GOVERNMENT OF THE UNITED SPATES
ENVIRONMENTAL PROTECTION AGENCY
M
0
N
SECOND CONFERENCE ON AIR QUALITY MODELING A
Y
MONDAY, AUGUST 10, 1581
MORNING SESSION
-------
GOVERNMENT OF THE UNITED STATES
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
ENVIRONMENTAL PROTECTION AGENCY
SECOND CONFERENCE ON AIROUAL1TY MODELING
MONDAY, AUGUST 10, 1581
MORNING SESSION
The conference was held in the Thomas Jefferson
Auditorium,-South Agriculture Building/ 14th Street and
f
Independence Avenue, S. W., Washington, D. C., Joseph
Tikvart,.Chief, Source Receptor Analysis Branch, Confer-
ence Chairman, presiding.
JOSEPH TIKVART . Chairman
RICHARD RHOADS - Panel Member
JAMrS DICKE . Panel Member
n. TTl^MAS I?EL"S Panel Member
NEAL R. GROSS
COURT REPORTERS AND TRANSDUCERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
ALSO PRESENT:
Environmental Protection Agency
WILLIAM COX
THOMAS CURRAN
BRUCE JORDAN
BERNARD STEIGERWALD
EDWARD F. TUERK
BRUCE TURNER
TRC Environnental Consultants
NORMAN BOWNE
Systems Applications, Inc.
_C. SHEPHERD BURTON
Schwartz and Conolly
STEPHEN CONNOLLY
Source Receptor Analy-
sis Branch
Monitoring and Reports
Branch
Chief, Ambient Stand-
ards Branch
Director, Office of
Regional Programs
Director, Office of
Program Management
Operation
Chief, Environment Op-
erations Branch 3
Vice-President
Vice President
President
'-'.-. ~ -. .-, /-. 2. - . r ". c
.
Forc-st Service
Vice President
Chief Meteorologist,
Rocky Mountain For-
est and Range Exper
iment Station
Hur.ton and. 'rilliam
LE'fTS KONTNIK
(202) 234-4433
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
ALSO ^RESENT (CONT'D.)
Pennsylvania Department o_f
Environmental Resources
JAMES SALVAGGIO
Los Alamos National Labor-
atory
MICHAEL D . WILLIAMS .
Mobil Research and Develop-
ment Corporation
STEVEN WISE
Federal Aviation Administra-
tion
Chief, Air Quality Plan
ning Section
Reserach Department
N. SUNDATARAflAN
Federal Highway Administration
HOWARD JONG ED YK
WILLIAM CARPENTER
I-Tational Oceanic and Atmos-
pheric Adnin i s tr at ion
U . S . Geological Survey
JO"!! "OLL
"u^loar R 3 -j u la tor Cc~.:r.i:~3ion
EAP.L :'J\RKEE, JR.
National Par?: ?3rvlce
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON. D.C. 20005
-------
ALSO PRESENT (COMT'D.)
Department of Energy
ROGER SKULL
4 State of_ Texas, Department of
Transoortation
5
RODNEY MOE
6
Oak Ridore National Laboratorv
7
ALAN WITTEN
8
Alcoa
9
ROBERT KOffil
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Salt .River Project
\-
DO:TALD itoo:i
Meteorolpcy Cor.inittee (TT-3) ,
.APCA
JERRY PELL
Utility T\ir Ragulatory
T37 V r-*
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
INDEX
Witness
Page
3 MR. RHOADS 6, 15
4 MR. TIKVART 7
5 MR. BOWNB 28
6 MR. COX 42
7 MR. TURNER 52
8 MR. POX 64
9 MR. JORDAN - - 1Q2
10 MR. CURRAN
11 MR. STEIGERWALD 120
12
13
14
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
CEJ
1
2
3
4
5
6
. 7
8
9
10
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
PROCEEDINGS
(8:58 a.m.)
MR. RHOYDS: Good Morning.
If you'll all take your seats, we'll begin. this
conference. In fact/ we'll begin it, even though there
are some people coning in the door.
My name is Dick Rhoads.
I must first announce a slight change in pub-
lished agenda. We had intended to open with a formal
welcoming address. Unfortunately, Kathleen Bennett, who
will he our new Assistant Administrator for Air, Noise,
and Radiation, in EPA, had to be in Ann Arbor today.
And Ed Tuerk, the current, acting Assistant
r
Administrator, had a las minute schedule conflict.
So in lieu of a formal address, let me, on be-
half of
,
the Units-:! States Environmental Protection Agen-
xrve all of you tc the Second -Triannual National
:~- 0:1 Mr Quality *:r.r.slirrg.
That's quite a title'isn't it, Joe. The Sec-
innual "'.r^tior.?.! Conference on Air Quality Modeling
I an pleased to recognize a wide diversity of
or^^niz^iio::^ participating in this Conference.
I seo representatives of several other Federal
Gcvcr:v.".2.-.t ajer.cies, ?--r.;±3 go\^rr..r.ents, local governments'',
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C 20005
-------
I believe we have representatives from both
2 the United States Senate and House of Representatives
staffs. As well as representatives of industry, many
3
4
5
And, possible, we should call this an interna-
b
.7
8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
consulting groups, several universities, and the scien-
tific community as a whole.
tional conference, because I notice the Ministry of En-
vironment of ONtario is represented.
I am sure that this wide diversity of views
will be of great value to the Conference.
I,-now, want to introduce Mr. Joseph A. Tikvart,
who is our Conference Chairman.
Joe is Chief of EPS's Source Receptor Analysis
Branch, with officer in Research Trianagle Park, or Dur-
ham, North Carolina.
And in case any of your are wondering, Source
?.sc-=?rt^r ."-.r.3-.lvsir- is our gobiil-^c.ygook term for air qual-
ity modelinc.
Joe v:Ill first provide us with some general
ir.f err?, tier, cr: th^ ccr.v.uct cf ths Conference.
riR. TTKVART: Hoed morning.
I ar. Joseph "i'cvart, Chief of the Source Recep-
tor ."r.3.lysis nr:\r.ch of the office of Air Quality Planning
and Ftc-..-.fl:ir.'!:5. I will hs your Chnirran for this
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
8
1 Conference or. Air Quality Modeling.
2 Participating with me on the hearing panel,
3 to solicit your view, and take your contents are Richard
4 P.hoads, who you've just me, v/ho is Director of the Mon-
5 itoring and Data Analysis Division; Toia Helms, Chief of
6 the Control Program Operations Branch; and James Dicke,
7 of the Techniques Evaluations Section; all of the Office
8 of Air Quality Planning and Standards.
9 I also would like to thank you for accepting
10
11
12
13
14
15
16
17
18
19.
20
21
22
23
24
25
our invitation to attend this Conference.
The dates and the purpose of this Conference
*
were announced in the Federal Register on June 12th, 1981.
I an looking forward to the generation and ex- .
change o.f many ideas over the next several days.
This Conference is being held in response to
re iUir-2.-i2.--.is :.:: faction 320 of the Clean Air Act.^ A Con-
f :.-.:-".-:= '.':. .v.i.r {v.r.lit;/ r.Dd-lirvr; is required at three-year
~~ ~ ~~~~--₯
intervals tc /-.:?.- stancUxrdize and improve modeling prac-
*
ticas v:i-hin air pollution control programs, such as pre-
rv.- deference r.elcl in December of 1377 addres-
£3d E.7.V a "I-tivrir^ Modeling Guideline." This year's Con-
far^r.c-2 "ill _:;cu3 en a najor probler?. in the utilization
of air ;'.::?'! i~y ~. tv."1 ? 1 s ror re <;:! ?.tory applications. That'.
i?, jivcr. th v. c.cc.'.raoj :" o::i~::iv.g :.10dels, ho^: should
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
decision makers incorporate uncertainty in their regu- .
latory programs. .
This year's topic is a result of statements made
by the public at several meeting on modeling guidance that
were held in October 1980. Through these meetings, EPA
had entered into a public hearing process to revise the
Guideline on Air Quality Models.
However, attendess at these, and subsequent meet
ings, made it clear that-model accuracy and uncertainty
were major and pervasive issues. These issues overshad-
owed the nore narrow problems of what specific models and
data bases should be applied to- individual air pollution
sources.
r ' '
Thus, EPA decided to delay revisions to its
modeling guidance and to direct the modeling conference
tc the issues associated with uncertainty.
Since thi technical, ccrr~.unity has net yet ccr.a
fully tc grips *.-;ith th'a quantification of model accuracy
and the treatment of uncertainty in air quality manage-
~ant activitos, additional background information was
thought to be necessary for a productive conference. A
focal i~oint -.-.as needsrl.
As a result, in May of this year, EPA sponsored
a T.;or\.i-/.,'\: on >ha col.-:-, of atrr.ostpheric models in rcgula-'-.
t~-ry L\vci..r!.?r. r.^!:ir.<7 that :-;as h^ld at Airlie House.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
The workshop was attended by, approximately,
40 invited individuals, ranging from lawyers to mathema-
2
icians. The attendees represented a broad rang of in-
3
terests. Their training, experience, and current respon-
4
sibilities made them highly qualified to deal with these
5
comolex technical and regulatory issues.
6
The suiraarv report from that workshop has been
7
provided to all who expressed interest in today's-Con-
8
ference. Additional cooies are in the rear of the audi-
9
torium.
We hope that this report will serve as a hon-
ing tool for your ideas and recommendations.
>
The Conference is designed to encourage an in
formation exchange, rather than solicit comments on spe
r
cific EPA proposals'. -
Topics to be addressed are the use of modeals
in r^gv.latcry prc^essss, and the accuracy and r-eiiabil-
v:c: ~.ro interested in rsccnr.andations for im
rr.er.ta, both in r.crieling procedures and. in regulatory pro
cessas, rith the go~.l of ensuring the .optimal use of air
quality r.odals in all programs which require their use.
H-pecific technical revisions to the Agency's
nof.eling -juidaline are not an agenda item. However, i-
r^^^.3 or^ienh^tl at thi-i Cc:-.for-?.nco . "ill be carefullv
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS.
1330 VERMONT AVENUE, NW
{202} 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
11
considered and factored, as appropriate, into EPA's over-
all modeling policy.
The guideline, when modified later this year,
will certainly reflect the discussions and suggestions
made at this Conference and the meetings in 1980.
We anticipate a public hearing on the guideline
early next year.
We have a full schedule for the next three days.
I would like to briefly- summarize it for you.
The first day of the Conference is organized to
bring you up to date on the issues of model accuracy and
the use of model uncertainty in decision making.
Tuesday morning is set aside for presentations
*
by other governmental agencies. Tuesday afternoon and
Wednesday are oner, for comments and presentations from
the general public.
Throughout ^his period, there is air.ple tirr.a for
questions and discussion..
We hnvo divided this morning into two panel
presentations. The first deals with model evaluation and
accuracy, with presentations by EPA, Electric Power Re-
search Institute, and the American Meteorological Society.
The presentations are followed by a period for
T" u r> _~ t - ^ r 3 a " 3 s r. s'.-.' T-. r s .
Ths 3-?co:v:~ pzrval is concerned with regulatory
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
12
1 aspects of the uncertainty problem and will summarize
.-
2 current EPA activities in this area. -
3 These presentations will, also, be followed
4 by. a question and answer session.
5 This afternoon will be devoted to a discussion
of the Airlie House workshop which was held last May.
7 Included will be a summary report, a panel discussion,
g and time for open exchange between the Panel, members and
9 the audience. _ ._
10 T7e have, specifically, invited those governmcnt-
al agencies identified in Section 320 of the Clean Air
12 Act to participate in this Conference.
13 We-have also tried to include any other agency
we knov.T"_that has an interest in air quality modeling.
The following agencies have requested time for
presentations and -.-.'ill appear Tuesday morning. They are
by th-2 Federal Aviation Aininistration, Mr. Sundatararr.an.
By the Federal Highway Administration, Dr. Jongedyk and
Carpenter. ~y the -National Oceanic and Atmospheric Admin-
istration, ".r. Jeff tor and Dr. Dra:-:ler. By the U. S.
Geological Furvey, ."!r. Gcll. .By the Nuclear Regulatory
Co-nission, "r. 'larkee. By the National Park Service, Mr.
Henderson, rsy the Department of Energy, Dr. Shull. And
;
by the Ft ate? o* ^ .?::?.?, "spartnant of Transportation, Dr.'..
14
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
I
2
3
4
5
6
.7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
.24
25
13
Prior to today, we have received request from
the following individuals to make oral presentations.
They are Alan VJitten, Richard Lcndergan, Richard Hanson,
Michael Basta, James Peterson, Robert Xohn, Donald Moon,
Jerri' Pell, Ray Wright, and Ralph Sklarew.
These presentations will begin late tomorrow
morning or tomorrow afternoon.
If there are any Government representatives or
members of the public who wish to make a presentation,
and I have not read your name, or you have not made arrang
ments this morning, please see Ann Asbill at the Regis-
tration Desk or Charlotte Hopper in the back of the audi-
torium.
f
I will announce an updated list of speakers
after l-.inch today. Also, a list of the names I have just
rco.:; is po.~-t3cl outside tha -auditorium for your further
'>-. re-iv.ir^c1. by thr Claan Air Act, a' verbatim .
transcript of these proceedings is being maintained.
The re^r.-t-r is Mike Ar.e.arsor. of the Meal R. Gross and
C qr.p a r. y , I r: c c _- p o rated.
3:_:r«kers are ?.ncoura:j3d to provide extra copies
of their iT-r.-santation for the convenience of the recorder
bo permit to d to enter
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
5
6
9
10
11
12
13
14
15
16
i.
17
18
19
20
21
22
23
24
25
T4
into the record any written comments they do not present
orally
The record will remain open for written state-
ments and consents until September 14, 1981.
The transcript and all written statements will
be maintained in Docket Number A-80=46, in the Central
Docket Section, Mail Drop A-130, of the Environmental
Protection Agency.
The address Waterside Hall, 401 M. Street, S.W.,
Washington, D. C., 20460
If you would like individual copies of the pro-
ceedings of this Conference, the verbatim transcript,
please contact Mr. Anderson, the recorder, directly.
The comments and discussion during this Confer-
ence will bo informal and non-adjudicatory.
*7hile several longer presentations are sched-
uled for today, indivi-5.ua! presentations on Tuesday ar.c
r."aar.a3d-?.y should, crer.arally, be limited to 10 minutes.
"her. makinr your presentation-, please give your
vrittar. statements to the recorder, ar.d sumirarize your
remarks, if they are lengthy.
Come to ths oodiun for your presentation. Pro-
jection equipment, namely, a 33 millimeter slide projector
ar.d ar. .cverhnr.a pro jo-tor, are available.
r. t , -i.'7. ./.:"! t if y
'
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
your organizations, and "your address, both verbally and
on any written statement. If, at any appropriate time,
you have a question or a brief observation, go to the
nearest microphone, either one of the two on the floor
or the one, here, on the podium. Cleaxly state your
name and affiliation for the recorder, before you pro-
ceed.
I should also like to briefly mention facili-
ties. Rest rooms are avilable down the side wings of
this building. Men's to my left and lady's to my right.
.Public telephones are also available in those wings.
In wing three, there is a very nice careteria
which is available for breakfast and lunch. If you wish
torgo outside to eat, you are on your own.
?»nn asbill, at the Registration Desk, will try
tc'bc of assistance vrith ?.:Y.'~ problems or questions you
:;o'.; -.:- would like to proceed with the formal
agenda as you have it.
?,r.:1 with that, I'd liha to. re-introcluce "r.
Rhcaas , vho .-/ill cive us an ovt-rviev of the background
*. r\ *- V* *" """ '"'"';* 1 "-*"" «N ' * -^ O
(202) 234-4433
".^. T:o.Y~>3: Don't v/orry. I'm not going to
:; : .:.-2 c'Ztor rach ar.-J every speaker. I was
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 70005
-------
16
1 just filling in for our missing Mr. Tuerk and Ms. Ben-
2 nett.
