceedingly unlikely in nature. A
possible solution would involve an automatic override to prevent successive
duplication of random numbers for short-term evaluations.
See further comments contributed after the conference by M. Smith,
Section 3.
2.7.9 Review and Comment on Pasquill's Recommendations for Interim
Changes to the Pasquill-Giffovd Curves1
1. No change should be made to the present "PG-curves" for O . It was
felt that Pasquill's recommendations regarding o" could not be generally
D
implemented at this time.
2. In regard to adjustments to stability classes to account for surface
roughness and the urban heat island effect, a practical compromise was
recommended: In an urban area the stability class should be changed
by one unit in the direction of unstable to account for the combined
effects of surface roughness and the heat island.
3. A correction should be made in the Gaussian model which takes account
of the plume dimension at an appropriate distance downwind from the
release point (see Section 2.7.6 of this report) but not necessarily
following precisely Pasquill's suggestion.
2.7.10 Use of Models in Urban vs. Rural Areas
It was noted that CRSTER is intended specifically for rural
calculations (i.e., relatively smooth surface without heat island effects).
-------
-108-
For urban cases (rough surface and urban heat island effect) the
appropriate version of RAM should be used in lieu of CRSTER. However,
the group emphasized that» the EPA should be certain that the RAM
algorithm has been validated by comparison with field data.
If CRSTER were to be used in urban-type situations the stability
class should be changed by one unit toward the unstable.
-------
-109-
FOOTNOTES FOR SECTION 2.7.
1. Cramer, in a post-conference contribution, see Sec. 3.7.2, emphasized
that this statement should not be interpreted to mean that the power-law
exponents currently used by the EPA are not subject to change. He
suggested some possible modifications.
2. Ibid. Cramer feels that the PG stability categories predicted by
STAR are very satisfactory provided that mixing height, vertical temper-
ature gradient, and the wind profile exponent are also specified for
each combination of wind speed and stability.
3. In his post-conference contribution, H. Cramer (see Sec. 3.7.2) has
prepared a well documented argument for modifying the PG 0Z curves,
this argument deserves careful attention.
4. The group was cognizant of the fact that the original az curves for
stability A was based on direct measurements, i.e., observations of
vertical distribution, only out to 100m and on indirect measurement,
i.e., reduction from observations of ground level distribution, out
to 800m (see Ref. 1).
5. In his presentation, Moore advocated that az and av be expressed as
analytical/empirical functions of source height, sampling time, and
surface roughness, as well as stability, wind speed and downwind dis-
tance. See Ref. 4 and Sec. 3.9.5.
-------
-110-
REFERENCES FOR SECTION 2.7.
1. Pasquill, F., Atmospheric Dispersion Parameters in Gaussian Plume
Modeling. Part II. Possible Requirements for Change in the Turner
Workbook Values, Environmental Sciences Research Laboratory, Office
of Research and Development, U.S. Environmental Protection Agency,
EPA-600/4-76-030b, June 1976.
2. Briggs, G.A., Lectures on Air Pollution and Environmental Impact
Analysis, AMS, 1975, TD883.A23.
3. Moore, D.J., A Simple Boundary-Layer Model for Predicting Time Mean
Ground-Level Concentrations of Material Emitted from Tall Chimneys,
paper presented for Ordinary Meeting of the Institution (Thermodynamics
and Fluid Mechanics Group) in London, November 1974.
4. Moore, D.J., Calculation of Ground Level Concentrations for Different
Sampling Periods and Source Locations, Atmospheric Pollution, pg. 51-60,
Elsevier Scientific Publishing Company, Amsterdam, Netherlands.
5. Briggs, G.A., Diffusion Estimation for Small Emissions, Environmental
Research Laboratories, Air Resources Atmosphere Turbulence and Diffusion
Laboratory 1973 Annual Report, USAEC Report ATDL-106, National Oceanic
and Atmospheric Administration, December 1974.
6. Van der Hoven, I., A Survey of Field Measurements of Atmospheric Diffusion
Under Low-Wind Speed Inversion Conditions, Nuclear Safety Vol. 17, No. 2,
March-April 1976.
-------
-111-
2.8 GROUP II-5: VALIDATION. AND CALIBRATION
PARTICIPANTS
N. E. Bowne W. A. Perkins
N. deNevers P. M. Roth
J. L. Shapiro
Moderator/Reporter: P. M. Roth
CONTENTS
2.8.1 Principles
2.8.2 Considerations
2.8.3 Other Issues
2.8.4 Proposed Mechanism for Meeting Requirements
2.8.5 Example Model Evaluation Information
2.8.6 Levels of Model Performance and Quality
-------
-112-
2.8.1
While the desirability of providing model recommendations to the
user community is well recognized, it is clearly unwarranted to offer such
recommendations when supporting information is inadequate or unavailable.
Consistent with this philosophy are the following requirements:
- The specification that models must meet certain standards of
performance if they are to be endorsed in the Guideline for
general use.
- The acceptance of model verification as a common practice in
overall model use where performance is not established.
To provide an orderly basis for determining model performance, we foresee
the following needs:
*
- Prescribed standards for model performance
- Specified model verification procedures
- Established performance evaluation procedures
In order to meet these needs, we strongly recommend that:
- An effort be mounted to establish the appropriate
standards and procedures (under the auspices of EPA).
- A continuing function be established within EPA expressly for
the purpose of model evaluation.
- Established standards and procedures be revisited on a. regular
basis.
See Section 2.8.4.
Recognizing that these recommendations do not satisfy the short-term
needs of model users, we recommend that the following practices be adopted
on an interim basis until the needed functions are put into practice:
Also of concern is the extent of generality of model applicability.
In other words, there is a need to place constraints on the range of
model applicability over which the performance standards are expected
to be met.
-------
-113-
- the extent of evaluation undertaken for models mentioned in the
Guideline be indicated in a manner consistent with Section
2.8.6.
- areas of applicability of the models and constraints (areas of
non-applicability) be listed.
Examples of the type of information to be supplied are shown in Section
2.8.5.
2.8.2 Considerations
We believe that attention must be given to the following considerations
in attempting to prepare guidelines for model evaluation practices and
procedures:
- Verification should be viewed in two contexts - general and site-
specific usage. The distinction between the two contexts should
be maintained wherever appropriate, giving recognition to potentially
differing verification requirements.
- Performance standards should vary in stringency and content, giving
recognition to:
- the differences in models
- variations in desired level of accuracy (compare screening
use and refined use of models)
- loftiness of performance goals. (This gives recognition of
the two-step process of (1) selecting a model with a certain
potential predictive capability and (2) establishing the extent
to which its capability is realized. See Section 2.8.6.
- Verification activities should be carried out with overall project
and information-gathering-goals in mind, i.e., give proper consi-
deration to:
- the problem to be resolved
- the nature of the Standard to be met (its form, its averaging
time, etc.)
- Verification goals should be in concert with potential level of
achievement insofar as it is limited by the accuracy of data input
to the model and data to be compared with predictions.
-------
-114-
- Obligations of the user (in contrast with the control agency) should
be specified
- Recognition should be given to the trade-offs between costs and
benefits prior to mounting any substantial verification program.
2.8.3 Other Issues
We wish to call attention to certain other issues that bear on model
evaluation:
- the "veriflability" of a model - Can a particular model, as a
practical matter, be evaluated? In some cases, it cannot.
- the permissibility of calibration - Under what conditions is the
practice warranted?
- the differences in verification requirements for long-term versus
short-term predictions
- the importance of evaluating models under conditions which "stress"
the model, i.e., which are designed to most readily uncover suspected
flaws or shortcomings
- the differing evaluative requirements for relative (e.g., rollback)
versus absolute (e.g., Gaussian plume models) predictors
- indication of the types or categories of statistical procedures
to be used in model evaluation, at a minimum
We believe that each of these issues should receive full consideration in
preparing a guideline document.
2.8.4 Proposed Mechanism for Meeting Requirements
Clearly a well defined mechanism must be established as quickly as
possible to meet the requirements outlined above and indeed to insure in a
larger sense that suitable models are properly developed and rationally
applied.
The Clean Air Act of 1970 requires modeling as a tool to achieve the
intent and goals of the Act. Accordingly EPA has the total responsibility
for all aspects of modeling from concept through application. Since the
-------
-115-
requirements for validation and performance as described herein are a key
part of the overall modeling problem, EPA has the responsibility to carry
out these tasks. It is a major effort on a continuing basis requiring addi-
tional internal support together with assistance by grant, contract and
consultants. In short, EPA must establish within its organization a group
responsible for carrying out the tasks described; to do so on an ad hoc
basis is totally inadequate.
There are a number of advantages in having an identifiable group
within EPA with total responsibility for the modeling effort including:
1. Model evaluation, performance, and applicability can be more readily
achieved and documented and the results made available to all parties.
2. Results of model applications can be evaluated and retained in a data
bank for use in subsequent model improvements.
3. Uniform methods can be established for processing new models.
4. A higher degree of standardization can be achieved in the use of models
both for enforcement and air quality management.
This proposed mechanism would provide for a periodic review of models in use
as well as those proposed for adoption and probably would involve formal
meetings with outside experts. However, the effectiveness of such formal
meetings would be greatly enhanced by having the benefit of the results of
a continuous effort within EPA.
2.8.5 Example Model Evaluation Information
Examples of evaluation information for models listed in the Guideline
may include the following information.
CDM - An urban multiple-source model that has been subjected to
limited evaluation for cities. Results indicate that
SO calculations are made more accurately than those for TSP.
Correlation coefficients ranged from to for SO
and correlation coefficients ranged from to for TSP.
CRSTER - A single point source model for the calculation of the dis-
tribution of short-term concentrations around the source.
Studies at four power plants indicate the maximum concentrations
-------
calculated are within a factor of 2 of the observed concentrations
at local monitoring stations.
Examples of limitations are:
CDM is an urban model that should not be applied to rural situations.
RAM is a multi-source short-term model that has not been evaluated.
It is composed from algorithms such as the Gaussina plume model and
plume rise equations similar (identical?) to GRSTER and in a similar
application could be expected to give similar results.
None of the models listed have been validated in complex terrain.
Use of differential receptor heighrs has not been validated.
2.8.6 Levels of Model Performance and Quality
We should recognize that models operate at varying levels of performance
and quality, and should be tested (validated, verified) against the standards
of the various levels. We suggest that there are roughly four such levels,
which in descending order are:
Top Level. A model in this level will predict concentration of any pollutant
at any time, any place, any averaging time, with uncertainty equal to the
uncertainty in the input data (emissions, meteorology). An example of this
type from physics is F = ma. For that model we truly can predict with
accuracy equal to input data accuracy, for all size and time scales, excluding
relativistic and uncertainty principle limitations.
To validate a model in this category would require comparing its time
and space resolved predictions with equally time and space resolved observa-
tions, using all sorts of meteorological inputs.
Second Level. A second level model would not claim to give accurate time
and space resolved answers but would attempt to predict time-resolved dis-
tributions of results which were comparable to observed ones. For example,
the observed mean and s.d. of the concentration at any point should agree
with the predicted one, within the uncertainty of the input data. An example
of this type from biology is mendelian genetics, which predicts the distribution
of properties (quite accurately) but generally does not predict, for example,
which seed will contain which properties. After the fact we have worked out
-------
-117-
a great deal of "first principles" to explain the distribution, but that came
after the fact.
Third Category. A third category would be models for which we have plausible
scientific bases, but for which we have not yet been able to show that the
models do truly make accurate predictions. For such models presumably a
verification of their predictive powers is possible, but not necessarily
easy. An example from physics of this type is the existence of "black holes."
Current theory says they may exist, while direct observation currently seems
possible but not easy.
Fourth Category. A fourth category of models are ones for which the scientific
bases is questionable but perhaps plausible and for which experimental veri-
fication is unlikely or impossible. A non-related example is the way we all
raise our children; we have poorly-founded but plausible ideas of how one
should do it, and no effective measures of their performance.
If we accept the idea that there are such categories of models, we can
enunciate the following ideas:
1. Although models in lower categories may have uses for screening purposes,
for disputed regulatory decisions one should use as high a category model as
is available.
2. There should be little effort devoted to models of a lower category if
a higher category model is available for the same task.
3. The category into which a model falls should be determined on the basis
of validation and testing, rather than any other way.
4. The validation requirements for the various categories should be different
from category to category. For the first category the requirements must be
for space and time resolved correspondence between calculated and observed
concentrations. For the second category it would be correspondence between
computed and observed space (but not time) resolved means and standard devia-
tions of the concentration distributions. For the lower two categories,
validation within the category would seem unnecessary; the objective of
validation would be to move the model to a better category.
5. EPA should be asked to classify existing EPA models into these categories
and indicate what levels of performance they believe can be reached in the next
five years for each pollutant and each averaging time.
-------
-118-
This whole set of ideas is summarized in Table 2.8.1.
Supplementary Material
A comment on the use of unverified models for making relative predictions
may be found in Sec. 3.10.10.
-------
Table 2.8.1. Categorization of Models by Validation Technique
Category
Description
Air Pollution Models which
may fall into this Category*
Proposed Validation
Technique
Time and Space Resolved
Source-Receptor Models
Comparison Between
Time & Space Resolved
Observed & Calculated
II
Statistical Property
Only Models
(Spaced Resolved)
Single Source Gaussian
Plume (e.g., CRSTER) ?
Multi-Source Gaussian?
Comparison of Statis-
tical Properties of
Space Resolved
Observed & Calculated
III
Models with Plausible
Scientific Bases, but
Currently not Validated
to Above Levels
Most CO Models,
Any Model Requiring
Empirical Calibration,
Photochemical Models
Move to Upper Category
if Validatable
IV
Models with Less Plaus-
ible Scientific Bases,
Probably not Validatable
Rollback
Not Validatable?
These choices are to some extent arbitrary and as shown some model fall in more than one category
depending on use.
-------
-120-
-------
-121-
3 SUPPLEMENTARY MATERIALS
This section contains commentary, opinions, and other written material
submitted by the participants both during and after the conference. Except
for retyping and minor editorial corrections, no changes have been made
in the content or expression of these contributions.
Sections 3.1 - 3.8 contain material germane to the working group
reports (Sections 2.1 - 2.8, respectively).
Material which is of more general interest has been placed in Section 3.9,
Comments which refer to the issues discussed in the plenary sessions of
the conference and additional issues are contained in Section 3.10.
-------
-122-
3.1 GROUP 1-1
3.1.1 Description of Texas Climatologiaal Model (TCM)
Submitted by Richard A. Porter at conference.
TEXAS CLIMATOLOGICAL MODEL
Reference: Porter, R. A. and Christiansen, J. H.; "Two Efficient
Gaussian Plume Models Developed at the Texas Air Control Board."
Proceedings of_ the 7th NATQ/CCMS International Technical Meeting
on Air Pollution Modeling, Airlie House, Va., September, 1976.
(Copy attached.)
Christiftisen, J. H. and Porter, R. A.; Users Guide to the Texas
Climatological Model. Texas Air Control Board, Austin, Tx, May, 1976.
Abstract: The TCM is a climatological model that predicts long-term
arithmetic mean concentrations of nonreactive pollutants from point
sources and area sources. The TCM is conceptually similar to the
Climatological Dispersion Model (COM) but incorporates design features
that reduce the model run time by as much as two orders of magnitude.
Equations: See references cited above.
-------
-123-
A. Source-Receptor Relationship
1. Arbitrary locationMB^ for each point source. Unlimited number
of sources.
2. Arbitrary location and square grid width for each area source.
The model will allocate area sources into a uniform square grid.
3. Receptor location is arbitrary grid (max. 50 x 50).
4. Release heights for point sources are accepted for any height.
5. The area source algorithm (Gifford-Hanna) does not consider
height of release.
6. Receptors are at ground level.
7. No terrain difference between sources and receptors.
B. Emission Rate
All sources have a single average emission rate for the averaging
time period (i.e., month, season, year).
C. Chemical Composition
One, two, or three pollutants are treated simultaneously.
D. Plume Behavior
1. Plume rise calculated according to Briggs (1971) neutral/unstable
equation.
2. Effective stack heights less than 10 meters are considered 10 meters.
3. Effective stack heights greater than 300 meters are considered 300
meters.
4. No plume rise for area sources.
5. Down-wash and fumigation not considered.
E, Horizontal Wind Field
1. Climatological approach
2. 16 wind direction
3. Mean wind speed calculated for each stability class from the joint
frequency function of stability, wind direction, and wind speed.
4. Wind speed corrected for physical stack height (same as CDM).
F. Vertical Wind Speed
Assumed equal to zero.
G. Horizontal Disperion
Assumed to be uniform within each 22.5 degree sector (same as CDM).
-------
-124-
H. Vertical Dispersion
1. Gaussian plume.
2. 6 stability classes (Pasquill-Gifford-Turner) A, B, C, D-Day,
D-Night, E+F.
3. No provision for variation in surface roughness.
*• Chemistry/Reaction Mechanism
Exponential decay according to user input halflife (same as CDM).
J. Physical Removal
Same as I above.
K. Background
Background may be entered by calibration coefficient for each
pollutant.
L. Boundary Conditions
Perfect reflection assumed at ground. Mixing height not a factor
because investigation shows no effect for typical climatology using
the CDM .47L total mixing scheme.
M. Emission and Meteorological Correlation
Emissions not varied.
N. Validation/Correlation
1. Model is self-calibrating with input of field receptor observations.
2. High correlation achieved of observed to calculated values for
Houston TSP 1975, Houston S<>2 1972, Dallas TSP 1972.
0. Output
1. Arithmetic mean concentration for the averaging time of the Clima-
tological input data and emission data (one month to one year).
2. Any combination of the following outputs are available:
a. Listing of concentration for an arbitrarily spaced square grid
of up to 50 by 50 elements.
b. A print plot of the grid concentrations.
c. Punched card output for isopleth maping (same as CDM).
d. A listing of the five high contributors to the concentration
(by Z concentration) at each grid point.
-------
-125-
P. Activity
The TCM has been widely applied for evaluation of new source impact
upon existing air quality and for evaluating the impact of growth in
urban areas. The model is in use by more than 50 industrial firms,
environmental consultants, and government (federal, state and local)
agencies throughout the U. S. and Canada. Speed of operation (up to
2 orders of magnitude faster than the CDM) and convenient output formats
have made this model popular with a wide variety of users.
-------
-126-
3.1.2 Description of Texas Episodic Model (TEM)
Submitted by Richard A. Porter at conference.
TEXAS EPISODIC MODEL (TEM)
References: Porter, R. A. and Christiansen, J. H.; "Two Efficient Gaussian
Plume Models Developed at the Texas Air Control Board." Proceedings of
the 7th NATO/CCMS International Technical Meeting on Air Pollution
Modeling, Airlie House, Va., September, 1976. (Copy attached.)
Christiansen, J. H.; Users Guide t£ the Texas Episodic Model. Texas Air
Control Board, May, 1976.
Abstract: The Texas Episodic Model (TEM) is a short-term (10 minute to 24
hour averaging time) Gaussian Plume Model for prediction of concentra-
tions of nonreactive pollutants due to up to 300 elevated point sources
and up to 200 area sources. Concentrations are calculated for 1 to 24
scenarios of meteorological conditions, averaging time, and mixing height,
Equations: See references cited above.
-------
-127-
A. St>urce-Receptor Relationship
1. Up to 300 arbitrarily located point sources.
2. Up to 200 arbitrarily located area sources.
3. A uniform square receptor grid of arbitrary spacing with up to 50
by 50 rows or columns.
4. Terrain assumed flat.
5. Unique release height for each source.
6. All receptors at ground level.
B. Emission Rate
Unique emission rate for each source.
C. Chemical Composition
One, two, or three pollutants treated simultaneously.
D. Plume Behavior
1. Plume rise according to one of six equations from Briggs selected
according to stability and distance from source. Effective stack
heights less than 10 meters are considered 10 meters. Effective
stack heights greater than 2000 meters are considered 2000 m.
2. Mixing height penetration factor (P) is a user input. If effec-
tive source height (H) is greater than P times the mixing height
the plume escapes. Otherwise the .47L mixing scheme from Turner's
Workbook is used.
3. Does not treat down-wash or fumigation.
E. Horizontal Wind Field
1. User supplied stability, wind speed, and direction for the averaging
time period (10 minutes to 3 hours) or for each 3 hour period to
build a 24-hour day.
2. Power law variation of wind speed with release height (same as COM).
3. Steady state wind for each scenario.
F. Vertical Wind Speed
Equal to zero.
G. Horizontal Dispersion
1. Semi-Empirical Gaussian Plume.
2. User supplied stability class for each scenario (Pasquill-Gifford-
Turner) .
3. Turner (1969) dispersion coefficients.
4. No adjustment for surface roughness.
-------
-128-
H. Vertical Dispersion
1. Semi-empirical Gaussian plume.
2. User supplied stability classes (Pasquill-Gifford-Turner) for
each scenario.
3. Turner (1969) dispersion coefficients.
4. No adjustment for surface roughness.
I. Chemistry /Reaction Mechanism
Exponential decay with user supplied half-life.
J. Physical Removal
Same as I above.
K. Background
May be input with calibration factor.
L. Boundary Conditions
1. Lower boundary: perfect reflection.
2. Upper boundary: reflection from top of mixed layer by the
scheme of Turner (1969) except as described in D.2 above.
M. Emission/Meteorological Correlation
User supplied values of wind speed, wind direction, stability class,
mixing height, ambient temperature for each scenario up to 24
scenarios.
N. Validation/Calibration
1. Limited validation with observed vinyl chloride observations.
2. Calibration by user supplied coefficients (A, B) so that
Xcal - A + BX predicted
0. Output
1. Concentration mean for each receptor grid point for averaging
times of:
a. 10 minutes
b. 30 minutes
c. 1 hour
d. 3 hours ^
e. 24 hours (based on eight 3-hour scenarios.)
-------
-129-
1. Output is available for from 1 to 24 scenarios in the following
formats:
a. listing.
b. print plot.
c. punched cards for isopleth maps.
d. culpability list of the high five contributors to the concen-
tration at each receptor grid point.
-------
-130-
3.3.3 Minority Report on Application of Multi-Source, Urban Model
Submitted by A. Boyer at conference.
MINORITY REPORT
The Application of Multi-Source, Urban Models
The application of complex multi-source urban models is to be primarily
the responsibility of control agencies. The results of multi-source simulations
may be used by control agencies to assess the effects of changing source
strengths, control measures, or changing land-use patterns. Multi-source sim-
ulations may also be used by control agencies for amending guidelines for the
application of single source models in areas where many single sources inter-
act.
Individual sources sould not be expected to do more than apply single
source models. In those cases where industries or members of the public choose
to apply multi-source urban models control agencies should encourage such
activity by providing urban source inventories.
-------
-131-
3.1.4
Description of the Environmental Research & Technology_, Inc. Model EETAQ
Post Conference Submission by B. Egan
Abstract: ERTAQ is a steady-state sector-averaged Gaussian plume model
that calculates concentrations of up to six pollutants from an
unlimited number of point, line, and area sources. The model may
be operated in either the "sequential" mode to calculate one-,
three-, or 24-hour concentrations for analysis of historical "worst-
case" impacts or in the "climatological average" mode to calculate
long-term averages for periods represented statistically by stability
windroses. Dispersion coefficients may be user-specified. In the
sequential mode, a fourth class of source, "tall stacks", is available
that provides for optional use of distinct dispersion coefficients
more representative of this class of source. The model may be
applied in both flat and hilly terrain. Up to 128 receptor points
may be specified, at each of which the user may specify background
concentrations as well as calibration factors. The contributions
of individual sources to selected receptors may be isolated at the
user's option. In addition, program options are available for
user-specified input format, storage of output files, and manipulation
of the results of intermediate computations.
Equations:
X = u glg2
For sources other than "tall stacks", at user's option, the crosswind
dispersion function g, may be sector-averaged over 22.5° by
= 1_
gl ~ 2x tan
For "tall stacks", the crosswind dispersion function gj is given by
the statistically "expected" value within the 22.5° sector for
receptors within the downwind sector, i.e.
1 . ,x TT/16^
SI = TT x/8 erf (o-y HT };
for receptors in the sectors adjacent to the downwind sector,
i I - er
This formulation avoids the difficulty of using centerline one-hour
values when accumulating concentration estimates for multiple-hour
averages.
-------
-132-
The vertical dispersion function g~ is given by
U>u<£f]*?,
tu. * J" •
where L = mixing depth
H = height of plume centerline above the ground-level
receptor.
Terrain Correction (tall stacks only):
H = H +4H - T~ x Min (z - z H +AH),
o 1 i o i>
wherein
H = height of plume above terrain at receptor
H = height of stack above stack base
AH = plume rise
Z = topographic height of receptor (above sea-level)
Zg = topographic height of stack base (above sea-level)
To = stability dependent, user-specified terrain correction
factor.
A. Source-Receptor Relationship
Unlimited number of point, area, line, and tall-stack sources at
any locations.
Up to 128 receptor points at any selected locations.
Unique topographic elevation for each receptor.
Receptors must be at ground level.
B. Emission Rates
Unique emission rate for each source that may be varied according
to diurnal, weekly, or monthly scheduling.
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1970) final plume-rise formulas
Stack-tip downwash (Gifford) for tall stacks
If plume height exceeds mixing height, concentrations further
downwind assumed equal to zero.
Plume and mixing depth both respond to terrain obstacles (see Equations)
-------
-733-
E. Horizontal Wind Field
Wind direction constant at all heights over all space.
Wind speed varies with height according to user-specified power-
laws dependent on stability class.
F. Vertical Wind Speed
Assumed equal to zero except as implied by terrain correction
factors.
G. Horizontal Dispersion
Gaussian plume sector-averaged in various ways depending upon
application. 5 stability classes used with user-specified dispersion
coefficients; different classes of sources may have different
coefficients. "Urban" and "rural" options.
H. Vertical Dispersion
Gaussian plume.
5 stability classes used with user-specified dispersion coefficients;
different classes of sources may have different coefficients.
"Urban" and "rural" options.
Option for initial vertical source dimension.
I. Chemistry/Reaction Mechanism
Not treated directly (see J).
J. Physical Removal
Half-life decay factors.
K. Background
May be specified for each receptor or for all receptors. May be
calculated if appropriate emissions inventory is input.
L. Boundary Conditions
Perfect reflection at the ground and at the top of the mixing
layer. Mixing height follows terrain with correction factor
(see Equations).
M. Emission and Meteorological Correlation
None specifically, but see B. Emission Rates (above).
-------
-134-
N. Validation/Calibration
Calibration option available which involves external determination
of linear calibration coefficients; slope and intercept may be
applied in subsequent runs. Comparison with observations made in a
number of studies.
0. Output
Concentration values at each receptor. Output routines search
complete data sets for high values and identify time periods of
interest.
-------
-135-
3.2 GROUP 1-2
3. 2.1 Comments on Section 2. 2. 5 Submitted by /?. Wevodau After the Conference
EN-I02S
ESTA9USHEOI802
E. I. DU PONT DE NEMOURS & COMPANY
INCORPORATED
WILMINGTON, DELAWARE 19898
ENGINEERING DEPARTMENT
LOUVIERS BUILDING
March 9, 1977
Dr. D. M. Rote
Argonne National Laboratory
9700 South Cass Avenue
Argonne, IL 60439
Dear Don:
I offer the following comments on the first round report of
Working Group 1-2:
Item 6 - "On the Question of Enumerative Vs.
Statistical Use of the Estimates of Short-Term
Concentrations" __
I agree with the consensus opinion that a statistical
approach is theoretically more valid. However, I
believe it is very important that the detailed inves-
tigation on the precise nature of the statistical
approach be completed prior to endorsement of this
approach. At this time, I favor the enumerative
approach as recommended in the draft guideline. My
main concern with a statistical approach involves
the difficulty in fitting air quality data, whether
measured or predicted, to a suitable cumulative
frequency distribution function.