3 We are convening this Conference on Modeling .
4 Uncertainity during a period of general uncertainty in the
5 entire air pollution control community. But it is also
a period of great opportunity and challenge. General
uncertainty exists because the Clean Air Act is in the
process' of revision. Opportunity and challenge exist be-
cause EPA is in the process of reevaluating our old ways
of doing business, and developing more efficient, effec-
11 tive, and scientifically supportable ways of protecting
12 the environment in which we all live.
I don't believe it is productive for this Con-
ference 'to-engage in conjecture over possible changes
to the Clean Air Act. Rather I reconr.end that, for the
16 purposes of this Conference, v;e assur.e that the two fun-
damental state^itts of the current' Clean Air Act rcm-airvr
By fundamental strategies, I refer to technology bassc".
emission standards for, primarily, new sources, and air
_ quality mar.rijcr.2nt for both no1.; and existing sources.
.. The purpose of technology based standards for
new scurcas is primarily to minimise nc'/ pollution and
22
23
24
25
hence maximize the potential for future expansion and
-jrc"."th. r.lth'uj!". air .reality ir.y.act i3 a consideration
:n .:.vt.-.'-li3\i:-.:- ::cc^;--::3 c-;y Va;:-^ c-'..-.-.-1 r.rds , detailed
i^""*1 ?. CPOSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
17
modeling plays a relatively minor role.
The purpose of Air Quality Management, on the-
other hand, is to directly protect -the public health and
welfare as defined by ambient standards, and to prevent
significant deterioration of air quality however it may
be defined. Air Quality Management does require a direct
assessment of the impact of emission sources on air qual-
ity.
There are, generally, only two ways to quanti-
tatively assess the impact of emission sources on air qual
T that"is 7~Ambient Monitoring and Air Quality Model-
ing.
Ambient Monitoring is a valuable too, but it is
expensive, time consuming, and of somewhat limited util-
ity. In this country, we spend approximately $30 million
per yt»ar or. air quality monitoring. It typically costs
any./hare frcr, 5 to 50 thousand dc-Ilars to set up and op-
erate a single monitor, and normally requires at least
one year before we obtain useful data.
Monitors usually cannot be located at the pre-
cise sit of maxir.un air quality irr.pact, and cannot record
snail incremental emission impacts (on the order of the
currant PSD increments). Of most importance, monitors
c?.nnct assess the of feet of futura actions. They cannot
evaluate th-a effect.! vones? of prcv.osecl omission limits
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
.20
21
22
23
24
25
or alternative control strategies, nor can they predict
the impact of proposed new sources.
Ambient Monitoring is appropriate for evalua-
ting air quality trends, for assessing urban or mult-sourc<
air quality, and for development and evaluation of air
quality models. But since monitors will not be able to
replace models, air quality models will be indispensible
tools for the forseeable future. We therefore need to
find the most effective ways to use these tools. o
For the next several days, we will be discus-
sing ways to. use a wide range of models., including single-
point, source models, complex source models, mobile source
models, area source models, photochemical models, and .__ '
special purpose models-which address visibility impacts
and long-range transport. '
'Then discussing ncclels, -.-;o rr.-iist.recogiu.se that
the inpu-t-d-~.r£-. u37cr-.33 ar. inhoroiit part cf any nodol, .ang^.
her.se, t/.-c- accuracy cf the input data has a direct.; impact
on the accuracy of the .v.odel output. We know that emis-
sion source parir.etors (and particularly mobile source
psrarr.ete.r-;} cftsr. contain substantial errors. We know that
the 7jual:.ty c-f metec-rolojical c.-'za can vary widely, and
that iar-;3 cliiforences in r.c-c'.iil output result depending
upon v.-ho-th-^r thr> -r.etsorol.o -;icr.l vlnta i.~: obtained frcra a
^r-Mstir-t^r! z::-c-Lt? r:.':±~ r.-.l^ -;i~r.l station or fro a
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202)234-4433 WASHINGTON, O.C. 20005
-------
19
distant National Weather Service Station.
2 The perfect data base does not exist. Improv-
2 ed data are becoming available, but" development if im-
proved data is expensive and time consuming/ and we must
_ recognize that the limited resources of regulatory a-
_ gencies will often preclude acquisition of data of the
b
_ high quality generally used for research. The optimum
balance must be achieved between the quality'of the input
data, the quality of the basic model algorithms, the re-
. liance we must place on the model output, and the ulti-
.. mate use of the model output.
12 "^ use, and must use,' models for a wide variety
of purposes-in air quality management. For example, we
13
. use models to make yes/no decisions on construction of
14
, , new sources at specific locations. We use them to decide
15
16
17
18
19
20
21
22
23
24
25
or. the need for controls more stringent than requireo by
the conventional technology- based standards. ?7e use the;?:
to dec it" c- ".\'hi -h .specif ic sources within a group of sources
require controls, and v:hat levels of control required.
"e use tho:r. to decide the relative effectiveness of al-
.ternative control strategies for entire metropolitan areas
and rr..-y soon be uuir.g then tc decide strategies for large
nulti-state regions.
These decision?- often have tremendous economic '
: -onru-n t^l ir.p3.cr. r. Often (probably toe often)
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-44331 WASHINGTON, D.C. 20005
-------
o
«7
13
14
15
16
17
18
19
20
21
22
23
24
25
20
these decisions are based almost solely on the results
of air quality models models which we all know to
have varying degrees of uncertainty. Uncertainty and
ways to accomodate it in the regulatory process are the
major thenes of this Conference.
5
This morning Bruce Turner will provide us with
6
an overview of EPA's rather extensive research program
to develop new and improved models. At least-my script
says Bruce Turner xvill do it this morning. He's sitting
on an airplane at Raleigh-Durham. He nay speak to use
later this morning, or early this afternoon. However,
>
he will be here. He will out line EPA's research program.
But I submit that regardless of how much re-
sea-rch effort we devote to developing and improving mod-
els and their data bases, some degree of uncertainty will
alvays exist in the.output. ' .... »
."ns -:.f our current yroble^s is that -our. knov?~--«cg
ledge of t>.e cl^.^rea is uncertainty cf models is fa-£ from
perfect, sr.d -far less than what it should be. There are
several rs:.32r.5 for this. ;>:ith the implementation of the
Clean Air Act cf 1970, -.cclels vary suddenly became pri-
mary re7ul?>tory tcols. Undsr the extreme urgencies of
that period, regulators, industry, and environmentalists,
^i i -' V <-* t "3 ** ? ^! .**. "^T r* 0'""* ^~ -"^ J ***** Z. *." T %*''*'. ~ " J)\* "^ 3T rv^- J^Sx S V7'"** IT *"* H \r ^ i 1 ~*
ab.*.0 au. th."': - 1 ";.-.*":. *!"*:r-" r*: -**"'-;-".1 the accuracy of the r.oc-cls
MEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
t330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
as of little significance, because regardless of accur-
acy, they wore by far the herst, indeed the only, methods
«
for predictina air auality impact.
3 "
After the initial surge of use in the early
1970's, great effort went into improving the internal air-
5
gorithms of the more widelv accepted models,and effort
6
into developing new models to handle problems which were
7
not amenable to solution by existing models. Still, not
8
much attention was devoted to quantifying the accuracy,
y
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
because each improvement in the models was 'generally con-
ceded to be an improvement in accuracy and,therefore, .
tended to reduce the problem.
Especially within the last few years, however,
it^.has become generally recognized that, despite the best
efforts of the scientific community, even our best mod-
al.3 will ccr.tir.u-2 to have inaccuracies or uncertainties
- --^553
f icier, t -;-.;;r.itude that (1) the uncertainty should be quan
tifiocl ?,nd c'ocunontod, and (2) the uncertainty should
be e:-:v.-licity ci'r.siilarecl in decisions v.'hich are based upon
The r':ajor i^nec!imont to quantifying uncertain-
ty ha? baeri IE.C!; of suiuc.blo: r.ata bases. If one defines
uncertainty as the different botv.*een the model output
anr. sor.-? .-.bo-r-l-.-.Ti-i "tr;;-.!-., " the:: in order to quantify
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
un certainty, one must quantify, that is, measure, the
"truth."
Under our current regulatory structure the dif-
ficult "truth" which we must somehow measure is the second
highest ambient concentration at any location and at any
time. This measurement theoretically would require an
infinite number of monitors operating for an infinite
time period in all possible types of terrain, all possible
weather conditions, and- all possible source configurations
and operating conditions. This is, obviously, imprac-
tical. However, this measurement, even in a practical
sense, still requires a very large number of monitors
operating for extended periods in the type of terrain
r
and with the source configuration of interest. The quanti
fication of modeling uncertainty is a resource intensive
task, and it is very time consuming. It requires care-
ful study design and meticulous data analysis and inter.-
pretation. "But it is a task which is absolutely essen-
tial, if we expect to explicitly consider modeling uncer-
tainty in our regulatory decisions.
Later this morning, Norm Bowne will discuss
the Electric Power Research Institute's very estensive on
going project to develop a set of data bases for evalu-
i
aticn and analysis of model uncertainty.
Also, Bill Cox will discuss EPA's ongoing multi--
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
year program to evaluate moclel accuracy, and Doug.Fox
2 will discuss the American Meteorological Society's on-
3 going work on model evaluation. These three model eval-
uation programs, as well as several other programs which
5 will be discussed during this Conference, should signif-
- cantly improve our ability to quantify and understand
the uncertainty of the mooels which are used for regula-
tory purposes. .
Given that models are essential regulatory tools
and given that it will be possible to determine the uncer-
.. tainty inherent in models, or at least in specific mod-
el applications, it is also our task to consider how this
._ - information can be used to improve our regulatory deci-
tj siens.
14
For example, we know that most conventional
15
16
17
18
19
20
21
22
23
24
25
models are mora accurate when predicting long term aver-
ages than vhsn predicting short tern (such as 3-hour or
24-hour) averages. This means .that when the air quality
regulatory constraint is based upon a short term average
(such as the current short term PSD increments and the
current short term ambient standards) the impact of model
uncertainty is inherently magnified. Unforotunately,
most cf tr.e health effects on which the ambient standards
are really oriented or based on the effects of short-term
pollutant dosages a subject which will be-discussed
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1 by Bruce Jordan later this morning. It would be diffi-r
2 cult to develop long-term surrogates for those short term
3 effects, although Bern Steigerwald -will discuss some pos-
4 sible regulatory mechanisms which we are considering to
5 alleviate some of these problems.
6 Also, for example, we know that most convention-
7 al models are more accurate when predicting the upper
8 percentiles of the air quality distributions -(such as
9 the 90th or 95th percenfeile) rather than when predicting
10 maximum or second maximum values. Certain regulatory
.. approaches such as statistically based standards, whrch
12 will be discussed by Tom Curran^ a little bit later, and
._ the allowance of more than one exceedance of a standard,
are also being considered to alleviate some of these prob-
lems.
15
The fundamental issue, however, is something I
16
17
18
19
20
21
22
23
24
25
have not yet mentioned. rveri, -.fter we have improved our
models to tha point of 'high accuracy with minir'.um uncer-
tainty even, after we have altered our regulatory pro-
grams to -rLninizG the effect of that uncertainty on our
»
regulatory decisions even after wehave evaluated our
T.ousls sv.c-h that v;e know explicitly what the residual
uncertainty is what should x\?e do with that knowledge?
He:: '^D .'
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
25
For a simplistic illustration, let us hypothe-
size that our model says a source must have an emission
linit of X to just barely attain the standards. And
further hypothesize that the uncertainty about X is plus
or minus 20 percent. What emission limit should be es-
tablished? Should the limit be X plus 20 percent, thus
minimizing the risk that we spend too much for controls?
Should the limit be X minus 20 percent, thus minimizing
the risk that public health be .jeopardized? Or should
the lir.it be simply X, in seme attempt to equalize the
risk of each impact? This is not an easy decision with
even such a simplistic example. Imagine the complexity of
the decision if we were faced with a real world example
r
in which both oublic health and millions of. dollars are
'
at stake
I'll Leave these questions hanging for the next
^ever."! ~ay^ sr.cl ~cnthr; l^uc hopefully not for years',
To h.-of us ;.. rc;:;-.rr.- f--r -this Conference, a group
of ercrer'is ccr.vc'r.e-"! a v:crkr".cp' sr.v.:ral nonths age cut-
sic!:; cf "aili: r:~tc:\ to 3COV2 cuL nnd discuss the major
issues involving model accuracy and uncertainty in the
regulatory proa.io3. I 1-n lie /?. that r.ost of you received
a surcr.ary of th-nir deliberations. This elite group did
noL fi,..: ;I! tV..i c--j--..r^; '-.;:. t they c!id a good jcb of find''
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
and some of the group will present and discuss their find-
ings.
I said the workshop group did not find all the
answers. Please recognize that the answers do not exist.
They will have to be created. The impact of emissions
on air quality if inherently a stochastic or probabalistic
process. Hence by definition there is no single answer
to the technical issue. Similarly, there is no single
answer to the public policy issues.
Ue need sorie creative, ideas from this Confer-
ence. We - need some innovative approaches, tte need some
sound technical facts. And most of all, we need some
collective thinking focused on the fundamental issues.
p
If v:e can accomplish this during the next 2*s days, then
wo car. consider this Conference a success. And I look
for' r'l t? u.-2!-pi;-..': you nake this Conference a success.
::?.. TIITVVV:1: X£.;:t en tha agenda we-' have for
you a r:an-3.1 discus-ion cr. th^ issues of model accuracy,
d2t £TT~. 1" £ b .'. C .". 3 , ,"i n C'. JL T: '-.''- 5-' -".. 'J-:\ t; .* OTIS .
T-*2'T.*s got c- panel of four individuals, all of
vhic'- / -"-.o !'.ivo don-3 a f=-.ir a?-.cunt of r.:crk in this area,
and I tl.ink so.;/- ve:"- interesting and informative infor-
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
27
us on the stage, we have a place for you.
Those are Bruce Turner, who is Chief of the
Environmental Operations Branch. Hopefully, Bruce will-
. be here sooner or later. He was warned. Second, Mr.
_ Norm Bowne, who's vice president of the TRC Environment-
o
al Consultants, who will be speaking for the Electric
6
_ Power Research Institute. Norm,would you come forward,
please.- Third, Mr. William Cox, who's a staff scientist
in the Source Receptor-Analysis Branch. And, fourth,
Mr. Douglas Fox, who's Chief Meteorologist of the Rocky
Mountain Forest- and Range Experiment Station, with the '
U. S. Forest Service, who will 'be speaking for the Amer-
ican Meterological Society.
Bruce Turner was scheduled to be first. He
14
will now be last. And, if you don't mind, Norm, would
15
vou be willing to oroceed.
16
Norm Bov.TxG, vice ^resident, TRC Environjr.entai-
17 * -
Consultc-nts, speaking 'for th-2 Electric Power Research
Institute.
19 . .
r-Jould vou olease hold vour nuestions and cbser-
20 ' '
vations until after all the panelists have spoken. We
£ 1
will have anple tine for questions,and answers, and ob-
servations at that tine. But please hold, them until after
23
-"re"2.r.t?.tior:3 have been r^ade.
24
25
NEAL R. GROSS
(202)234-4433
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
u
18
19
20
21
22
23
24
25
. . MR. BOWNE: My name is No man Bowne. I am Vice
President and Chief Consulting Scientist for TRC Envir-
onmental Consultants, Inc., 125 Silas Dean Highway,
Wethersfield, Connecticut. I am appearing today to pre-
sent preliminary results from the Electric Power Research
Institute's Plume Model Validation Project. I will re-
fer to the- as EPRI and PMV in the future. While the work
reported here was sponsored by EPRI, any conclusion I pre-
sent are my own and do not necessarily represent conclu-
sions or opinions of EPRI or its member utilities.