In model applications to evaluate compliance with air
quality standards, I favor use of the second highest
value I agree that fumigation, stagnation, thunder-
storm" downdraft and terrain-induced mixing, none of
which are considered by CRSTER, can be important
factors in potential short-term violations. However,
substitution of the highest computed value for second
highest value is not an appropriate alternative in
these situations. This approach skirts the issue.
-------
-136-
2
March 9, 1977
Dr. D. M. Rote
In these abnormal situations alternative procedures
which do consider such factors should be employed
(or developed) instead of using the highest computed
value from a model which does not consider these
factors .
I believe the report is accurate and representative. You have
done an excellent job in coordinating and reporting the opinions
exposed during our workshop sessions.
Very truly yours,
ENGINEERING SERVICE DIVISION
R. I. Wevodau
RIW:braw
-------
-137-
3.2.2 Comments on Short-Term Analysis and on M. William's "Rationale for
Elimination of the Maximum of the Second Highest for Modeling Purposes"
C3.2.3 below).
Submitted by R. Porter after the conference.
Comment 1; Lake shore fumigation should be included in consideration of any
short-term standard. Also Briggs (1969, p 51) recommends that plumes will
not escape in a lid situation unless the calculated plume rise Ah is greater
than 2 times the mixing height. Normal fumigation and thunderstorm downwash
are too short in duration to influence a 3-hour mean.
Comment 2: The problem with the second high standard for modeling is that
we are attempting to predict concentrations too far out on the tail of the
distribution already. 1/365 is even harder to predict than 2/365. Any
probabilistic scheme should include a test of the goodness of fit of the
calculated data to the assumed frequency distribution. (See the submission
to the NATO committees) . Results in Frankfurt showed that if the data
does not fit the distribution absurd numbers are generated for the second
high. See Section 3.10.1 for a more expanded discussion of this topic by
Porter.
-------
-138-
3.2.3 Rationale for Elimination of the Maximum of Second Highest for
Modeling Purposes
Submitted by M. Williams after the Working Group 1-2 meeting.
Since the modeling used will not consider model uncertainties, the
observed second highest concentration at a point may indeed be somewhat
greater than the standard if the second highest model value is permitted to
approach the standard. It should be noted that CRSTER does not calculate
concentrations during fumigation or stagnation. It is also possible that
thunderstorm downdraft circumstances or rapid terrain induced mixing may
result in high concentration in the real world. The model does not reflect
such circumstances, thus it may not represent the actual frequency of high
concentrations. Thus, a procedure which uses model calculations for the second
highest does not assure compliance with standard. In order to provide a
greater margin of safety, i.e., to provide greater assurance that the
observed concentrations will not exceed the standard when the model
estimate does not, the quantity to be compared with the standard should be
the estimated maximum concentration.
-------
-139-
3.3.2 Description of the SAI Reactive Plume Model
NAME OF f'X)DEL: Reactive Plume Model (RPM)
Source of Model: Systems Applications, Inc.
950 Northgate Dri ve
San Rafael, California 94903
Sponsor: California Air Resources Board
Sacramento, California 95825
References: Liu, M. K., M. A. Yocke, and P. Mundkur, "Numerical Simu-
lation of Reactive Plumes," 68th Meeting of American Institute
of Chemical Engineers, Los Angeles, California, November 1975.
Liu, M. K., D. Durran, P. Mundkur, M. Yocke, and J. Ames, "The
Chemistry, Dispersion, and Transport of Air Pollutants Emitted from
Fossil Fuel Power Plants in California," Draft Final Reoort submitted
to California Air Resources Board. SAI Report ER 76-18, April, 1976.
Type of Model: Lagrangian—for either single point or areal source.
Special Feature of Model: This model is designed to estimate concentra-
tions of reactive species downwind of a single point or areal source
of pollutants. Assuming the pollutants are well mixed in the vertical,
this model is more suitable for plume fumigation and trapping
conditions.
Status of Model Development: Operational
CHARACTERISTICS OF THE MODEL FORMULATION
Model Equation: Mass conservation
dC_.
dt ^ 'chem
Plume Rise: Input to model.
Turbulent Dispersion: This model contains two options: either the
measured plume width and plume depth (as a function o, downwind
distance) of the .classical Pasquill-Gifford methods (Turner, 19o9)
may be used. Plume dispersion is determined either by the clas-
sical Pasquill-Gifford method (Turner, 1969) or from the observed
plume width and plume depth as a function of downwind distance.
Provision has been made for entrapment of background pollutants.
-------
11-22
-140-
Wind Shear: Wind velocity at plume height must be used. Although
a simple correction according to a power law can be easily
incorporated, no treatment of wind shear is currently in the model.
Terrain Interaction: No current treatment but simple consideration
(similar to the Valley model) can easily be incorporated.
Chemistry: This model is written in a modular form which can accept
any chemical kinetics submodel with a maximum of 50 reaction steps.
The kinetic mechanism that is currently in this model is a mod-
ified version of the Hecht-Seinfeld-Dodge mechanism (1974) for
hydrocarbon-NO-SO? system.
^\ f— ' •
Spatial Scale: Medium scale (~tens of kilometers)
Temporal Average: Short and Medium (hourly averages).
DATA BASE REQUIRED
Source: Stack location
Either stack emission rates or initial pollutant concentrations
within the plume.
Meteorology: Wind speeds
Stability class (or plume width and plums depth)
Radiation intensity
Other: Kinetic mechanism
Ambient pollutant concentrations
CHARACTERISTICS OF THE MODEL COMPUTATION
Computer Language: FORTRAN IV
Computing Time: 10-50 PCU seconds (CDC 7600) for a typical run
Storage Requirement: -145,000 actual large core.
IBM 370/168 Compatibility: Yes
KNOWN MODEL APPLICATION: This model was applied to the following seven
point sources (Liu et al., 1975, 1976; Tesche et a!., 1976):
Moss Landing Power Plant, Monterey, California
Los Alamitos Power Plant, Los Angeles, California
Haynes. Power Plant, Los Angeles, California
Mobile Oil Refinery, Los Angeles, California
Four Corners Power Plant, Farmington, New Mexico
Hobbs Power Plant, Hobbs, New Mexico
Jefferson Power Plant, Jefferson, Texas
-------
-141-
S.2.2 Description of the DEPICT Model
Reference; l) Sklarew, R.C., and J.C. Wilson,
"Applications of DEPICT to the Garfield, Navajo, and
Ormond Beach Air Quality Data Bases"
Science Applications, Inc. Report prepared for
Southern California Edison, July 19T6
2) Sklarew, R.C., Wilson, J.C., and Frabrick, A.,
"Evaluation of Air Quality Models Point Source Models"
Science Applications, Inc., July 1976, under contract
to the California Air Resources Board, Sacramento, CA.
Abstract; The DEPICT (Detailed Examination of Plume JBnpact in Complex
Terrain) is a 3-dimemsional eulerian numerical point source
model. The model calculates the temporal and spatial
concentrations of inert or reactive pollutants in flat or
complex "terra in. The model is modular in design and has the
ability to update algorithms in an efficient manner. The
model currently uses either the Eschenroeder l6-step or the
Hecht-Selnfeld 39-step chemical mechanisms for the reactive
pollutants. The model is applicable to assess air quality
impact for point sources located in rural environments.
Equations; Conservation of Specie Equation
A. Source-Receptor Relationship
Point sources only; (10 maximum) uniform grid squares, user defined grid.
Sources can be treated as ground level or elevated releases.
Receptors are located at center of grids.
Receptor locations are 3-dimensional.
B. Emission Rate
User specified emission rate for each pollutant for each point source.
Emission rates can vary hourly for each source.
-------
-142-
C. Chemical Composition
D. Plume Behavior
Plune rise calculation is based on the work of Briggs with simple
modifications to obtain estimates of inversion penetration.
Four cases are considered:
1} vholly unstable atmosphere
2) deep ground base inversion
3) elevated stable layer
if) shallow ground base inversion.
E. Wind FLov Field
The model calculates a three dimensional wind field using the f olloving
equations.
~ o
< di
2) U-
3) *^
Where {> is the perturbation velocity potential, and^are tramstaission coefficients
based on temperature profile or stability classes (1-7) Wind observations are
projected upward frcm the point measured through the portion of the grid
without any measurements based on the power law £/- ^>(.^»|
The initial horizontal wind components are calculated using a
interpolation schsoe of the vertical profiles for each layer.
-------
-143-
F. Vertical Dispersion
The vertical diffusivlty parameters are calculated based on the
algorithms of Smith and Howard.
where U = wind speed at point of interest
= standard deviation of the wind vane fluctuation and is
dependent on stability class.
= turbulence scale length (in meters) and depends on height
above ground and stability.
G. Horizontal Dispersion
The horizontal values of diffusivity are calculated using the relationship
K^ocK*
where o depends on stability class.
H. Chemistry and Reaction. Mechanism
The node! has the option of using either the Eschenroeder 16-step or the
Hecht-Seinfeld 39-st®P mechanism.
• Physical Removal
The l6-step mechanism includes a reaction between NOg and participates. The
Hecht-Seinfeld has no physical removal process.
J. Background
Treated as an hourly input for the chemical mechanisms.
K. Boundary Conditions
Lower boundary (surface of earth) perfect reflection.
Upper boundary - see pluae rise,
-------
-144-
L. Emission and Meteorological Correlations
User supplies hourly values of vind speeds measurements (surface and aloft),
nixing height measurements, stability fields, and emissions.
M. Validation/Calibration
Preliminary validation of the DEPICT Model has been completed for the following
areas:
1) Garfield-aaelter
2) Navajo Generating Station (SO )
3) Ormond Beach Generating Station
(SF6, NO, B02, 03)
N. Output
The model predicts the hourly temporal and spatial concentrations for each
grid for inert or reactive pollutants.
-------
-245-
3.3.3 Description of the SAI Urban Airshed Model
Post Conference Submission by Philip M. Roth
References: (a)
(b)
Roth, P. M., S. D. Reynolds, P. J. W. Roberts, and
J. H. Seinfeld (1971), "Development cf a Simulation
Model for Estimating Ground Level Concentrations of
Photochemical Pollutants," Final Report and 6 Appendices,
APTD 0908-0914, Systems Applications, Inc., San Rafael, CA.
Reynolds, S. D., M. K. Liu, T. A. Hecht, P. M. Roth,
and J. H. Seinfeld (1973), "Urban Airshed Photochemical
Simulation Model Study-Volumes I-III, EPA-R4-73-030 a-h.
Systems Applications, Inc., San Rafael, CA.
(c) Reynolds, S. D. et al. (1976, 1977), "Continued Research
in Mesoscale Air Pollution Simulation Modeling," Volumes
I-IV (EPA 600/4-76-016 a-d) and Volumes V-VII (in review
by sponsor), Systems Applications, Inc., San Rafael, CA.
Abstract:
The SAI Urban Airshed Model is a fully three dimensional
grid-based model capable of predicting the spatial and
temporal distribution of both inert and chemically
reactive pollutants. Basic inputs to the model include
the specific meteorological, emissions, and chemical
characteristics of the region of interest. Predictions
for up to 13 pollutants may be obtained, including CO,
SO?, 03, N02, NO, four hydrocarbon classes, total aerosol,
PAN, HN02, and H202. A 42 step kinetic mechanism is em-
ployed to represent the" pertinent chemical phenomena.
The atmospheric diffusion equation is solved numerically
on the three dimensional grid to predict the dynamic
changes in pollutant concentration levels over a period
of up to a few days. The model is applicable to the
examination of regional air pollution problems, such as
the evaluation of alternative emission control straegies.
Equations:
3C.
S,
-------
-146-
where:
c. = concentration of specie i
ZT,7 = horizontal components of the wind
w = vertical component of the wind
KH = horizontal turbulent diffusivity
KV = vertical turbulent diffusivity
R. = rate of formation of specie i by chemical reactions
S. = rate of emission of specie i
A. Source-Receptor Relationship
1) Emissions from line and area sources are apportioned to
each grid cell and are assumed to be emitted at ground level.
2) Each point source is treated separately. The stack height
and plume rise determine the cell in which the emissions
enter the grid.
3) Ambient concentrations are calculated for each grid cell.
Each concentration represents a spatial average over the
volume of a grid cell.
B. Emission Rates
Emissions are calculated external to the main program using
EPA or other appropriate emission factors. All emission
rates can vary in time.
C. Chemical Composition
Both inert and reactive species are considered. The thirteen
pollutants for which predictions are made include: CO, SOU, NO,
NOa, 03 four hydrocarbon classes, PAN, H202, HN02, and total
aerosol nitrate, sulfate, and organic).
D. Plume Behavior
A point source plume enters the modeling gri^l at a height
calculated from the actual stack height plus the plume rise
given by the formula of Briggs. Consideration is given to the
determination of whether the plume penetrates an elevated
inversion layer, and, if so, whether it breaks through the layer.
-------
-147-
E. Horizontal Wind Field
Horizontal wind components are calculated external to the main
program using objective analysis techniques in conjunction with
avaliable data taken at the surface and aloft. The wind field
is fully three-dimensional, thus allowing for the treatment
of wind shear effects. Temporal variations are also considered.
F. Vertical Wind
The vertical wind component in each grid cell is calculated
using the continuity equation and the horizontal wind component
inputs.
G. Horizontal Dispersion
The horizontal turbulent diffusivity (KH) is assumed to
be a constant.
H. Vertical Dispersion
The vertical turbulent diffusivity (Kv) varies in space and
time and depends on the height above the ground, wind speed,
surface roughness, and the atmospheric stability class.
Algorithms developed by Liu et al., and Lamb et a!., are
included in the model.
I. Chemistry/Reaction Mechanism
A generalized lumped mechanism consisting of 32 reaction steps
developed by Whitten et al., is employed to describe the
chemical interaction of organics, NOX and 03. Organics are
segmented into four groups determined by the following carbon
bond characteristics: single bonds, relatively reactive (fast)
double bonds, slow double bonds, and carbonyl bonds. Six
reaction steps are included to treat the oxidation of S02- The
formation of nitrate, sulfate, and organic aerosol products is
parameterized by four reaction steps.
J. Physical Removal
Surface deposition of species is treated in each ground-level
grid cell. The rate of deposition depends on the type of
vegetation or ground surface in that cell.
-------
-148-
K. Background Pollutants
The influence of background pollutant concentrations is treated
in the initial and boundary conditions of the governing equa-
tions.
L. Boundary Conditions
The pollutant flux for each species must be specified at
all points on the boundary (both horizontal and vertical) where
the wind is blowing into the modeling region. The boundary
condition at the ground incorporates the influence of ground-
level emissions as well as surface removal processes.
M. Emission and Meteorological Correlation
Emissions and meteorological inputs to the model are compiled
in special data preparation programs. These programs are
tailored to each application of the model in order to provide
an effective interface between the existing observational data
and the emissions and meteorological input file needs of the
SAI model.
N. Validation/Cali brati on
Evaluation studies using an early version of the SAI Model
developed in 1973 were carried out for Los Angeles, Las Vegas,
and Denver for CO, NO, N02, 03 and hydrocarbons. The latest
version has been applied to Denver and is currently being
adapted to St. Louis, Los Angeles, and Sacramento.
0. Output
The model produces gridded maps illustrating the spatial
distribution of one-hour-average pollutant concentrations
over the entire region of interest. Vertical concentration.
profiles and predictions at air monitoring stations or other
user-selected sites may also be displayed. The model output
may also be interfaced with contour plotting routines to
generate diagrams of concentration isopleths.
-------
-149-
3.3.4 Description of the Environmental Research & Technology, Inc. Model ARTSIM 2.0
Post Conference Submission by B. Egan
Reference: "Lagrangian Photochemical/Diffusion Model", i-nvironmental
Research £ Technology, Inc. (In preparation), May 1977.
Abstract: ARTSIM is a trajectory-oriented model intended for regional
application. It simulates chemistry and diffusion in a moving
polluted air mass. The chemical model contains 54 reactions and
explicitly treats four hydrocarbon classes, namely, alkenes, alkanes,
aromatics and aldehydes as well as photochemical oxidants, SCL, and
sulfate. The model computes pollutant concentration as a function
of height and time. It contains 3-modules: (1) trajectory generation;
(2) source emissions; (3) chemical/meteorological.
Equations:
|f= |f UCU.t) H ) * »{e)
wnere c = vector of pollutant concentrations
K(z,t) = diffusion coefficient which varies with height
and time
R(c) = vector of chemical reaction rates
A. Source-Receptor Relationship
Area and point sources are used. Elevated point sources may be used.
Generalized input structure allows use of source emissions data at
various levels of resolution. Source-oriented and receptor-oriented
trajectories may be specified.
B. Emission Rate
Rates must be specified for each primary pollutant. Hourly rates
for traffic and stationary sources required.
-------
ARTSIM - page 2 -150-
C. Chemical Composition
54-reaction chemical model including NO, NCL, 0_, S0~, sulfate, alkenes,
alkanes, aromatic hydrocarbons, and aldehydes.
D. Plume Behavior
N/A
E. Horizontal Wind Field
Hourly u- and v-components must be input for trajectory calculations.
F. Vertical Wind Speed
Not included.
G. Horizontal Dispersion
Assumed negligible
H. Vertical Dispersion
Diffusivities are specified at up to 10 vertical levels with arbitrary
time resolution. Any atmospheric stability class can be used for
various times of day.
J. Chemistry - Reaction Mechanism
See Section C
J. Physical Removal
Physical removal is simulated by a variety of ground boundary conditions
and/or chemical reactions in kinetic model.
K. Background
Background concentrations can be specified as initial or boundary
conditions.
-------
ARTSIM - page 3 -151-
L. Boundary Conditions
a) Source boundary condition at ground:
3c.
K(o,t) —;JY = - 0. (t) , i denotes ith species, 0. (t) = source flux
b) Constant-concentration boundary condition:
c = constant
i
c) Absorption at the ground:
K(o,t) | = vd C? ,
Vj = deposition velocity
n = power of c^
d) Top boundary condition (impermeable boundary)
—^- = 0 , for all i.
M. Emission and Meteorological Correlation
N/A
N. Validation - Calibration
a) A. Q. Eschenroeder, J. R. Martinez, and R. A. Nordsieck,
"Evaluation of a Diffusion Model for Photochemical Smog
Simulation", General Research Corporation, CR-1-273, Oct. 1972
b) AeroVironment, Inc., "Las Vegas Valley Air Quality Study",
April 1976
c) AeroVironment, Inc., "Truckee Meadows Basin Air Quality Study",
April 1976.
-------
- ARTSIM - page 4
N. d) Stanford Research Institute, "Present and Prospective San Francisco
Bay Area Air Quality", December 1974.
0. Output
Species concentrations as functions of time and height above ground.
Time resolution is arbitrary down to 1 minute. Variable vertical mesh
spacing may be used.
-------
-153-
3.3.5 Description of the Environmental Research & Technology3 Inc. Model LAPS
Post conference submission by B. 3gan
Reference: R. A. Nordsieck. "A Local Air Pollution Simulator (LAPS);
Volume I, User's Guide," Environmental Research £ Technology,
2030 Alameda Padre Serra, Santa Barbara, California 93103,
1977 (In preparation),
Abstract: LAPS employs numerical techniques to calculate concentration
fields downwind of single or multiple concentrated sources. These
sources may be individual point sources at various heights, an area
source strip on the ground at arbitrary orientation with respect to
the wind, or combination of the two. The model uses the Lagrangian
or trajectory approach with lateral dispersion. Steady or unsteady
conditions may be modeled and the averaging times associated with
the calculated concentrations are related to the averaging periods
of the input emissions and meteorology. Vertical mixing conditions
are simulated using time- and space-varying eddy diffusivities. Up
to seven pollutants may be modeled simultaneously with optional
equilibrium chemical coupling between the species.
Equations/.
The basic transport equation modeled by LAPS is:
8c 8c
3lf + U^x
where
c = pollutant concentration
u = wind speed in x-direction
D (z) = lateral eddy diffusivity at z
y
D (z) = vertical eddy diffusivity at z
z
Q(x,y,z,t)= source-sink term to model pollutant emission fluxes
and simple chemical reactions
-------
-154-
LAPS - Page 2
In the Lagrangian formulation, the coordinate system of the air
parcel is oriented with x in the direction of the wind and moves at the
wind speed u. Thus, the wind speed relative to the moving coordinate
3c
system is zero and the term u— is removed from the equation. LAPS
oX
solves the reduced equation by a combination of finite difference tech-
niques in the z and t dimensions and multiple superposition of an analytical
solution for Gaussian spreading from a finite-width source in the y and t
directions.
A. Source-Receptor Relationship
Up to 10 point sources at arbitrary locations.
Unique stack height and efflux parameters for each stack.
Separate uniform area sources allowed over entire region and/or within
a finite-width strip located and oriented by user input.
Specific receptor locations not currently calculated, but a ground
concentration map is provided and vertical profiles arc
available at user specified time intervals.
B. Emission Rate
Unique average emission rates of all pollutants for each point and
area source.
C. Chemical Composition
Up to seven chemical species can be specified by name and molecular
weight.
D. Plume Behavior
Br'iggs (1971) final plume rise formula for neutral or unstable
conditions with u > 3.1 mph.
Downwash not treated.
Plume rise limited to inversion height unless stack penetrates base
of inversion layer.
-------
LAPS - page 3 ~155-
E. Horizontal Wind Field
Wind speed is uniform over y and z, but may be varied with time.
Vertical mixing induced by wind shear can be modeled through adjust-
ment of the vertical eddy diffusivities.
F, Vertical Wind Speed
Neglected.
G. Horizontal Dispersion
Lateral dispersion is treated using an analytic solution for one-
dimensional diffusion from a finite-width source. Each cell
in a row at each vertical station is treated as an isolated
source which diffuses laterally for one time step. The results
of these calculations are superposed to give the complete solution
in each row of cells.
The lateral eddy diffusivities can be varied with height and time.
H. Vertical Dispersion
Vertical dispersion is modeled using a Crank-Nicolson finite difference
formulation for vertical diffusion with variable eddy diffusivities.
The vertical eddy diffusivities can be varied with height and time.
I. Chemistry - Reaction Mechanism
Equilibrium chemical coupling of NO, N02> and 03 is optional.
First order conversion of S02 to sulfate is optional.
Governing rate constants and conversion rates are input parameters.
J. Physical Removal^
Fallout not modeled.
Rainout not modeled.
-------
-156-
LAPS - page 4
J. (cont'd)
Impaction of particulates or complete chemical uptake at the ground
is optional for any species.
K. Background
Vertical profiles of background concentrations are specified by the
user for each species.
L. Boundary Conditions
Optional time varying surface fluxes of each species, independently
specified for a source step and for the remaining area.
Constant ground concentration may be specified for any species to
enable calculation of surface uptake.
Reflection coefficient at sidewalls may be set by the user.
M. Emission and Meteorological Correlation
N/A
N. Validation - Calibration
Thus far, validations have consisted of comparisons with various
analytical solutions; for example, for steady-state Gaussian
plumes and step changes in flux or concentration boundary con-
ditions. In each case, it has been possible to achieve accuracy
in the 10% and under range, which is certainly within the accuracy
of real-world input data. No comparisons with measured data
have been attempted as yet, owing partially to the scarcity of
good measured data, either roadside CO profiles or point source
plume concentrations collected with simultaneous car counts or
pollutant emissions and meteorological data.
0. Output
Vertical concentration maps giving concentrations at mesh points in
an array of up to 10 vertical stations and 20 horizontal stations
at user-specified time intervals. (Optional)
-------
LAPS - page 5 -357-
0. (cont'd)
Ground concentration naps showing ground concentrations at up to
20 stations normal to the wind direction at user-specified time
intervals along the air trajectory. (Optional)
Ground concentration contours - A printer-plotted symbol nap derived
from the array of ground concentrations delineates up to 10 user-
specified contour levels. The user nay select which species are
to be plotted and specify different contour levels for each
selected species.
Concentration vs. distance plot - This option produces a printer-plot
of ground concentrations vs. distance along the trajectory for
up to five user-selected species.
Ground concentration crossplot - This optional output gives ground
concentration profiles normal to a roadway (i.e. a source strip)
at user-specified intervals along the road.
-------
-158-
-------
-159-
3.4 GROUP II-1
3.4.1 Description of the Environmental Research & Technology3 Inc. Model EGAMA
Post conference submission by B. Egan
Reference: Egan, B.A. and J.R. Mahoney, 1972a: "Numerical modeling
of advection and diffusion of urban area-source pollutants."
J. Appl. Meteor., 11. 312-322.
Egan, B.A. and J.R. Mahoney, 1972: "Applications of a
numerical air pollution transport model to dispersion
in the atmospheric boundary layer." J. App1. Meteor., 11.
1023-1039.
Abstract: The Egan-Mahoney advection-diffusion model (EGAMA) simulates
the dispersion mechanisms of grid-cell emissions using a numerical
solution to the basic tracer equation. The model utilizes moments
of the concentration distribution within each grid cell in a computation
scheme designed to virtually eliminate numerical, pseudo-diffusion
effects. The capability of this model to treat spatial and temporal
variations in the wind and diffusivity profiles allows for specialized
adaptations for a wide range of modeling, applications, e.g., near-
field dispersion of highway sources, fumigation episodes due to
sources located near land-water interfaces, and long range transport
problems.
Equations: The basic mass conservation equation for a pollutant species
in a planer non-divergent flow field may be written:
3c_ „ 3c vi£ + JL(K-^)+Q-R
aT - " u ax ' v 9y sz l 9z } *
where C = pollutant concentration
U = component wind speed in mean (x) direction
V = cross wind speed component
K = vertical eddy diffusivity
Q = emission rate
R = Removal or production rate (specific form
depends upon mechanism)
A. Source-Receptor Relationship
Emission rates may be assigned to each grid cell. Average concentrations
within each model grid cell are computed by step-wise integration
for evenly spaced time intervals. Two or three dimensional grid
systems may be specified.
B. Emission Rate
Time and spatially variable emissions within each grid cell can
vary with time and represent average values over the geographical
space encompassed by a single grid cell.
-------
-160-
C. Chemical Composition
Transformation and decay as well as surface deposition of the
contaminant species under examination may be incorporated. These
processes are of primary concern for long range transport applications.
D. Plume Behavior
No plume rise formula per se is incorporated by the model. Emissions
are assumed well mixed in the vertical within each grid cell.
Preliminary plume rise calculations (e.g., Briggs formulae) can be
performed to establish the vertical location of each emission
source.
E. Horizontal Wind Field
Steady-state or spatially/temporally varying ambient winds may be
specified throughout the grid system. The speed at any point is
considered constant during a given time step. The effects of
obstacles on altering the flow field may be simulated.
F. Vertical Wind Speed
Normally assigned zero initial value. Vertical wind speeds resulting
from flow over obstacles or downwash are internally generated (for
example return flow circulation in depressed highway sections).
G. Horizontal Dispersion
Turbulent diffusivities can be specified to simulate horizontal
dispersion. In applications with large grid cells, horizontal
dispersion is often neglected compared to the transport terms.
H. Vertical Dispersion
The vertical diffusion component is simulated by a conventional
forward-time, center-difference technique modified so that variable
grid spacing can be specified in the vertical. In regions where
parameters or concentrations vary rapidly with height, resolution
and accuracy can be improved by smaller vertical grid spacing.
I. Chemistry/Reaction Mechanisms
Program presently allows incorporation of two specie reactive
chemistry. Modifications are underway to expand capability to
multiple species.
J. Physical Removal
Surface deposition of the contaminant species modeled is simulated
at the lower boundary of the grid system. Other removal processes,
including chemical transformation and decay can also be incorporated.
-------
-262-
K. Background Concentrations
Can be specified as initial conditions throughout the grid system
and in the form of boundary values which are advected into the
computational region.
L. Boundary Conditions
Winds, diffusivities, and pollutant fluxes at top and bottom of
grid system may be specified.
M. Emission and Meteorological Correlation
Could be specified.
N. Va1idation/Ca1ibration
Extensive validation/calibration study performed for highway
applications (2D-version) limited validation of 3D version performed
in long range SOX modeling study.
0. Output
Concentration fields at all grid locations at specified time intervals
or time averaged.
-------
-162-
-------
-163-
3.5 NO SUPPLEMENTARY MATERIALS
-------
-164-
-------
-165-
3. 6 GROUP II-3
3.6.1. .Description of TAPAS Model
Submitted by D. G. Fox at the conference.
TAPAS
Reference: (a) Fosberg, M.A. and D.G. Fox. "A Topographic Air Pollution
Analysis System" to be submitted to Atmospheric Environment,
(b) Fosberg, M.A. and D.G. Fox. "An Air Quality Index to Aid
in Determining Mountain Land Use". In Proceedings of the
Fourth National Conference on Fire and Forest Meteorology,
Nov., 1976, USDA For. Serv. Gen. Tech. Rep. RM-32, p. 167-
170, 1977.
(c) Fosberg, M.A., D.G. Fox, E.A. Howard and J.D. Cohen.
"Nonturbulent Dispersion Processed in Complex Terrain",
Atmos. Env. 10, p. 1053-1055, 1976.
(d) Fosberg, M.A., W.E. Marlatt and L. Krupnak. "Estimating
Airflow Patterns Over Complex Terrain, USDA Forest Service
Research Paper No. 162,'16 p., 1976.
(e) D.G. Fox, G. Wooldridge, and others. "An Experimental
Study of Mountain Meteorology". In Proceedings of the
Third Symposium on Atmospheric Turbulence Diffusion and
Air Quality. Amer. Met. Soc., Raleigh, NC, 1976
Abstract:
TAPAS combines a simulation of the wind field over mountain-
ous terrain with a Gaussian derived diffusion model. The
diffusion model is employed in each grid cell of the calculation
in order to provide an estimate of the mixing conditions
within these cells. These conditions are combined with
the Pollutant Standards Index such that a maximum allowable
emission is calculated. These in turn represent an
atmospheric constraint for planners to work with.
Equations: U)
Wind Model
(a) Cressman objective analysis
(b) Potential flow over topography
(c) Influences of surface temperature and roughness.
See reference (d)
(2) Dispersion Model
-------
or
u -
o
-166-
tv\eav\ diWaev^ i* £&cU cell
0
V
= ^- * |0
Q
b T
1 v\uw o
f
Ql - £MAiiiJoiA v-afc cf life- MSiCLuH ^o /• cie-lcr iu i m u
-------
-167-
TAPAS
A. Source-Receptor Relationship
(a) Sources are evaluated on the basis of their instantaneous
emission rate (rag/sec) for/fvalues and in terms of their
total emission (mg) over the standard time period for the
"Upvalues.
(b) There is no distinction made between point, line and area
sources.
(c) There are no specific receptors, Analysis is for ground
level concentrations.
B. Emission Rates
(a) Calculated from EPA emission factors external to the model.
(b) Model provides the total allowable emission within each grid
cell (ranging from .25 km2 to 9 km2) to achieve a preselected
level of air quality.
C. Chemical Composition
Only non-reactive pollutants are treated.
D. Plume Behavior
No explicit treatment of plume behavior.
E. Horizontal Wind Field
The wind component calculates an overall driving wind by
Cressman objective analysis. This is then altered by
topography within the restrictions of potential flow. The
potential flow is corrected by surface temperature and
surface roughness considerations.
F. Vertical Wind
The calculated wind is constrained to follow the terrain at
the surface. The rate of change of vertical velocity
is explicitly calculated as the divergence.
-------
-168-
TAPAS
G. Horizontal Dispersion
A Gaussian formulation is altered to include the effects of
mass divergence ( see ref. c) on horizontal dispersion.
This effect is coupled with the values of
-------
-169-
TAPAS
0. Output
A matrix of the allowable emissions within each grid cell is
output for each individual ambient air quality standard, i.e.
1 hr.CO, 8 hr. CO, 24 hr. TSP, or for any other preselected
level of air quality, i.e. for 24 hr. TSP ^= 100,
75 /e/m3 or <^ = 8,7* =75 /Jg/m3 (class I)
= 17,^= TS^g/m3 (class II)..
The model is able to accomodate selection of different values
of V at different grid points so that class I, II and III
areas can be analysed together.
-------
-170-
3.6.2 Validation Data on the VALLEU Model
Submitted by Herschel H. Slater, post-conference
Valley Model: Comparisons of Observed and Estimated
Concentrations and Related Observations
The Valley Model was initially used to estimate the impact of
emissions from single sources on elevated terrain. There are few data
available to evaluate this or any other model or analytical routine in
rough terrain situations, because: (1) it is difficult to locate and
operate monitoring equipment in complex terrain; (2) the representa-
tiveness of meteorological data is often uncertain; and (3) some short-
term standards are addressed to rare events. Sensing the rare event may
require a very highly reliable, continuously operating monitoring
program.
The dramatic example of the latter is provided by data collected by
the Kennecott Copper Corporation at two of their sampling sites on
January 20, 1976. The bearings from the main stack to the monitors were
within 5° of each other (about 250° True). One monitor was about 2.7
miles (4.5 km) from and 300 feet (100 m) above the stack top. It meas-
ured a 24-hour S02 concentration of 2.71 ppm. The second monitor was
about 3.0 miles (5 km) from and 1100 feet (350 m) above the stack top.
It measured 0.02 ppm for the same 24-hour period. The maximum concen-
trations at monitors with elevations near or below the base of the
stacks were at most 10-22% of the highest concentrations measured on the
hillside. (It is probable the concentrations at the monitors near the
elevations of the stack bases were caused by sources other than the
pollutants emitted from the stacks.) It is quite apparent that the
effluents from the stacks were contained in or below an inversion and
lay along the hillside above the base and below the crest.
-------
-171-
Comparisons of estimated and observed 24-hour sulfur dioxide con-
centrations measured at sites located at elevations greater than the
stack top of a nearby source are shown in Table A-l. In five cases, the
emission rate was well documented. To make the concentration estimates
it was assumed that stable conditions (F) and light wind speeds (2.5
mps) described the dispersion conditions.
Quantitative comparisons have been made which digress from the
original purposes and applications of the Valley Model; e.g., compari-
sons made for 1-hour averaging times. Also, observations exist which
can be only qualitatively related to the plume impingement concept of
the Valley Model. Some examples which may provide some useful insights
follow.
Lantz, Hoffnagle and Pahwa^5' developed 1-hour estimates from the
Valley Model and compared estimated and observed concentrations from the
Navajo Generating Plant Study. Using 4 sets of meteorological inputs
they developed 12 data-pairs. The ratio of estimated to observed
maximum 1-hour concentrations ranged from 0.7 to 2.4.
Slowik and Pica^ used the Valley Model to estimate concentrations
at a sampling site on Laurel Ridge, Pa., near the Conemaugh Generating
Station. Comparisons were made with one year of 1-hour sulfur dioxide
concentrations. The data were screened on the basis of wind directions
collected at a wind sensing site on Chestnut Ridge, 14 km from the
monitoring site. The study presumes that the wind observation on
Chestnut Ridge defines the plume trajectory in all cases. If this
-------
-172-
TABLE A-l Comparisons Between Estimated Maximum and Second
Highest and Observed 24-Hour S02 Concentrations.
Estimated Concentrations Are Based On Valley Model
Assuming F Stability and 2.5 mps Wind Speed.
Location
Crusher* '
Lower /9x
Lake UJ
S1te,,x
106 w
S1tem
107 IJ;
C-H111°'4)
Jones /i x
Ranchu'
Source
Garfleld
Smelter
Garfleld
Smelter
Navajo
P. P.
Navajo
P. P.
Anaconda
Smelter
Morencl
Smelter
Miami
Smelter
Period
4/15/73-
1/31/74
2/01/74-
1/31/75
3/08-
12/16/75
1/1-25/76
10/1/74-
2/17/75
10/1/74-
2/17/75
24-Hour Concentration^9'
Estimated
2480
2480
1.18 ppm
1.18ppm(b)
36
25
Observed
Max
2564
6130
2.66 ppm
2"T1 MMMt\CJ
.71 ppm
32
30
2nd
High
2473
3130
1.20 ppm
2.14 ppm(c)
19
15
Ratio
Est/Obs
Max
1.0
0.4
0.4
0.4
1.1
0.8
2nd
High
1.0
0.8
1.0
0.6
1.9
1.7
Although some data have been acquired at C-Hin near the
Anaconda smelter, unresolved uncertainties with the data
make It Inadvisable to apply them for validation purposes
1975
1974
1975
15490(b)
8610(b)
8610(b)
2547
2042
2642
2416
1760
1548
6.1
4.2
3.3
6.4
4.9
5.6
(a) Mlcrograms per cubic meter except where otherwise Indicated.
(b) Emission rates not well-documented.
(c) Reliability of air quality data not ascertained.
4-77
-------
-173-
very tenuous assumption is accepted, then the ratio of estimated to
observed 1-hour concentrations was 40 for the maximum hourly, 65 for the
second highest and 30 for the 90th percent!le concentration.
Keen observers of plume configuration have long noted that plumes
from the stacks of sources located in hilly terrain usually have greatest
impact on higher terrain. Studies of the plume travel of the Clifty
Creek power plant and observations near the Widow's Creek facility
confirm this observation in a qualitative sense. (The Navajo study
provides quantitative confirmation.) Observers do not agree that a
particular plume configuration such as high-wind fumigation, inversion
break-up fumigation or impingement is associated with greatest ground
level concentrations. The critical configuration is influenced by the
characteristics of the source, its site, the frequency of weather events,
the location and orientation of nearby terrain features, and the operating
schedule of the facility.
Scorer^ ' has described several circumstances which support quali-
tatively the assumptions on which the Valley Model is based; namely, in
hilly terrain the highest 24-hour concentrations caused by pollutants
emitted from a stack whose top is lower than the elevation of nearby
terrain features occur under light wind stable conditions. Scorer
describes plume characteristics of a number of sources. He states that
in Bohemia (p. 36): "The gases from large power stations drift with
very little dilution above the height at which they were emitted, but
below the mountain tops, until they impinge on the hillsides and damage
the pine trees " He describes a situation in New Zealand where,
-------
-174-
due to steeply sloping nearby mountains, air pollutants are trapped in
calm weather and often impinge on the hillsides. Further, on page 46,
he cites the coastal area of Lebanon where "...the plumes from the
factory chimneys at the coast impinge on the rich fruit-growing fields
which slope up steeply inland..."
fo\
The classic Trail, B.C. study of Hewson and Gillv ' suggests that,
on occasion, higher concentrations occurred on the slopes of the Columbia
River Valley, above the valley floor but below the crests of the sur-
rounding hills, than elsewhere. •
fq)
An oft cited tracer study by Start, Dickson, and Wendellv was
conducted in Huntington Canyon, Arizona. Based upon the single line-of-
best-fit through observed vs. estimated concentrations from the four 1-
hour valid tracer tests conducted during stable conditions, within a 48-
hour period, the authors conclude that the dilution in the Canyon was 15
times that to be expected over flat terrain. Estimates were made by the
bivariate Gaussian formulation, using the flat-terrain Pasquill-Gifford
diffusion parameters corresponding to the stability observed in the
Canyon. However, when the same data are expressed as the ratio of the
estimated concentration to the study-period maximum observed concen-
tration at each of the four sampling arcs (at distances about 2.2, 2.8,
4.2, and 6.2 km), we obtain values of 0.95, 1.1, 2.6, and 1.0, respectively.
Relative to many modeling results, this is an excellent relationship.
It is interesting to note that each maximum occurred at an end sampler
of the respective arc. How adequately this set of data represents the
dispersion conditions when the highest concentrations of the year occur
has not been resolved.
-------
-175-
REFERENCES FOR SECTION 3.6.2.
1. Office of Air Quality Planning and Standards (AEROS data system),
U.S. Environmental Protection Agency, Research Triangle Park, N.C.
2a. Eeaney, R.J.: Statement on Behalf of Kennecott Copper Corporation,
December 16, 1975.
-b. Kennecott Copper Corporation: Air Pollution Emergency Episode,
January 21 through 24, 1976. KCC Technical Report (undated).
3. Rockwell International;Meteorology Research, Inc.; and Systems
Applications, Inc.: Navajo Generating Station Sulfur Dioxide
Field Monitoring Program. Vol. IV: Ground-Level S02 Measurements.
September 1975.
4. Anaconda Copper Company Observational data.
5. Lantz, R.B., G.F. Hoffnagle, and S.B. Pahwa: Diffusion Model
Comparisons to Measured Data in Complex Terrain. Third Symposium
On Atmospheric Turbulence, Diffusion, and Air Quality. AMS, Raleigh,
N.C., p. 476, October 1976.
6. Slowik, A.A. and G. N. Pica: A field comparison of the EPA "Valley"
Model, "Half-Height" Model and a suggested new model in complex
terrain under stable atmospheric conditions. Interim Report
No. 1, Pennsylvania Electric Co., August 31, 1976.
7. Scorer, R.S.: Pollution in the Air - Problems. Policies and
Priorities. Rutledge & Kegan Paul, London and Boston, 1973.
8. Hewson, E.W. and G.C. Gill: Part III - Meteorological Investigations
in the Columbia River Valley, near Trail, B.C. Report Submitted to
the Trail Smelter Arbitral Tribunal (by R.S. Dean and R. E. Swain)
U.S. Bureau of Mines, Bulletin 4537 1944.
9. Start, G.E., C.R.Dickson and L.L. Wendell: Diffusion in a Canyon
within Rough Mountainous Terrain. NOAA TM ERL-38, August 1973.
-------
-176-
3.6.3 Comments by D. Henderson
Submitted after the conference
\
o UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
,,4 ^ ' REGION VIII
I860 LINCOLN STREET
- "'ER. COLORADO 8O2O3
March 23, 1977
REF: 8AH-A
Mr. Albert E. Smith
Energy and Environmental Systems Division
Argonne National Laboratory
9700 South Cass Avenue
Argonne, IL 60439
Dear Al:
As a follow-up to our recent telephone conversation concerning comments
on the EPA Modeling Guideline Working Group II-3, I am submitting the
following suggestions. My first suggestions are related to use of
the EPA Valley Model in complex terrain. These are followed by sug-
gestions on the written statements prepared by Bruce Eagan, Douglas
Fox, and Sumner Barr.
As I indicated in the workshop I have used the Valley Model differently
than presented by Herschel Slater at the workgroup. I have attempted
to use local meteorological conditions which would take into consideration
the frequency of occurrence of the particular stability, wind direction
and wind speed frequency distribution. This allows more versatility
in applying the model, but requires consideration of persistence of
the prescribed meteorological condition.
Several suggested changes to the Valley Model User's Guide were given
by me to Herschel Slater and Ed Burt on the telephone. I suggested
that in the introduction a statement similar to that given below be
included.
"The EPA Valley Model is a modified Gaussian technique designed for
making ground concentration estimates for plumes emitted from area
sources and elevated point sources. Experience indicates that maximum
24-hour ground level concentrations frequently occur in elevated ter-
rain when plumes are contained in a stable layer below the height
of the terrain with the flow blocked under light wind conditions.
This condition was given primary consideration in the Valley Model
development for short term estimates."
-------
-177-
Th e following comments pertain to the submission referred to at the
beginning of this letter.
On page 11 at the end of section (ii), the statement is made that
the box model approach is a conservative screening method. Without
specifying averaging times, and how the top of the box is defined
one cannot generally conclude that a box model is conservative.
The results of the box model are very sensitive to the designation
of the top boundary.
At the bottom of page 11 and also in one of Dr. Eagan's sections,
physical modeling is proposed as a useable technique for source im-
pact analysis in complex terrain. The guidelines should point out
that the frequency of ocurrance and persistence of meteorological
conditions are not determined 1n physical modeling techniques.
Several other suggestions were given to you on the telephone, but I
believe the above covers those items I agreed to write you about.
I hope these suggestions will be helpful to you.
Sincerely yours,
Donald Henderson
Regional Meteorologist
-------
-178-
3.6.4 Comments by M. Will-Loans
Submitted after the conference.
O n t he occasions of which I am aware, the use of a plume half height
correction w ould have produced drastically lower values than those observed
or those predicted by Valley. These cases involve stable flow toward terrain
as high as the approximate plume height.
With respect to the use of the principle of conservation of mass to con-
sider what flow situations may be possible, it is very important that ^overly
simplistic models not be used. There is a tendency on the part of some to
use very simple models which assume that the wind is approximately uniform
in direction the over than lower 2000 feet of the atmosphere. Under stable
conditions this assumption is frequently false Furthermore, the use of this
assumption may suggest restrictions on plume behavior which do not, in fact,
exist. We have found cases where a stable plume reached a distant ridge top
(55 km) with little or no standoff distance. The ridge was approximately a two
dimensional feature oriented at 45° to the direction of travel. Winds were ap-
proximately 3. 5 m/sec.
With respect to alteration of diffusion rates we have found that terrain
elements do not appear to influence the diffusion of stable plumes unless they
are at full plume height. As long as there is no intervening terrain between
an obiect and the source, it seems that no alteration in diffusion rates is justi-
fied for elevated plumes during stable conditions.
-------
-179-
3. 7 GROUP II-4
'6.7.1 Comments on the Group II-4 Discussion of the
a Curves
%
Submitted by M. Williams after the conference.
Unfortunately I was not a party to the discussion which produced these
conclusions so I do not know what basis was used to produce the stated con-
clusions. However, I have reviewed data which does relate to the question
at hand. First, the Central Electricity Generating Board has reported on
measurements near to power plants with stacks of 137 meters and 183 meters
I,
in height. The calculational technique was that of the Pasquill H <•»;*» method
which is similar to the Turner method except that the J-*.-5 are a little
smaller in the Pasquill technique and the plume rise is calculated differently.
The plume heights calculated through the Kr^i* method are significantly
smaller than those calculated through Brigg's plume rise used in CRSTER.
Measurements near (1.4 km) the plant with the taller stack generally gave
values comparable to those predicted, the exception being that during A
stability with winds of 2.9 m/sec a 3-minute concentration of 58 pphm was
predicted while the measured value was only 32 pphm and the highest measur-
ed value was only 36 pphm. Using a Brigg's plume rise with Turner disper-
sion parameters the calculated value is 40 pphm which appears to be in good
agreement. Thus this evidence indicates that farily good agreement is ob-
tain.d if the' Turner values are taken as 3-10 minute values. The same docu-
ment suggests that the maximum one-hour values are from half to three-
quarters of the 3 minute values. Thus the Pasquill curves may overpredict
-------
-180-
by 33 to ] 00% if used as on--hour values at distances of about 1.4 km. This
would suggest that the corrections should be applied to rs^, not 0\ . In addi-
tion to the ground-level measurements there are measurements of actual
plume dimensions. These indicate that vertical spreads will be well in ex-
cess (more than twice) of those expected for Class B occurred even though
the wind speeds were too high (5.8 m/sec) for Category A. This data does
not seem to support either the replacement of Category A by Category B or the
linear extrapolation technique. Both of these procedures would lead to dra-
matic underpredictiontfor low wind speed, near plant cases.
•
The measurements do suggest some underprediction at 5 km. Further-
more- some of GX for categories B-D were greater than given by Turner
in addition to the- English work there is work in eastern Montana which
sts that the Turnrr C, curves are . appropriate. These data are based
• 6 C <^ «.
on work v.ith a silver tracer. I have enclosed Figures 5 and 6 from this
v. or .
Finally, the Lappes study also suggests the importance of looping type
?:tua:ions". Tbi- \\ork shew? that the highest ground level concentrations en-
countered occurred close in. There were two occasions where it looked as
thoue'n the- plume was underooinc classic looping type behavior near the Key-
£.
stone plant.' In both cases peak values were comparable to those predicted
with a Erigg's plume rise-Turner dispersion parameters combination. Fur -
mere, repeated instances of high concentrations were found at 11:00 a.m.
-2-
-------
-292-
These data suggest that while the computational procedures may require
some revision it would be imprudent to change the 0^ curves without much
more careful analyses. With the existing data as strong an argument can be
made for changing the ff^ curves as can be made for changing the *%. curves.
The suggested changes for A stability might veyy well lead to drainatic under-
predictions.
The recommended changes for the other categories also appear to be
.; "• - " ' •'• *
unsupportable. Certainly the measurements of dispersion during stable con-
ditions in the Southwest show a much different behavior. TV A experience al$«
indicates behavior similar tq that found in the ffouthwest. Experiments re-
ported by Slade also are inconsistent with the proposed changes.
1! A. !-5artin and F. R. Barker, "Further Measurgttents Around-
Stations I-III", Atnospheri c Envir&nment, ypl. 7, 1973,
:-.Power
pp |T-37.
?. Jar.es A. I'.einbscb, arlin B. Super, John T. M^rtland, "Dispersion Fron
An Elevated Source Over Clostrip, Montana"*.; raper #7p-?6.6 presented
^ the 66th Annual Meetinr of the Air ?cliuticr Centre! Association
in Boston, June 1975. • -
3. F. Peeler, Jr and L.E. TTi^iejrer, "Dispersion from Tall Stacks:; An; Evaluation
ir. Proceedings of the Second'International Clean Air Congresp edited by
Enruund 8: Beery, pi01i9-i056,
-3-
-------
-182-
Francis A. Schiermeier and Lawrence E. Kiemeyer, "Larre Power Plant Effiueri
~tudy \IAPPES) Volume I-Instrumentation, Procedures, and Data Tabulations
(1968)", U. S. Dept. of H. E. W., June 1970 pp
-------
-183-
75-26.6
Figure 5. Comparison of the Colstrip vertical diffusion
coefficients to those of Turner and Bovme's urban curves.
il 11 11 MI i i
eon.
Figure 6. Comparison of the Colstrip horizontal diffusion
coefficients to those of Turner and Bowne's suburban curves.
14
-------
-184-
3,7.2 Supplementary Comments on Discussion Topics of
Working Group II-4
Submitted by H. Cramer after the conference adjourned.
1. VERTICAL PROFILES OF WIND SPEED
I am in agreement with the working group' s statement that the representa-
tion of the increase of wind speed with height employed in the EPA models by means
of a power law with the exponent varying according to stability categories is satis-
factory. However, this statement of the working group should not be interpreted to
mean that the values of the power-law exponents now used by EPA are fixed and not
subject to change. For example, wind-profile exponents are known to vary not
only with stability category but also with wind speed and surface roughness. Spec-
ifically, it may prove desirable to vary the present values of the power-law expon-
ents for the C and D stability categories with moderate and high wind speeds and
with large changes in surface roughness parameters.
2. CONFLICTS AMONG SYSTEMS FOR ESTIMATING STABILITY CATEGORIES
In my experience the STAR program incorporating Turner1 s system using
hourly surface observations to assign Pasquill stability categories is very satisfac-
tory provided it is recognized that other meteorological parameters such as the
mixing height, vertical gradient of potential temperature and wind-profile exponent
must additionally be specified for each of the various combinations of wind-speed
and stability categories. Of these parameters, the mixing height is by far the most
variable and should not be considered fixed for any stability category, especially
neutral. (One of the basic deficiencies of the Pasquill-Gifford a curves is that
z
they contain implicit mixing heights — see comments below under Topic 3.) In my
view, the opinion expressed by the working group that the STAR progrem tended to
predict unrealistic ally high frequencies of occurrence of neutral stability reflects
-------
-185-
a lack of understanding of the importance of taking the mixing height into considera-
tion when considering the behavior of tall-stack plumes during neutral stability.
The STAR program output yields a stability classification that strictly applies in
the first 10 meters or so above the surface. If the plume stabilization height is in
a stable layer above the surface layer, which is frequently the case for tall stack
plumes, the plume behavior does not and should not correlate with the behavior ex-
pected on the basis of the stability category assigned to the surface layer. The
STAR program stability classification must be supplemented by detailed specifica-
tion of the wind and temperature profiles along the vertical from the ground surface
beyond the height attained by the upper edge of the plume as it travels downwind.
As shown in Figure 1, the meteorological inputs required for application of the
Gaussian plume model to tall stacks are thus directly related to vertical profiles of
wind velocity and temperature (and in some cases to the vertical profiles of humidity
and turbulent intensity as well) representative of a large reference air volume. This
reference volume, which includes the source and the points on the ground at which
concentrations are to be calculated, extends vertically to the top of the mixing layer
and has horizontal dimensions large enough to contain significant ground-level
concentrations.
I would like to register my complete and irrevocable opposition to the AT
method for determining stability categories. Even if one ignores the very consider-
able and insurmountable practical difficulties in measuring and interpreting small
vertical temperature differences of the order of 0.1 C, the AT observations are
strictly applicable only over the measurement height interval which is generally
100m or less. In working with tower AT measurements and following the AEC
(ERDA) guidelines relating AT to stability categories, I have never been able to
obtain results that appear to be reasonable on the basis of any conventional criteria.
On the other hand, use of the Turner and STAR program procedures yields results
that appear to be consistent with the wind measurements made on the towers and
other conventional criteria.
-------
z = H = Depth of surface mLxins layer ~ 1 kilometer
h m
v = x = Maximum downwind distances ~ 100 kilometers
Jh h
zh = Hm
-
RH{x,y,z,t}
I {x.y.z.t}
Figure 1. Schematic representation of the reference air volume.
-------
-187-
3. VERTICAL DISPERSION ESTIMATES
As we have discussed over the telephone, the first sentence on page 7 of
the 2 March First Round Report of Working Group II-4 was typed incorrectly and
should be changed to read:
"In their present form, the Pasquill-Gifford crz curves are
unsuitable for calculating the ground-level concentrations produced
by tall stack emissions. Specifically, the a curve for A stab-
ility results in large overestimates of the short-term maximum
ground-level concentrations (1-hour to 24-hour averages) and in
large underestimates of the distances to the maximum concentra-
tion compared with observations."
There are two basic deficiencies in the Pasquill-Gifford a curves that
z
make them inherently unsuitable for describing the vertical dispersion of tall stack
plumes:
• They refer specifically to the vertical dispersion of plumes
from sources located at or near ground level and thus con-
tain the effects of the large vertical gradients of atmospheric
density (temperature gradients) and turbulence near the air-
ground interface
• They are principally based on measurements of vertical
plume dispersion made at distances less than 1 km from the
source; the portions of the curves extending beyond 1 km are
extrapolations and were originally intended only to serve as
rough approximations to vertical dispersion from ground
sources at these longer distances
The light lines in Figure 2 show the Pasquill-Gifford curves for the A, B,
C, D, E, and F stability categories. The strong curvature of the A, D, E and F
-------
-188-
DOWNWIND DISTANCE (km)
FIGURE 2. Pasquill-Gifford az curves (light lines) and suggested modi-
fications (heavy lines).
-------
-189-
curves reflects the influence of the vertical gradients of temperature and turbulence
in the air layers close to the ground on the upward vertical dispersion of plumes
under very unstable conditions (A stability) and under neutral or slightly stable to
stable conditions (D, E and F stability). Both measurements and theoretical rea-
soning indicate that the downward dispersion of plumes from elevated sources
toward the ground surface is characterized by cr curves that plot as straight
z
rather than curved lines on double logarithmic paper (i.e., cr K x) as shown by
Z
the heavy lines in Figure 2 labeled A' through Fr. The upward dispersion of plumes
from elevated sources depends on the vertical temperature gradients and turbulence
in the air layers above the plume stabilization height.