Air quality models, or more accurately, atmos-
^
pheric dispersion models used to predict air quality have
been elevated to the role of quantitative, analystical
devices for estimating various environment impacts of
pollutants. The use of existing models to predict con-
centrations for material emitted from tall stacks^has
i
called into-question the accuracy and reliability of these.
models in various terrains. A statement frequently heard
is that plume models predict surface concentrations to
within a factor of two. But rigorous evaluations to dem-
onstrate that models are capable of predicting concentra-
tions with a factor of two or any other factor are rare.
EPRI recognized the need for model evaluation
and developed PMV as the most comprehensive program of
plume concentration measurements ever undertaken to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBER'S '
1330 VERMONT AVENUE, NW
(207) 234-4433 WASHINGTON, D.C. 20005
-------
1 to determine the accuracy of models'typically used to j
2 assess the impact of fossil fueled electric generating
3 facilities. EPRI formed an advisory committee of scien-
4 tists from the EPA, MOAA, National Laboratories, univer-
5 sities and utilities to plan and review the PMV program
at each step including the results I will describe to-
7 day.
8 . Our conclusions are that existing Gaussian plume
9 models are not capable jof_ accurately predicting maximum
ground-level concentration or position of the maximum
or. -an hcur-by-hour basis. However, when comparing the '
12 maximum predicted concentration for the four months of .
13 data examined to the maximum concentration observed dur-
ing the same period, the agreement is within 25 percent
Predicted plume widths were half those observed. Dis-
1C tar.ce r»r..:l direction to rax5.rv.im concentration T.-:ere fraauent
ID
17
18 A technical 'report, P'rellr.ir.ary Results from
19 the L?^I Flu'^.3 r'o'.lal Valiant ion Project Plains Site,
ftrt
gram, is r.ttached as Exhibit A.'to my testimony. I will
22 lir.it rv.y cescripticn of the proqrarr. to those measurements
23
24
25
that v:^r2 actually used in our analyses in the report.
r'j/se r?t::- th^.t r'"-'V is an cngoirvj project and rr.orc in-
^:r--.~L~:\ -..-ill bi avail.-Ms in f-.i- future, not only for
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
. 7
8
9
10
.11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
the level-terrain plains site but"also for sites in more
complex terrain.
The four main objectives of the Plume Model Val-
idation project are:
1. Establish by statistically rigorous proced-
ures the accuracy and uncertainty of model predictions
of concentrations of airborne pollutants emitted fron tall
stacks,
2. Assess the performance of a given model over
a rang of meterological, topographical, and source conci-
tior.s,-
3. Create and archive an extensive data base
of measured power plant plume behavior with the accura-
r
cy of the data certified through external, third-party
audits, and
.
4 . develop ancl validate ir.proved plume models
for tc^Gcrrsphiccil, rr.etacrolcgical ancl source conditions
that ars r.n.-.ic'aquatily treats-3 by .e:: is ting models.
T<'c- ex.=nin-2 the accuracy cf nodel pradictions
by testing the r.cde-1's ability to predict certain features
of the ground-level concentration-. The features we have
exarr.ir.-a- are t.h-2 r.axir.u." concor.traticn, the postion of
the ma:-:i-un cciiC2ntration, and the vidth of the plume.
Tn tha hnliaf that the -n^-sis would be more \
tr,?."t:.h: e ir. ?. rlr1: rv.r?. 1 ~.r.?~ '."it.h n single sov.rcc, the
NEAL R. GROSS
. COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Kincaid Generating Station near Springfield, Illinois
.was selected. It is owned by the Donmonwealth Ediwon
Company. The plant is a baseload facility located in
central Illinois and is a coal-fired, mine-mouth plant
employing two 660 megawatt generators. Both units are
i
vented through a single, 187-meter stack with an exit
diameter of 9.0 meters.
Two types of concentration measurement programs
were conducted. Standard air quality monitors were us^ed
to measure sulfur dioxide concentration at distances be-
tween 3 and 30 kilometers, the distance'range of antici-
pated -maximum impact. These monitors operated continuous-
ly to provide a detailed record in time at 38 locations.
r-
The"second concentration measurement program used a trac-
er gas that sampled by portable monitors at 200 locations.
The trao3r V.-TIS rsleasod through the stack. ~Th'c~lAtensive
tracer r jdsursr.er.t program -.-as conducted during three "w
i- ;..pril-"a;; 19°° and a-rr.in for three weeks in July 1980.
Those monitors operated six to nine hours each test day
to ^rovicle a' r.etp.ilecl record of the ground level pattern
of concer.trc.tion.
Ths rov."tir;2 ?iir ~ucilit" r..oriltors *.-z-ro dis''"i'^
uted as 3'hcr.:n in Figure 1. All of the figures may be founc
at the brr': c^ rv/ ].rei:ar-j-f. tcsti-.ony. The tracer monitor''-
25
NEAL R. GROSS
COURT REPORTERS AND TRAMSCRIBERS
1330 VERMONT AVENUE. NW
(202J 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
to bracket the ground-level concentration maximum.
Winds and temperatures were measured during -the
routine program at the Plains Site on a 100 meter meteor-
ological tower at four levels. The tower, solar and ter-
restrial radiation equipment were located at a site lo-
cated about one kilometer east of the Kincaid Plant.
Other meteorological measurements made at the site includ-
ed routine weather observations and vertical .wind and
temperature soundings.
During the intensive periods, meterological
measurements made during the routine program were contin-
ued, but .the measurement frequency was increased. Wind
and temperature soundings were made on an hourly shcedule
when tracer was being released.
Continuous measurements were made of stack-gas
emissions and stack temperature at the 137-meter level
of the st.-c!-:. The S00 emission rate, exit velocity and
terrporature v^re calculated also from the plant operat-
ing inforr.atior. , daily fuel consumption data, hourly
electrical load c:\ta, ar.d daily coal analyses.
a
Cunlity of the routine field measurement data
has bo^r. assessed through external audits on a quarterly
basis Jb\ ascertaining compliance :;ith commonly accepted
procedures and the challenging the instruments against :
,202)234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
1
2
o CRSTER - Selected based on its regulatory status
o
4
5
power plants.
o MTJLTIMAX - Selected in a modified form CEQM, as.
^ TT»^\ /
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Four Gaussian models were selected for evalu-
ation. They are:
as an "approved" model in the EPA Guidelines. CRSTER is
commonly used in the licensing applications for isolated
6
7
a model- that predicts values exactly the same as CRSTER,
but at user specified receptor locations. This has al-
lowed the P}TV project to predict CRSTER-equivalent con-
centration-values at sites corresponding to actual field
monitoring locations.
o TEM (Texas Episodic Model) Employs a unique
r
feature to adjust the.horizontal dispersion parameters
for various observation averaging times. TE.*1 was includ-
acl in the ?*?* project to investigate the effects of this
o "1ST! -- Chosen because it. contains a 'unique algc
rithr for G.~tir.;-,tinq plume rise.
Evaluation of TE"1 and !'?DM revealed that cer-
tain feature- of those models affocted" accuracy of the
prediction of concentration. TEM predicted concentra-
tions, especially plune-center-line gro'ond-level concen-
tratirr.r, t'::.~t ."::e tec Ic1-' becnvsc of the dispersion ad-
justr-.-.t. T'\ :j "c-^-- -".r^: cl-1 r.ct ~.~.rri* th-o plvme to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
to break through stable layers near .the top of the stack
and ground-level concentrations were predicted as much as
ft
20 times the highest concentration observed during the
3
entire program. These models are dropped from further
discussion here and the rest of my remarks pertain to
5
the EPA CRSTER model as reflected in the CEQM version.
o
Four performance measures were selected to char-
acterize the observed and predicted tracer concentration
8
patterns from the 200-station network for model evalua-
9
tion:
10
1. Magnitude of the maxinun (highest) concentration
12
13
14
15
16
17
18
19
20
21
22
23
24
25
for each hour,
2. Magnitude of the highest concentration for each
hour at a given distance from the source,
3. Location (distance and direction) of the high-
est concentration by hour, and '
'4. Flurr.e wif.th, that is, cro.~sv-in.-cl standard devia-
tion of rrcui-.c: Isv2l concentrations at a given distance
from the source.
Statistical ccr~.pari5.~ns of .rhssrved and pre-
dicted: sulfur dio;-:idi conc-2- tra-ticns v:ore performed for
the continuously operating 28-station network. They were:
1. C -:>. .--.1-2 the 30 hi-hest observed values with the
30 highest v-reclicted values, regardless of location or
-(CorAj-ir? tha -;-_-;::;r cr.c'i o .c the cunulativo froquenc
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
35
distributions of observed and predicted concentration.)
2. Compare each of the 30 highest observed values,
with the corresponding predicted value at the same loca-
tion and time, and compare the 30 highest predicted values
with the corresponding observed values.
3. Compare the highest observed value for each hour
with the highest predicted value for the same hour re-
gardless of location.
4. Compare the set of all observed values with" he
set of all predicted values for all hours and stations.
Either the observed or predicted value -was required to'be
>
above' the threshold level of the monitor (15 ppb) to elim-
inate comparison of insignificant concentrations.
r .
Model calculations were made without any attempt
to .fit the c*3.ta just as they would be used to assess a
prc^.o52c. sourer.. Ti7->. usid measured emission "rate's,,* stack
zor>"-irritur^i ?.:v:! velocities rather than assur.iiivj a co
3t2.r.t r^uc vl-._ch :r.v..st ?;..: cl^ne -Tor a proposed source.
Th2 meteorological data preprocessor from the EPA CRSTER
code was emv-.loy.2cl to forr.'.at the Springfield netorological
r.ata for input to the model.'
Insults croT. both the routine network, using
sulfur dioxide as the pollutant of interest, and from
:
t":~ c;v.?. c:." ]~.:-:::\ ~lour:.::<. t::.-.-?3J r..--t./ork \;erc used to clc- "
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Prediction of distance to maximum concentration
for each one hour tracer test exhibited no predictive ' -
ability as shown in Figure 2. If there were predictive
ability the plotted points would be distributed along the
diagonal line. The largest differences between predicted
and observed distances occurred for slightly unstable
(Category C) and neutral (Category D) stability condi-
tions. The maximum concentration was usually observed
closer to the source than_predicted for these stability
classes.
Prediction of direction to maximum concentration
is dependent on the wind direction used to drive the model
The results for the 10-meter wind direction at Springfield
are illustrated in Figure 3. Most of the points are be-
low the diagonal line of perfect agreement. Considerable
scatter is also obvious. Home 'ir.proverner.t is achieved
by using the v.-incZ direction chsarved at the Tito 130 rr.et-
ars abcr.'o the ground as shovr. in Figure 4. It is not sur-
prising that the 100-r.eter wind is a better indicator
for a plu-s that typically achieved heights of 300 to 500
4
:r.Gtors above the ground.
I'Tidth .of the plume on the surface was examined
by corvarir.g-the observed standard deviation of the cross-
arc cor.cetraticn pattern with the model standard devia- :.
tieri, si-,-ia-y . Figure 5 illustrates the plume width
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
37
results. All but 2 of the 50 points indicate observed
plume- width larger than predicted. The observed concen-
tration pattern indicated a plume two and a one-half
tines wider than predicted at 3 km from the source. Fig-
ure 6 shews a continuation of the sane bias at 7 km from
the source. The average difference in plume width was
a factor of 2.2 at this distance.
Up to this point we have shown the model has
a bias in prediction of direction to maximum concentration
vrhen using 10-r,eter wind observations, a bias in predic-
tion of plume width and no ability to predict distances
__ , >
to maximum concentration for hourly patterns. The next
comparison that I will show you is predicted and observ-
r
ed maximum sulfur dioxide concentration from the routine
monitoring network. T7e paired the highest concentration
observed in th3 2S-st?.ticn fixed air quality hetv-'ork with
t'.is highest predicted value for the same hour over the""1
nat-.-.-ov!-:. Th.\t is, tho pradicicc! and observed concentra-
tions are fror, the sane time? but not necessarily the same
location. Figure 7 shovs the general disarray of this
comparison. But, please note on feature,. the highest ob-
served ar.d highest predicted concentrations are both a-
bout 100 cpb l?.rger than the second highest values. Fig-
ure R 3',-ovr? that bet:-. '-''.2 30 h:'.~=-.?ist predicted concentra-
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
between 150 and 250 ppb, although the observed values are
in a nuch tighter cluster. But at no tine did one of the
30 highest predicted values correspond to one of the 30
highest observed values. In fact the highest observed
value associated with the high predicted values was only
100 ppb. A similar result is indicated for the observ-
ed high concentrations. This indicates no capability
for model prediction on an hour-by-hour basis.
There were some unexpected results. Figure^9
shows the maxirvan sulfur dioxide concentration observed
at the around plotted against emission ^rate for the 30 .
highast observed events. The coal burned is relatively
constant in sulfur content so the emission rates reflect
the plant load. Apparently increased plume rise associ-
ation with larger plant load ballances the increased sul-
fur sr-iosiDr.T -?.nrl produces little change in highest ground
' rtT.c5~ tr~^ I.T.. .71'ur? 1~ illustrates tha ruu;inu.*T- ro-l
-
c'~ server tr-.^-a jonj-entration, that is ground concentration
dividsc1. ~y e~.i-?sior> rate, as a function of distance from
th-?. str.r::. - "c-'.^l ~:---.^iccir;-.~ cf naxinun concentration
sl'.r.r ar. r.'v^u^t cli"b ~o a r^xir.u;-. at 1 or 2 kiloneters and
a de.cre~.sc ?.t r/r-^-.ter distances. The observed tracer
data ir.uicates a r.sxir.uT. first occurring at 3 km and very
1-*+-!.-> .^'--i.-.^ -~ h -' v"v - - * -,'',-.=: »-> - rl i qf-qnr--- n-P 70 l- i 1 -
«, V-^L. «. v^....t»-^_ _ -l_Vj_.,,. ^ .' ._.»'_ . ta>_/ .*. V.t.LOlvk.lll^xV^ V.J*. ^UW J v. JL X ,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
, 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3 be seen in the sulfur dioxide data. Figure 11 compares
4
5
stations within 3 to 5 kilometers of the stack. The high-
6
est point are ovsrpredictions. Finally, at distances of
5 to 10 kilometers shown in figure 12, there is systematic
o
but small overprediction. Finally, the data observed.and
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Another indication of the riddel's inability
to predict consistently as a function of distance may
the 30 highest ranked predicted and observed values.
Data points are paired by rank, not location or time, for
predicted at locations between 15 and 20 kilometers were
grouped as shown in Figure 13 and underprediction is indi-
cated.
."'.odels are used to predict highest or second
highest concentration for regulatory purposes and they
are not used to predict impacts in real time. We have
demons:: rated a total lack of accuracy in the real-time
preclicr.io". , :.:v>t in vi>i:.'ir.c; the high concentration there
o b-"= boaruls in both the r.\cclel predictions and
d concentrations. The comparison of highest
center Lin:; predicted cor.csntrntior.E with, highest observed
ccncar. t.v.it icr.2 for 33 t^st hours are shewn as a cumula-
tive distribution function in Figure 14. For these one-
h~ur r:v.csr. tssirs, t'::-= roclol ur-cor^riuictec^but the upper
encl of the distribution indicates reasonable agreement. .
'» ?i;-ilar cc'"j--ri3o;i of th-i cumulative frequency
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
.