In the absence of elevated temperature inversions that restrict upward
plume growth, there appears to be an approximate linear relationship between ver-
tical plume dispersion and cr . For example, Pasquill (1974, p. 202) cites experi-
Z
ments by Hogstrom (1964) in which smoke puffs were released at a height of 50 m
with the result that a encreased linearly with travel distance out to about 300 m
z
in all stabilities; at longer distances, a tended to increase less rapidly with dis-
Z
tance possibly because of a restriction on further upward expansion by an elevated
stable layer at the top of the mixing layer. Pasquill (1974, p. 200) aslo cites
experiments by Hay and Pasquill (1957) involving continuous tracer releases at a
height of 150 m which showed a linear relationship between vertical spread (o^)
and the travel time to downwind distances of 500 meters, the maximum distance
at which measurements were made. Briggs (1975, p. 36) notes that recent studies
of tall stack plumes show az approximately linear with distance in the most un-
stable categories.
The strongest evidence of the inapplicability of the Pasquill-Gifford
-------
-190-
max
2Q
7T6U
max
where H is the plume stabilization height and both cr and a are evaluated at
the distance x of the maximum ground-level concentration. Assuming all
parameters to be fixed except a and
-------
-191-
stability places the maximum ground-level concentration about twice as close to
the stack as the modified cr curve and yields a maximum concentration twice
£*
as large as the modified &z curve. Similarly, the P. G. o"z curve for D stability
places the maximum ground-level concentration about five times farther from the
stack than the modified a curve and yields a maximum concentration four times
smaller than the modified az curve.
Table 2 shows the GZ/VV ratios for all the Pasquill-Gifford curves at distances
from 0.1 to 10 kilometers. Note that the Cg/a ratios for C stability are approxi-
mately constant with increasing distance (which reflects the condition that az « x)
while the ratios for other stability categories either increase with distance (A and
B stability) or decrease with distance (D, E and F stability). Assuming that the
vertical dispersion of tall stack plumes toward the ground when the plumes are con-
tained in the surface mixing layer requires that 0.6) or underestimation (vz/vy < 0. 6) of
J
the maximum hourly ground-level concentration. It follows that the P. G. crz
curves lead to large overestimates of the maximum ground-level concentrations for
A and B stability and to large underestimates for D, E and F stability. Similarly,
the P. G. crz curves lead to large underestimates of the distance to the maximum
ground-level concentration in A and B stability and to very large overestimates of
the distance to the maximum ground-level concentration for D, E and F stability.
The degree of overestimation or underestimation in a particular stability category
is directly related to the plume stabilization height. The minimum plume stabilization
heights by stability category at which these effects become significant are approxi-
Veil and Hoult (1973) in a study of SOg observations from the Keystone plant found
that a value of crz/a = 0. 6 correlated best with hourly maximum ground-level con-
centrations during unstable conditions.
-------
-192-
TABLE2
RATIOS OF az/o- FROM THE PASQUILL-GIFFORD CURVES
Distance
(km)
0.1
0.2
0.5
0.7
1.0
1.5
2.0
3.0
4.0
5.0
6.0
8.0
10.0
Stability A
Category
0.53
0.57
0.96
1.32
1.97
3.39
5.05
8.49
B
0.54
0.57
0.63
0.66
0.69
0.74
0.79
0.85
0.91
0.96
1.00
1.07
1.14
C
0.59
0.60
0.58
0.58
0.58
0.58
0.58
0.59
0.59
0.59
0.59
0.60
0.60
D
0.60
0.56
0.50
0.47
0.44
0.41
0.38
0.35
0.33
0.30
0.29
0.27
0.26
E
0.60
0.56
0.49
0.46
0.42
0.38
0.35
0.31
0.28
0.26
0.24
0.22
0.20
F
0.57
0.54
0.47
0.44
0.40
0.36
0.33
0.28
0.26
0.23
0.21
0.19
0.17
-------
-193-
mately 100m (A stability), 300m (B stability), 45m (D stability), 22m (E stability)
and llm (F stability).
Confirmation of the effects described above is readily found in the results
of model validation studies sponsored by EPA. For example, Lee, et_aL_ (1975)
describe a validation study of the EPA CRSTER Model (which uses the Pasquill-
Gifford curves) that involved the application of the model to four power plants. In
each case, there was no significant positive correlation between concurrent calcu-
lated and observed hourly and 24-hour average SC^ concentrations. For a year of
data, the CRSTER Model tended to overpredict the maximum observed hourly SO2
concentrations and to underpredict to maximum observed 24-hour average concen-
trations. The poorest model performance was at the Canal Plant, which is located
near Cape Cod Bay and consequently has a much greater frequency of occurrence
of D, E and F stability than the other plants studied. Table 3 shows a comparison
of hourly SO2 concentrations observed at monitor stations in Tacoma, Washington
downwind from the 172-meter stack of the ASARCO copper smelter with two sets of
concurrent calculated concentrations. One set of calculated concentrations was
made by means of the short-term Gaussian plume model described by Cramer, et al.
(1975) which, except for the use of crz curves similar to those shown by the heavy
lines in Figure 2, is practically identical to the EPA CRSTER Model. The second
set of calculated concentrations (Pasquill-Gifford) was made by means of the same
short-term model except that the Pasquill-Gifford curves were used. Differences
in the two sets of calculated values in Table 3 are essentially due to the differences in
the a curves because the same source and meteorological inputs were used for both
z
sets. For all the cases shown in Table 3, the average ratio of calculated and ob-
served concentrations is 0.11 for the model calculations using the Pasquill-Gifford
curves and approximately unity for the model calculations using the modified curves.
The principal explanation is that the Pasquill-Gifford curves for D and E stability do
not allow the plume to come to the ground at the distances of the monitors. This is
the same result found by Lee, et al. (1975) at the Canal Plant. The modeling
10
-------
TABLE 3
COMPARISON OF CALCULATED AND OBSERVED
HOURLY SO2 CONCENTRATIONS
Case
No.
2
3
4
6
7
8
Monitor
N26th and Pearl
Reservoir
Highlands
N26th and Pearl
N26th and Pearl
Reservoir
N26th and Pearl
N26th and Pearl
N2Gth and Pearl
N26th and Pearl
Reservoir
Observed
Concentration
(ppm)*
(1.23)
0.27
0.73
0.40 (0.46)
0.56 (0.68)
0.62
0.30 (0.22)
0.26 (0.26)
0. GO (0.46)
0.32 (0.21)
0.37
Calculated Concentration
(ppm)*
Pasquill-
Gifford
0.03
0.00
0.06
0.12
0.00
0.00
0.04
0.01
0.00
0.00
0.00
Cramer, et al.
(1975)
0.88
0.04
0.44
0.25
0.61
0.64
0.30
0.23
0.26
0.44
0.39
Ratios of Calculated and
Observed Concentrations**
Pasquill-
Gifford
(0. 02)
0.00
0.08
0.30 (0.26)
0.00 (0.00)
0.00
0.13 (0.18)
0.04 (0.04)
0.00 (0.00)
0.00 (0.00)
0.00
Cramer, et al.
(1975)
(0.72)
0.15
0.60
0.63 (0.54)
1.09 (0.90)
1.03
1.00 (1.36)
0. 88 (0. 88)
0.43 (0.57)
1.38 (2.10)
1.05
Pasquill
Stability
Category
D
D
D
C
E
E
D
D
D
E
E
^Numbers enclosed by parentheses are concentrations measured by the ASARCO SO2 monitor at
and Pearl.
**Numbers enclosed by parentheses are ratios of calculated and observed concentrations for the ASARCO SO2 mon-
itor at N26th and Pearl.
-------
TABLE 3 (Continued)
Case
No.
9
12
13
14
19
Monitor
N2Gth and Pearl
N26th and Pearl
N26th and Pearl
McMicken Heights
Meeker
Meeker-Brown
Meeker
Meeker-Brown
Meeker
Meeker-Brown
McMicken Heights
Tukwila
N26th and Pearl
Reservoir
N26th and Pearl
Observed
Concentration
(ppm)*
0.37 (0.67)
0.31 (0.21)
0.25 (0.25)
0.50
0.28
0.31
0.38
0.42
0.59
0.49
0.30
0.41
0.42 (0.35)
0.50
0.27 (0.17)
Calculated Concentration
(ppm)*
Pasquill-
Gifford
0.00
0.06
0.00
0.25
0.04
0.06
0.06
0.07
0.06
0.05
0.06
0.06
0.00
0.00
0.07
Cramer, et al.
(1975)
0.64
0.11
0.31
0.52
0.28
0.35
0.42
0.50
0.56
0.53
. 0.09
0.10
0.87
1.00
0.10
Mean Ratios
Ratios of Calculated and
Observed Concentrations**
Pasquill-
Gifford
0.00 (0.00)
0.19 (0.29)
0.00 (0.00)
0.50
0.14
0.19
0.16
0.17
0.10
0.10
0.20
0.15
0.00 (0.00)
0.00
0.26 (0.44)
0.11 (0.12)
Cramer, et al.
(1975)
1.73 (0.96)
0.35 (0.52)
1.36 (1.36)
1.04
1.00
1.13
1.11
1.19
0.95
1.08
0.33
0.24
2.07 (2.49)
2.00
0.37 (0.59)
0.97 (1.00)
Pasquill
Stability
Category
D
A
E
D
D
D
D
D
D
D
C
C
E
E
B
I
V-4
Oi
I
*Numbers enclosed by parentheses are concentrations measured by the ASARCO SOa' monitor at N26th and Pearl
**Numbers enclosed by parentheses are ratios of calculated and observed concentrations for the ASARCO SO2 mon-
itor at N26th and Pearl.
-------
-196-
techniques, emissions data, meteorological data and the air quality observations
referenced in Table 3 are described in detail in the report prepared for EPA by
Cramer, et^al. (1976).
13
-------
-197-
REFERENCES
Briggs, G. A., 1975: Plume rise predictions. Paper presented at the AMS
Workshop on Meteorology and Environmental Assessment, Boston, Mass.,
September 29-October 3, 1975.
Cramer, H. E., H. V. Geary and J. F. Bowers, 1975: Diffusion-model calcula-
tions of long-term and short-term ground-level SC>2 concentrations in
Allegheny County, Pennsylvania. H. E. Cramer Company Technical
Report TR-75-102-01 prepared for the U. S. Environmental Protection
Agency, Region in, Philadelphia, Pennsylvania. EPA Report 903/9-75-018.
NTIS Accession No. PB-245262/AS.
Cramer, H. E., J. F. Bowers and H. V. Geary, 1976: Assessment of the air
quality impact of SO^ emissions from the ASARCO-Tacoma smelter. EPA
Report No. EPA 910/9-76-028. U. S. Environmental Protection Agency,
Region X, Seattle, Washington.
Hay, J. S. and F. Pasquill, 1957: Diffusion from a fixed source at a height of a
few hundred feet in the atmosphere. J. Fluid Mech., 3, 299.
Hogstrom, U. 1964: An experimental study on atmospheric diffusion. Tellus, 16,
205.
Lee, R. F., M. T. Mills and R. W. Stern, 1975: Validation of a single source
dispersion model. Paper presented at the 6th NATQ/CCMS International
Technical Meeting on Air Pollution Modeling, Frankfurt/Main, Germany,
24-26 September 1975.
Pasquill, F., 1974: Atmospheric Diffusion (Second Edition). Ellis Horwood
Limited, Sussex, England, 429.
Weil, J. C. and D. P. Hoult, 1973: A correlation of ground-level concentrations of
sulfur dioxide downwind of the Keystone stacks. Atmospheric Environment,
T 707-721.
14
-------
-198-
3. 7. 3 Comments on Report of Working Group II-4
Submitted by D. Bruce Turner and L. E. Niemeyer after the conference.
Group II-4 expressed concern with the adequacy of the Pasquill-
Gifford vertical dispersion coefficients for unstable conditons when
applied to sources with tall stacks. Nevertheless, this distinguished
group did not see fit to make a recommendation for changes at this time.
The Pasquill-Gifford curves have stood the test of over a decades' use.
The prudent course of action is to continue to use the time-tested factors
until such time as the data which are now becoming available as sampling data
from the vicinity of facilities with tall stacks are organized, until they
are subjected to full scientific review and scrutiny and until the
scientific community agrees that current values for the vertical dispersion
parameters result in wholly unsatisfactory estimates. It is our current
task to encourage, enhance and aid a prompt scientific analysis and review
of this matter. To our knowledge such activities are underway by several
atmospheric scientists and by professional scientific societies.
-------
-199-
3.8 GROUP II-S
No Supplementary Materials
-------
-200-
-------
-201-
3.9 GENERAL SUPPLEMENTARY MATERIALS
3.9.1 Use and Formulation of the Hanna-Gifford Model
February 25, 1977
U.S. DEPARTMENT OF COMMERCE
National Oceanic and Atmospheric Administration
ENVIRONMENTAL RESEARCH LABORATORIES
Post Office Box E
Oak Ridge, Tennessee 37830
Mr. Kenneth Brubaker
Energy and Environmental Systems Division
Argonne National Laboratory
9700 S. Cass Avenue
Argonne, Illinois 60439
Dear Mr. Brubaker:
I've had a chance to review the section on my model on pages
2.18 through 2.21 of your report "Descriptions of Air Quality
Models and Abstracts of Reference Materials." There are a
few corrections that should be made. I suppose these problems
are basically my fault, since we have published our model in
a series of articles, rather than in a comprehensive user's
guide.
You and the EPA break up our model into a short-term and a
long-term model. In reality, we never intended such a division
and believe the model is equally applicable to averages from
20 or 30 minutes on up. Historically, we began in late 1969
with the equation
/2 1 (Az/2V^
V (7 o(l - W
t = i
(1)
where az = axb. After applying this equation to several gridded
urban areas, we discovered that the formula
X = C(Qo/U) (2)
works just fine for most areas, where
f1 —
(3)
-------
-202-
The "constant" C theoretically equals about 600, 200, and 50 for
stable, neutral, and unstable conditions, respectively. So, you
see we do have theoretical expression for C, in contrast to the
statement made in your review. Further, we always caution that
(1) should be used in place of (2) whenever the local emission
Q0 is much less than the Q's of neighboring grid blocks. Also,
it has always been recommended as part of our model that strong
point sources (there are usually 10 or 20 in an urban area) be
treated separately using the standard plume model.
Of course, whenever good observations of concentrations are
available, the constant C in equation (3) or the expression
/tTiKAx^^-lVUCl-b)) in equation (1) should be replaced by
a calibrated value. The diurnal variation of C has always been
questionable, but through the analysis of much CO data from
several states I have recently developed a diurnal curve for
C. This will be reported in the open literature during the next
few months.
We have also found from studying observed pollutant concentrations
that the calibrated C varies with the pollutant, being highest
for CO, lowest for S02, and intermediate for suspended particles.
There are several hand waving arguments for these differences.
Other modelers at the Nordic Hills workshop reported exactly the
same behavior with their models.
In 1973 I extended this model to Include the photochemical
pollutants NO, N02, and oxidants. The seven-step reaction
mechanism proposed by Friedlander and Seinfeld was used, although
any kinetic mechanism could be plugged into the model. Predictions
of this model were compared with predictions of other models in
the Los Angeles basin, showing that our model was just as good
as the others. In this case as well as in the other reprints
that I have enclosed, we test or validate our model extensively.
Because of the ease with which our model is applied, I can
confidently state that it has been validated in the open
literature much more often than any other urban dispersion
model. Whenever a new set of observations comes out, we test
it.
The following outline is my suggestion for a revision to pages
2.18-2.21 of your report:
References:l)Hanna, S. R. A Simple Method of Calculating
Dispersion from Urban Area Sources. J. Air
Poll. Cont. Assoc., 12_f 774-777 (Dec. 1971).
2)Gifford, F. A. and S. R. Hanna. Modeling Urban
Air Pollution. Atmcs. Environ., _7» 131-136 (1973).
3)Hanna, S. R. A Simple Dispersion Model for the
Analysis of Chemically Reactive Pollutants.
Atmos. Environ., 7_, 803-817 (1973).
-2-
-------
-203-
Abstract: This is basically an area source model that can
be applied to any size grid square. It can be used in
conjunction with the Gaussian plume model, which is used to
treat the largest point sources in the region. Chemical
reactions and physical removal mechanisms can also be
incorporated into the area source model.
Equations: For grid squares in which the local area sources
emissions are much less than those in neighboring grid squares:
for area sources across which emissions are nearly uniform
X = C(Q/U)
where
^ _. 1 _ y-1
If good data on x» Q» and U are available, C can be estimated
or "calibrated" with these data.
A. Source-receptor relationship
Uniform grid squares defined by user.
Receptors and area sources at ground level.
Receptor at center of grid square.
Point sources at any location.
B. Emission rate
User-specified for each grid square or point source,
emissions not time-varying over the period of interest.
C. Chemical composition
Define the normalized concentration
X±* = XiU/CQi (4)
When chemical reactions are not important, then according
to eq. (3), X* = 1.0.
Assume that C = Ax/Z, where Ax is the width of the region
and Z is the height of vertical dispersion. For illustration,
use the Friedlander-Seinfeld (1969) seven step photochemical
kinetic mechanism. Then
—3—
-------
-204-
[0 .] = 6 - 6 a constant
-> INOJ
la [NO]* = - I - [NO,]* [RH]« a
~ ! + P^01* IRH]*
(5)
(8)
- ONOJ*/[NO]')
where t* = tU/Ax and a, X, 9, and y are rate constants. These
equations can be rewritten, using the reference, for any kinetic
mechanism.
D. Plume behavior
Gaussian plume model with Briggs1 plume rise for point
sources.
E. Horizontal wind field
User-supplied wind speed and direction over a 16 point wind rose.
No variation of wind with height.
Constant winds for each calculation.
F. Vertical wind speed
Assumed equal to zero.
6. Horizontal dispersion
Narrow plume approximation for area sources.
Power law ay = axb for point sources.
H. Vertical dispersion
oz = axb assumption used, with a and b from Smith (1968)
or Briggs (1973).
-4-
-------
-205-
I. Chemistry and Reaction Mechanism
Friedlander and Seinfeld (1969).
Any mechanism could be input, however.
J. Physical removal
1/(1 + C(vd/u)) factor applied to concentration
prediction, where v
-------
-206-
3 9.2 Recommended Changes in Draft Guidelines
Submitted by Michael D. Williams
On page 6
I believe that the second highest of all estimated concentrations should
be used. Thus on page 6 the relevant sentence would read:
"Thus, emission limits which are to be based on an averaging time of
24 hours or less shall be based on the second highest of all estimated concen-
trations (plus a background concentration which can reasonably be assumed to
occur with that concentration; see section on background concentrations). "
This is consistent with the protocol between EPA and Salt River Project
with respect to the measurement program for the Navajo Generating Station
Sulfur Dioxide Field Monitoring Program.
I note that if the highest, second highest concept is to be used, then
meteorological data equivalent to the life of the facility should be used rather
than for a five-year period. It seems likely that the second highest at an in-
dividual point in a year will increase as the number of years tested.
Page 19
Under item (4), with respect to an area with meteorological or topo-
graphic complexities, I believe this item should read:
"(4) If the meteorological or topographic complexities of the region
are such that the use of any available air quality model is precluded, then the
model used for control strategy evaluation may be limited to a rollback model30,
if a dense monitoring network is in place and has been operating for at least
two years. In this context a dense monitoring network would be one in which
all points expected to receive high concentrations on the basis of models" appli-
cable to the terrain are well covered. "
I believe this change is justified because a rollback model based on a
limited sampling network which does not address the points of controversy
is likely to be much worse than best estimates based on existing models.
Page 29
The last sentence in the first paragraph should read:
-------
-207-
"The receptor grid must allow sufficient spatial detail and resolution
so that the location of the maximum or highest, second highest concentration
is identified for all areas. "
The phrase "which are generally accessible to the public" could, under
some interpretations, be limited to roadside areas. The mountain climber
struggling with a difficult pitch probably needs clear air too. Furthermore,
the intent of the standards is to protect more than just human life. Finally,
the health of vegetation on a steep slope not generally accessible to the public
may be very important for esthetic or soil holding purposes.
Page 30
After the second sentence in the second paragraph, insert the following
sentences:
"In cases where the impact of elevated sources on high terrain during
stable conditions is to be considered the relevant winds are those measured
at expected plume height rather than ground level or near ground level. In
cases where the frequency of A stability conditions is important, methods
should be used to either make direct measurement of the frequency through
the use of a bivane or adjust the frequency given by the Turner categoriza-
tion scheme to provide a better estimate of the actual frequency. "
During stable conditions the ground level inferences of stability, wind
speed and wind direction are likely to be so poor as to make their use unac-
ceptable. Table I reports all days of stable conditions (at plume height) re-
ported in Schiermeier's "Large Power Plant Effluent Study (LAPPES) Volume
HI. Instrumentation, Procedures and Data Tabulations (1970)" January, 1972.
-2-
-------
-208-
TABLE I
Date
April 20
April 22
April 2 7
April 28
April 30
May 5
May 8
May 9
May I'l
May 15
Oct. 14
Nov. 9
Time
0708
0700
0635
0800
0830
0700
•800
0700
0630
0830
0800
0900
Height
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400
Sfc
50 m
400
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Sfc
50 m
400 m
Speed
(mps)
2.2
4.3
12.0
0.0
2,6
12.1
-------
-209-
Exhibit #1 also shows that surface winds tend to be poor predictors.
With respect .to the second line work by Lun«uand Church "A Comparison
of Turbulance Intensity and Stability Rates Measurements to Pasquall Stability
Classes, " Journal of Applied Meteorology, Volume 11, June, 1972; pp. 663-
669, illustrates the large discrepancy between stability as indicated by fluctua-
tions in the vertical component of the wind and by Turner stability categories.
Finally I have difficulties with both CRSTER and \jAlley. The principal
defects in CRSTER are that fumigation is not considered and that reduced
plume rise with a capping inversion is not considered. I note that a group
(Dames and Moore) modeling for utilities in the Southwest has suggested that
the plume rise may be restricted to 2/3 of its normal value. There is a case
in the Lap^se study in which the plume rise was about one-half of its expected
values (May 4, 1970, near Ho«%er City). I believe some changes could be in-
corporated into CRSTER which would permit better predictions in the case of
limited mixing situations.
With respect toMdlley, I believe a number of modifications are reason-
able which would improve its prediction capabilities. These can be drawn
from experience at the Navajo Plant plus aircraft measurements of stable
flow dispersion in the Navajo region, Four Corners region and the TV A.
For more distant travel I recommend that time of travel be considered.
The principal elements of the model changes are described below.
-4-
-------
-210-
Modifications to CRSTER
Schiermeir's LAPPES study carries a case where the mixing layer
clearly inhibited plume rise. On May 4, 1970, ground level concentrations
were consistently about 0. 5 ppm at 3 km downwind between 10:30 and 10:45
(p. 80). Helicopter temperature profiles gave a strong inversion at 765
meters (p. 218); pilot balloons indicated an average wind speed between stack
top and 750 meters of 2. 2 to 2. 5 m/sec (p. 244). Plant operational parameters
(pp. 287-288) gave one unit at 2126 gm/sec, VS = 21.3, T = 149° fi, DT = 137° (L,
while the other unit was at 2556 gm/sec, VS = 21. 3, T = 158, DT = 106. With
a stack height of 244 meters and stack radius of 3.65 meters, the expected
stack height will be 1380-1535 meters. Under these circumstances CRSTER
would give a zero concentration because the expected height would be greater
than the height of the mixing layer. A B stability calculation using the actual
height 765 meters would give x = 1376 ug/m3 as opposed to a measured maxi-
mum of 1493 ug/m3. The predicted maximum would occur at 4. 5 km. On the
other hand, an A stability calculation would give 3742 at 1. 2 km. Thus, for A
stability one would have to use an assumption that one -half of the plume pene-
trated the stable layer and was trapped. On this basis I suggest that
He = L if L i Hs + . 5 _fc he
however, for A stability
(1) Qe = Q+Q (L - Hs- >5 Ah ) .
. ^ £
Ahc •> s '
Replacement for Valley
Ihe results of the Navajo study indicated that the maximum 3 -hour con-
centraliu~ ~ould be represented by:
(2) y. -~
J. Tr u J"-
with the values considered as 10-minute averages and extrapolated to 3 hours
via a technique:
-.T-
Note that the equation for'X does not use plume reflection. The parameters
cry and oz sure. Turner values for E stability. However, aircraft data in TVA,
Four Corners and the Navajo area suggest that the distance dependence is in-
correct. Instead a form:
0
-------
-211-
is suggested. Thus I recommend
rfy = 140. 6 x • 5* ) oy, o*z in meters
) x in kilometers
Oz = 68. Ox -21 )
Finally, I believe •$••'*.« of flight should be considered. Thus the model
would calculate the concentration expected at the receptor at x and then examine
the wind directions for the V**« x/u. If the wind direction were reasonably con-
sistent during that time the calculated concentrations would be presumed to
have occurred at the receptor. Otherwise, they would not. Work by Start,
et al. in Utah suggests similar relations apply to unstable and neutral condi-
tions.
-------
O O 0 O O
^ T 1 ? *-
O O O
o
O
. n
O O
4OOO
••Illi lili
•I
35OO'
. . .11..li.. Him
3OOO
250O
j '
i . .1.. I -Nil
2OOO'
.--••• •••iiil.il
15UO'
. • i.. .. •-•- M--
1000'
5OO
.lii-iii-- • •-!- 11
-# Cases
SURFACE
.lllllll.. .
O O
o o
o o
r» \o
Direction
Figure 6. The arrays of direction by height for 70 pilot balloon runs
from surface to 4000 feet above San Juan Plant Site, March 15-23, 1972.
-23-
-------
-213-
3.9.S Conments on the RAM Urban Model
itft-*Sinffer <^Auieoro/ogists, >jnc. DATE: February 4, 1977
TO: (MESy Model Study File (Maynard E. Smith)
FROM: JRM , ,
RE: The RAM Urban Model
Howard Ellis made many points in his study of the RAM Model, a
few of which deserve comment in regard to the Argonne Conference.
RAM is an urban model designed to calculate maximum 24-hour con-
centrations on a grid using a year's worth of data. It could be used
for 1-hour and 3-hour predictions. Criticisms of it follow:
1. The model has not been validated. An extensive validation
study must be done before the predicted results can be con-
sidered realistic.
2. o'z values are excessively high for Pasquill A and B (they
are equal for these two classes). The az curve for Pasquill
A and B is not as extreme as the CRSTER Model. However, the
•effect would be a doubling or tripling of the predicted max-
imum concentration compared to our modeling of tall stacks
for 1-hour periods under very unstable and unstable conditions.
3. These o"2 values were developed from the St. Louis tests in .
the mid-60*s, using low-level tracer releases and do not ap-
pear to be applicable to even low-level sources, let alone
elevated sources. In brief, <$z is considered inversely pro-
portional to the crosswind integrated concentrations observed
in these tests. Ellis documents several studies showing that
the urban plume centerline is lifted by a mean upward wind
component, leading to very low surface CIC and the resulting
inflated
-------
-214-
3.9.4 Corrments Regarding EPA Models
ENVIROPLAN, INC.
An Environmental Planning Company
47 ORIENT WAY. RUTHERPORD. N J. O7O7O
2O1 935-5O98
February 16, 1977
Mr. Maynard E. Smith
"Smith-Singer Meterolegists, Inc.
134 Broadway
Amityville, New York 11701
Dear Maynard:
Concerning your 1/29/77 letter, I would like to stress
the following points concerning the E.P.A. Models no\7 in use.
1. RAM Model (Urban Version). This model was used
for the first time in E.P.A. regulation setting in Ohio and
has seven serious problems. The enclosed evaluation adequately
describes our concerns as well as recommendations for a Revised
RAM Model for urban areas. I and many of the electric utilities
in Ohio with plants in urban areas will be very grateful if you
would read the enclosed critical evaluation, perhaps discuss it
with me prior to the meeting, and then forcefully present the
criticisms you concur with at the 2/22 meeting. Perhaps the
most, serious problem with Urban RAM is the use of vertical
dispersion rates that are greatly in excess of Pasquill-Gifford
vertical dispersion rates and that lead to predicted maximum
short-term concentrations from power plants that are tv;o to
four times previous predictions using Pasquill-Gifford dispersion
rates. I consider the use of Urban RAM as presently constituted
to be the single most serious problem with present E.P.A. models
for urban areas.