21
22
23
24
25
distribution of hourly observed and predicted sulfur di-
oxide concentrations is shown in Figure 15. Again there
is reasonable agreement at the upper end of the frequen-
cy distribution although, in this case, the model over-
predicts-. One perplexing problem that needs further exam-
ination is the overprediction exhibited by sulfur diox-
ide data and .underprediction exhibited by tracer data.
Is it due to different spatial density of sampling? We .
will have to examine the remainder of our data in an....ef-
fort to answer that question.
We have examined a limited number of three-hour
«
. >
average sulfur dioxide predictions at this point. The
preliminary results of the ranked pairs of 30 highest
predicted and observed concentrations are plotted in Fig-
ure 16. A slight underprediction of the high values is
indicated axcspt for .tha single highest value. ___ 4
Tn surT.r-trv our cr^lirrinary results from a tall-
. stack, buoyant source in flat terrain indicate that EPA
>
air quality prediction nodels exhibit significant scat-
ter or rr.ngs of observed concentrations associated with
predicted concentrations and scn-3 bias or error in aver-
a~e values. ?.ias was apparent in predicting direction
to naxir.u.". concentration v:hen using a low level wind di-
rection, and ir. praclictir.^ rlur.e "ic?th at the ground.
There '.-'r.s bi.-:?: ir. t.':rr.3 cf ever;" "rc^.ictirig dict.ir.co to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS '
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
maximum concentration for near neutral stability. There
was so much scatter for hour-by-hour prediction of high-
est concentration that any bias is -completely masked, but
certainly there was no skill in hour-by-hour predictions.
Relative concentrations decreased as plant load increased,
and highest ground-level concetration was essentially
the sane over the range of load conditions examined.
Relative maximum concentration was virtuallyunchanged
with distance between 3-and 20 kilometers. Finally, there
appears to be a bound on highest concentrations in both
the models and-the atmostphere as indicated by compari-
son of highest predicted .and observed concentrations.
These results indicate that the EPA CRSTER mod-
r
el may predict the maximum hourly concentration with an
accuracy of plus or minus 25 percent, but the location
arvcl tirr^e cf actual occurrence will be completely incTepen-
cli-t cf the prediction. These results imply that use__
of the nodal in a multiple source situation or added to
short-term background is not' feasible because of errors
in position, plurva width and tiding.
Thank you for the opportunity to appear today.
I will ivj happy to respond to your questions.
'IR. TIT-7ART: Thank you, Itora. '
I1 J. . !!':: to prco~e:"; -.;ith our second panelist '-.
the.;. "'-a-'j ''{l"a.a- Tox, c'-a.'"'7 scientist on t!\e Source
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
42
Receptor Analysis Branch.
Bruce, would you join us on the stage, please.
MR. COX: I would like to begin by reviewing
-the results of several studies that have been recently
completed to evaluate the reliability of models. After
that review, I would like to.briefly describe EPA's plans
to more systematically evaluate the performance of all .
models being used'or proposed to be used for regulatory
purposes
Before going over the details of these recent
model evaluation studies, I would like to make two gener-
al observations. The first observation, summarized on
the first .slide, is that these studies seem to confirm
r
what leading atmospheric scientists have stated for some
time; that is, (1) models are reasonably reliable in es-
timating th-s magnitude of highest concentrations occur-
ring scr-.-3tir.-3 ar.c! sor.evhere v.'ith an areas; (2) models
appear to c;o a. better job of estimating concentrations
that arc avoragacl in tirr.a or s~sco; and (3) plume mod-
els perform yccrly in predicting concentrations at spe-
cific locations, primarily due to uncertainty in know-
ing precisely vhsre the plume is located for any given
hour. Much of the current criticism concerning the ac-
curacy cf no civil 5 usually focusos on their inability to
7:oli.-.::". y .: r \:: -..-.: _ z?:.c.-.~.::.-" ::L :::.: "t specific locations
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
in times. However, this criticism overlooks the fact
that in ir.ost regulatory uses, the specific location and
exact tine of high values are of little importance. Only
the magnitude of high concentrations is used to determine
regulatory needs and models currently used for regula-
*>
tory purposes are nuch more reliable in this type of an-
alysis. In fact, as you've heard, errors in peak-esti-
typical, now these errors are certainly well within tH*e
often quoted factor-of-tv;o accuracy that has long been
recognized for these models. For this reason, statements
*>
on modeling error must be interpreted very carefully and
conclusions..reached only in the context of how model
f
nates are used-by the-decision maker.
The socond general observation is that todate
thari he-, s been a no'table lack of consistency arAcng the
various oturHas vir.h re.5pc.ct to the statistical techniq-c
ar.r. '.jr^.':>.i ~al presentations used to describe no del" per-
f or.-.-iC-r.c;:. As a r-sult, ir. h=-s-beer. c.if-ficult to inter-
cor.Vvr.re cli^ rer^ulca frcr-. uifferer.t studies v;ith a great
dc-gree cf confidence. This prcblen will be sorr.ev.-hat al
levi^t^^ ir. t'.\^ future ;.:hen the -..'ir-reccrr.raeridecl perform
ance measures are *:idely user! in nodel evaluations. In
~"'.. -.-ill r-r-ly '-.-.. ..:'-'-_ 0:1 J:a r.'-T. i^asares in our
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
As I just mentioned, several fairly recent stud-
ies have been completed including the EPRI "study you just
yeard about that supply information about the accuracy
of -.riodels. Each study was designed to allow comparisons
of available monitoring data with model estimates to see
how well the r.odel performed in estimating actual meas-
ured poilutatnt concentrations. -
The first study was performed under contract
to evaluate the performance of the CRSTER Model. In this
I study, SO data collected in the immediate vicinity of
two power plants located in rolling terrain were used ,'
in the evaluation. These two plants are the Stuart and
"u^skingum plants both located in Ohio near the Ohio Riv
er. Hourly S0n emissions for each of these plants were
computed fror-. hourly avaraage r.cr.thly fuel sulfur and
t>.3 'r.c-;swntt generation values for each unit. The re-
rults, i'.-.cwri in the s^jcn:! viawr.raph, -.re based on SC
2
air quality data collected frotr. four stations at I-Iusking-
u.~ and fivs stations around th'o Stuart' plants in 1973.
.T-.lth~u::>: a nu:-.v.er of p^rfcrr-ance indicators
were ccputed to describe how well CRSTER performed, a
couple 3f sirv.-lc statistics shrv; this fairly clearly.
Slide 2, the second vic-wgraph, corparas the 99th per-
C2ntil^s of observed a:v1 ::sti-r?.ted one-hour and 24-hour
i-.v^rc^ji- ."? lr-.V;:l;.. '!"... c,: rvccr':r^.t: "r:s shown correspond
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202)234-4433 WASHINGTON. D.C 20005 ;.
-------
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
-!
i
to the average values occuring at the individual stations.
As you can see, the ratios for the one-hour concentra-
tions were somewhat closer to one than for the 24-hour
concentrations. While the one-hour results indicated no
significant tendency for CRSTER to under- or over-predict,
the 24-hour 99th percentile values were underestimated
by about 40 percent. One other point is worth noting, and
that is correlation coefficients between estimated and
observed SO concentrations at specific sites were gen-
erally low. This indicates that plume models such as
CRSTER ara not now as reliable in tracking concentration
variations from one time period to another as they are in
estimating maximum concentration levels.
r It is interesting to compare these results for
the Stuart arvd Muskingum plants with the results of EPRI's
study. Viev.-graph 3 shows a comparison of estimated and
obsorvafi c~-7~r.tr-tions for several points on the frecjuen-..
cy distribution of tha tracar data set collected at Kincaid,
These statistics, indicate that CRSTER slightly underesti-
mates hrurly measured values but thase differences are
relatively small ranging from about 10 percent fot the max-
imum concentration up tc about 30 percent for the upper
Comparisons of rcf.el estimates and measured con-\
centrati.?-r.=; fr"~ the routine FOO r. -f^rk show similar
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
,202, 234-4433 WASHINGTON. D.C. 20005
-------
. results. In viewgraph 4, the cumulative frequency
distribution of observed and estimated concentrations for
A
482 hours is disolayed. The distribution of estimated
3 ' '
values parallels the observed levels, especially for the
4
70th through 98th percentiles.. The model appears to
5
slightly overestimate the higher concentrations; the med-
6
ian of the 30 hichest predicted values exceeds the med-
7
ian of the 30 highest observed values by about 10 percent.
8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Okay, viewgraph 5 summarizes one other important point?
that seems worth emphasizing from the EPRI Study. The
summary of findings for that project indicate that overall
CRSTE'R showed no systematic bias of over-prediction or
under-orediction of highest concentrations and that
CRS'TER demonstrated the least bias when compared with
tvro other plune models.
As expected, correlation coefficients-between
oLoer-'Sc; and pr»:'iccec! concentrations vere generally.
indicating that CP.ST7R was not particularly useful for
predicting ths hov.r-by-hour history of observed S0? or
tracer conc-intr-.tioris.
The second example-I have involves an evalua-
tion of the ?A" r.oclel using St. Louis S02 data collected
as part of EPA's RAPS project in 1976. The model was
.U50-;1 to -^Fti-.r.t-.e hourlv :.verar?« "G0 concentrations at
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
meterological data as inputs. The hourly averages of
measured and predicted SO values were then used to cal-
culate 302 concentrations for various averaging times
from three hour averages up through annual averages.
Okay, viewgraph 7 shows a.comparison of the
composite 13-station cumulative -frequency distribution
for hourly averages of predicted and observed SO con-
centrations. Overall, the agreement is very good when
viewed across the entire network. Where frequency dis-
tributions from, individual stations were examined, it
- appeared that -concentrations at the center city sites
are slightly overestimated. Further analysis of this
indicated that this over-prediction is probably related
to the '.ray emissions from area sources are handled and
not, necessarily, to the model itself.
"h-en hcur-by-hour comparison between estimat-
r-f; an?. :"'.' sur sJ. ~C~ are examiner., the matchup is not near-
ly as 'joo.-l. r.s expected, the correlation coefficients
between cbssrvsc. and predicted concentrations were gen-
oralli" l;.sr, t:.r,n C. 3', ir.u:'.catir.-j reasonably poor agree-
ment. I.e.--; pairvise correlations are fairly typical of
plume oriental models, r.s I mcr.ticr:cd, and reflect the
problem cf accurately accounting fcr plume positioning
Or. -2 'ir ^r.i.l.:"it ." .:";. m
merge:!, and that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
1
levels improved as averaging time increased. In other
i
words, on an absolute basis, the RAM model estimates for-
3
quarterly and annual averages were closer to measured
levels than they were for hourly and daily concentrations
5
However, on a percentage basis, it appeared that differen-
6
ces between nodel estimates and measured values were not
7
a strong function of averaging time.
8
The next viewgraph, slide 7, summarizes the
*/
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
is that differences between measured and predicted SO
2
or results and conclusions from these evaluations. The
first conclusion is that peak short-term concentration
levels are estimated fairly well when the precise loca-
tion and.time period are ignored. The error in estima-
ting highest concentrations was typically in the 10 to
40 percent range depending on the model and averaging
time.
The second conclusion is that each model was
less successful in reproducing measured concentrations
at a given station and time period as indicated by the
gnnern.lly lov? correlation coefficients between measured
and estimated concentrations. 'Both RAM and CRSTER had
low correlations betv.-eon pairs of measured and estimated
values- probably since transport winds crucial for plume
rr.odals arc not knovrn vith precision.
Our cvorall ccr.".li:-sdor. fron these limited nunbo:
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
-49
of evaluations is that models seem to perform reasonably
2 well in estimating concentrations that have been of great-
0 est concern to regulators and decision makers.
o
As I mentioned earlier, EPA has developed a
plan to perform a systematic evaluation of models used
. for regulator" burooses. The purpose of these evaluations
b ~
_ will be to document the strengths and weaknesses o-f each
model using the measures of performance recently recom-
_ mended by the AMS.
y
In the past, the availability of good data bases
and the time of regulatory demands have'limited our eval-
uations. Now, however, several comprehensive data bases
collected by private industry and government, are avail-
13 |j
able which should be reasonably adequate for estimating
14
th-a accuracy of both proposed and recommended models. As
15
pirforr.^nco evaluations are completed, "independent: tech-
paa-r reviews are plar.nsc1.. Wizh the completion- o-f-«cg
the^- rsvisvs, results vill b- i.-.cluc'.ad in E?A's model-
18
19
20
21
22
23
24
25
ir.g gui^e:ines.
~ho next vievrrrr.ph, slice B,. shows the models
that are to be evaluated. The r.cdels are. grouped into
eight categories that v.'ill he evaluated sequentially be-
ginning with the evaluation of rural models. In all,
the ii-jht c":to:jc-i'ic:.- of r.-.Ciljls include, as you can see,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
50
source comples; (4) reactive pollutants, (5) mobile
source, (6) complex terrain, (7) visibility, and (8) long-
range transport. Each of the models in a given category
will be evaluated together using the sane data sets so
that relative comparisons of model performance can be
determined.
Right nov; work is underway to evaluate poten-
tial data sets needed for the evaluations and also to
develop a generalized data base structure to accommodate
multiple data-sets and model categories. Selected data
bases meeting nininum criteria in terms of representative-
ness and quality will be created and installed on EPA's
UNIVAC computer for subsequent use in the model evalua-
f
tions. .
Along v;ith the data base review, we are build-
in-house expertise in (1) applying performance statistics
co actual data a^t^, (2) gain-ing knowledge about :«cclel~
ner.Torr.anc2, and (3) identifying problems related to
interpretation of performance r.easures. Initially this
involves th,~ pr^-.rration and testing, of computer soft-
.ware to calculate p3rfcr~ance' measures. Existing soft-
;~-;:~ ;:?.r':^--;cs ars !"«ing U22c1 to perform limited analyses
of hov.T r.oc!2l performance measures respond to random var-
ic.t..Lcr:.;"- ;_;". r.'.cr..";! input ^-^i."£T.~..2tGrs suc.i as trcxnsport v.rinds
. . .., . , -. -i .. T c .., _ r. i_ _ v i - ^v ,-*.-. 4. . -, -1 ^- ,,_,.7
_'...-»... -.. .. ^ , J. . '. . ^ V- -. . . v- 1? _^.- - -i. -1- O . . *S 4. tl \* U .*. \^- C4 J. *-/^.* 'fci * . C* rj
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
on model performance caused by data uncertainties, when
we may be able to suggest elementary standards for model
performance some time in the future.
We expect that it will require several years
to fully evaluate the models shown. A contractor has
begun work to evaluate the category of rural models which
we hope will be completed by late this year. Detailed
planning is also underway for evaluating the category of
urban models, and hopefully, these performance evaluations
will be completed in calendar year 1982. Depending on
"the" availability of resources, models in the other six
categories will be completed in subsequent years on a
schedule consistent with the availability of data bases
r-
and resources. . .
A number of issues will come up as more exper-
ience is gainsd in evaluating models. The problems of
space ar.d time correlations?- of. model estimates and meas"-
urei values can seriously affect confidence statement on
r.oclel perforr.Gr.C3 statistics. For this and other reasons,
no one really kncv.-s yet how to interpret performance sta-
.tistics nor doas any one realy knov: how to translate per-
formance statistics into measures of uncertainty that can
be readily used by the decision maker. The active partic-
Grs ar;JI statisticians is going to be noed:
£' Jt, ._ W I i C -. ^-N.
::ar. '".- better understood an
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON, O.C. 20005
-------
I ' treated effectively
2 As I indicated earlier, we plan to incorporate
3 the results of these performance evaluations into a scien-
4 tific peer review of models used or proposed to be used
for regulatory purposes. EPA is now discussing poten-
tial mechanisms for conducting the peer review with the
A?IS. V7e should know more about how this will proceed as
the rural model performance evaluations are wrapped up
Q i earlv next vear.
9 |
10 To avoid any unacceptable delay in having at
». least-some--understanding of how models perform*, EPA is
developing under contract a report on model accuracy that
should be available in early 1982. This report will pro-
f
vid= us "with a summary statement on the accuracy of each
nodel for vhich oerformance information is available.
15
,. The the extent "osaibla, these limited statements will
16
17
18
19
20
21
22
23
24
25
v.ro st-.-.tia liirs cr: the form ar.d type reccrr.mer.cled by the
?-V", rerc-j.-isir..^: that sxtsnsivo rsanalysis of old data
bciS23 is i~cossi~'le in this s'lort tir^s frame.
Th-rf: vou
*'.R. TII7/AP.T: Thank you, Bill.
T\\3 next rranelist is .v.r. Bruce Turner, who is
Chief of the I^r.vircnrr.ental Operations Branch of the Office
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17 '
18
19
20
21
22
23
24
25
52
acitivities.
MR. TURNER: Thank you, Joe. I am very pleas-
ed to be here and see friendly faces that I couldn't see
through the fog in Raleigh-Durham this morning.
The work that I am reporting is conducted by
the Meteorology and Assessment Division. The Division
is part of the Environmental Sciences Research Laboratory
and is located in Research Triangle Park, North Carolina.
Most of the meteorological talen in the division is provid
ed by scientists on assignment to EPA from the National
-Oceanic-and"Atmostpheric Administration (NOAA), Depart-
ment of Commerce.
The major extramural research effort being con-
r
ducted is the Complex Terrain Modeling Program consisting
of field studies and theorectical model development. The
conduct of the study is viewed as a five-year program,
starting with a small, relatively ideal, terrain feature
and. then progressing to more complicated site configur-
ations r.ora tyy>ic?.l ~f source-siting problems.
Iii ssehinc; a conical shapei -terrain faature
on a plar.a, Cinclsr Cone ?,utte in Idaho, about 100 meters
high '-/as selc-ctof-. The size is amenable to more dense
sampler coverage for a set budget than a larger feature,
:.co:il shape is ex tc-rely useful for comoar-
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
towing tank.
Field v/ork was completed in September 1930 on
flow visualization experiments used in planning the 18 .
tracer experiments which were run in October and November,
1980.
Of 99 hours of tracer releases during the first
15 experiments, 54 hours were identified as having the
tracer plume move.toward the hill with, the tracer sampling
network in good operation-. .For 38 of these hours the
plun-,e centerline went over the butte depicted schematical-
ly ir. this diagram (Figure 1). For 16 hours the- plume
was below the top of the butte, shown schematically here
(Figure 2.) ...
r
These data are being used to test selected mod-
els including a potential flov; nodel.
.'lost-of the continuing v/ork the rest of this
y2~r will consist of analysis, of -the c'ata collected last
fall ancJ the proper archiving of that data for continu-
ing use. r.o field work will b'e conducted this calendar
year.
Future plans point toward a cooperative field
study ir» the fall ,of 1932 with other interested parties
in order to utilize the funds available most.expeditious-
i.,
j~ \ .
Corelc-nsr.ting tha terrain v:ork being acccr.pl isho
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
1
2
3
4
5
6
7
54
by contract is an in-house effort in fluid modeling.
In this program a series of experiments were
performed in the towing tank to duplicate the field ex-
periments for one hour in particular (0500-0600 case 206)
representative of very stable atmospheric conditions.
Figure 3 shows the measured tracer concentra-
tions for t-e field study.
8
tests -are :
Gross stability classifications are not suffi-
cient fcr~characterizing concentration distributions.
However gross classification would appear to be reasonable
for predicting the values of maximums (but not locations).
f
Surface concentration distributions around the
hill are extremely sensitive to changes in wind direction.
The location of the -axi-urn concentration shifted throught
;-\n an-le of 'rproxi~a::2ly GO ccrrre-io locking from the
source vith a shift of only! 10 dc-^ro'es in wine! direction..
Because of the absence of Icxtf frequency fluc-
tuations _ir. wine spsoc1 and direction in the tank, the
concentration distributions observed in the tank were ex-
ceec.ir.-^'ly r.arr.v.:; maxirr.ur. ccr.cantraticns were 5 to 10 tine
larger thn:i these observed in the field,
"'.n 3ttvjr"."t '."e.o r'.ctdc. to oivul.it-s the lcv.T fre^uen
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
I (202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
8
9
10
11
12
13
14
15
16
17
IS
19
20
21
22
23
24
25
from tov;s done at a series of- discrete wind speeds and win<
directions. The resulting pattern is shown in this fig-
ure (Figure 4). This attempt was moderately successful
in that 30 percent of the model concentrations were with-
in a factor of 2.5 of the field concentrations. The high-
est model concentration was also a factor of 2.5 higher
than th3 field m-.ximum.
This study demonstrates the considerable use-
fulness of physical modeling for simulating dispersion
under stable conditions.
A comprehensive evaluation of four photochem-
ical a-ir quality simulation models has been performed.-*
The four models are the Photochemical Box Model (PBM);
the Urban Airshed Model (UAM), also referred to as the
sr,l-AlRS"ZD TiOdsl; the Langrangian Photochemical .Model
(L?*I) , also call-id the T^.T-ELFTAR nioclel; and the Liver-
. . r.-_^jg
*»
characteristics and level of deto.il modeled, v:ere run in
pr.rall.jl usir..; ."-ti frcr. the r.j-jio.ial Air Pollution Study
or1, for r'.ie sta-xy from the tv.'o-yaar period.
:.: tc7t£. 'MV "T:;, t'.-.:' l"^."." '-/as found to qross-
ly' uii "!.-:i'ii >i.:.a^o c:;iCa.'it anv.1 cvcre^tir.ate ^10, "Q^/ and ,
z'.'.-.: /.I-;".:.'! :^ nc': -a^ur'lv transfsrvablc to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS-
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11-
12
13
14
15
16
17
18
19
20
21
22
23
24
25
to other computers, is not likely to be suitable for fur-
ther intensive investigation.
Simnary statistics for model performance of
oxidants are found in this figure (Figure 5) which has
two sections. The top section shov;s comparisons between
predicted and observed maxiirm, at the time and place of
the observed naxinum noted with the subscript S to indi-
cate "specific". The bottom section gives comparison
of observed maximum with predicted maximum (regardless
of time for P3M and regardless of tiine or position for
UAM) noted with the subscript I to indicate "independent".
The average difference is used as a measure of bias. The
mean absolute difference is a measure of scatter. The
independent estimates will be equal to or higher than
the specific estimates so that the average difference
vrill be cor. c r.ore nsgative in c:c-ing from specific to in-
c " ^ *i c* * " **
The Photcchar.ical Bo:;- "cdel is seer, to overes-
j ti-ate concentrations in the -rvca'r. , both for the specific
c.nc*. for the- ina3j.-2nf.3nt r^xi-.a, v.'ith soir.ewhat larger scat-
ter for tl'.o independent naxir.a. The Photochenical 3cx
"odal i.s r.ost applicable to st.ignr.tion conditions v;hcn
the 2r.tl_-e urban areas is nest lively to act like a huge
rea^tc
(202) 234-4433
The Urban Air?h-;: Vrclcl hns -~ bias to:;ard
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
I
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
underestimation an'd the scatter-decreases'when considering
the independent maximum.
Although the Langrangian Photochemical Model
has very small bias, the scatter indicates both large
over- and underestimates for particular cases.
It should be noted that these measures should
be examined relative to the magnitude of the observed
concentrations shown at the bottom of the table. Since
the spatial resolution is greater for the UAM and LPM
than for the PBM the observed concentrations over the
smaller areas are higher. Scatter is about one-third the
mean concentration for the PBM and about one-fourth the
mean for the UAM for independent maxima.
* A technical report, "Evaluation of Four Urban-
scale Air Quality Simulation Models," is being reviewed
ar.5 should ha available from !-7TI5 about the first of the
year. Also, f.:c papers v;ill be presented at the San An-
Evaluaticr. vor'-: on an additional 1C days data
is ur.cla:.".;?.-_ for the three mc-cal- including r.cdificaticns
which ~.;r- improve their performance. Completion of this
vrorh is expected ir. early 1982.
I:-, beginning a prcgram of research on long range
transport, deposition and removal by precipitation, the.
Ion" t = rm' r. ot?c 1 !!*."R"-"A? clcivolc^o-:; bv r-T^I International
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
I'" v:as adapted and applied to-Eastern--!Io-rth America'in 1979.
2 The adapted model, EMAMAP, v;as used to calculate month-
3 ly, seasonal, and annual distributions of sulfur dioxide
4 and sulfate concentrations and-wet and dry deposition.
5 Model calculations were based on emission data that in-
c eluded both specialized data prepared for the Sulfate
Regional Experiment (SURE) and the EPA's National Emis-
sions Data System (NEDS). Emissions are compiled on a
gridded format with kO kilometer resolution.
10
21
22
23
24
25
Comparison have been made between calculated
.. and measured SO and sulfate concentrations considering
12 squares 140 kilometers on a side, for 6 months between""
August 1st, 1977 and October 31st, 1978. Measurements
lo
consist of about 50 EPRT-SURE sites for SO_ and sulfate
14 ' *
. plus additional data stored in SAROAD.
15 |
, This fiqure (Ficuro 5) shows the calculated and
16 ' " . .... »
r.sasureci ror.thly so., concer.trr.tiDr.G for August 1977.
~'2 rar.xir'.'.'ir. of 59 r:icrogr?.T.s per cubic meter occurs' in
scv.tlv.:23tarr-. Ponrsrlvania. The -^.easurad maxinun of 47
The- nsxt figv.ro (Figure 7) shews the calculat-
ed ancl :'.i.-i5,:ra:1 ronc'.ily avorag?'1 sulfate concentrations
for August 1977. 7\ote that tho ~-?.xir.Viun is predicted to
;:^ t'r.-,- ;^r:-> of or b?;L;.'.:, tho .-.::- Irun predicted is tho
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
same as the measured, but they do occur at one grid square
away from each other.
Research is continusing along the direction of
making a multi-layered model for SO , sulfate, and NO
2 'X
(the current ENAMAP is one layer). Under the memo of in-
tent between the U. S. and Canada, four U. S. models and
three Canadian models are being evaluated with a common
data base for January and July 1978. We are'responsible
for the evaluation of ENAMAP..
A major in-house effort continues to be the de-
velopment -of -the Regional Photochemical model as part of
"EROS (Northeast Regional Oxidant Study). This model,
being formulated in.modules, is designed to simulate short
f
tern (one to three hours) mean concentrations of NO, -JO2/
0 , CO, PAN and four groups of hydrocarbons. The domain
of th3-nodal is approximately 1,QGQ kilometers on a side.
It ',:ill initially bs appliacl -to the northeastern U. S.
for comparison v/ith data resulting from NER03. Currently,
Ccita gathered in 1979 are being, utilized. Budget cuts
have clclavod the availabilitv of data Gathered in 1980
-* - . -*
until early next year.
The r.\oclel contains four layers for vertical
resolution. The surfaces that separate each of the layers
are variable- in both s-./ace and ti~io, and are calculated in
t~ ':?. :o into account ir. an
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
cptir.im manner many physical processes,'""including:
horizontal transport and turbulent diffusion,
buoyant rise of urban plumes,
4 rnsso and larger scale vertical notion induced by
5 terrain, and divergence,
low-level stratification of surface emissions at
night,
cumulus cloud venting of pollutants,
subgrid scale chemistry effects resulting from point
and line emissions into the 18 by IS kilometer grid cells,
.. surface deposition, washout, and rainout.
The objective of this effort is a validated
13 nodel which can be used to assess the impact of oxidant
control olans on regionallv transported ozone from one
14
urban center oh others. The regional scale modeling pro-
1O
, i gra.Ti is .-"-si-na^ in a mar.nor which allows for the logi-
16 i . >
17
18
19
20
21
22
23
24
25
c-.l rl-sv^lcp-^nt , ref i"2;:;er.t, ov~lunti.cn, and verification
- - -^
of uh-2 ~c-~.j:l ar.c! its subccr.poner.ts; and provides a fra.~n-
'.-ork fror. which ncf.el research and dovelopnent can evolve
to r*.23t futv.i:;- r ~: rv.ir -r.c r.t z ccr.cc-rnir.g long-range trans-
port. T>..^ r.ocel is constructed In a generalised form
with modular components describing the various physical
ar.cl chsr-ical processes cc-rurrnng on the regional scale.
Such -a frsnevor:: allows for refinements, deletions, and
"» f*r> cr c; o
v' *
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
Ti O
*"*
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
re-tooling of the basic model. - " -
The development of the regional oxidant model-
is scheduled for completion in 1984 and is considered as
phase one in a multi-phase program covering other issues
relevant to regional modeling.
The addition of modules to treat the transport
and fate of fine inhaled particulate matter will begin in
1982 and should be completed inl!985.
Other possible _additions include: long distan-
ce transport and fate of sulfates and nitrates, regional
patterns of. regional acid precipitation, and regional. _
patterns of visibility degradation.
Additional models are included in UNAMAP (Ver-
sion 4) which was made availble by NTIS in March. An
eight-page handout giving brief paragraphs of descriptions
for each of the 21 models is available.
A rvrir.cinal research effort pertaining to the
~od3lir:g of tha impact of point' sources is developrr.ent
of a disparsior. -chsr.e utilizing fluctuation statistics
as ir.at. Tho v^r'-: is proceeding and all available dis-
persion studies which include fluctuation measurements are
hsirvj utilizer! in the d3XTclcpr.ent of the method.
T>To are anxiously awaiting the availability of
the rpRT v.ata for the I'incaid study as this will be the \
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
A
of the RAM Using the RAPS Data'Base from our contractor
3
on the project, we recognized that sorte additional work
could complement the usefulness of those reports. That
5
work emphasizing model performance near the extremes,
for the second-highest cnce-a-year values, has been com-
pleted and will be presented later this month at the 12th
8
International Technical fleeting on 7u.r Pollution Model-
«/
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
t*.
the scheme.
Upon recipt of the final reports on Evaluation
ing anc" Its Application sponsored by !IATO. Some results
are 3ho'.:n in t'.ie Summary Table (Figure 8) . Since there
'.\ras one station, 104 that suddenly started measuring high
concentrations during the last two months of the year, we
r
suspect a change in emissions, the network second high-
»
est concentration is greatly underestimated when all sta-
tions arc- ::icT.u-!ec". Tha green ratios are estimates over
-.-3z-3'\- *-.-- : 3 . :.-:'' for L'.v= three hour period is .39. And
for :- 2-?-V.r/.:r --rioc5 is .2' s^ov:ing very severe un-.-
,
clsresti-stG.-, ct the r.of.el. 'Tl-.en station 104 is neglect-
ed, a:-..", or.ly t'i3 ether 12 stabler." are cor.sidc-rocl, the
r.\e?isuroa conctantratiDns con-s clovr. in the range of the
:\~5el .?-^t.i--o-.t-25, hue still rcrjul-ts in under^stir-.atas for
the 3- to 2 -'-hour averaging tir.a .
NEAL R. GROSS
COURT REPORTERS AND TRANSCMSERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
* s; .t 63
» f^ '.-
indicate that when station 104 is eliminated the model
2 overestimates the annual concentration, having a ratio
3 of 2.21, overestimating by more than a factor of 2.