20 Use of Pasquill Stability Class A as the Most
Extreme Vertical Dispersion Rate. My concerns with this issue
and the necessity of using a less extreme vertical dispersion
rate under the most unstable meteorological conditions are
presented in the appendix to the Muskingum River Plant report
prepared by Enviroplan in late 1976. Your office has a copy
of this report.
3. Prediction Modeling in Complex Terrain. Two issues
are of special importance here:T5the assumed point of closest
approach of the plume centerline to ground-level as the terrain
rises, and 2) the enhancement of the turbulence and dispersion
as the plume approaches elevated terrain. The Egan approach to
issue 1) of reducing the effective stack height by half the
-------
-215-
Mr. Maynard E. Smith
February 16, 1977
Page 2
increase in ground elevation seems more reasonable than using
the E.P.A. CRSTER Model's full ground displacement procedure
of subtracting the entire difference in ground elevation
between receptor and stack base from the effective stack
height. Progress on this issue and issue 2) would be helpful
in eliminating some of the very major conservatisms in E.P.A.
prediction modeling with CRSTER.
4. Proposed Use of Five Years of Meteorology in the
Prediction Modeling. If E.P.A. adopts this procedure of using
five individual years of meteorology in modeling, it is important
some measure of actual operating rates rather than constant
maximum possible operating rates be used. One excuse for using
constant maximum operating rates is that only a single year of
meteorology is being used for analysis. If the number of years
of meteorology increases, some adjustment in the assumption on
operating rates is needed to produce reasonable predicted con-
centrations.
Finally, I would certainly welcome any future
opportunities there may be to participate directly in future
meetings*of this type.
Sincerely,
ENVIROPIAN, INC.
Dr. Howard M. Ellis
President
HME/pl
Enc.
ENVIROPLAN. INC. • An Environmental Planning Company •
-------
-216-
3.9.5 Descriptions of C.E.G.B. Air Quality Models
Central Electricity Research Laboratories —
Description of C.E.G.B. Air Quality Models
References Papers by Moore D.J. listed in EPA-600/4-76 030a, also:
Scriven, R.A. and Fisher B.E.A. and Fisher B.E.A. 1975 Atms Env. 9^
49 and 59 and 1063.
Fisher B.E.A. and Maul P.R. and Moore D.J. 1976 Proceedings of Symposium
'Systems and Models in Air and water Pollution* Institute of Measurement
and Control, London.
Moore, D.J. 1975 Proc. Inst. Mech Eng. 189, 33. and 1976 Atmospheric
Pollution 51-30 Ed. M. Benarie, Elsvier.
Abstract
The C.E.G.B. models include
(i) Gaussian models for predicting maximum g.l.c.s. from single
sources, the vertical spread being related to an average
vertical diffusivity, conservation of emitted material assumed.
(ii) Diffusivity profile models for calculating medium and long
range effects including wet and dry deposition and chemical
'. reactions.
-------
-217-
Gaussian Model
A. Source-Receptor Relationship
Point source only, twin stacks assumed separate for plume height
calculation single source for g.l.c. calculation. Multi-flue stacks as single
sources. Receptor measuring points usually at 2 m above surface.
B. Emission Rate
Take from station load and fuel data.
C. Plume Behaviour
Flume rise calculated form C.E.G.B. plume rise equations. Growth
due to relative motion taken into account in calculating location of maximum
g.l.c. but not its magnitude.
D. Horizontal Wind Field
Wind speed and direction measurements made to about 2 stack heights
on adjacent tall towers. Value of B based on measurements of crosswind spread
and fluctuations in measured wind direction.
2
E. Vertical Dispersion o = LX.
- - J - r T~~ "-••*• — ' Z
L is a function of:
Height of source and/or height of mixing layer (H,h) (m)
Free stream wind speed (U) (ms )
Lapse rate above stack top (36/3z) (Km )
Sensible heat flux due to (i) solar heating of „
ground. (E.) (Wm )
o
or (ii) advection effects
or cooling of _
cloud-tops (Ej) (Wm~ )
Surface roughness length (Z ) (m)
Coriolis parameter (f) (s )
Assumed independent of sampling the T for T > 3 min.
2 22
F. Lateral Dispersion a = LX + B X .
B is a function of free stream wind speed and sampling time T(s) (for
given source location and wind direction).
-------
-218-
G. Emission and Meteorological Correlations
The calculations give ensemble mean values of maximum g.l.c. under
a restricted range of meteorological conditions and emission rates when plume
material remains within the mixing layer. Vertical wind effects and interaction
with the top of the mixing layer cause variations about the ensemble mean and
lead to a scatter with values falling within a range 0 to 2 times the ensemble
mean value.
Terrain effects are not included as the sites studied are mostly
flat.
Calculations refer to SO- but could be applied to any other
conservative emission with negligible fall velocity.
Background is estimated from observations. The model may be used
to make estimates of concentration out to about 3 times the distance of
maximum g.l.c.
Sampling periods from 3 minutes to 1 year.
H. Validation/Calibration
Comparison with 3000 hours of observations at 2 power stations.
r.m.s. residuals
-------
-219-
B. Horizontal Wind Field
Average wind direction calculated using available meteorological
data (wind or pressure field).
C. Vertical Wind Speed - None.
D. Vertical Dispersion
K profile, K proportional to height in layer of depth Z.., constant
»<
from Z, to h, zero above h. Constant value of h consistent with values of L
in gaussian model (K.. = -*- where L and K are both independent of z).
E. Crosswind Dispersion
Based on gaussian model for hourly average and wind rose or
distribution of wind trajectories for annual average.
F. Surface Deposition
Use is made of effective deposition velocity (V ) at the top of
O
the surface layer, which is accurate for long range transport.
V .
g
For near and middle distance transport exact K model uses V at ground.
O
G. Removal by Precipitation
Can be included by suitable decay constant for hourly or daily
average. Precipitation included statistically with constant expected duration
of dry and wet periods for annual average.
H. Chemical reactions
Conversion of S02 to SO, included.
Background; Man-made contribution may be calculated if source inventory available.
I. Validation/Calibration
Flux of S02 leaving U.K. compared with aircraft measurements of
vertical profiles for middle distance transport.
Dry and wet deposition of S over W. Europe compared with OECD
network. Observations of concentrations of S02 and SO^ at distances up to
80 km from groups of power stations compared with aircraft cross-sections.
-------
-220-
G.l.c. of S0_ and SO, at distances up to 150 km from urban
complex.
J. Output
Variable to suit requirement - e.g. surface deposition pattern,
g.l.c. pattern, SO. flux.
-------
-221-
3.9.6 Description of the Air Quality Short Term Model
Reference:
Abstract:
Equations:
"Air Quality Short Term Model," Illinois Environmental
Protection Agency, Springfield, Illinois, January 1976.
The Air Quality Short Term Model (AQSTM) is a steady
state Gaussian plume model for estimating concentrations
of relatively stable pollutants for averaging times from
an hour to a day from multiple point sources in .level or
complex terrain. Concentrations can be computed to simu-
late inversion break-up fumigation, lake shore fumigation,
and atmospheric trapping.
1) Contribution from each upwind point source
Q
cr u
2) Trapping
Q
'27T OyLu
i
~X2
ry \
^ay
2~
3) Fumigation
— __ «xp
^Truery,hr
•\
\
y
i",.J
.2
wh«r«i h,: H -f 2CTZ
4) Continuous Lake Shore Fumigation
-------
STABLE AIR
>v_>ta**^^<>«fcA^«^A--*»-*^-.*--A-A
>
Co
•>
•'
I
Figure 2. Lake Shore Fumigation.
-------
-223-
A. Source - Receptor Relationship
Arbitrary location for a maximum of 200 point sources
Up to 900 receptors located on uniform rectangular grid
Unique release height for each point source
Unique separation for each source-receptor pair
Unique topographic elevation for each receptor and source
Receptors must be at ground level
B. Emission Rate
Unique constant emission rate for each source
C. Chemical Composition
Treats one or two pollutants simultaneously
D. Plume Behavior
Briggs (1971, 1972) final plume rise formulae
Does not treat downwash. If plume height exceeds mixing height, maximum
plume height is limited to the mixing height. Treats fumigation by: (a)
user specifying specific height of limiting lid and averaging time, or (b)
user specifying rate of rise of inversion breakup (mixing height) *»nd starting
E. Horizontal Wind Field height.
User specifies hourly wind speed and direction
Wind speed corrected for release height based on power law variation.
Constant, uniform (steady-state) wind assumed. In complex terrain, plume
allowed to rise % the distance between the base of the stack and the height
of a ground-based receptor. Plume assumed to remain at constant height above
F. Vertical Wind Speed ground following initial rise.
Assumed equal to zero
G. Horizontal Dispersion
Dispersion coefficients from Turner (1969); no adjustments made for
variations in surface roughness or travel time.
User-supplied hourly stability class
Averaging time adjustment according to Turner (1969)(to one hour).
H. Vertical Dispersion
Dispersion coefficients from Turner (1969); no adjustments made for
variations in surface roughness, averaging time or travel time
User-supplied hourly stability class
I. Chemistry/Reaction Mechanism
Not treated
-------
-224-
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: user-input mixing height used; perfect reflection
assumed
Permits user to input continuous non-horizontal boundary layer
or rising boundary layer.
M. Emission and Meteorological Correlations
N/A
N. Validation/Calibration
No calibration option provided
Direct application of Turner (1969) procedures
0. Output
Average concentration, source contributions at each receptor for
total period of interest
Individual point culpability list for 5 maximum receptors
P. Computer Time Requirements
14 sources with 800 receptors requires 8 seconds CPU time
(IBM 370/168)
-------
-225-
3.9.7 Application of Air Quality Models Under the Ontario Environmental
Protection Act
Submitted by A. E. Boyer: post-conference.
The design of pollution control equipment and the
administration of environmental protection legislation are both
complicated by the wide variability of air quality resulting from
the interaction of pollution source characteristics and limiting
meteorological variables.
From the point of view of source design, it is often
desirable to design facilities to operate under whatever range of
meteorological and operating conditions are likely to occur during
the life of the source. An evaluation of limiting meteorological
conditions, however, often shows that these limiting conditions
occur a small fraction of this time; to design facilities to
operate under all conditions, including these infrequent happenings,
may result in greatly increased design costs.
The development of environmental legislation requires
resolution of the need for laws which are easily understood and
enforced and at the same time, cover a wide range of complex
possibilities. These two ends are not easily resolved.
The development of several air quality simulation models
provides tools for the analysis of the interaction of important
source and meteorological conditions. The problem then is how to
apply these models within the legal and engineering limitations
noted above.
The Ontario Environmental Protection Act provides the
framework for agencies and pollution sources to function within
a simply enforced law based on a point source model. This Act
also allows for redefinition of these simple controls in the
event subsequent ambient air quality does not meet desired goals.
-------
-226-
An example of the application of both simple point
source and complex urban air quality simulation models within
Ontario legislation is illustrated by Boyer and Shenfeld in,
"Atmospheric Impact of Coal Firing", Power Magazine, March 1975.
The question of converting an urban power generating station to
an all coal-fired operation is examined. In this example, the
Ontario Ministry of the Environment applied a multi-source urban
model to the question of air quality impact under the assumption
of the interruption of natural gas consumption by industrial and
commercial users, limiting meteorological conditions and con-
version of a large power generating station to an all coal-fired
operation.
The conclusions of the above study are less important
here than the methodology and the division of responsibility
between the individual sources and the control agency, in this
case, the Ontario Ministry of the Environment.
Each of the sources contributing to the S02 levels under
study had previously been required to meet short-term (30-minute)
air quality criteria for various pollutants. Compliance with
this regulation can be satisfied by estimates of impingement
concentrations based on a simple point source model. Certain
frequently occurring meteorological conditions are specified for
the basis of this estimate.
The responsibility of the control agency is then to
estimate the combined effects of multiple sources under a wide
spectrum of weather and operating conditions. This is accomplished
in a variety of ways, including the application of urban air
quality simulation models. ^»2
If after application of the urban model, the control
agency decides that the combined effect of multiple sources all
meeting the regulation for a single source is unacceptable, then
the single source limits may be changed by the issuance of new
guidelines for the application of single source models. Other
options may be included in the amended guidelines as, for example,
the use of supplementary control systems.
-------
-22?-
Working within this framework, each source is assured
that so long as ambient air quality objectives are met, the
single source requirements with which they are forced to comply,
will not be made more stringent.
The advantage of the guideline described above is that
it allows for enforcement based on a simple straightforward point
source model, while at the same time, allowing for a more detailed
assessment of complex multi-source model applications by the con-
trol agency. In addition, while the various evaluations are in
progress, it is in the interest of both the agency and the various
sources that ambient air quality goals are met.
The procedure suggested above is similar to the use of
emissions offset regulations used by the Environmental Protection
Agency. They are different in that the Ontario guidelines make
allowances for not only emission offsets to achieve desired air
quality, but also variations in stack design, emission temperature
and velocity, and in some cases, may include options for supplemental
controls under limiting meteorological conditions.
1. Boyer, A.E. and Shenfeld, L. "Atmospheric Impact of Coal
Firing", Power, March 1975.
2. P.S. Wong, K. Heidorn and D. Yap, "Modeling Sulphur Dioxide
Levels in the Sarnia Area", Water, Air and Soil Pollution
(1976) 407-414.
-------
-225-
5,3.8 Comments Regarding CRSTER
t wwc.
134 BROADWAY. AMITYVILLE, N. Y. 117O1 TEL: 516-691-3395
March 21, 1977
Dr. John J. Roberts
Deputy Division Director
Energy and Environmental Systems
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
Dear Dr. Roberts:
Your letter of March 4 about the Specialist Conference on the
EPA Modeling Guideline suggested that individual letters be sub-
mitted for the Appendix where participants have comments which are
pertinent to the sessions, but were not actually a part of pro-
ceedings.
In Session II-4 the randomizer system used in the CRSTER model
was discussed, and the working group decided that the method should
be studied to see whether wind direction variability is adequately
represented. This letter is a partial response to that suggestion.
The CRSTER model provides for a random adjustment to the wind
directions which are typically reported only in 10° intervals by
the National Weather Service. This procedure is supposed to simu-
late the real wind fluctuation from hour to hour.
\
The effect of the system on calculated ground-level concentra-
tions has been evaluated. Table 1 shows the concentrations directly
downwind of the three-hourly mean wind under three conditions;
(a) with no direction fluctuation, (b) with a JIl° shift, and
(c) with a +4°, -5° shift. The latter is the maximum variability
possible in the existing CRSTER system. Only with the maximum fluc-
tuation is the mean ground-level concentrations reduced appreciably.
Most modeling systems based on real data use a method sumulating
a more rapid lateral plume diffusion than allowed by CRSTER. TVA
uses Oy values which are larger than Pasquill-Gifford; both Smith-
Singer and the state of Maryland use wind shear terms; Cramer assumes
more rapid diffusion while the plume is rising after release.
-------
-225-
Dr. John J. Roberts March 21, 1977
Two studies are suggested to EPA to define a better randomizer
system:
1. Study real hourly mean wind direction fluctuations
to determine the typical fluctuations over 3-hourly
and 24-hourly periods.
2. Examine the oy values implied by the TVA, Maryland
and Smith-Singer systems. These data should show
how the randomizer might be improved.
Sincerely yours,
«ayna¥d E. Smith
Mark L. Kramer
John R. Martin
MES:la
Enclosure
-------
EFFECT OF WIND DIRECTION RANDOMIZATION ON GROUND-LEVEL CONCENTRATIONS
Wind
Speed
(m/sec)
Wind Categories:
Fixed Direction
Stability
Unstable
Neutral
(Hypothetical 300-meter stack)
o
Three-Hour Concentrations Along 180° Radial (ug/m )
i — • - - ^ ;
Fixed Minimum Maximum
Distance Direction Variation Variation
(km)
2 97 96 84
5 769 761 649
10 453 448 371
15 261 258 209
5 3 3 .1.5
/ 10 264 212 •' 116
' 15 535 497 219
ies: 1st Hour 2nd Hour 3rd Hour
ion 360° 360° 360°
ation 360° 359° 001°
ation 360° 355° 004°
tNO
CJ
aiW*, vjrn
-------
-231-
3.9.9 Cormsnts On the Need for Model Aaaupaey
Post Conference Contribution by J. L. Shapiro
In several different parts of the report, references are made
to the required accuracy for uniformity of validity of models. For
example, the Preface refers to an implicit "level of accuracy desirable. "
In Section 2. 5. 4 "the group generally agreed this assumption is not
uniformly valid. " • There are other references to this subject, both
explicit and implicit.
While some attention was directed to this by Group II-5, with an
indication that performance standards of models should be related to the
"loftiness of performance goals, " the overall tenor of the report does not
adequately reflect this view. It appears that the validity of models as
evaluated by the workshop participants is readily destroyed by citing
instances where the models' predictions were exceeded. In general,
the group is relatively satisfied with models that over-predict as
compared to models that under-predict.
This traditional approach obviously is necessary in many
applications but it is also obviously highly over-restrictive in some
cases, particularly where PSD limits are on the border line of
significance. In such cases, models should be adopted which are the
best available and which most closely represent an "expected value"
with the recognition that some cases will over-predict while others
will under-predict. As experience is gained and we learn more about
the details, we would expect that such models will be refined to reduce
the magnitude of the errors.
-------
-232-
-------
-233-
3.10 POLICY ISSUE SUPPLEMENTARY MATERIALS
3.10.1 On the Use of Statistical Techniques for the Prediction of Second
High 24 Hour Concentrations
Post Conference Contribution
Richard A. Porter3 P.E.
Air pollution standards styled as levels "not to be exceeded more than
one time per year" require a probalistic approach in modeling. However, models
that are currently available (listed in the draft guideline) are not suitable
for determining the second high value. Only the RAM model is available with
a statistical post processor, but the RAM model contains no method for testing
the model-generated data for goodness of fit to the proposed log-normal dis-
tribution. Predictions of values at the extreme end of the probability dis-
tribution function are very sensititive to the fit of the data to the proposed
distribution. Order of magnitude errors may be encountered when the data is
forced to fit the wrong distribution.
Many investigators have examined the frequency distribution of air
pollution concentrations at air pollution monitoring facilities. A wide range
of skewed frequency distributions have been fitted to the empirical data. The
distributions proposed include Weibull (Barlow, 1971; Milokai, 1972) and neg-
ative-binomial distributions (Prinz and Stratman, 1966); but by far the most
extensively fitted distributions are log-normal (Bencala, and Seinfeld, 1976;
Knox and Pollack, 1974; Shoji and Tsukatani, 1973) in the multiple source ur-
ban environment and exponential (Gifford, 1959; Scriven, 1965; Gartrell, 1966)
in an environment dominated by a single source. In the case of the urban
environment the work of R. I. Larsen (Larsen, 1970; 1971) based on analysis of
seven years of continuous monitoring data from USEPA monitors in urban areas
is the most widely cited investigation of the log-normality of air pollution
concentration. Studies by P. J. Barry (Barry, 1971; 1975) support the theory
that receptors,dominated by a single source,record concentrations of pollu-
tant that are exponentially distributed. The work by Barry is based on sev-
eral years of data using ARGON-41 injected in the plume of a power plant.
(Porter and' Christiansen, 1976).
In the paper "Predictions of Annual Sulfur Dioxide Concentrations for
Frankfurt An Main, Federal Republic of Germany, Aug., 1971 to July 1972
(Porter and Christiansen, 1976) , a frequency distribution of concentrations
-------
-234-
was generated using the Texas Episodic Model for 242 receptor points in the
Frankfurt area. Two methods were used to estimate the parameters of the
log-normal distribution: 1) the graphical method of Larsen (Larsen, 1971),
essentially the concept used in RAM; 2) the Delta- Log-normal distribution
(Atehison and Brown, 1957), a three parameter log-normal distribution that
accepts zero variates. Tests for log-normality using the Kolmogorov-Smirnov
Statistic (Lillienfors, H. W., 1967) failed at 73 of the 242 receptors for
one or both of the methods used to estimate the log-normal parameters.
All but two of the receptors that failed the test for log-normality
were located on the edge of the urban area. Both theory and observed data
suggest that the frequency distribution of concentrations at a receptor due
to a single point source is exponential (Gifford, 1959; Barry, 1975). In
theory, a receptor located in the center of a uniform area source will have
a log-normal distribution of concentrations (Bencala, and Seinfeld, 1976).
One can expect that receptors that are located in urban areas (away from the
influence of a strong single source) will have log-normal distributions, and
receptors that are influenced by a single source will have exponential dis-
tributions. However, the frequency distributions of concentrations at recep-
tors that do not fit in one of the above categories are probably a mixture
of the two distributions. The experience with the Frankfurt study cited
above indicated that unreasonably large values for the second high concentra-
tion were predicted for non-urban receptors that had low predicted mean con-
centration. These receptors failed the test for log-normality. It is sug-
gested that any model that uses statistically processed data to estimate con-
centrations should include a test of the data for goodness of fit to the pro-
posed distribution.
Alternative Methods
The alternative to the statistical approach is to use the "worst case"
estimate for the concentrations. The RAM model can be used to survey a spec-
ified period of meteorological record to determine the worst case condition
in terms of predicted concentration. Or a short term model (such as the Texas
Episodic Model) can be exercised on selected scenarios of meteorology that
can be expected to result in the highest concentrations for the sources in-
volved .
-------
-235-
REFEPENCES FOR SECTION 3.20.1
1. Barlow, R.E., Averaging Time and Maxima for Air Pollution Concentra-
tions, NTIS AD 727/413, ORC 71-17 (1971).
2. Milokaj , P.G., Environmental Applications of the Weibull Distribu-
tion Function: Oil Pollution, Science, 176:1019-1021 (1972).
3. Prinz, B. and Stratman, H., The Statistics of Propagation Conditions,
Staub, Vol. 26, Number 12, pp. 4-12 (1966).
4. Gifford, F.A., Statistical Properties of a Fluctuating Plume Model,
Proceedings of Symposium on Atmospheric Diffusion and Air Pollution,
Advances in Geophysics, 6:117-137 (1959)
5. Scriven, R.A., Some Properties of Ground Level Pollution Patterns
Based on a Fluctuating Plume Model, CEGB Lab Note No. RD/L/N 60/65
(1965).
6. Gartrell, F.E., Control of Air Pollution From Large Thermal Power
Stations, Revue Mensuelle 1966 de la Sec. Beige des Ingenieurs et
des Industrials Bruxelles, pp. 1-12 (1966).
7. Larsen, R., Relating Air Pollutant Effects to Concentration and
Control, Journal APCA, 20:214-225 (1970).
8. Larsen, R., A Mathematical Model for Relating Air Quality Measure-
ments to Air Quality Standards3 USEPA Pub. No. AP-89 (1971).
9. Barry, P.J., Uses of Argon 41 to Study the Dispersion of Effluent
from Stacks, Nuclear Techniques in Environmental Pollution, Inter-
national Atomic Energy Association, Vienna, pp. 241-255 (1971).
10. Barry, P.J., Stochastic Properties of Atmospheric Diffusivity,
Atomic Energy of Canada Ltd. AECL 5012 (1975).
11. Shoji, H. and Tsukatani, T., Statistical Model of Air Pollutant
Concentration and Its Application to the Air Quality Standards,
Atmospheric Environment, Vol. 7, pp. 485-501 (1973).
12. Atchison and Brown, The Lognormal Distribution, Cambridge Univer-
sity Press , (1957) .
13. Lillienfors, H.W., On the Kolmogorov-Smirnov Test for Normality
With Mean and Variance Unknown, J. Am. Stat. Assoc. 62:399-402
(1967).
14. Porter, R.A. and J.H. Christiansen, Predictions of Annual Sulfur Dioxide
Concentrations for Frankfurt An Main, FRG, Aug; 71 to July; 72,
Practical Demonstration of Urban Air Quality Simulation Models,
NATO/CCMS Air Pollution Pilot Study in print (1976).
-------
-236-
15. Bencala, K.E. and J.H. Seinfeld, On Frequenot Distributions of Air
Pollutant Concentrations, Atmospheric Environment, Vol. 10, pp. 941-
950, (1976).
-------
-237-
3.10.2 Methods for Estimating Levels Not to Be Exceeded or Not to Be
Exceeded Move Than Onae Per Year
Distributed at Conference
Richard A. Porter
The sound scientific approach to estimating compliance with standards
styled "not to be exceeded more than xxx times per xxx" is a probabilistic ap-
proach. However, none of the models that are now widely available contain
the statistical post processor necessary to make such a probability state-
ment. : Therefore, some method must be stated for estimating compliance with
such short-term standards without having the necessary statistical post
processor. There are two possible methods: 1) survey of historic data;
2) estimation of worst case from meteorological and source considerations.
Survey of Historic Data
This is the method used by RAM. The question becomes : What period of
record should be used ? Using more than one year of data causes problems with
computational time. Using only one year of data limits the number of meteor-
ological conditions examined. Practical considerations can decree the per-
iod of record examined. Less than one year of data would be unacceptable.
Estimation of Worst Case
An experienced modeler can estimate the worst case conditions when
only a few sources are being considered. Such an option is very economical
and should be available as a tool for estimating worst case.
JThe paper Predictions of Annual Sulfur Dioxide Concentrations for Frankfort
Air Main Germany, Aug. 71 - July 73. 7th Technical Meeting of the NATO/CCMS
Committee. (Porter and Christiansen, 1976), discusses the problems associated
with frequency distribution determination.
-------
-238-
3.10.3 Position on Recently Proposed Amendments to the Clean Air- Act
(Draft; prepared by the Committee on the Meteorological Aspects
of Air Pollution of the American Meteorological Society and
not yet approved as of the date of the conference by the Amer-
ican Meteorological Society Council.)
Distributed at Conference
Bruce A. Egan
The 94th Congress considered, at length, proposals to the Clean Air Act
of 1964, as revised in 1970. Many issues were discussed and presented to
Congress and its committees in consideration of these amendments; a joint
House-Senate conference bill was also prepared. Congress adjourned without
passing any amendments. The Committee on the Meteorological Aspects of Air
Pollution of the American Meteorological Society feels it necessary to estab-
lish certain issues of policy regarding the use of meteorological knowledge
in air pollution studies. Inasmuch as these policy issues were addressed in
the Clean Air Act Amendments, we feel it is important to comment on areas
where meteorological expertise is required.
Existing legislation and Environmental Protection Agency regulations
have established National Ambient Air Quality Standards (NAAQS) designed to
protect the health and welfare of the general public by not allowing air
pollution levels to exceed certain values for different averaging times.
Under these regulations, the establishment or continuation of any source of
air pollution is allowed if it can be shown that pollution from that facil-
ity will not add to the pollution burden by an amount which would cause an
excess of the NAAQS. A second type of regulation adopted by the Environmen-
tal Protection Agency addresses the emission of pollution from new industrial
and other sources. New Source Performance Standards (NSPS) require that a
facility must restrict emissions of certain criterion pollutants per unit of
production using Best Available Control Technology (BACT). Thus there are
in existence two types of standards, one which addresses ambient air quality,
and a second which requires that each new source be controlled so as to lim-
it, within prescribed amounts, its pollution emissions to the atmosphere.
The Supreme Court of the United States has determined (Sierra Club
vs. Ruckelshaus) that the Clean Air Act Amendments of 1970 empower the Envi-
ronmental Protection Agency to issue a third type of regulation to prevent
-------
-239-
the significant deterioration of air quality and in particular to preserve
existing clean air regions. Regulations were promulgated by the Environ-
mental Protection Agency in 1974 which provide for the classification of
areas within the U.S. according to allowable increments of deterioration.
In Class I (pristine areas) no significant deterioration of air quality
would be allowed. Class II areas would allow very little deterioration
and Class III regions would allow air quality to deteriorate to the NAAQS.
A number of methods or 'models' have been used historically by federal,
state, and local agencies, by industrial sources and others for calculating
the ambient air quality resulting from various configurations of air pollu-
tion sources. Mathematical models, when used responsibly with good meteoro-
logical data, have proven invaluable in analyzing the effects of new construc-
tion on ambient air quality.
This committee is particularly concerned that language in certain of
the versions of the Clean Air Amendments proposed to the 94th Congress and
in the Conference Report (House of Representatives Report No. 94-1742, Sep-
tember 30, 1976) included wording to the effect that a model or a group of
models would be designated by the Administrator of the Environmental Pro-
tection Agency to evaluate the impact of new and existing sources of emis-
sions on ambient air quality.
The term 'model', with respect to air quality considerations, has
been taken to mean a relationship, analytical or empirical, which relates
the ambient air quality as measured at a receptor to the emission of mate-
rial from sources influencing that receptor. Such relationships can be
established and have, in fact, been validated on many occasions, if meteo-
rological conditions are adequately known. While there is much active re-
search in progress to improve the current state-of-the-art of such models,
the present models do not permit accurate predictions of ambient air qual-
ity when meteorological conditions are not known. Further, many meteoro-
logical phenomena which are responsible for threats to short averaging time
standards are not amenable to simple mathematical treatment.
Our specific concern is with the concept that there should be currently
established one particular model which will be capable of analyzing all con-
ceivable situations. From a professional meteorological point of view, we
-------
-240-
feel this is not a tenable position to support. The scales of atmospheric
motion which are responsible for dispersing pollutants vary from site to
site and exhibit quite variable behavior. While meaningful averages can
generally be calculated and established from data, there is no model cur-
rently available which has a proven reliability for a range of commonly en-
countered topographical and meteorological situations of importance. The
diversity of meteorological, geographic, and site-specific conditions which
exist is such that very careful choice of a model is important and atten-
tion must be given to whether a given modeling approach adequately handles
the situation of concern. Models, after all, are simply an extension of
professional opinion and abilities on the part of the model maker. To in-
cautiously or casually apply a model to a specific situation without care-
ful meteorological and other professional consideration is not appropriate
for problems having significant economic and social implications. While
it may be desirable from a legal point of view to adopt one uniform method
for calculating air quality, it is a scientifically indefensible and un-
reasonable procedure, given the non-uniformity and complexity of the atmos-
phere, and the wide variability in meteorologically relevant geometric fac-
tors.
The above discussion has been implicitly considering the pollutants
for which non-deterioration standards are presently being proposed, namely
sulfur dioxide and particulates. When one considers the problems of mod-
eling the transport and transformation of other pollutants of concern such
as sulfates or oxidants which may involve atmospheric chemical reactions
and other physical phenomena still not well understood (for example, dis-
persion of pollutants over long distances) then the arguments are even more
compelling to not advocate the adoption of uniform modeling techniques.
The state-of-the-art simply does not justify it.
The atmosphere is a large but finite resource. Used responsibly, it
can provide a disposal mechanism for many of the pollutants of our modern
society. Responsible use, however, includes continual consultation with
professional meteorologists whose knowledge and experience will provide
guidance on the best current methods for evaluating the effects of emissions
on air quality, and who can provide even more reliable methods as our under-
standing of atmospheric phenomena increases.
-------
-241-
3.10.4 Consistency and Standardization
Distributed at Conference
Bruce A. Egan
Considerable discussion at the conference centered on the issues of
needs for consistency and standardization and the compromises inherent in
adopting standardized models.
Standardization of approach was stated as a concept which would great-
ly relieve the work load presently within the various EPA regional offices
and support facilities with respect to reviews of SIP revisions, New Source
Reviews, PSD, etc. Standardization was also stated as an objective expressed
to EPA by various industry groups who were concerned that different EPA re-
gional offices had different approaches and this may result in variations in
required emission limitations for what otherwise could be considered similar
source situations. Therefore, differences resulted simply as a matter of
location. A case where two regional offices disagreed on the methodologies
for the same source affecting both regions was cited as indicative of the need
for a consistent approach. It was stated that new sources needed to be
assured that, after construction was started, the ground rules would not
change so as to require still more stringent emission limitations. Section
318 of the September 30, 1976 House of Representatives Conference Report was
cited as requiring that a modeling conference be held and that..."special
attention shall be given to appropriate modeling necessary for carrying out
subtitle C of Title I (relating to prevention of significant deterioration
of air quality)."
Standardization of modeling approaches has the potential for creating
other problems. The present state-of-the-art does not permit the identifica-
tion of a single specific model which would be most appropriate for most
sources. Site by site analyses which include room for professional judgment
regarding choice of modeling techniques and meteorological parameters is
generally required. They are especially appropriate for the analysis of
siting and control alternatives of sources where large capital investments
are involved. The problem with the use of a specific standard model is that
if this model is non-conservative (that is, it is considered to be appropriate
for a "most probable" concentration estimate) it can be expected to err
sometimes by overpredicting and sometimes by underpredicting. Over predictions
-------
-242-
would suggest emission limitations which are beyond that called for in order
to maintain ambient air quality standards and therefore not cost-effective;
underpredictions would suggest emission limitations which would result in
concentrations in excess of the standards - therefore not responsive to EPA
regulations. Thus, a standardized approach if used to define emission
limitations which would result in concentrations close to an air quality goal
(NAAQS or PSD increments) has hidden costs associated with potential "over-
kill" and "underkill" on a site by site basis.
A second problem associated with standardization as discussed above
is that it tends to discourage the advancement of our scientific understand-
ing of air pollution phenomenon. A standardized approach would have resolved
the debate of the two regional offices cited above but the resolution might
not have satisfied either's concerns about the specific technical issues at
hand. Science advances more responsively when technical differences are
aired and a judgment passed on the basis of the technical arguments.
A solution proposed in these guidelines which meets the concerns out-
lined above is to move toward the adoption of standardized models for screen-
ing purposes. These models would be purposefully made to produce "conserva-
tive" estimates of ambient air quality levels. They would be used to iden-
tify and efficiently deal with "non-problems." If, for example, the conser-
vative model indicated that a new source would not threaten air quality goals,
a quick approval could be granted. If the model indicated a potential pro-
blem, then a refined modeling and analysis approach should be defined which
would attempt to produce more realistic estimates of the actual impact. The
required approach could involve the collection of more complete data, more
applicable (site-specific) meteorological parameterization and model valida-
tion efforts. The guideline document can then provide guidance on the choice
of generic models (basic model types) which are thought to be most appropriate
for the various source types and for various combinations of source config-
uration, local meteorology, and local topography.
Discussion concluded in the topic of providing in the guideline docu-
ment a strong statement regarding the rights of sources to choose alternative
modeling approaches for the refined types of estimates. The predictions which
result from the use of alternative models must be defensible in a scientific
sense implying adherence to known physical principles and the presentation of
-------
-243-
model validation data. A corollary to this discussion was the recognition
that if a new model were to prove superior to the screening model from an
accuracy point of view (adjusting for the built-in conservatism of the
screening model) then the new model might be a candidate for upgrading the
standardized model at a future date. A conclusion was the recognition of
the need to establish a very good data base in the public domain for purposes
of evaluating alternative modeling approaches.
-------
-244-
3.10. 5 Statistical Evaluation of Compliance
Distributed at Conference
Noel de Nevers
In evaluating the modeling results to determine compliance with short-
term AAQS (not to be exceeded more than once per year) or PSD (not to be
exceeded), the test shall'consist of computing the predicted concentration
at the worst receptor point, for each applicable period (1 hr., 3 hr., or
24 hr. where applicable) for a period of not less than years and longer
if data are available and using the most probable modeling techniques. The
resulting computed values shall be fitted by a suitable cumulative frequency
distribution function. The distribution shall be considered to demonstrate
compliance if the cumulative distribution function indicates that the second
high short term standard will not be exceeded twice in a calendar year, more
than one year in each years, and that the PSD increments will not be
exceeded more than once per years.
This version should be regarded as a draft version. The final form is
incorporated in Section 2.2.5.
-------
-245-
3.10.6 Statistical Evaluation of Compliance (Revised) *
Distributed at conference by Noel de Nevers
In evaluating the modeling results to determine compliance with
short-term AAQS (not to be exceeded more than once per year) or PSD (not
to be exceeded), the test shall consist of computing the predicted concen-
tration at the worst receptor point, for each applicable period (1 hr.,
3 hr. or 24 hr. where applicable) for a period of not less than _1 year
and longer if data are available (but generally not more than 5_ years) and
using the most probable modeling techniques. The resulting computed values
shall be fitted by a suitable cumulative frequency distribution function.
The distribution shall be considered to demonstrate compliance if the
cumulative distribution function indicates that the (second high) short
term ambient air quality standard will not be exceeded twice in a calendar
year, more than one year in each xx years, and that the PSD increments
will not be exceeded more than once per y_ years.
(The above assumes we MUST use the same format (second high) as
is used in the definition of the AAQS. We understand that the format was
chosen to correspond to the realities of ambient air quality sampling.
The realities of modeling are different. Thus for modeled compliance -
we would prefer the appropriate section of the final sentence to read
"... indicates that the short term ambient air quality standard will
not be exceeded, on the average, more than once per calendar year, and
that the PSD . . . "}
We have purposely left the x and y in the above paragraphs undefined.
We believe that a thorough study of the consequences of possible choices
of those values should be made before values are assigned.
*
This version should be regarded as a draft version. The final form
is incorporated in Section 2.2.5.
-------
-246-
3.10.7 Comment on Statistical Guide for Compliance
Distributed at Conference
W. A. Perkins
It is common engineering practice to use statistically computed cri-
teria for structural design to meet extreme meteorological conditions. Drain-
age as flood control improves and wind loads on structures are two examples.
Appropriate meteorological parameters are treated by special statistical
procedures to estimate conditions for specified return periods ranging from
10 to 100 years (10% to 1% of the distribution function). These estimates
often exceed the period of record by a factor of two or more.
Meteorological conditions required for atmospheric transport models
are very different from those used in the above examples. However, the con-
cept of using a statistical approach to determine compliance is analogous to
established practice.
-------
-247-
3.10.8 Validity of Hour-by-Hour Estimates of Air Quality (The Use of the
Nowoast in Pollution Potential Forecasting)
Distributed at Conference
A. E. Boyer
A practice developed in pollution potential forecasting has been to
analyze current conditions so as to prepare a consistent picture of air
quality and meteorological data in map format. This process is very similar
to the meteorological exercise called "map analysis." In the case of the
weather map, selected reports of pressure, temperature, wind, precipitation,
and other variables are used to estimate the continuous initial distribution
of air mass and frontal characteristics.
In the case of pollution potential forecasting, an exercise referred
to as the "nowcast" consists of analyzing initial values of wind velocity and
shear, continuous temperature gradients, stabilities, along with measured and
simulated air quality values to estimate the continuous initial state of
velocity and thermal structure of the atmosphere.
The nowcast is an attempt to use a limited amount of air quality and
meteorological data in the vicinity of a pollution source to present a snap-
shot or analysis of the continuous distribution of air quality and meteor-
ology in three dimensions. It is based on the concept that if one is given
the height of an elephant at eye level, it is likely that one could estimate
other dimensions of the elephant. In a similar fashion, given a limited
number of meteorological, source, and air quality measurements, a skilled
meteorologist using mathematical simulation models can fill in the gaps
between the reporting points. This initial state analysis is then the basis
of decisions for the need for environmental control in the coming hours. It
is a corollary of meteorological predictions that no prediction can be more
detailed or more accurate than the initial state analysis upon which it is
based. It is for this reason that the concept of the nowcast or of the
initial state analysis of air quality and meteorology is so important.
-------
-248-
3.10.9 A Comment Concerning the Use of Unverified Models as "Relative
Predictions "
Post-conference contribution submitted by Philip M. Roth
Oftentimes there is doubt as to the accuracy of a particular model
as an "absolute predictor" due to shortcomings in its formulation. On
occasion such models are used instead to estimate differences in concentra-
tion between a base case and a case of interest. Insofar as we can see,
there is no evidence to support the hypothesis that a model whose absolute
predictive capability is in doubt will, in general, perform more reliably
when used as a relative predictor. Hence, those notes of caution raised
concerning the use of models as absolute predictors, i.e., the need for
model evaluation and verification for a variety of relevant conditions,
should apply when these models are used as relative predictors as well.
-------
-249-
3.10.10 Comments on Various Issues
134 BROADWAY. AMITYVILLE, N. Y. 117O1 TEL: 516-691-3395
March 25, 1977
Dr. John J. Roberts
Deputy Division Director
Energy and Environmental Systems
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
Dear Dr. Roberts:
This letter summarizes several ideas which we consider
pertinent to the issues raised at the recent Specialist's Con-
ference on the EPA Modeling Guidelines. We would like to have
it included in the Appendix.
1. Screening Procedures
Over the past decade the effort required for the approval
of any facility emitting pollutants has grown enormously.
Much of this effort is good because it represents genuine
concern about environmental problems, but some is totally
unnecessary paperwork and computation. We strongly favor
the use of screening techniques to simplify approval proce-
dures .
As recommended by Working Group 1-2, it should often be
possible to reach a "yes-no" decision about a new facil-
ity from a limited set of calculated hourly mean concen-
trations. These should cover a wide range of hypotheti-
cal wind and stability conditions, but need not involve
any site data. If none of these test calculations exceed
a specified concentration level, the facility could be
accepted without further study.
To apply such a system one should know how hourly values
are related to 3-hourly and 24-hourly concentrations.
The recent paper by Martin and Reeves* should be helpful
to EPA in developing such a screening procedure. They
show the following ratio.s apply for a typical group of
power plants:
* Martin, J. R. and Reeves, R. W. , Relationships Among Observed
Short-Term Maximum Sulfur Dioxide Concentrations Near Coal-Fired
Power Plants. 1977 Annual APCA Meeting (in press).
-------
-250-
Dr. John J. Roberts
March 25, 1977
Q 77
= n 28
Maximum 3-Hour Concentration
Maximum 1-Hour Concentration
Maximum 24-Hour Concentration
Maximum 1-Hour Concentration
Maximum 24-Hour Concentration _ Q 37
Maximum 3 -Hour Concentration
2. Air Quality Standards
Although we understand why EPA is reluctant to revise
the established air quality standards, a number of the
Conference participants favored a change in expressing
the standards so that more reasonable and reliable
statistics could be employed. Revising the air quality
standards so that maximum values are not involved would
be a great improvement for the following reasons :
a. The maximum or second maximum, whether computed or
observed, is an inherently controversial quantity.
One can always argue that the examination of a
longer period of record or a more adverse set of
assumptions would reveal a higher concentration than
had been discovered to date.
b. A standard based on extreme values tends to involve
spurious data. Improper recorder operation, computer
errors or even one untenable assumption may suggest
an apparent violation where none exists.
c. Protecting against an absolute maximum is not a real-
istic objective. All human activity involves risk
and it is reasonable to balance the severity of the
event against the probability of occurrence. Total
elimination of risk is an unreasonable objective.
There is no technical reason why the short-term standards
cannot be redefined so that the limiting values fall with-
in the normal statistical population. Careful choice of
new limiting values would still prevent the existing max-
imums from being exceeded very often, but it would elimi-
nate the controversial "never" concept.
For example, one might specify in a new 3-hour ly SC^ stand-
ard that a concentration of 200 /ig/m^ should not be exceeded
in more than 1% of the cases. Based on typical probability
distributions this would insure that a l,300Aig/m3 level
would seldom be exceeded.
-------
-251-
Dr. John J. Roberts March 25, 1977
3 . Alternative Models or Algorithms
The participants emphasized the basic conflict between
the need for more consistent regulatory treatment and
the need to retain flexibility. In striving for con-
sistency, EPA Guidelines should not stifle alternative
approaches to solving air quality and emission problems.
If they do, certain situations would be treated unrea-
sonably, and both research and imaginative review would
be discouraged.
On the other hand, capricious acceptance or rejection
of alternatives at the state and regional levels could
destroy regulatory consistency, a primary objective of
the Guideline.
We recommend a review procedure for alternative algorithms
in which the systems would be evaluated by the EPA techni-
cal staff in Raleigh. Once an alternative system has been
approved, it would no longer be necessary to justify its
use for each subsequent application in any region, " other
than to assure that the system is appropriate for a par-
ticular problem.
The term "equivalent" should not have the same stringent
limitations as would be true of air quality measurements.
Diffusion modeling is much less precise, and there is no
certainty that a given EPA algorithm is significantly bet-
ter than some other alternatives. The limits of accept-
ability should therefore be correspondingly broad.
4. Rollback
The initial reaction of most of the Conference participants
was to reject rollback entirely as a modeling technique.
It seems to us, however, that rollback still has a place in
control strategy. It is appropriate when applied to a large
group of very similar sources which cannot be modeled or con-
trolled in any other way. Residential heating and automo-
biles are good examples of this type of source.
Rollback should not be used as a substitute for a more
erudite analysis merely because the problem is complex, the
modeling is difficult, or because it is time-consuming.
Application in these circumstances is precisely what has
given rollback a bad name.
5. Background Concentrations
The definition of background concentrations is still an
unsettled problem. However, there are serious flaws in
-------
-252-
Dr. John J. Roberts March 25, 1977
two methods which are frequently proposed.
a. For an isolated source, or closely-packed group of
sources, the proposed background is the maximum
concentration at a nearby air quality station
determined when the source cannot affect it. This
definition suffers from the same drawbacks as the
maximum in an air quality standard. Instrument
problems and very unusual meteorological circum-
stances may be responsible for this single value.
It is erroneous and overly conservative to assume
that this maximum background exists throughout the
region and on all time scales. A significantly
lower value should be selected.
b. The draft Guideline suggests an optional method to
determine background when no monitors are near the
source. This method consists of using as background
an average monitored concentration from sites in
similar topographic and climatological settings .
The flaw in this technique is that the average moni-
tored concentration at one location is influenced by
its own local sources. It does not represent back-
ground even at the point of measurement. Transferring
this concentration from one location and calling it
the background at another site compounds errors .
6. Use of Measured Air Quality Data
Accurate air quality data, regardless of the agency re-
sponsible for measurement, must play a more important role
in assessing compliance with standards and in model vali-
dation.
As mentioned earlier, the frequency distributions of air
quality data are more valuable than isolated maximums .
They provide a more complete picture of the impact of a
pollutant on the area. Distribution curves will reflect
the effect of emission changes more faithfully than maxi-
mum values, since they are much less sensitive to minor
variations in meteorology or measurements.
Another important use for the data distributions is in
the validation of diffusion models. Comparing the pre-
dicted and observed concentration distributions is more
-------
-253-
Dr. John J. Roberts
March 25, 1977
effective than relying upon matching the observed and
predicted maximums , especially for short-time scales.
Sincerely yours,
Mark L. Kramer
"John R. Martin
MES/MLK/JRM:rg
-------
ERT
-254-
3.10.11 Corrments on the Use of Proprietary Models and Some Examples
Post conference submission by B. Egan
ENVIRONMENTAL RESEARCH &TECHNOLOGY, INC.
CONCORD,MASS. • CHICAGO • LOS ANGELES • WASHINGTON. DC
REF: AQSD-1428
25 April 1977
Dr. John J. Roberts
Argonne National Laboratory
9700 South Cass Avenue
Argonne, Illinois 60439
Dear Dr. Roberts:
I would like to offer some supplementary material relating to the
issue of proprietary models and the inclusion of descriptions of models
in the Report of the Specialists Conference on the EPA Modeling Guideline.
ERT feels that it will be very important for EPA to clearly articulate
its position regarding its intended exclusion of proprietary models in
the Modeling Guideline recommendations. In particular, a clarification
should be made regarding whether the intent is to discourage the use of
proprietary models by EPA regional offices or whether the intent is more
generally to not accept the results of analyses performed by anyone on
the basis of proprietary model computations.
As a consulting firm with business in the areas of model development
and application, ERT is concerned about the implications of the latter
possible intention. We feel strongly that, just as patents and copyrights
protect investments in other commercial areas, establishing a proprietary
status to models or algorithms which have commercial value, provides a
means of encouraging and protecting private investments in the development
of advanced modeling techniques. Our company, for example, over the
years has invested well over one-half million dollars in internal development
projects relating to model development. We believe that private investments
have historically contributed substantially to advancing the state-of-
the-art of modeling and we, therefore, wish to discourage EPA from
"legislating away" incentives for further advancing our capabilities in
the area of modeling air pollution impacts. In our opinion, proprietary
status does not at all preclude the possibility that a model (including
computer code) can be reviewed and tested in detail by concerned parties.
Having stated our feelings about the principles involved, as a
matter of practice we recognize the potential benefits of having a
number of our models more widely appreciated. We are, therefore, offering
for inclusion in the Report of the Specialists Conference, the enclosed
descriptions of models which ERT currently utilizes in air quality
studies. We have summarized the main features of four of the models in
accordance with the format Argonne used in the referenced document:
Descriptions of Air Quality Models and Abstracts of Reference Materials.
696 \flRGINA ROAD CONCORD. MASSACHUSETTS 01742 (617) 369-8910 TELEX 923 335 ENVIRORES CNCM CABLE; ERTCON
-------
Dr. John J. Roberts 27 April 1977
REF: AQSD-1428 -255-
Versions of each of the models have been considered non-proprietary by
ERT in the past. If it becomes a prerequisite for inclusion in the
modeling guideline, ERT would offer the current versions as non-proprietary
also.
I would suggest that the ERTAQ description be included in the
section on multiple-source, set 1 pollutants: the ARTSIM and LAPS models
be included in the section on set 2 pollutants and EGAMA be included in
the Long Range Transport and Loss Mechanism
For more general interest we have enclosed a tabular description of
computer models utilized by ERT for a variety of applications. An
important point is the recognition that a number of important environmental
assessment problems require special models which include detailed treatment
of effluents as a function of unique source configurations or thermodynamics
(e.g. cooling towers, gas turbines). Other applications, such as SCS
(Supplementary Control Systems) feasibility studies, require specialized
programs which utilize system operational input data and provide output
information which includes considerations of a variety of engineering
and operational constraints.
I hope these materials will be of use to you in the final report.
jrely yours,
CAJL*S^
A. Egah" /
Manager
Air Quality Studies Division
/mk
Enclosures
Section 3.1.4
+Sections 3.3.4 and 3.3.5
Section 3.4.1
-------
-255-
TAHULAR SUMMARY OF SONlli KliY ERT, INC. AIR QUALITY M011HLS
Model
Key Elements/Features
Typical Applications
ERTAQ
• Gaussian Plume in Vertical
• Sector Crosswind Averaging
• Climatological Wind Rose Input
• Briggs Final Plume Rise
• Turner (Pasquill/Gifford) o
• Point, Line and Area Sources
• Climatological Mixing Depth
• Regional Studies
• "Background" AQ
• Seasonal/Annual Averages
• Arbitrary Source-Receptor
Geometry
• Level Terrain Only
• Low Level Sources
• Transportation Links
PSDM
• Double Gaussian Plume (ASME
coefficients)
• One or Two Point Source Sites
Only (£ 10 stacks per source)
• Sector Centerline Concentra-
tions Only
• "Weather Condition" Inputs
Designed for stack height
determination
Isolated Source DEC
Feasibility Studies
Identifications of "Worst
Case" Meteorology
Identification of Maximum
Ground Level Concentrations
EGAMA
• High Resolution Numerical
Simulation Model based on
Conservation of Mass Equation
• Eddy Diffusivity (K Theory)
vs. Gaussian Dispersion
• Time and Spatially Variable
Meteorological Inputs Allowed
• Requires independent determina-
tion of flow fields (Done
internally for Highway Appli-
cations)
• Forward time-step "Marc' ing"
Computation Routine
• Highway Impact Studies
(Especially well suited for
near roadway AQ estimates)
• Special Studies:
Downwash
Street Canyon
Terrain Effects
Fumigation
• Research Tool
Long range transport
Chemical Transformation
Sea/Lake Breeze Flows
ATRAJBOX
Air Trajectory Transport
Model with well-mixed BOX
approach
Linear Chemistry and Depo-
sition processes
Tracks transport of air from
its origin to receptor for a
specified time
• Regional Transport of S02
and SO.
• Ideal for predicting darly
mean concentrations at a
receptor for an episode
-------
-257-
TABULAR SUMMARY OF SOME KEY I-.RT, INC. AIR QUALITY MODELS continued
Model
Key Elements/Features
Typical Applications
COOLTOKR
• Based on Convective Rise of
Cumulus Cloud Formations
• Moisture Parametrization, Phase
Change and Plume Merger
Considerations
• Meteorological Data Inputs
Cooling Tower Studies: Visible
Plume Length, Icing and Fogging
Potentials
DEPOT
• Drift Deposition Model which
Includes Consideration
of Plume Dynamics, Droplet
size and Mass Distributions,
evaporation effects on drop
fall velocity
• Cooling Tower Salt Deposition
Predictions
PSDM-2
• PSDM modified to incorporate
fall velocity and ground
deposition rate
Particulate AQ studies
FUMIG
• Based on EGAMA and Bounda -y
Laver Parameterization
• Time dependent inversion
breakup fumigation
• Sea Breeze Fumigation
SULFA3D
• Three-Dimensional Adveci-ive-
Diffusive Grid Model
• Linear Deposition and Transfor-
mation processes with first
order Chemistry
• Long range regional transport
• Regional Sulfate Predictions
MONITOR
• Dispersion Model inputs used
to rank useful 1'ness of
alternative monitoring
locations
Optimal Monitoring Site
Location
PROBL
• Dispersion Model inputs
used to generate statistics
for probability analysis of
• SCS Feasibility Studies
-------
-255-
TABULAR SUMMMIY OP SOM1I KI-Y URT, INC. AIH QUALITY MODELS continued
Model
Key Elements/Features
Typical Applications
RATE BOX
• One-Dimensional Sector/Square
Model with well-mixed BOX
Approach for Point/Urban Area
Sources
• Finite-Difference Numerical
Solution
• Includes Washout and Rainout
Point or Urban Scale Sulfate
Predictions
Ideal for local SO-, SO. and
H2S04
DIFDEP
(Diffu-
sion/Dep-
osition)
CRSVAL
DIFKIN
• Same as Plume with TWO Tracer
Equations for SO. and SO.
• Linear Chemistry and Removal
Processes
Point Source Sulfate Problems
Acid Rain Predictions
• Modification of EPA CRSTER
Models for Complex Terrain
• Incorporates soi:e features /
of EPA-Valley Model
• Gaussian Plume r.oncentration
profile with Multiple Images
Estimates of hourly ground
level concentrations due to
large point source in areas
of complex terrain
National and State Ambient
Air Quality Standards Studies
• Finite Difference solutions
based on mass conservation Eqn.
• Chemical kinetics and the
upward spread through a
series of vertical cells
• Vertical diffusion coefficients
Regional maximum oxident
concentrations
"Background" 0, concentration
Transportation strategy
evaluation of 0- concentration
ASTEC
• Gridded Area Sources
• Gifford/Hanna Concept
«
• Computationally Inexpensive
• Regional Model (Flat Terrain)
• Multiple Area Sources (Homo-
geneous Emission Geometry)
PLUME
• Rising Plume Deflected by
Crosswind
• Morton, Turner and Taylor
Local Similarity Hypothesis
• Conservation Equations and
Entrainment Coefficients
• Meteorological Sounding Data
Inputs
• Gas Turbine Plume Studies
• Unconventional Stack Geometry
or Effluent Characteristics
• Final Plume Rise Predictions
-------
-259-
3.21 Descriptions of Air Quality Models
This section is an exact reproduction of Sections 1 and 2 of the
notebook entitled "Descriptions of Air Quality Models and Abstracts of
Reference Materials" prepared by ANL staff and distributed to the parti-
cipants prior to the conference. A supplement to that notebook was prepared
and distributed at the conference. It has also been reproduced here.