4 Another evaluation study is underway related
5 to extremes, specifically, again, second-highest once-a-
6 year concentrations. This time for point source model-
ing. Data for this study was obtained through technical
8 reports prepared by contractors for EPA. In addition to
comparisons of model estimates to measurements for each
10 monitoring station, study of the maximum second-highest
11 concentration for each network was examined. For 12 net-
12 work-years of data for the 24-hour averaging time, the
13 - model underestimates the network second-highest 8 out of
14 12. Only one value, an overestimate, is off by more than
15
16
17
18
19
20
21
22
23
24
25
a factor of tvo. For 11 network -years of data for the
3-hour a<-£ racing time, the model overestimates The net-
work sorcnd-!?.i jhest 7 out of 11. Two values, both
y more than a factor of two. '"'
r£3, ?.r3 cf
It takes a lot of data to be able to look at
th.'2 ~'.'.t. ~?~.\- . 7.n d "oil wil.l note that there are just vcrv
few men suras that you can mak'e with 12 network-years of
data, "j-.-.t that represents an awful lot of effort on the
part of the people doing the sampling. And 'I than': them
I-..!::'.-.criczl record of noclol
NEAL R. GROSS
COURT REPOk.c^ AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
O'i
. performance these studies v;ill aid in assessing model
usefulness in impact-studies.
i
Thank you.
3
MR. TIKVART: Thank you, Bruce.
4
Our last speaker on this panel is Dr. Douglas
5
Fox. Doug is Chief Meteorologist of the Rocky Mountain
6
Forest and Range Experiment Stations, with the U. S. For-
7
est Service.
8
This morning, Doug is speaking on behalf o.f 3
the American .".eterolocrical Society, and is reporting on
10
the AMS-FPA-cooperative agreement.
11
Doug.
12
MR. FOX: Thank you, Joe.
13
I notice that there is at least one change since
14
ve were here talkincr in 1977. ' And that is that the podi-
15 - " '
urn is nov or. ~r.3 ri-rrht side -of the sta-^e instead of on
16 " ' " *
>.ft £ic:~ of th3 sta.--5 vhcre it v:as before. I don't
17
kr.ov; if that's a sicrr.ificant chancre or not.
18
It's- =1 real pleasure for me to be here to pre-
19
sent the ro^-lcs en the America:'. '!-3ierolocrical Societv-
20
21
22
23
24
25
~T-A ccpr.-jrr.tive agreement on .air quality modeling.
As m?.ny of you knov;, 7PA and the AMS entered
into this cooperative effort in September of 1979 for the
purpOD-2 o" yirovidir.c; tc^hr.icr.l .?.esis';;.p.ce and o. civ ice to '.
r?.' f o:"-"?1:?'" ;- -.:--* i7;.;;i?-'" .of r1?:5.?" in~ mandate:! by the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Clean Air Act.
In particular, the AMS to the following four .
points, number one, conduct a .review of, and make recom-
mendations on previous work concerning air quality models
accomplished by EPA.
Two, undertake a general review of the state of
knowledge in air quality modeling.
Three, offer suggestions concerning recommend-
ed air quality models and criteria for their selection.
And, four, evaluate data base requirements for
use with models.
The AMS motivation for this cooperative effort
is straight forv;ardward. namely, to attempt to provide
a funnel through which the atmospheric sciences commun-
ity might more directly feed new scientific information
into EPA programs and policies.
7hs cooperative agreement, which has been unclor-
v;ay for so~e 13 months., had bee?, under the direction of
a 3"^.ll st^e.rir-- ccmittes consisting of two former chair-
men or; tl\-~- .'.!!? Corciittee or. Turbulence and Diffusion,
r.fmsly Harry 1 Sanderson and Stavs TTr.nna, and two former
chairmen o.r the, ?-."!S Cormittee on Air Pollution, namely,
ny^alf and ^rv.ce T-r^n.
. Dr. Free3, ^rhite. Fred v:ould you stand up,
hr.ov you hate to -do that, but achncvlVro yourself,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, O.C. 20005
I
Fred
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
has served as staff director of -this a-ctivity on a part-
time basis. And that's a snail statement that doesn't
»
convey anywhere near the time, effort, sweat, and blood
that Fred has given to us. And I think the whole commun-
ity of air quality modeling is well served by his efforts.
Larry Neimeyer, Herb Slater, and Joe Tikvart
have been our principal EPA contacts,
The steering committee met 20"times during the
course of this project, to date. We have formally util-
ized tha advice of over 40 individuals, and 100's of
others on an informal basis.
There assistance an." thoughtful suggestions are
gratefully appreciated.
Ne have- prepared a report to EPA documenting
sorr.2 of these activities. Unfortunately, the report-is
ir. ths final stages of review and net available at this
o
T"e have asked *'r. Tikvart to send all registrant
f this? -.-seting a copy of our report when it becomes a-
The -V.S r.c.T.bers of the stesr.ing co;;LTiittee have
atte;-;-::t-rd to draw a fine- line b^tvreen offering advice
or. sciar.ti.~ic aspects of air quality modeling and becom-
ing involve:! in regulatory decsons.
7:vr ?::?.-'plD, ':* r.v.'o "c*- b-.-en involved in the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
drafting of a review of.proposed new guidelines. Nor
have we commented on any specific EPA modeling decisions.
TTe took no formal part in the Airlie House workshop, which
:tr. Tikvart "enticr.ec! earlier.
We have, howeverr felt it necessary to comment
on EPA's use of scientific information, the EPA's pos-
ture toward acceptance of new ideas, and on some of the
knotty problems confronting regulatory modeling. -
In this regard, our report, probably, will not
represent a totally objective, unbiased, purely scientif-
ic reflection of the state of the art in air quality mod-
eling. Rather it represents a collection of thoughts on
>
the subject, colored to a large extent by on-going debates
about the appropriate use of models,'.and modeling, in
r
the management of air-resources.
T3 are prepared today to read the recommendation
that 3:re c'.r?.ftac ir: ov.r report. 3ufc, first", we~v.*ish to
ur.-.:l^rscore tYu~ scir-r.tific validity of using dispersion"
In a ro?.l sense, it is this management, or the
res:! for r.anage.v.oiit, that drives the 'use of air quality
modeling in regulatory applications.
"hile scisr.tists have been attempting to pre-
dict ground-level concentrations of pollutants, for many
yonrs, t':o Ic-gal r.ar.r.atc ':.o make rpcc.ific economically
- :nifi ?-.:-_ .ocif-'.^.-.s, i-r-.-of.';r. ^-jcv. ; vedict ior.3 , has
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.grown increasingly important as --the -Clean Air Act has
.
evolved to its current status.
There is overwhelming scientific support for
the hypothesis that the ground-level concentration of a
.pollutant can be related to the amount and nature of e-
missions, the meteorological elements, especially the
wind speed and direction and atmospheric stability, to
the chemical transformations, and to the amount of atmos-
pheric turbulence present in the intervening space be-
tween the emission and the measurement point.
However, nearly all the parameters of this very
complex system are stochastic, or randon, in nature.
Thsy are also poorly sampled, or measured, for any par-
tiqular application, and must be predicted to make pro-
jections of what might happen in the future.
"These facts lead to uncertainty in, and inaccura-
cy of, any particular concentration prediction.
.** -
Tha scientific reality., which we wish to under-
score, is simply that models will also predict, a 'number
which has so.- degree of. uncertainty associated with it.
It is expected that better models and better
data will reduce the uncertainty, but not eliminate it.
is uncertainty does net,
in our mine's, eliminate the usefulness of modeling. In-
V.VDG::, to th~ c:-:"ant that ~ouels reflect our best
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
I
2
3
4
5
6
7
8
9
10
11-
12
13
14
15
16
17
18
19
20
21
22
23
24
25
understanding of the relevant; physical processes, they
represent a scientifically sound and objective means for
taking into account difference in source geometries and
topographic and ~3t3orological settings and assessing
arubient air Duality irapact.
Thus-, models provide a logical and environmen-
tally equitable basis for decision making.
We feel that an appropriate scientific response
to the reality of uncertainty is reflected in your recom-
r-.entions. ^nsically, tho reconrr.endations represent three
r^f2nsT"s lv . lin^s of thought.
»
"a-isly, one, improved research is needed to re-
duce the uncertainty of modeling estimates.
*. Two, 7.PA must strive to quantify the uncertain-
r.~.r;'::n should he coc
""-.? ::ecor.-:-.2;u1ations are nrssanted, rcn.Thly, in
^pacific. ."-.n d, :-.;.;, I'll recid cur specific recorrr'.onda-
w .' . * U *
"v.~bnr n-e, V2 rice "~-ir.d that EPA strongly en-
^ or ^ "* t "">-'r- i ^ '^ °
.t^,r- '
c-ric c?iv"or.-3.1r>n models for assis-:
to 'he nancrcront of
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
j!
Ss
r
of air quality. Based on our review we have concluded
that these models are useful and appropriate technical
tools, but the interpretation of model predictions should
include a consideration of statistical conficence inter-
vals.
Number two, -we recommend that EPA continue and
enhance research and development activities to increase
our understanding of the following topics/ which I empha
size are not, necessarily, listed in order of priority.
A. Long-range transport and transformation
of pollutants thought to be involved in acid precipita-
tion.
B. Dispersion in regions of mountainous ter-
F
rain. "
C. Modeling of chemically reactive pollutants
of concern, to the formation of ozone and other ir.porte.nt
at-T.cst-heric ccr.s titutents V" - ''-
D. ".Ode ling of the atmostpheric transport,
cl^ostions , and. ultimate fate of toxic materials in the
And E, identification of representative meteor
ological information.
The scientific basis for models and the quanti
fication of modeling accuracy vail require efforts that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20003
-------
1
2
3
4
5
6
. 7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
| ' 4" '" ' 71
Current Rederal research in this area, support-
ed primarily by the EPA, DOE, N'OAA, and the National
Science Foundation, is not sufficient to keep up with
ever increasing and changing demands by those who use
the results.
Upgraded research should be the first priority
for the future.
Our recommendation number three. We recommend.
that EPA take greater initiative to anticipate the di-
rection of regulatory modeling needs. Researchers must
be nore sensitive to the regulatory context of their pro-
ducts and ara encouraged to provide a finished product
in a timely fashion.
- f
Regulators must be willing to incorporate new
improved procedures when they become available.
To- bring researchers and regulators together,
;."3 reco:-.;-.er.d joir.tly conducted workshops or other pro-
cedures, dr.~v.:'.nc- heavily on scientists outside E?A, to
improve the U3~ of nev: schientific findings and to pro-
vide r.utually acceptabl.?. direction to EPA programs.
Our re corur.er.dat ion number four. T-*e recommend
that r-?.\ adept ar.d issun specific guidance, for example,
as a part of the rSD regulations, relating to meteoro-
logical" data.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
. ?: -72
> *
meteorological data inputs in air quality models is stres-
2
3
4
5
6
7
8
9
13
14
15
16
17
18
19
20
21
22
23
24
25
Suggestion from a recent workshop held by EPA,
which a review of which is published in the February
1931 issue of the Bulletin of AMS, focus on the import-
ance of on-site measurements of turbulence intensities,
wind velocity, temperature gradient, and mixing depth.
We endorse the recommendations of that workshop.
Number five, we recommend that EPA adopt -a~spe-
cific set of measures to statistically evaluate new and
existing model performance. One sugges.tiong of such mea-
-- "*
sures is based uoon results of a workshoo conducted under
10
11
12
the Ai'S-EPA-Cooperative Agreement in September 1980 at
f
Wood's Hole, Massachusetts. Results of this workshop
I
wer-3 published in the .May 1931 issue of the Bulletin of
W-.ich, >,y the "ay, I understand is .be-in-g.-4.3»suec!,
':33n '-".iiju~S'ia by previous speakers here this morning.
::umber six, we racornend that EPA review all
nuclei3, ir.cladir.:; those currently reccmr-.ended in the mod-
eling guideline, and these being considered for regula-
- C C\ 110 "1S
This review should h-2 based upon a statistical
a!x;v~ j.:~'. ':~ . r= ;:.;;^l;.i ficr.lly VA£5o:l ;.:;:-: rovic-./.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
73
The peer review is intended as an evaluation
cf the scientific nerits of the r.oclel. We believe that
the peer review should have the following characteristics.
A. Adequate financial and staff support should
be-provided to allow the conduct of the statistical eval-
uations, and to ensure feedback of review findings to model
developers.
B. !To less' than five reviewers, knowledgeable
in the science of r.odeling and in the regulatory appl^-
caticns of r.odels, should be used.
C. Reviewers rust be provided the results of
the statistical performance evaluation, before they con-
duct the--.rest of their review.
--. " And D, reviexvers should consider all models
in a given category. All the'rural models. All the
urban r.oJ.slr . Etcetera. .._.__,
T-'e further rcccr:.r.--nd that a group distinct
frori those responsible for regulatory modeling within
*>
r?A.as3-u::.o cvc-rsll responsibility for this review.
?,e-err--ar.d?.tier, r.u^her seven. We rscorr_~end that
C?A rsfccuo its -efforts to achieve consistency in model-
i^? ap?----"t-.-r>s. Th-2 *ocus should be shifted rtway from
issues o.". specific nur;arical agreement between alterna-
tive ^io::-i"I.s (and tho ur>e of ?. limited set of models) , \
r.r.rl tc,'.:=,:~~( i=~uos cf ::c")S "'.;. t c:\ce of r-.prroach to modeling
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
probler.s.
Ue are concerned that EPA has overdrawn its
mandate to specify r.odels with reasongalbe particulatiry
to imply numerical agreement. Instead, EPA should de-
valope a review process that is consistent, for example,
in terns of meteorological data requirements, types of
dispersion nodal outputs used/ and the like.
This review process should also treat new mod-.
else responsibly, recognizing the admonition to EPA-con-
tained in Judge Robinson's Alabama Power decision, that
in questions of modeling controvarsey -- and we quote
^
Judge"Robinson, "I?A should have the should move to
adopt the more accurate procedure."
. ;. Our recommendation number eight. We recommend.
that E?.A recognize differences' between tv/o numerical re-
3u!ts as V^-in-j significant only if they are outsiae a
c-nfic^co ir.t^r--::! str^isticr.ll--- constructed, about tha
For -T-.-cample, mc.ial c-roonents, such as plume
ris-2 of ci£-7..ers:.or. coefficients, may be based on differ-
ent theories cr data sets, but still may produce results
that agroc- within the expected accuracy of the method.
Our recommend?ticn number nine. T7e recommend
that ?.??-. cc-nsif'.-sr adopting a procedure to evaluate com-
-"1
tai^'lnrc'.s on the bncis of
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE, NW
WASHINGTON. D.C. 20005
-------
more robust statistics than the highest second-highest
value. For example, consideration of a NAAQS violation,
or PSD increment consuption, could be reformulated to
3
. the 95th percentile,
4
This approach would necessitate an adjustment
5
of the numerical value and/or the manner in which a stand-
6
ard is calculated to ensure that emission limitations,in
general, would not be more or less stringent than they
8
are presently.
10
11
12
13
14
,15
16
17
18
19
20
21
22
23
24
25
9
Such adjustments may' need to differ for excep-
tional circumstances. For example, those involving com-
plex terrain settings.
The basis for this recommendation is that more
frequently occurring meterological events can be pre-
dicted with greater confidence than rarely occurring e-
vsr.ts.
T7e su-;cest that not only would this char.-je per-
nit -.c.v-l"; to predict the ir.or-2 frequently occurring e-
vcp.ts with greater accuracy, but also- could allow a pre-
diction -- a "reduction in the quantity o.f meteorological
clata' ^r-ii'sr.tly require.! as input: to models.
The procedure would have tho additional bene-
fit of r-2c.u^in:; the su'.: jectiv-.- ju.vjor:ontal aspects about
what cor. - titutes realistic worst case conditions. ;
r ^ ::-"-. ::.^^Lon- r.unbsi 10. "2
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
76
recommend that EPA call for the use by decision makers'
of more of the results presently available from modeling
~ or models to supplement the predictions of the high-.-
est second-highest value, or a substitute as discussed
above.
Specifically, current modeling practice forces
the calculation of concentration for a large number of
time periods, and at a large .number of receptor points, '
in order to identify the highest second-highest values.
This output provides valuable supplemental in-
formation to the decision maker on the frequency dis-
tribution of high values, as well as the geographic area
of .exposure,.-and the relationship to meterological pat-
r
terms over different time intervals.
This information, now largely ignored, should
4- * -i ri
_».c >_<
ecision makir.q.
.-,., 1, ,, u - « -r _ f
w*i *.'!; tLCt-t..^. w .
I have a fe*--: "or ox act copies of thaoe comments,
if ar.yone i-; -jrticur.rly interested in receiving them,
r.« during the various breaks,
I do, hov.-^vo.r, want to emphasize that the rsc-
;r.3 \:hich I read to you are still undergoing
?-:':.-:-. lc;1 ar.rl thorough revic'-:. Seme changes may \
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
(202) 234-4433
-------
1
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
r,
77
Thank you for your attention.
MR. TIKVART: Thank you, Doug. And thank you
to all the panelists for your presentations. I think .:
they were all quite informative.
If you'll all take a seat, I think that you
can use the microphones in front of you to answer ques-
tions, or to respond to discussion items.
I would like to begin the discussion, this morn-
ing, of the issue of modeling accuracy by asking a ques-
tion of Doug Fox.
That is, I believe in two of the recommendations
you mentioned, there's indication of confidence intervals
around model estimates. Does the AMS report make spe-
r
cific recommendations 'on how these confidence intervals
should be calculated or presented? Or would you care to
make a general observation on that?
MR. FOX: T'*G"11, the- report does not r.^ke any"
specific recommendations on that. I guess the thing I
vrculd say is there is still a fair bit of uncertainty
associ^tcc! with evaluating model -performance, itself.
Quantifying the methodology and the procedures
are fairly straight forward from a statistical perspec-
tive. But I think we don't have very much experience
*
with using the various kinds cf techniques.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON D.C. 20005
-------
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
before anything specific is recommended there's a fair
bit of evaluation and study done, and various types of
techniques are tried.
Ue don't have any specific recommendations .
MR. TIKVART: Okay.
Does anybody else here on the haring panel have
a question or comment for the speakers?
MR. DICKE: I'd like to adress a couple of com-
ments to Norm Bowne on the EPRI report.
T.-7ith regard to the EPRI plume model validation
project," I'n a little confused by sorae of the "findings .
that Mr. Bowne stated, and curious as to why some of the
other statements were not made.
r
I nade about four observations that, inaddition,
somewhat piggyback on Bill Ccx's presentation, and I'm
vender ing if ?7orm would be- -willing to respond to some
Fir at, I wonder if wa're being reasonable in
faulting atrv."oh3ric dispersion -ode Is for poor skill in
2sti:r.;it-ir.'j concentrations .of specific location and time.
For exsr.ple , we've heard ssveral cases, even, this morn-
ing, that tV.o ability to of r.cdel users to specify
transport, v.-ir.cls, and the rate of atmospheric dilution,
in .ir.y ir.rr?:-. 3:-st cf tir.s, is certainly limited. \
n T .-. - '-'-.-, >-... -'--" 1 i~ c;" -''.v.'.'c. indic.Vc sis cle^.r 1"
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
79"
that due to the random nature of the atmosphere we should
not have great expectations for the accuracy of individ-
ual concentration estimates.
As a result, I wonder if we may not have re-
invented the wheel by finding limitations that, in actu-
ality, were taken for granted fay early workers in atmos-
pheric dispersion.
And, in addition, I believe that the EPRI re-
port states that spacial density of the monitoring net?-
work is insufficient for evaluating the predictive skill
on an hour "by"hour basic.
Secondly, I was concerned about the finding of
the maximum-, concentrations were independent of emission
rate. And I thought that was a little bit counterintu-
itive, because we all know that if emission
om a source
are reducsd to zero, c-hac. there can't be any- .rrrc-uru! level
:. _>-.:. Vr c
/ i_
?-.! IP. revie--.---ir.-r that full EP7I report,. £ carre
he conclusion that this independence was a little
bit pre.-r.ture, and that .?. nuch -ore thorough analysis ol
t!-.2 clr.ti were r.jocec., especially in terras of the width
the error bars
r.'.e of tha factors that need to be further
Lra the ral.^.ti:::.-''.\r-.3 ar:::"'^ lc-r.rl, emissions,
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
with the highest observed concentrations. <
In the same vein, I wonder if the relationship
between maximum concentration and downwind distance might
not also ba questionable, because of ihe'limited number
of data points presented in the EPRI report. I'm wonder-
ing if they may not be insufficient to state'with any
confidence that the observed relationship is significant-
ly different from the predicted, using a classical dis-
persion model.
And, of course, Norm did not present any crit-
ical information on the effects of sample size.-and the '
variable meteorological conditions.
And, lastly, I was curiour as to why two of
the most important findings of the EPRI study were not
mentioned", as Bill Cox did. And I'd like to reiterate
those that.for the highest concantration estimates, the
._, .,--.. ._-._..- . -T? n^-cU-l ='-..'-.-=.-'* ;>:-> s'-s^e"~>>i-~ =»t*-c>£-n
of. ov-sr-prs'-'lic-ion or un^er-prediction. And that, on
balinco, th.-t Cr.^7.-. nc-onstratoc the ' least amount of
,~;.~.s for j-1"'c;r?.z. ? -".r.~, ^ic-.z .'M.";h .".'cr.cDnt^'citj.on vciiX'os.
.Me'.-, if there arc ar// of these fcur particular
points, "or-, that you'd likn to respond to --
I!?.. DICKH: -- ycu C2:: -lo so now, or., of course,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5 are not professional modelers. And I did that to demon-
_ strata to them something that we've known for a long time,
b
7
And I think that in my presentation I demonstra-
te
ted why you couldn't. Primarily, wind direction and dis-
«7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
81
MR. BOT7ME: I'd like to respond right now, Jim.
First, yes,I think it's reasonable in faulting
the ability of ncdels to predict a specific time and lo-
cations the v;ay I did today, because much of our audience
that you can't predict specific time and location^
tance tc* rr.axirnur.; concentration. Errors that are part .
"of" the stochastic nature of the model. .And I "did that
. >
for that reason, to bring out- the point.
.-Second, the spatial density being inadequate.
r
Certainly, it was not adequate from the standpoint of
really defining the plume using the sulfur dioxide net-
'"ork, because there are' only 25 stations., - __ »
i:cv:-?vor, the scstial density is entirely acle.-
quate for the tracer net'.'ork. IT3 had 200 monitory. And
in evory test that v-e used in cur analysis, wo bracketed
the plu-.s, h-cth the coir.g from cne si els of the plume to
the 'other and being sure that the maxirrum concentration
-..-.-.=; a-Pt-.^v- '- '-!-! f* ".:»- =»-<~ .-,:-..-'' l-.^-^^-m *- \ 5 ~\ p c: *- r-Tf Clf t"hpi
k . M» O ^M. . - ^^ *. ^>.>*v _ u o v- «. v^ * . ,.'-i *_ ^- ^J ±. , ^«. « i * JL C* O k> C«JL \«. \J J. L.I * tJ
ground l^vel naasurenents.
?ccrs-.r., you point cut the iii'Iapsndor.ce of the \
:o.e -2~ i ." 3 icn ra te .
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
82
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
And you are right, we need more data. However, that was
based on the high 30 concentrations that were actually
observed during that period of time. They covered a rathe::
wide range of emission rates, from about three-auarter
load on one unit to full load on two units.
And I don't say that'that will that partic-
ular event will occur at every power plant. But it just
happened to occur, here, at Kincaid. I did riot find it
a result that I expected." And, therefore, thought it
worthy of presentation.
The -eterological conditions over the highest
30 observed concentrations were relatively well distrib-
uted in the unstable range of meteorlogical conditions.
Third, there you are concerned with the maximum
concentration versus distance. Figure there are insuf-
licior.'j. cl:;::a. "-2 will have r.cre data to look at within
t'..': r.-^:-:^ .sly tenths. That <:;{£ h-asac! on ap^roxirr./v-ily '1:30
tihjtu Vere actually us-3c, v.vre v.-c-
C... 3 .. .."» ~ T^£ -f^ -* . -.c-v w-*i SG: ^ "/ C "*» Oil""
per.s:- thr-.t those were the highest cc:i-
: -..v.cle a.ita S3t. ?o even thoucii we
t'v":. ."rir^t 130 hours, or so, of joint prediction
s
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
*
83
1 Finally, two inportant findings that you said
2 I overlooked. No over or under prediction on the part
3 of CRSTER. I did say that CP.3TER was plus or r.irms 25
4 percent of the highest concentration. And/ I thought that
5 I implied that CRSTER gave the least bias by throwing
0 out the other two models that we described before we went
7 on to then by describing their bias.
8 But I agree that at the high end of the concen-
9 traticn -- and I'm still perplexed by bur data on why we
lO 'over-predict for SO, and underpredict for SP-. And we're
H not done yet.
12 MR. TIK7ART: Do any of the panelists have any
13 additional observations, or questions, of another panel-
14 ist that they wish to make?
15 ?!o. Then, we're open to questions ,or cbserva-
<- ^ --- .- .S.-.-J.X. A-:-.~ ~. ; :":"'" P. "'"- ^.-"' e^-7.-v »--) --5« *-"~.-a -; ;l^v-,->
16 " '-» - - - -"' *- - - " -- " ..«.>,>.-
,TT - .^. 4 ;. _
18 ' = ---'--
*". 3 "" ~ """' --_ ^".", "."Q'jiT' ." ;.: ~ ' i:on cr c'"^rr'Z"*7ticn^
21 GDr.tl^r.-.sr. -- t'.:o centler.en, here, en the left,
22 'ir-1:. :-.\-.-Ur v.v , p?.ca = o.
no I thirJ: it's or.3. Just n;oke sure you speak
j-1 ,. . ,i. -.-.-
24 v-r ----- - -"
,,,^ .^m-.. ^ t ^^^.\» -
25 -' - >: - :'-- - - '- ........ :' '^~ - ' '"'''
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON. O.C. 20005
-------
!i
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
«.
17
18
19
20
21
22
23
24
25
84 I-
or possibly Jin Dicke. What I'd like to express regard-
ing models to be evaluated, sone I'm not familiar with.
I wonder if you could identify. In particular the long;
Z short Z models, RTDM dash WC, plumes, 3141-4141.
MR. TIKVART: Before you go too long. Jim,
why don't you answer these one at a time.
MR.'DICXE: Okay. These models represent, by
and large, all but maybe a couple of the models which
were received by EPA in response to the solicitation for
non-EPA models. And these happen to be the acronyms that
the model developer uses.
MR. TIKVART: Start over again, please. One
at a time.
. f
MR. DICKE: "Long Z short Z is a pair of models
developed, primarily, by the H. E. Cramer Company for
use in was tern Pennsylvania, essentially for P.egion Three
cf EPA. --'-.. \
MR. GOODIM: P.TK-! -r-7C .
MR. DIC"E: That is a model prepared by En-
vironr.ar.t Research Technology, Incorporated. RTDM stands
for rough terrain dispersion model. WC is for worst case.
MR . r.OOD ri : PI U.-V3 .
MR. DICKE: Plur.e Five. I'm sorry. That's
PLur-.o Fivo. That's a ~oc?c-l prepared by Pacific Gas and
i ,--> ?; r: r-. i- * * f- -- «;
pretty scon
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20003
-------
i' i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
85,
we may have Plume Six. And I guess some time there was
a Plume Four.
MR. ^OODIN": 3141, 4141
:*?.. Die:*": If you are familiar v/ith the work-
book-for comparison of air quality models, you know that
there are a series of four numbers by which you classify
applications, etcetera. Dispersion models. Enviroplan
developed a number of different models to look at vari-
ous situations. And 3141-4141 represent two different
models which essentially look at particularly terrain
applications. So those two models happen to fit the
workbook of comparison of air quality model nomenclature.
MR. GOODIN: Visibility models.
MR. DICKE: To my knowledge, that is the title,
which, again, ERT gives to their visibility degradation
l. 7IC:^: That is a'rioc^l from Rockwell, In-
I'~ r.ot sure of the c::act -name. But it's
:hi7h :':-.sy c:.vs especially to their long r«?nge
r3, those that a.::7 r.ot EP7. models, are models that were
Register so-
1?30.
'2021 2344433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS.'
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
5
g
7
8
9
86
Approximately 30 models from developers outside
of EPA vrsre submitted.
Lev;.
MR. KO!7T!iIK: I!y name is Lewis Kontnik. I'm
12
13
14
15
16
17
18
19
20
21
22
23
24
25
with Hunton and T-Tilliams, representing a utility air reg-
ulatory group. -And I have a question
MR. TIKVART: Lev;, will you kind of face, par-'
tially, so that everybody can hear you.
MR. KO:iT:,TIK: Sure. Lew Kontnik. Hunton ariS
"."illiams, representing a utility air regulatory group.
I have a question for Mr, Rhoads and Mr. Cox
having to do with the EPA plans for model accuracy de-
termination. And that is, basically, what kind of pro-
r
vision is "?A naking for public access and comment in
tl-.Et i^v^str'.-ation?
"y.-'-.ctly r.^v t::s excuse nn. Ask your q
I v:a:;t to he sure I ans*;er the
r'eli, raybe I should leave it
.-as vr.at provision is EJ\\ naking
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
V <'£,.:* ' /
-------
. determination exercises that were described here?
. MR. TIKVART: Okay. You mean the model eval-"
z
uation olan that Bill Cox discussed?
3
MR. "ONT'TTK: Yes. And other plans. If there
4
are others that I'n not asking about.
5
MR. TIKVART: Okay. -That's a two phase plan.
6
The first of which involves,stepwise, each category of
models.- And the results of that evaluation,'for each
8
cateogry will be made public, one way or other. Exactly
*7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
87
how I'n1. not positive yet. And we will accept public com
msnt on that. Obviously, the results of that will have
an effect on what' nodels we recommend for specific appli
cations. That work is underway now. The first set of
f
models, hopefully, .the rural models, hopefully, will
be avilable for public review shortly after the first
is bas"
"!o" ru:.c!:ly v:e e.q. the athar categories of ir.cxiels
ss:" or. resources available to us. And I think, lit-
erally, we're talking about several years here.
The second phase is a peer _ review. And also
the results of that peer review will open' for public re-
vie--,; an:! corrvsnt. I thir.k that peer review will properly
have to lag about six months after the completion of- the
5 1 a t i .; l'. A o « 1 : ? - r f o :cr ?. n c .1 3 v a. 1 u a t i en.
"o ~;i -,"!t?.r.J. :.:?. r^I r-aso " :Mi.cl
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
"
-------
88
and comment the results of these. They will be factored
3
4 yet.
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
in to our recommendations on models for specific appli-.
cations. But the entire mechanism hasn't been worked out
Okay. - -
!!r. Youngblood.
MR. YOUNGBLOOD: I'm Phil Youngblood. Director
of air programs for Conoco, Incorporated. And I have
a question about the general area of the presumption that
a single source model, specifically CRSTER, does pretty
well, if you compare predictions versus observations on
an unpaired basis, but doesn't do to v;ell in real space
and real time. ' '-'-
There were a couple of allusions made well,
given that, my question is what are the implications of
this fsilv.r- to cor-oiro to veil in real space and tirr.e,
what are thi- ir.pl icat ions of tfhcit tc a, say, a multi-
source model, such as PAH or MPT^R.
Anc1 Bill Cox, I think, I'n not sure, but I'd
like t-- ash you, first of all, will this be looked into
in the validation program that you described. And was
it alr:-a;'1y lco!;e:"I into in the case of the HAPS data that
you presented this morning? :'
T wasn't tec sura if you wera shewing data which
r,'.\~::zc. t'.\?.t, .:t a ^iven rcc:.~:-tor, if you disregard the tine
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17"
18
19
20
21
22
23
24
25
89
of occurrence, that short tern concentrations, predicted
versus observed, you can throw time. We're doing well.