Section 3 of the notebook, which has not been reproduced here, contained
abstracts of the references cited in the draft Guideline.
-------
-260-
-------
-261-
SECTION 1
GENERAL
-------
-262-
Introduction
This section contains two outlines, the first listing the aspects of
atmospheric dispersion which simulation models must treat and the second
providing a list of features defining applications to which models are applied.
The first outline of model characteristics provides the framework for
the model descriptions in Section 2. It also provides a common basis for
comparing and evaluating models.
The second outline of features of model applications is intended to aid
in the identification of specific features characterizing situations related
to various issues to be addressed.
-------
-263-
Model Characteristics
I. General considerations
A. Method as implemented
B. Basis for parameterization
II. Specific elements
A. Emission characteristics
1. Source-receptor relationship
a. Downwind distance
b. Orientation
c. Release height-elevation
d. Horizontal location
e. Ground level vs. elevated
2. Emission rate
a. Spatial variation
b. Temporal variation
3. Chemical composition of emissions
B. Transport characteristics
1. Plume behavior
a. Plume rise
b. Fumigation - inversion breakup
c. Downwash
d. Looping
e. Trapping
2. Horizontal wind field
a. Shear (variation in speed and direction)
b. Periodic variations (seasonal, diurnal)
c. Terrain effects (e.g., channeling)
d. Persistence
3. Vertical wind speed
a. Terrain effects
b. Lake breeze
4. Horizontal dispersion
a. Stability
b. Surface roughness
c. Averaging time as compared to transport time
5. Vertical dispersion
a. Stability
b. Surface roughness
c. Mixing height
d. Elevated vs. ground level sources
C. Removal and transformation
1. Chemistry and reaction mechanism
a. Secondary production
b. Chemical removal
2. Physical removal
a. Dry deposition
b. Precipitation scavenging
c. Resuspension
-------
-264-
D. Background
1. Long term
2. Short term
3. Directional dependence
E. Initial conditions
F. Boundary conditions
1. Ground
2. Mixing height
3. Sides in complex terrain
G. Correlations
1. Emissions
2. Meteorological parameters
III. Validation
A. Sensitivity analysis
B. Field studies
IV. Calibration - confidence limits
V. Requirements for implementation
A. Resource requirements
1. Personnel
2. Monetary
3. Computer
B. Data requirements
VI. Output
A. Averaging time
B. Spatial resolution
C. Source culpability list
D. Frequency distribution
E. Special output (e.g., amount deposited
-------
-265-
Features of Model Applications
I. Pollutant characteristics
A. Production
1. Primary
2. Secondary
3. Resuspension
B. Removal
1. None
2. Chemical
3. Physical
C. Chemical identity
1. Single substance
2. Mixture
II. Averaging time
A. Long-term: (month, season, year)
B. Short-term: 1-24 hrs. (1, 3, 8, 24)
III. Source characteristics
A. Number
1. Single
2. Multiple
B. Geometry
1. Point
2. Area
3. Line
4. Combination
C. Release height
1. Ground level
2. Elevated
D. Temporal
1. Constant
2. Time-varying
IV. Transport characteristics
A. Geography
1. Simple terrain
2. Complex terrain (rough terrain, lake breeze)
3. Urban
4. Rural
B. Range
1. Short
2. Meso
3. Long
C. Transport time
V. Output
A. Mean
1. Specific location
2. Arbitrary locations or isopleths
B. Frequency distribution
1. Specific location
2. Arbitrary locations or isopleths
C. Maximum at specific receptor or overall maximum
VI. Activity
A. Overall planning
B. Source-specific review
-------
-266-
-------
-267-
SECTION 2
DESCRIPTIONS OF MODELS
-------
-268-
Introduction
This section provides brief descriptions of the models suggested for use
in the guidelines. Rollback has not been included. A brief discussion of the
limitations of rollback is provided in Reference 30 in Section 3. The format
of each description follows the outline of model characteristics given in
Section 1. The descriptions are in alphabetical order.
-------
-269-
APRAC-1A
APRAC-1A
Reference: Nos. 34, and 35 in the guideline.
Ludwig, F.L. and R.L. Mancuso. "User's Manual for the APRAC-1A
Urban Diffusion Model Computer Program." Prepared for Division
of Meteorology, Environmental Protection Agency, under Contract
CAPA-3-68C 1-69) (NTIS PB 213091), Research Triangle Park, North
Carolina 27711, September 1972.
Ludwig, F.L. and W.F. Dabberdt. "Evaluation of the APRAC-1A
Urban Diffusion Model for Carbon Monoxide." Prepared for Di-
vision of Meteorology, Environmental Protection Agency, under
Contract CAPA-3-68 (1-69) (NTIS PB 210819), Research Triangle
Park, North Carolina 27711, February 1972.
Abstract:
Equations
APRAC is a model which computes hourly average carbon monoxide
concentrations for any urban location. The model calculates
contributions from dispersion on various scales: extraurban,
mainly from sources upwind of the city of interest; introurban,
from freeway, arterial, and feeder street sources; and local,
from dispersion within a street canyon. APRAC requires an
extensive traffic inventory for the city of interest.
UL
%tn 22. £
32 kn\ -h> 1000 km
i-b-- /.(>.. i
r * / __!U — l*i — I
J uau \ f-bf /
o
-------
-270- APRAC-lfl
~$-l\ u)livcU/ X = WI'I.CA
^ = ketaU-
-------
.271- APRAC-1A
APRAC-1A
A. Source-Receptor Relationship
User specifies set of traffic links (line sources) by providing link
endpoints, road type, daily traffic volume
The traffic links may have arbitrary length and orientation
Off-link traffic allocated to 2 mi x 2 ml grid
Link traffic emissions are aggregated into a receptor oriented area
source array
The boundaries of the area sources actually treated are 1) arcs at
radial distances from the receptor which increase in geometric
progression, 2) the sides of a 22.5 sector oriented upwind for
distances greater than 1000 m, and 3) the sides of a 45 sector
oriented upwind for distances less than 1000 m.
A similar area source array is established for each receptor
Sources assumed at ground level
Up to 10 receptors
Receptors at ground level
Receptor locations are arbitrary
Four internally defined receptor locations on each user-designated
street are used in a special street canyon sub-model
B. Emission Rate
Daily traffic volume for each link and off-link grid square is input
and modified by various factors to produce hour-by-hour emissions
from each link
Link emissions aggregated as described above: sector area
source contributions obtained analytically
Off-link traffic emissions on 2 mi grid are added into sector area
sources
In street canyon sub-model, a separate hourly emission rate is
provided by user for the link in question
C. Plume Behavior
Does not treat plume rise
Does not treat fumigation or downwash except in street canyon sub-model
In street canyon sub-model, a helical circulation pattern is assumed
D. Horizontal Wind Field
Hourly wind speed and direction in tens of degrees is input
No variation of wind speed or direction with height
Constant, uniform (steady-state) wind assumed within each hour
E. Vertical Wind Speed
Assumed equal to zero except in street canyon sub-model
Helical circulation assumed by street canyon sub-model
-------
.272- APRAC-1A
F. Horizontal Dispersion
Sector averaging uniform distribution within sectors
22.5 sectors beyond 1 km
45.0 sectors within 1 km
G. Vertical Dispersion
Semi-empirical/Gaussian plume
6 stability classes; stability class determined internally from user-
supplied meteorological data [modified from Turner (1964)]
Dispersion coefficients from McElroy and Pooles (1968), modified
using information in Leighton and Ditmar (1953)
No adjustments made for variations in surface roughness
Downwind distance variation of a assumed to be ax for purposes of
doing analytic integration
In street canyon sub-model, empirical function of wind speed and street
width and direction is used
H. Chemistry/Reaction Mechanism
Not treated
I. Physical Removal
Not treated
J. Background
Box model used to estimate contribution from upwind sources beyond
32 km based on wind speed, mixing height, annual fuel consumption
In street canyon sub-model, contribution from other streets is
included in background
K. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: perfect reflection; ignores effect until concentration
equals that calculated using box model; uses box model (uniform
vertical distribution) thereafter
Mixing height determined from morning radiosonde data as follows:
midnight to dawn: constant at pre-dawn value obtained using
minimum urban temperature
dawn to sunset: afternoon maximum temperature used to obtain
maximum height; hourly values obtained from surface temperature
variations
sunset to midnight: linear interpolation with time
L. Emission and Meteorological Correlation
Emissions a function of hour of the day and day of the week
Meteorological parameters are functions of hour of the day
-------
-273- APRAC-1A
M. Validation/Calibratlon
No calibration option provided
Some documented validation experience available
N. Output
Hourly concentration values at each receptor
Frequency distribution based on hourly values can be obtained
-------
-274-
AQDM
AQDM
Reference: No. 26 in the guideline.
TRW Systems Group. "Air Quality Display Model." Prepared
for National Air Pollution Control Administration under
Contract No. PH-22-68-60 (NTIS PB 189194), DHEW, U.S. Public
Health Service, Washington, D.C., November 1969.
Abstract:
Equations:
The Air Quality Display Model (AQDM) is a climatological
steady state Gaussian plume model that estimates annual
arithmetic average sulfur dioxide and particulate concentra-
tions at ground level. A statistical model based on Larsen
(1969) is used to transform the average concentration data
from a limited number of receptors into expected geometric
mean and maximum concentration values for several different
averaging times.
Sources '•
k-l f-l
W
X ^
u,L
C=
«//S-f«nc€ be-kueen reiepb
low ha*
/c center-tine
-------
-275-
AQDM
AQDM
A. Source-Receptor Relationship
Arbitrary location for each point source
Arbitrary location and size for each area source
Up to 225 receptors located on uniform rectangular grid
Up to 12 user-specified receptor locations
Unique release height for each point, area source
Unique separation for each source-receptor pair
Receptors at ground level
No terrain differences between source and receptor
B. Emission Rate
Point sources: single rate for each source
Area sources: single rate for each source
Each source treated by effective single point source
approximation
No temporal variation allowed
C. Chemical Composition
Treats one or two pollutants simultaneously
D. Plume Behavior
Holland (1953) formula, with adjustment for stability
No olume rise calculated for area sources
Does not treat fumigation or downwash
If stack height plus plume rise is greater than mixing height,
ground level concentration assumed equal to zero
E. Horizontal VHnd Field
Climatological approach
16 wind directions
6 wind soeed classes
No variation in windspeed with height
Constant, uniform (steady-state) wind assumed
F. Vertical Wind Speed
Assumed equal to zero
G. Horizontal Dispersion
Climatolonical approach
Linear interpolation between 22.5° sector renter!ines;
center value calculated by sector averaging procedure
(narrow plume approximation)
Averaging time = 1 month - 1 year
-------
-276- AQDM
H. Vertical Dispersion
Semi-empirical/Gaussian plume
5 stability classes (Turner, 1964)
Neutral stability split internally into 60% day, 40% night
Dispersion coefficients from Pasquill (1961) and Gifford (1961)
Neutral dispersion coefficients used for all neutral and stable classes
No provision for variations in surface roughness
I. Chemistry/Reaction Mechanism
No provision for treatment
J. Physical Removal
No provision for treatment
K. Background
Input single constant background value for each pollutant
L. Boundary Conditions
Lower boundary (ground): perfect reflection
Upper boundary (mixing ht): no effect until a ^.47H
(occurs at x=x.) z
for x > 2x, uniform mixing
in between-linear interpolation
transition region used
M. Emission and Meteorological Correlation
Hind speed, direction, stability
correlated via wind rose
Emission rate - not correlated
with any other factor
Non-sequential (climatological)
limited correlation
Mixing height adjusted according to stability class:
Class A - 1.5 x afternoon climatological value
Class D (night, internally divided)
average of 100 meters and afternoon cl imatological value
Class E - assumes 100 meters
N. Vali dation/Calibration
Calibration option available
Substantial experience but limited documentation
-------
-177-
0. Output
1 month - 1 year averaging time simulated (arithmetic mean only)
Arbitrary averaging time by Larsen (1969) procedure
(typically 1-24 hr)
Assumes
1) lognormal concentration distribution,
2) power law dependence of median and maximum
concentrations on averaging time
Up to 225 gridded receptor locations, 12 arbitrary locations
Individual point, area source culpability list for each receptor
-------
-278-
CDM
COM
Reference: No. 27 in the guideline.
Busse, A.D. and J.R. Zimmerman. "User's Guide for the Climato
logical Dispersion Model." Publication No. EPA-RA-73-024
(NTIS PB 227346/AS), Environmental Protection Agency, Research
Triangle Park, North Carolina 27711, December 1973.
Abstract: The Climatological Dispersion Model (COM) is a climatological
steady-state Gaussian plume model for determining long-term
(seasonal or annual) arithmetic average pollutant concentra-
tions at any ground level receptor in an urban area.
A statistical model based on Larsen (1968) is used to trans-
form the average concentration data from a limited number
of receptors into expected geometric mean and maximum con-
centration values for several different averaging times.
Equations : lA <, (,
StcWk
0.8 L
* cix
-------
-279-
CDM
COM
A. Source-Receptor Relationship
Arbitrary location for each point source
Area sources equal uniform grid squares
Receptor location arbitrary
Arbitrary release hts for pt. and area sources
Unique separation for each source-receptor pair
Receptors are at ground level
No terrain differences between source/receptor
B. Emission Rate
Point sources: single rate for each source
Area sources: single rate for each source
area integrations are done numerically one 22.5° sector
at a time; sampling at discrete points defined by
specific radial and angular intervals on a polar
grid centered on the receptor
Day/night variations in emissions, same variation assumed
for all sources.
C. Chemical Composition
Treats one or two pollutants simultaneously
D. Plume Behavior
Only Briggs (1971) neutral/unstable formula used.
If stack height + plume rise is greater than mixing height,
ground level concentrations assumed equal to zero.
Alternative to Briqgs - input value of plume rise times wind
speed for each point source.
No plume rise calculated for area sources.
Does not treat fumigation or downwash
E. Horizontal Wind Field
Climatological approach
16 wind directions
6 wind speed classes
Wind speed corrected for release height based on power law
variation exponents from DeMarrais (1959)
Constant, uniform (steady-state) wind assumed
F. Vertical Mind Speed
Assumed equal to zero
-------
-280- CDM
G. Horizontal Dispersion
Climatological
Uniform distribution within each of 16 sectors
(narrow-plume approximation)
Averaging time = 1 month to 1 year
H. Vertical Dispersion
Semi-empirical/Gaussian plume
5 stability classes as defined by Turner (1964)
Neutral stability split into day/night cases on input
Dispersion coefficients taken from Turner (1970)
Area sources - stability class is decreased by 1 category from
input values (to account for urban effects)
Neutral dispersion coefficients are used for all neutral
ajid stable classes.
No provision for variations in surface roughness
I. Chemistry/Reaction Mechanism
Exponential decay, user-input halflife
J. Physical Removal
Exponential decay, user-input halflife
Always applies the same rate constant
K. Background
Input single constant background value for each pollutant.
L. Boundary Conditions
Lower boundary (ground): assumes perfect reflection
Upper boundary (mixing height): no effect until dispersion
coefficient equals 0.8 of the mixing height, uniform vertical
mixing assumed beyond this point.
M. Emission and Meteorological Correlation
Wind speed, direction, stability correlated via wind rose
Mixing height is adjusted according to stability class:
Class A - 1.5 x afternoon climatological value
Class D (night) - average of morning and afternoon climatological values
Class E - morning climatological value
Emission rates: day-night variation allowed; all sources vary
by same factor
Non-sequential (climatological) limited correlation
-------
CDH
N. Validation/Calibration
Limited validation experience
Calibration option available
0. Outnut
One month to one-year averaging time simulated (arithmetic
mean only)
Arbitrary averaging time by Larsen (1969) procedure
(typically 1-24 hr.)
Assumes
1) lognormal concentration distribution,
2) power law dependence of median and maximum
concentrations on averaging time
Arbitrary number and location of receptors
Individual point, area source culpability list for each receptor
Point, area concentration rose for each receptor
-------
-282-
CRSTER
CRSTER
Reference: No. 13 in guideline.
Abstract:
Environmental Protection Agency. "A User's Guide to the
Single Source (CRSTER) Model." Office of Air Quality Plan-
ning and Standards, Research Triangle Park, North Carolina
27711, 1977. (In preparation)
CRSTER is a steady state Gaussian plume technique applicable
in uneven terrain. The purpose of the technique is to:
1) determine the maximum 24-hour concentration from a single
point source of up to 19 stacks for one year, 2) to deter-
mine the meteorological conditions which cause the maximum
concentrations, and 3) to store concentration information
useful in calculating frequency distributions for various
averaging times. The concentration for each hour of the
year is calculated and midnight - to - midnight averages
are determined for each 24-hour period.
Equations:
T 4= 1,6
X=
> (.6 L
« O (
cUss 7
L =
V4 - (stock
f
ba.se of
-------
-283- CRSTER
CRSTER
A. Source-Receptor Relationship
Up to 19 point sources, no area sources
All point sources assumed at the same location
Unique stack height for each source
Receptor locations restricted to 36 azimuths (every 10°) and
5 user-specified radial distances
Unique topographic elevation for each receptor; must be less
than stack height
Receptors must be at ground level
B. Emission Rate
Unique average emission rate for each source
Monthly variation in emission rate allowed
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1971, 1972) final plume rise formulas
Does not treat fumigation or downwash
If plume height exceeds mixing height, concentrations further
downwind assumed equal to zero
E. Horizontal Wind Field
Same as RAM
F. Vertical Wind Speed
Assumed equal to zero
G. Horizontal Dispersion
Semi-empirical/Gaussian plume
7 stability classes used; Turner (1964), Pasquil (1961)
Class 7: extremely stable, elevated plume assumed not
to touch the ground
Dispersion coefficients from Turner (1969); no further adjustments
made for variations in surface roughness, transport or
averaging time
H. Vertical Dispersion
Semi-empirical/Gaussian plume
7 stability classes
Dispersion coefficients from Turner (1969); no further adjustments
made
-------
-294- CRSTER
I. Chemistry/Reaction Mechanism
Not treated
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection at the same height as the receptor
Upper boundary: perfect reflection
Multiple reflections handled by summation of series until a = 1.6 x
mixing height z
Uniform vertical distribution thereafter
Mixing height is constant and follows topographic variations:
Taken from base of stack for determining whether plume punches through
Taken from receptor elevation for determining vertical concentration
distribution
Hourly mixing height obtained from radiosonde data using same
interpolation algorithm as RAM
M. Emission and Meteorological Correlation
Same as RAM
Monthly emission variation allows limited emission -
meteorology correlation
N. Validation/Calibration
No calibration option provided
Comparison with observations around at least 5 separate power plants
have been made
Additional work in progress
0. Output
Highest and second highest 1-hour and 24-hour concentrations at each
receptor for the year plus the annual arithmetic average at each
receptor
For each day in the year, the highest 1-hour and highest 24-hour
concentration values found in the field of receptors
Hourly concentrations for each receptor are output onto magnetic tape
for further processing, for example to obtain the frequency
distribution
-------
-285-
HANNA-GIFFORD
HANNA-GIFFORD
Reference: Numbers 28, 29 in guidelines.
Hanna, S.R. "A Simple Method of Calculating Dispersion from
Urban Area Sources." J. Air Pollution Control Assn., Vol. 21,
No. 12, pp. 774-777, December 1971.
Gifford, G.A., and S.R. Hanna. "Modeling Urban Air Pollution."
Atmospheric Environment, Vol. 7, pp. 131-136, 1973.
(also S.R. Hanna, private communication*)
Abstract: Two slightly different versions of the same model are de-
scribed. The first is a purely deterministic version which
does not require any local air quality data for its imple-
mentation. The form of the second version may be derived
from the first on the assumption that area source emissions
are uniform. For actual use however, it is recommended that
local air quality data be used to, in essence, calibrate the
model .
Equations
a,V>
See also Section 3.9.1.
-------
-286- HANNA-GIFFORD
HANNA-GIFFORD (Short-Term)
A. Source-Receptor Relationship
Area sources only; uniform grid squares; user-defined grid
Sources assumed at ground level
Receptors assumed at ground level
Each receptor is assumed located at center of a grid square
Unique separation for each source-receptor pair
B. Emission Rate
Arbitrary user-specified emission rate for each grid square
Emission rates assumed constant
C. Chemical Composition
N/A
D. Plume Behavior
Plume rise not treated
Does not treat fumigation, downwash
E. Horizontal Wind Field
User-supplied hourly wind speed and direction
No variation of wind speed or direction with height
Constant, uniform (steady-state) wind assumed within each hour
F. Vertical Hind Speed
Assumed equal to zero
6. Horizontal Dispersion
Narrow plume approximation; horizontal dispersion not treated explicitly
H. Vertical Dispersion
Semi-empirical/Gaussian plume .
Dispersion coefficient assumed of the form a = ax
Analytic integration of upwind area source z
contributions assuming each can be represented
as an infinite crosswind strip of width equal to
area grid spacing.
Dispersion coefficient parameters a, b modified from Smith (1968)
Stability classes modified from Smith (1968)
No adjustments made for variations in surface roughness
I. Chemistry and Reaction Mechanism
Not treated
-------
-287- HANNA-GIFFORD
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: mixing height assumed high enough to have no effect;
treats only effects of lower boundary
M. Emission and Meteorological Correlations
Emission rates, meteorological parameters all input by user on hourly
basis
N. Validation/Calibration
Calibration not used in previous applications. Some validation experience
has been published
In nearly all applications, a single set of dispersion coefficient
parameter values (a = 0.15, b = 0.75) has been used, corresponding
to neutral stability
0. Output
Hour-by-hour average concentration values at each receptor
-------
HANNA-GIFFORD
-288-
HANNA-GIFFORD (Long Term)
This model is intended for use in predicting an area-wide average
pollutant concentration, and is expected to work best for long averaging
times. The working equation is
X = CQ/u
in which
X = average concentration within a suitably defined region,
Q = average emission rate per unit area within the same region,
u = average wind speed in the polluted layer over the desired
averaging time, and
C = proportionality constant to be determined empirically for
each different region
Specific values of C for both S(L and TSP concentrations in a large number of
U.S. cities have been presented by Gifford and Hanna (1973). It is recom-
mended that local emission, meteorological and air quality data be used to
determine the value of C appropriate for the region of interest. If such
data is not available, an approximate average value of C = 225 has been
recommended for use in evaluating the true area source effect in the absence
of removal or decay processes.
-------
-289-
HIWAY
HIWAY
Reference: No. 36 in guideline.
Abstract:
Equations
Zimmerman, J.R. and R.S. Thompson. "User's Guide for HIWAY:
A Highway Air Pollution Model." Publication No. EPA-650/4-74-
008 {NTIS PB 239944/AS), Environmental Protection Agency, Re-
search Triangle Park, North Carolina 27711, February 1975.
HIWAY is a Gaussian plume model that computes the hourly con-
centrations of non-reactive pollutants downwind of roadways.
It is applicable for uniform wind conditions and level terrain.
Although best suited for at-grade highways, it can also be
applied to depressed highways (cut sections).
p
o r
X = ~uT } ^
Q
* CO -emission
/uv*4
or
lorvcj Uruyfk of line,
usiWi 4meoi'^Al ru/e
L
I "~ 1-rr.
1.6>L
for
^i
_}
= 2-
-------
HI WAY
-290-
HIWAY
A. Source-Receptor Relationship
Horizontal finite line, multiple line sources (up to 24 lines)
Straight lines, arbitrary orientation and length
One road or highway segment per run
Arbitrarily located receptors, downwind of the sources
Unique source-receptor distance defined
Arbitrary receptor heights
Arbitrary release heights
Cut section mode
Receptors cannot be located in the cut
Emissions treated as coming from 10 equal uniform line sources
at the top of the cut
Flat terrain assumed
Line sources treated as sequence of point sources; the number used is
such that convergence to within 2% is achieved
B. Emission Rate
Constant uniform emission rate for each lane
C. Chemical Composition
N/A
D. Plume Behavior
Not treated
E. Horizontal Wind Field
User specifies arbitrary wind speed and direction
No variation of wind speed and direction with height
Uniform, constant (steady-state) wind assumed
F. Vertical Wind Speed
Assumed equal to zero
G. Horizontal Dispersion
Semi-empirical/Gaussian plume
User specifies which of 6 stability classes to be used; Turner (1964)
Dispersion coefficients from Turner (1969); for
distances less than 100 m, dispersion coefficients from
Zimmerman and Thompson (1975)
Level grade mode - initial value of dispersion coefficient equals
3 meters
Cut section mode - initial value of dispersion coefficient approximated
as a function of wind speed
No further adjustments to dispersion coefficients are made
-------
HIWAY
-291-
H. Vertical Dispersion
Semi-empirical/Gaussian plume
User specifies stability class
Dispersion coefficients from Turner (1969); for distances less
than 100 m, dispersion coefficients from Zimmerman and Thompson (1975)
Level grade mode - initial cr =1.5 meters
Cut section mode - initial cr = function of wind speed
I. Chemi stry/Reaction Mechani sm
Not treated
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: perfect reflection
£) Stable conditions or mixing height greater than 5000 m:
assumes no effect, treats only reflection from ground
U) Other stabilities with a > 1.6 times mixing heights
assumes uniform mixing
Hi] Other neutral er unstable conditions: perfect reflection,
multiple reflections treated by summation of series
M. Emission and Meteorological Correlation
N/A; user inputs all specific parameter values for hour in question
N. Va1i dati on/Cali brati on
Validation studies have been published
0. Output
1-hour average concentration at each receptor
-------
-292-
PTDIS
PTDIS
Reference: Same as PTMTP.
Abstract:
PTDIS is a steady-state Gaussian plume model that estimates
short-term center-line concentrations directly downwind
of a point source at distances specified by the user. The
effect of limiting vertical dispersion by a mixing height
can be included and gradual plume rise to the point of
final rise is also considered. An option allows the cal-
culation of isopleth half-widths for specific concentra-
tions at each downwind distance.
Equations:
if
-------
~293~ PTDIS
PTDIS
A. Source-Receptor Relationship
Single stack of arbitrary height
Up to 50 receptors specified by user; all receptors at
qround level, below plume centerline, at user-specified
downwind distances
Flat terrain assumed
B. Emission Rate
Single constant value
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1971, 1972) plume rise formulae
Alternatively, one user-supplied plume rise value can be used
Does not treat fumigation or downwash
If plume height exceeds mixing height, ground level concen-
tration assumed equal to zero
E. Horizontal Wind Field
Wind directions implicit along source-receptor direction
Uses user-defined wind speed
No variation in wind speed with height
Constant, uniform (steady-state) wind assumed
F. Vertical Wind Speed
Assumed equal to zero
G. Horizontal Dispersion
Semi-empirical/Gaussian plume
Calculations for a single user-specified stability class,
Dispersion coefficients from Turner (1969); no adjustments
made for variations in surface roughness, averaging time
or travel time
-------
PTDIS
H. Vertical Dispersion
Semi-empirical/Gaussian plume
Calculations done for user-specified stability class
Dispersion coefficients from Turner (1969); no adjustments
made for variations in surface roughness
I. Chemistry/Reaction ftechanism
Not treated
J. Physical Removal
Not treated
K, Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: user-input mixing height used; perfect
reflection assumed
Multiple reflections numerically accounted for by summation
of series
M. Emission and Meteorological Correlations
N/A
N, Validation/Calibration
No calibration option provided
Direct application of Turner (1969) procedures
0. Output
Centerline, ground-level values of concentration and normalized
concentration (concentration x wind speed/emission rate) a , a
and plume height at user-supplied downwind distances y
Isopleth halfwidths of up to eight user-specified ground
level concentrations, at same downwind distances as above
-------
-295-
PTMAX
PTMAX
Reference:
Abstract:
Equations:
Same as PTMTP.
PTMAX is a steady-state Gaussian plume model that performs
an analysis of the maximum short-term concentrations from
a single point source as a function of stability and wind
speed. The final plume height is used for each computation,
A separate analysis must be made for each individual stack;
but the model cannot give the maximum concentrations of a
combination of stacks.
g
w.
-------
PTMAX
PTMAX
A. Source-Receptor Relationship
Single stack used
Determines downwind distance to ground-level maximum concentration
Flat terrain
Unique release heiqht
B. Emission Rate
Single value
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1971, 1972) final olume rise formulae
Does not treat downwash or fumigation
E. Horizontal Hind Field
Wind direction implicit along source-receptor direction
Calculations done for fixed, internally defined set of wind
speed values
No variation in wind speed with height
Constant, uniform (steady-state) wind assumed
F. Vertical Hind Speed
Assumed egual to zero
G. Horizontal Dispersion
Senri-empirical/Gaussian olume
6 stability classes as defined by Turner (1964)
Calculations done for dispersion coefficients from Turner (1969);
no adjustments made for variations in surface roughness,
travel or averaging times
H. Vertical Dispersion
Semi-empirical/Gaussian plume
Calculations done for 6 stability classes
Dispersion coefficients from Turner (1969); no adjustments made
for variations in surface roughness
I. Chemistry/Reaction Mechanism
Not treated
-------
-297- PTMAX
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: mixing height assumed high enough to have no effect
M. Emission and Meteorological Correlation
N/A
N. Val i dation/Cali brati on
No calibration option provided
Direct application of Turner (1969) procedures
0. Output
Maximum around level concentrations, distances to maximum
values and final olume heights for all stability classes
and wind soeeds
Averaninq time less than 1 hour (Turner, 1969)
-------
-298-
PTMTP
PTMTP
Reference: No. 17 in the guideline.
Environmental Protection Agency. "User's Network for
Applied Modeling of Air Pollution (UNAMAP). (Computer
program on tape for point source models, HIWAY, Clima-
tological Dispersion Model, and APRAC-1A), NTIS PB 229771,
National Technical Information Service, Springfield, Vir-
ginia, 1974.
Abstract:
PTMTP is a steady-state, Gaussian plume model that estimates
for a number of arbitrarily located receptor points at or
above ground-level, the concentration from a number of point
sources. Plume rise is determined for each source. Down-
wind and crosswind distances are determined for each source-
receptor pair. Concentrations at a receptor from various
sources are assumed additive. Hour by hour calculations
are made based on hourly meteorological data; both hourly
concentrations and averages over any averaging time from
one to 24 hours can be obtained.
Equations:
if
-------
-299-
PTMTP
PTMTP
A. Source-Receptor Relationship
Up to 25 arbitrarily located point sources
Up to 30 arbitrarily located receptors
Unique separation for each source-receptor pair
Unique release height for each source
Terrain assumed flat
Arbitrary height above ground for each receptor
B. Emission Rate
Unique constant emission rate for each source
C. Chetni ca 1 Compos i ti on
N/A
D. Plume Behavior
Briggs (1971, 1972) plume rise formulae
Does not treat fumigation or downwash
If plume height exceeds mixing height, concentration at
any receptor further downwind is assumed equal to zero
E. Horizontal Wind Field
Uses user-supplied hourly wind speed and direction
No variation in speed and direction with height
Constant, uniform (steady-state) wind assumed within each hour
F. Vertical Wind Speed
Assumed equal to zero
G. Horizontal Dispersion
Semi-empirical/Gaussian plume
User-supplied hourly stability class, Turner (1969)
Dispersion coefficients from Turner (1969); no adjustments
made for variations in surface roughness, averaging time
or travel time
H. Vertical Dispersion
Semi-empirical/Gaussian plume
User-supplied hourly stability class
Dispersion coefficients from Turner (1969); no adjustments
made for variations in surface roughness
-------
-300- pTMTp
I. Chemistry/Reaction Mechanism
Not treated
J. Physical Removal
Not treated
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: user-input mixing height used; perfect
reflection assumed
Multiple reflections numerically accounted for by summation
of series
M. Emission and Meteorological Correlation
User-supplied hourly values of wind speed, direction, stability
class, mixing height, ambient temperature (used in plume
rise calculations) are correlated
N. Validation/Calibration
No calibration option provided
Direct application of Turner (1969) procedures
0. Output
Hourly concentration, individual source contribution list at
each receptor
Average concentration, source contributions at each receptor
for total period of interest
-------
-301-
RAM
RAM
Reference: No. 32 in guideline.
Abstract:
Hrenko, J. and D.B. Turner. "An Efficient Gaussian-Plume
Multiple Source Air Quality Algorithm." Paper presented
to the Annual Meeting of the Air Pollution Control Association,
Boston, Mass., 1975. (Available upon request from EPA, OAQPS,
Monitoring and Data Analysis Division, Research Triangle Park,
North Carolina 27711)
RAM is a steady state Gaussian plume model for estimating
concentrations of relatively stable pollutants for averag-
ing times from an hour to a day in urban areas from point
and area sources. Level or gently rolling terrain is
assumed. Calculations are performed for each hour.
Equations:
from Siv\^U
^
T
* of ro-^j
Source i»\ Qt
I
tntt»*(
7
r-:r*
> l.fcL
u
-------
-302-
RAM
rU
-------
RAM
-303-
RAM
A. Source-Receptor Relationship
Arbitrary location for point sources
Receptors may be
1) arbitrarily located
2) internally located near individual source maxima
3) on a program-generated hexagonal grid to give good coverage to
a user-specified portion of the region of interest
Receptors all at same height above (or at) ground
Flat terrain assumed
Unique stack height for each point source
User may specify up to three effective release heights for area sources,
each assumed appropriate for a 5 m/sec wind speed. Value used for
any given area source must be one of these three
Unique separation for each source-receptor pair
B. Emission Rate
Unique, constant emission rate for each point, area source
Area source treatment-
Narrow plume approximation
Area source used as input; not subdivided into uniform elements
Arbitrary emission heights imput by user
Areas must be squares; side lengths = integer multiples of a
basic unit
Effective emission height = that appropriate for 5 m/sec wind
Area source contributions obtained by numerical integration
along upwind distance of narrow-plume approximation formulae
for contribution from area source with given effective re-
lease height.
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1971, 1972} plume rise formulas
Does not treat fumigations or downwash
If plume height exceed mixing height, ground level concentration assumed
zero
E. Horizontal Wind Field
lkp<; user-suDDlipH hniirlv wind soeeds
Uses user-supplied hourly wind directions (nearest IQO), internally
modified by addition of a random integer value between -4° and +50
WinS speeds corrected for release height based on power law variation,
exponents from DeMarris (1959); different exponents for different
stability classes, reference height = 10 meters
Constant, uniform (steady-state) wind assumed within each hour
F. Vertical Wind Speed
Assumed equal to zero
-------
G. Horizontal Dispersion
Semi-empiricalI/Gaussian plume
Hourly stability class determined internally by Turner (1964) procedure
six classes used
Dispersion coefficients from McElroy and Pooler (1968) (urban) or Turner
(1969) (rural). No further adjustments made for variations in surface
roughness or transport time
H. Vertical Dispersion
Semi-empirical/Gaussian plume
Hourly stability class determined internally
Dispersion coefficients from McElroy and Pooler (1968) (urban) or Turner
(1969) (rural). No further adjustments made for variations in surface
roughness
I. Chemistry/Reaction Mechanism
Exponential decay, user-input halflife
J. Physical Removal
Exponential decay, user-input halflife
K. Background
Not treated
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: perfect reflection
Neutral and unstable conditions
Multiple reflections numerically accounted for by summation of series
until o*^ = 1.6 x mixing height
Uniform mixing assumed in vertical thereafter
Stable conditions: ignore effect of upper boundary
Mixing height for a given hour is obtained by suitable interpolation using
data from soundings taken twice a day. Interpolation technique depen-
dent on mode of operation (urban or rural) and calculated stability
class for the hour in question as well as the stability classes for
the hours just preceding sunrise and sunset. See attached description
and figure
M. Emission and Meteorological Correlation
User supplies hourly values of wind speed, direction, mixing height and
other meteorological variables required for determination of stability
class and plume rise
N. Validation/Calibration
No calibration option provided
No documented validation or comparison with observational data
-------
RAM
-305-
0. Output
Hourly and average (up to 24 hours) concentrations at each receptor
Limited individual source contribution list
Cumulative frequency distribution based on 24-hour averages and up to
1 year of data at a limited number of receptors can be obtained from
special versions of RAM (RAMF, RAMFR)
-------
-306-
Two different mixing hciyhts are calculated by the preprocessor.
One is for basically rural surroundings; the other is for urban
locations. The user is given the option to specify which he wants
to use. Hourly mixing heights are determined from maximum heights
(MXDP) for yesterday (i-1), today (i) and tomorrow (i+1) jnd from
minimum mixing heights (MflDP) for toddy (i) c/nd tomorrow (i H )
(See Figure* 1.)
For urban mixing height, between midnight und sunrise; if the
stability is neutral interpolate belv.vm MXDP.. , and MX!'P. 'J, if
stability is stable use MflDP^ •'?•. Tor hours between sunrise and
1100, if the hour before sunrise w;>;, neutral, interpolii^f he-two LMI
HXL'P^j and MXDPi Q. For sunn*so to 1400, if the hour before sunrise
war; stable, interpolate between MKHP. and MXDP. (-'r. For 14UO i:o
sunset, use MXDP. ("5,,. For hours between sunset and inidirinht; i i;
stability is neutral interpolute between MXDP. and MXDP.+, ;6, if
stability is stable interpolate between MXDP. and MNDP. '7.
For rural mixing height between midnight and sunri;.., inU'.-polate
between MXDP1_1 and MXDP. ($ For hours between sunrise and 1400, if
the hour before sunrise was neutral interpolate between I1XDP. and
MXDP.'3X For sunrise to 1400, if the hour before sunns': was stable,
interpolate between 0 and MXDP. (To.. For 1400 to sunse1'., use MXDP.
iV. For sunset, to midiniiht, interpoljte Ivtwc, n MX1.1P. mil MXUP.+ ;j--
A listing and detailed description of input formats for the preprocessor
are given in Appendix A.
-------
TC-r.'.Y
JA1-' !
MXDP. , '.v.
"i ~ ' • .'•/ ^ V v f' p
(,._^1....L.J / -.
1 / *<&
\ / ^ \
"l - s \
y / _>
----- ^\
j.-.-.j. .
t
1 II
SR 14 SS
or,i
i
'"'"'i-i 1 ® ® ,«pp
...^,, — •— - — ^-— ^_ ' MALI .
(NEUTRAL) / fT Oj)
- • MXDPi + l
i
cb
7°
0
1—1
LU
=
/
/
/
SR
14
TIME
SS
FIGURE 1. DETERMINATION OF MIXING HEIGHTS
-------
-308-
VALLEY
VALLEY
Reference: No. 14 in guideline.
Abstract:
Environmental Protection Agency. "User's Guide to the
Valley Model." Office of Air Quality Planning and Stand-
ards, Research Triangle Park, North Carolina 27711, 1977.
(In preparation)
VALLEY is intended for use in calculating annual and maximum
24-hour average SOa and TSP concentrations from single point
sources in complex terrain. A climatological approach is
used in calculating the annual average. The maximum 24-hour
averages are calculated by assuming F stability and a wind
speed of 2.5 m/sec and are intended to apply to the situation
in which the plume impinges on a hill.
Equations:
l* 6 t
t ' Z Z Z
Xfcfm Q£
or
1C * o ff K > U
nse
t> * C
10
lo m
-------
-809- VALLEY
U- 2,-Sr vr\/s«c awd k= G^|>rt)pn.ftie v^cAtjr '(p/ reapW of i
If H>U S^ 14 = L.
-------
-320-
VALLEY
VALLEY
A. Source-Receptor Relationship
Arbitrary location for each point source
Arbitrary location and size for each area source
112 receptors on radial grid, 16 directions; relative
radial distances internally fixed, overall scale may
be modified by user; location of grid center defined by user
Unique release height for each point, area source
Receptors at ground level; ground level elevations above
mean sea level defined by user
Total number of sources less than or equal to 50
B. Emission Rate
Point sources: single rate for each source
Area sources: single rate for each source
Each source treated by effective point source approximation
No temporal variation allowed
C. Chemical Composition
N/A
D. Plume Behavior
Briggs (1971, 1972) plume rise formula for both point/area sources
Alternatively, a single constant plume rise value may be input
for any or all sources
Does not treat fumigation or downwash
If plume height exceeds mixing height:
A. for long-term calculations, ground level concentrations
assumed equal to zero
B. for short-term calculations, maximum plume height is
limited to the mixing height
E. Horizontal Wind Field
A. For long-term calculations
Climatological approach
16 wind directions
6 wind speed classes
No variation in windspeed with height
Constant, uniform (steady-state) wind assumed
User must specify wind speeds representative of
each class; these are not internally defined
B. For short-term calculations, specifically to predict
the second highest 24-hour concentration expected in 1 year:
Class F stability and 2.5 m/sec wind speed assumed with
user-defined direction. These conditions are assumed
to exist for 25? of the 24-hour period; an internal ad-
justment is made for this
-------
VALLEY
E. Horizontal Wind Field (Cqnt'd)
C. In stable conditions, in complex terrain, concentrations
for receptors located above the point of impingement are
obtained by linear interpolation between the value ob-
tained at the point of impingement and a value of zero
at a height of 400 meters above that point. The value
at the point of impingement is taken to be equal to the
value 10 meters below plume center! ine. For receptors
located below the point of impingement, the effective
plume height is equal to the height of the plume above
receptor elevation or 10 meters, whichever is larger.
The plume is assumed to remain at a constant elevation
following the initial rise .
In neutral or unstable conditions, in complex terrain, the
plume is assumed to remain at a constant height above
topography, following the initial use
No variation of wind speed with height
Constant, uniform (steady-state) wind assumed
F. Vertical Hind Speed
In stable conditions, assumed equal to zero.
In neutral and unstable conditions, assumed such that
the plume remains at a fixed height above terrain.
G. Horizontal Dispersion
cav™ pl« approbation) for calculating
center values of each of 16 sectors; linear mterpolation
hptween centerlines as in AQDM
Averaging time 1 month to 1 year for long-term calculations.
H. Vertical Dispersion
Semi -empirical /Gaussian plume
Urban mode: 1Qe/n
for al neutral and stable classes
topographic ejects
Rural mode:
SKi'B
s rs -
-------
-312- VALLEY
I. Chemistry/Reaction Mechanism
Exponential decay, user-input halflife
J. Physical Removal
Exponential decay, user-input halflife
K. Background
Not treated in any mode
L. Boundary Conditions
Lower boundary: perfect reflection
Upper boundary: perfect reflection
Neutral, unstable conditions - multiple reflections accounted
for by summation of series
Stable conditions - ignores effect of upper boundary,
treats only reflection from lower boundary.
M. Emission and Meteorological Correlation
Wind speed, direction, stability correlated via wind rose approach
Emission rate not correlated with any other parameter
Non-sequential; limited correlation
Mixing height adjusted according to stability class
Urban, long term:
Class A - 1.5 x afternoon climatological value
Class D (night) - 0.5 x afternoon climatological value
Class E - assumes morning climatological value
Rural, long term:
Class D (night) - 0.5 x afternoon climatoloqical value
Stable classes - ignores existence of any mixing height
(assumes no limit)
Short term calculations - input value is ignored, only F
stability is considered
N. Validation/Calibration
No calibration option available.
Some validation experience, but limited documentation
0. Output
Long-term mode:
Long-term arithmetic means, source contribution list for
each receptor
Short-term mode:
Second highest 24-hour concentration, source contribution
list for each receptor
-------
-313-
References
Briggs, G.A. 1971. Some Recent Analyses of Plume Rise Observation In: Proc.
Second International Clean Air Congress, Englund, H.M. and W.T. Biery (ed.).
New York, Academic Press.
Briggs, G.A. 1972. Discussion on Chimney Plumes in Neutral and Stable Surround-
ings. Atmos. Env. 6_, 507-510.
DeMarrais, G.A. 1959. Wind Speed Profiles of Brookhaven National. J. App.
Met. 16., 181-189.
Gifford, F.A. and Hanna, S.R. 1973. Modeling Urban Air Pollution* Atmos.
Env. 7., 131-136.
Larsen, R.I. 1971. A Mathematical Model for Relating Air Quality Measure-
ments to Air Quality Standards. OAQPS Publication No. AP-89 (NTIS PB
205 277) Office of Technical Information and Publications, EPA, Research
Triangle Park, N.C.
Leighton, P.A. and R.B. Dittmar 1952. Behavior of Aerosol Clouds within Cities,
Joint Quarterly Progress Reports Nos. 2, 4, 5, 6 (2 vols.), Contracts
DA-18-064-CMC-1856 and DA-18-064-CMC-2282. Stanford University and Ralph
M. Parsons Co. DDC Nos. AD 7261, AD 31509, AD 31508, AD 31507, AD 31510,
AD 31511, respectively.
McElroy, O.L. and Pooler, F. 1968. St. Louis Dispersion Study. U.S. Public
Health Service, National Air Pollution Control Administration Report No.
AP-53.
Smith M. 1968. Recommended Guide for the Prediction of the Dispersion of
Air-bom Effluents. 1st ed. The American Society of Mechanical Engineers,
New York.
Turner, D.B. 1964. A Diffusion Model for an Urban Area, 0. App. Met. 3, 83-
91!
Turnpr DB 1969 Workbook of Atmospheric Dispersion Estimates. PHS
Publication No. 999-AP-26 (NTIS PB 191482), Office of Technical Information
and Publications, EPA, Research Triangle Park, N.C.
Zimmerman J.R. and Thompson, R.S. 1975. User's Guide for HIWAJ a Highway
™r Pollution Model. EPA Publication No. EPA-650/4-74-008. National
Environmental Research Center, EPA, Research Triangle Park, N.C.
-------
-314-
-------
-315-
4 Materials Distributed to Participants Prior to the Conference
In order to prepare the invited participants for productive
deliberations, a number of letters and documents were distributed at least
two weeks prior to the conference. This section contains a listing of the
documents and reproductions of the correspondence.
The distributed documents were:
1. A draft of the "Guideline on Air Quality Models and Associated
Data Bases:, Source Receptor Analysis Branch, MDAD, OAQPS,
U.S. EPA, January 1977.
2. "Descriptions of Air Quality Models and Abstracts of
Reference Materials", a notebook prepared by K. Brubaker
and A. Smith from EES/ANL with assistance from the staff
of OAQPS/EPA, describing the technical details of the
models suggested in the draft guideline, and abstracting
the references cited therein.
3. "Turbulent Diffusion - Typing Schemes: A Review", F. A.
Gifford, Nuclear Safety, Vol. 17, No. 1, Jan.-Feb. 1976.
4. "Atmospheric Dispersion Parameters In Gaussian Plume
Modeling".
Part I - "Review of Current Systems and Possible
Future Developments", A. H. Weber,
EPA-600/4-76-030a, July 1976.
Part II -"Possible Requirements for Change in the
Turner Workbook Values", F. Pasquill,
EPA-600/4-76-030b, June 1976.
Reproductions of other correspondence follows:
-------
A
-316-
ARGONNE NATIONAL LABORATORY
We are pleased that you have agreed to participate in a workshop for
the peer-review of modeling guidelines proposed by the U.S. Environmental
Protection Agency. We realize that your schedule is full and appreciate
your willingness to assist in this important task.
The EPA Office of Air Quality Planning and Standards has prepared these
modeling guidelines for application in the review of new sources under federal
regulations such as those governing prevention of significant deterioration
and as an aid in the revision of State Implementation Plans. EPA as well as
the States must fulfill these important review responsibilities now; there-
fore, our challenge is to advise on the best approach to modeling air quality
impacts within the current state of the art and with consideration for required
data and resources. As participants in the conference, we should view the
forthcoming draft guidelines as a well-considered attempt by EPA to meet this
challenge. Hopefully, our efforts will build upon their draft to produce the
final version.
•
The conference will be held at Carson Inn/Nordic Hills near Chicago,
Illinois, from February 22-25, 1977. A tentative conference schedule and bro-
chures describing Nordic Hills are included. As you can see, our schedule is
full but the accommodations are excellent.
Also enclosed is a registration and housing form and a return envelope.
Please indicate the nights for which you will require lodging. Since the
conference does not begin until noon on Tuesday, February 22, you may find
it convenient to arrive in Chicago on Tuesday morning.
By February 8 you should have received a copy of the proposed guide-
lines and background information. Please contact Donald Rote [(312)739-7711
ext. 5266 or FTS 388-5266] or Albert Smith [(312)739-7711 ext. 3259, 3240 or
FTS 388-3259, 3240] if you have not received the material by this date.
9700 South Cass Avenue. Argonne. Illinois 60439 • Telephone 312-739-7711 • TWX 910-258-3285 • WUX LB, Argonne, Illinois
-------
-317-
Tlie costs of rooms and scheduled meals will be paid directly by the
sponsors. To cover the cost of meals outside the conference, per diem of
up to $15/day will be paid to eligible participants. As government agencies,
we cannot make cash advances. Forms for reimbursement of allowable expenses
(air fares, ground transportation, and other conference related expenses)
will be available at the conference. Please note that receipts are required
and that government employees can be reimbursed only if their expenses are not
covered by their agencies. If you wish us to purchase your airline tickets
in advance, please fill out and return the attached "Request For Airline
Ticket11. We must receive these ticket requests before February 4.
In addition to the expenses noted above, you will be paid an honorarium
of $150.00 per conference day.
A list of attendees has been included for your information.
Again, let me express my sincere appreciation for your attendance at
the conference. With your support I feel sure that the conference will yield
a workable set of guidelines.
Very truly yours,
John J. Roberts
Energy and Environmental Systems Division
JJR/blt
attachments
-------
-318-
Dear
I wish to thank you for accepting an invitation to participate
in the Modeling Workshop and helping us to develop more uniform and
consistent modeling guidance.
The choice of analytical procedures to assess environmental
impacts has become a significant, and often the paramount* issue in
the administration of the new source, nonattainment, and prevention
of significant deterioration policies. I believe that we have gathered
a distinguished and balanced group of scientists to consider this
important problem. Certainly the group represents a broad spectrum of
interests and concerns. But more important, you and other participants
are recognized not only for your scientific competence, but for your
appreciation and depth of understanding of the nature of the problems
that arise in administering a national environmental program.
I wish you success and have asked my staff and their associates
at Argonne National Laboratory to do whatever is necessary to help
you make the workshop a success.
Sincerely yours,
Walter C. Barber
D1 rector
Office of Air Quality Planning
and Standards
-------
-319-
UriC 1UI USEMH
ARGONNE NATIONAL LABORATORY
February 3, 1977
TO: Conference Participants
FROM: John J. Roberts
SUBJECT: Background Material for Conference on EPA Modeling Guidelines
Enclosed please find some information which you will find of use
in preparing for the upcoming Conference on EPA Modeling Guideline: a
draft of the guideline itself; descriptions of the models suggested for use
in the guideline; and excerpts from the references cited in the guideline.
Several additional references will be sent under separate cover.
If you intend to propose alternate models or submodels for
consideration by the conference, it would be most helpful if you could
document your position in a format similar to that used in Sec. 2 of
the enclosed Descriptions of Air Quality Models and Abstracts of Reference
Materials. Thirty-five (35) copies should suffice for the conference
participants; we will also have copying facilities available on the
premises at the conference. In addition we will be able to make 8* x 11
viewgraphs. An overhead projector for the viewgraphs and a 2 x 2 slide
projector will also be available.
We look forward to meeting with you and to discussing your ideas
for improving the guidelines.
JJR:tb
Enclosures
m n«;« fina-^q • Teleohone 312-739-7711 • TWX 910-258-3285 • WUX LB, Argonne, Illinois
9700 South Cass Avenue, Argonne, Illinois 60439 leiepnone oi
-------
-320-
MC-MM
ARGON NE NATIONAL LABORATORY
February 4, 1977
TO: Participants in the Specialist Conference on EPA Modeling Guideline
FROM: John J. Roberts
SUBJECT: Potential Conference Issues
We have some additional* information which may be of use to you
in connection with the up-coming Specialist Conference on EPA Modeling Guideline.
A number of issues and questions are likely to arise as you review the
draft and discuss it with your colleagues. Without focusing attention on
specifics and without the intention of biasing your thinking, we have
prepared a list of potential issues. This list is certainly not complete
and we welcome additions or other modifications. Furthermore, we do not
expect to discuss each of these issues in detail unless the participants
feel such discussion is warranted. The issues identified so far fall
into three categories:
1) Policy Issues
2) Guideline Contents
3) Guideline Format or Structure
The major objective of the conference is, of course, not to
simply identify issues but to reach consensus regarding the guideline
itself. This consensus could presumably take one or more of the following
forms as appropriate:
1) Statements regarding the adequacy of the guideline in
terms of the issues (majority and minority opinions may evolve).
2) Recommendations of specific changes and additions or deletions
to the guideline including both the content and format thereof.
3) Identification of issues which are regarded as important but
which can not be adequately addressed at this time because of
incomplete information or other constraints.
Again we look forward to meeting with you and to a fruitful
conference.
JJR:tb
*A copy of the draft guideline along with other materials is being sent
under separate cover.
9700 South Cass Avenue, Argonne, Illinois 60439 • Telephone 312-739-7711 • TWX 910-258-3285 • WUX L8, Argonne, Illinois
-------
-321-
Potential Conference Issues
1. Policy Issues
1.1 What should be the limits of the discretionary powers of users?
1.2 The guideline will be subject to a periodic review and updating
process. What should be the form and period of review?
1.3 Which concentration estimates should be used for specific policy
questions? That is, which short term estimate (highest, 2nd highest,
etc.) should be used for SIP and NSR?
1.4 Should pragmatic considerations (computer and/or manpower cost and
time) enter into model choice guidance?
2. Applications (or problems associated with specific applications)
2.1 Features which are associated with specific applications (for
general list of such features see the table in Sec. 1 of the
notebook prepared by ANL)
2.2 Importance of each of these features for specific applications
and the requirements for accuracy of treatment by models.
2.3 Criteria for deciding whether features such as downwash, complex
terrain, anomalous meteorological conditions, etc., deserve
special treatment.
3. Suitability of models specified in the guideline and criteria for
selecting alternatives or updating
3.1 Matching models to application requirements (qualitative
considerations)
3.2 Accuracy determination and specification
- What measures of accuracy should be used. How should they
be determined?
- What constitutes acceptable validation? (purpose, procedures,
documentation)
- Calibration - What role should it play?
What method(s) should be used?
Criteria for use?
3.3 Criteria for acceptability of alternatives to the proposed models
or portions thereof?
-------
-322-
4. Specialized technical issues
4.1 Characterization of turbulent diffusion
4.1.1 Typology of regimes
4.1.2 Algorithms within each regime
4.2 Boundary conditions
4.3 Formation and loss mechanisms - guidance on when to worry and
how to deal with
4.4 Estimation of short-term concentrations
4.4.1 Calculations of multi-hour concentrations
4.4.2 Statistics of 2nd highest value
4.4.3 Suitability of peak-to-mean ratio technique
4.5 Receptor points: density, location, grid or cell size
5. Data requirements - Representativeness (location, duration); Accuracy
(sampling method, quality control)
5.1 Air quality data
- Background estimation
- Model validation
- Model calibration
5.2 Meteorological data
5.3 Emission data
- Temporal and spatial resolution, accuracy
6. Structure and content of guideline
6.1 Appendices - Summary and/or documentation of models?
b.2 Guidelines for modeling photochemical smog?
US GOVERNMENT PRINTING OFFICE 1977 - 751 254
------- |