Bill, would you generally address that.
MR. COX: Yes. The point you're making is that
CRSTER plume models do not predict well at specific events
They dcn't predict veil specific hours, specific stations.
But and they do well at predicting maximum concentra-
i*
tior.s. T-7a've spurred within a factor of 10 to 40 percent,
25 percent mentioned by Mr. Bovne.
But using the RAPS data base, in which we looked
at concentration at specific stations, again, not matched
specifically at tine, but looking at frequency distribu-
tions , -they matched fairly well. When the slide that
I shoved, showed the average of the entire 13 stations
r.etwcrh, which is a still further averaging, process. But
lonkir.g r,t the in^ivic'v;^! 5r*qv.3r.cy rlistributicns, at
3^3cific s^"io-3, tho jir.irin-ri -.-?.re still fairly wall.
And by fairly '.:~-ll I rear, within, about 20 to 30 to 40
Y3T."T12LOCC: Of slicri tern highest concen
~~X: 7>i^,> ir? =hor^ torr. These are hour-r
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON, D.C. 20005
-------
90
1
2
3
4
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
.24
25
MR. COX: That's right.
MR. YOU-JGBLOOD: But "you're pinning it down
on .Location.
MR. COX: Specific locations. So what I'm say-
ing, you have a multi-source environment, you have some
cancellation of errors. And still you're doing a pretty
good job of estinacing that frequency distribution at
given stations.
MR. YOUNGBLOOD: Thank you.
MR. HAMBURG: My narae is Fred Hamburg. And I'm
wi^h Radiation Management Corporation.
I have tv;o general types of questions. The
first one refers to the models that were presented this
r
rooming. The question in iny raind was: are these models
that, were presented and discuss-rert, the sane as the nod-
* * - - 1 » .L. ,-* . . . 1 _v * . ""*... ,"* ? .. «« HV * c.' _, _ ^ _ T
-»_*" .-. -. ^ . .."..,- ' ^ ."" . .' _\- *- -1, 1 i » ^-.. '^. -» O *tmf » V O J, 4 * - ! . ^ 1_ ..^^^ \ «, ^ JL. w f
li":;. C:7."T."n, ';: h=~.r."'! tal/: ^.bc-ut field- progxT,r\s, validr,-
.iier a. As actually used, CRSmr:^ usually
3 to --o to l;h:? r. --.--- rt .=«. l-jort an:!, there
ferent '"-nil gann.
Ycur airport data are
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
91
data as compared with the use of a tower.
So I would.like to know whether or not there
2
are studies to show the validity of these two parallel
3
approaches. One in which you are using the Pasquill Gif-
4
ford determination of stability by means of airport data,
5
by means of the Turner algorithm. Or by means of a fi^lta
6
T off a tower. Or a sigma theta, or a sigma V or W, or
7
whatever.
8
Also are astability conditions included. How .
9
is mixing height determined. And so on. Now, this is
10
the first question that I have pertaining to whether or .
not CRSTER, as used in the field, is the CRSTEP. that is
12
being validated here.
13
MR. TIKVART: Let's.tackle that one first.
14
15 "~
*!R. TiOVr.TE: First, yes, CPSTER as used in the
16
field is ey.c-.ct.l" the- sar.5.'" TlVe i^ferr.ation I reported
17
en utilized Sor.inafield Airocrt data. The closest air-
18 .".*-.'
port. ..
19
TV; Peer la ir-»o>r a r'r!c.-t?. -.vsre used for the mix-
20
ing height.
21
^hore v,"-re astr.bilitv conditions. Most of them
22
wore caotured during the suruner last vear. And \\'c had
23 " \
5-c-cijil =tu;lie3 to rake certain that V.*G didn't miss any.
24
ri.-./:llyf in o'.;r full report, .ve do cor>p.?.rri the
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
. 7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
.24
25
; 92
neteorological tower data with the National leather Ser-
vice data. And, probably because we're not very far a-
part, and. it's nice and flat out there, there weren't
really any significant differences.
MR. TTKVART: Okay. Bill, maybe you can elab-
orate a little bit. But the evaluations that Bill Cox
referred to also was the standard version of the model
using, .primarily, National Weather Service data. Any
further qualification.
MR. COX: wall, the only qualification was that
v;e-used hourly estimates of emissions. Okay. And not'
MR. TIKVAP.T; Second we're going to have to
speed up. Second question.
MR. HA/3URG: Okay.
The saccnd question. I wss wondering whether
.--V, f- ^ >-'->,-> JT-,^*. <-'-.-- '- i- V. iT- T- .T> *
th?.t th-2 3hort-t3rr* predictions are not quite
ac- good 23 tl:.cy cught tc ho, frcr. the standpoint of re-
liability, ';:-.ithar or rot there's additional v;orl; going
or. v:it:; re^'arc. t:.- predictions for accidental releases
of eith.-r "^u^^rr'ous or nuclear material?
!"^.. TII'VAHT: "orr., or ?,rucG. Either one of
(202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON. D.C. 20005
-------
93
I Division to make available pop Barling techniques." .Arid
2 approaches using pop Barling techniques would allow
3 fairly rapid calculation of concentrations.
4 Of course, under accidental release conditions,
5 the estimate of the emission rate is usually the thing
6 that is so very difficult to try to put a handle on.
MR. FOX: The U. S. Weather Service has recent-
ly revised its techniques for calculating toxic corridors.
And uses the ocean breeze dry gulch equations instead of
the standard Pasquill-Gifford methods. This is the stand-
ard which, essentially, DOD is beginning to us.
12 MR. TIKVART: Go ahead.
13 MR. SADAR: My name is Tony Sadar. I'm repre-
.. senting the Allegheny County Health Department, Bureau
._ of Air Pollution. Control, in Pittsburgh, Pa.
10
_ This is for Mr. Cox. I understand you're eval-
. usting a variety cf models.: ''2Xs your one slide shoved,"
18
25
you h?.ve many rr.oclals that will be examined in the future,
19 !':-. v.-on-5-orir.c I clicln't see the Valley model included
2Q on t'vr-ra. Is that r.cdel still- htsinrr or in that still
endorsed at all? And is it going to be examined any fur-
-~ ~>
22
.... TIICYART: All the models listed there are
23
,....._ .. . _ ._ . isiclcr to be rc-f:.::ed models, "odels that'
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
itf 1 1 vvr^nv *M»
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
94
nodel was left off. Hovrever, I don't remember, Bill, are
Complex One and Two on that list?
IP rov ve>-=
. 1 . i \rf> «/- » .*. _ 0 »
MR. TIKVART. So Complex One and Two, which
in a cynical way, can be considered hourly versions of
Valley, I believe are on the list. So the answer
there are no screening techniques on the list.
1 MR. SADAT*: Thank vou.
MR. TIXVART: All right.
I
MR. MEYER: My name is Hod Meyer, with EPA's
Office of Air Oualitv Plannina and Standards.
**, >
And I have a cement and two questions.
The comment has to do with the need to, per-
haps, lock at the shorter tern meteorological-data, when
one is interested in predicting hourly concentrations.
_ . >
'.-:~3'':>-: b.isn Coin"; sone '..'crh to exnnine ozone
fc-.-.-.^tio-.. r-j-;.". v.-fj b^.^ically h;:v£ found that,if one Icoks
C-.!i i.\ ;vir.V:t-. -'-_" r".in'.i15 c-ccou'iui^ic o- t.iG v/ind "vrlocity,
T-r^- vir. " : u_; , ith 3c"^v'».-\t c.l;T^vi/:>.:-.t t'r-r jectoriss than
if sor.vior.a .~.voraccs this over ar. hour period.
'.n:" so, a-^ain, ~ trin!; ^.~rt of the problem and
uncertainty tV.at arisen in predicting hourly concentra- .
tions or -:-;_ ..c-llut>./.'; ^cul,", pc;:ha:j3, be alleviated sorre'-
-..'..-.'.: '.:;- *.-:'::"- r.t r~"~ r '....': ':.- .:' ..: avoiracri-ic: tir.-.es, wit'
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS -
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
95
meteorological parameters.
The first question is really for Mr. Bowne.
And that is I'm surious. You mentioned that the 100.'
meter wind measurements were a much better indication
of the trajectory of a power plant plume than the 10
meter ones were. Do you have any feel at all as to wheth-
er or not that is because you're looking solely at an
elevated source. That is to say, if one were interested
in bottling and area with area sources as with point
sources, would you still feel that the 100 meter winds
were a more accurate indicator of trajectories that would
be involved.
And then the second question is, if you answer
- f
to the first- is yes, do you feel, based on your experience/
there would ba any way that one could use 10 meter winds
c.rsl than, somehow, draw inferences about what the 100
-\2ter vir.cls '-ould look 111:5*. '~~
!!?.. BO'-7!;S: Ths answer tc the first question
is no. Thn 100..meter wind, I think, was the best esti-
rr.?.t n tv,".t '.:?. had. hero, bacause 'it-v"?r the one that was
the closest to the altitude of pluir.e travel, in this par-
t * cu 1 ^ r ccj" "*
I think that the wind that's closest to the :
ctl'-ltu:*2 c." -.:> 1 ur.'.e' travn] is -robablv the v;ind direction
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
13
14
15
16
17
18
19
20
21
22
23
24
25
-96
And I don't know of any x^ay to project 100
neter wind directions fron 10 meter wind directions, rou-
tinely.
MR. TIKVART: Next.
MR. GREYFORD: My name is Dick Grey ford.
t
MR. TIKVART: I'm sorry. I can't hear you.
MR. GREYFORD: My name is Dick Greyford. I'm
with Foster Wheeler Energy Corporation.
I have sone questions. You're evaluating
programs. I'd like to know, will they be available to
us through U*7AMAP? And will they replace programs al--
ready in U^TAMA??
10
11
12
MR. TIKVART: The answer to the question is
r
as the models are evaluated, and as decisions are made
as to the relative performance of the models submitted,
-- and thos3 v.-lthlr. u*:?,;:?-.P, decisions will be made as
* .- = -1- ?
^1 ^fc K, *J \. _
tai;;ir.:r r.':-ut rebels submitted, by private consultants,
in r-.ar.y rr.-;:--, r-nc1 thv/ r'-iy, or mny not, want their mod-
el in T~:A: ::.?. They may v.-ant to market it individually.
j -To I *.:ould sc-.v v.'hsthcr or not the model be-
'*
i. corr-33 -r-.rt of u:-T?-.*:?.r depends en its regulatory usefulness,
hov well it ;;-:;rfcrms ralativc tr> ether models, ancl v:hc-th-
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
. 7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"y-f
UNAMAP, or whether he wants to market it himself.
hs long as the model -can be made generally avail
able, that is our only requirement for whether or not the
model is released.
And that can either be UNAMAP or privately.
Lou.
MR. SHITTFELD: My name is Lou Shenfeld of the
Ontario 'Ministry of the Environment.
"ly first question is. addressed to Norm Bowne.
Can you comment why, -with such a commendable
effort to measure air quality over this space, up to 20
kilometers fro- a source, why there is such little ef-
fort to locate the plume? That is determine the wind
r
firild over the area? Especially, if you are planning to
vj.Li.lo.t~ h-3 r.ioc.ol on an hour by hour basis.
Mr. r.*: '::.".: ":11, :irst, ;;?. t I reported on
vas the application of the ncdel in its reoulntory set-
tin j her-?. , -.-.aro it's uaec" vith airport data. But we
hi-.v; .:.:!:. r :_.:::: ^..". en, yrt, tlvj v.re cf the ether pieces
O" vir. 1 .- j--i;--'..?.T.t that ^;e had cu!: there. The soundings.
<2 the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
because we consider that to be a scientific evaluation of
the model, while what we have reported on to this point
is an operational evaluation of the nodel, where we're
evaluating the model with the same kind of wind informa-
tion, or meteorological information that is usually a-
vailable.to the user, who wants to project the impact of
new plant.
MR. SKEIIFELD: On a regulatory basis, the mod-
el isn't used to predict ..concentration at a certain point,
or at a csrtain time, is that correct?
-*!?.. BO'JTTE: In the usual method of application
by PP.-., it is merely a maximum..concentration over seme
time period, over a fairly wide range of distances from
the source.
. MR.. SHESFELD: On that bases, the model did
*r;s1**3n*-'^"? P"^'-^, r:** ^"T^I^C 2 S T^^TT* *»*"*
V Ct Jl. -1. V* 14 ^f ^f f L. **. ~~*-J \^ «~ * ,. -*. 1 * Irl *3 t- * r * ^_- i. >«. V_ 4 . hv 9
so fa
\~ >J ..*...-_ ^,
You die! nctr.ention a/tine frame for your vali-
dation .of Icn-j
(202) 234-4433
^n^e transport models.
n'-TSR: IV; r.ot sure that I really knew
ir.:: v:it:- r2--;!\rd to a specific model, F:*
-^r.thly v.->^-.:.s.? th.~.t they are trying to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
99
look at in terns of both air concentrations and in terms
of deposition. And that is the primary time frame for
the validation v.-ork and evaluation v:ork that is being
done jointly v:ith the Canadians, as I understand it.
On the basis of the regional model, I do know
that the basic time frame is one to three hourly periods
for the current cxidant work. And I vrould assume that
for long range transport, in the future, that concentra-
. 3
tions over a longer time period would probably be used.
But hose vould probably result from arithmetic addition
of the shorter time frame concentrations".
>
I'm not positive about that one, Lou.
MR;~TIKVART: Can we move onr please.
r
!!R. SIIEMFELD: Thank you, very much.
.'!?.. TIKYART: ' I'd like to stay reasonably on
»
'
tc t'i-t re~£.in-
.;c -..a-stic:-.:-.-3/
.3 othar -:-::. ~rtur.it ies to discuss, so -o
ricon. There'll
e :?e brief.
*T.. ?r::~r-;i:: ^.v-c-rvjc S^hs'.-.-o. Petco 7r.vircr.mental
This is for Mr. Pox and :?r. Tikvr.rt. TTo\^ will the recom-
r.sr.f-iti-".- that tho AVS-r?7. panel put torrethsr be imple-
r.er.ted bv r?-\?
i
:'.?.. TI?"VA"7: "o -.-.ill ha--e to formally receive
> ~ - - " r> v .-.
(207) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
100
how. But we will make an effort to respond to those rec-
ommendations, once they are submitted to us. And I sus-
pect at the time of the hearing on the revised, guideline,
\
which will be sometime in the first part of 1982, we will
present a response to the recommendations at that time.
Don.
MR. MOO!;: Don Moon. Salt River Project, Phoen-
ix. Short question, for Bruce, I guess.
I-'e've addressed, primarily, the flat terrain
comparisons today. And we've confirmed that they're ac-
*
curate .within a factor of two. ^7hat do you feel-, right'
new, at this stage of the gams, is the accuracy in com-
plex terrain?
r
!!T. TURNER: "l think that is going to depend
upon the averaging time, a-^c.in. ?':. "; Irt c f tk:r. n.r^- r.ct. And some -deviate
by r>ore "lik:: L\ factor cf 1C.
r^a were we raa-
NEAL R. GROSS
COURT REPORTERS AMD TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
101
r
. IIR. TIKVART: Okay. Thank you. Thank you, to
all the panelists. And thank you, for the good questions,
z
*7e now have a fifteen minute break scheduled.
3
Don't wander too far. We will start promptlv at 11:35,
4
(thereupon a morning break
was-taken at 11:19 a.m.)
6
(11:33 a.m.)
7
MR. TIKVART: The next panel will be concern-
ed with incorporation of uncertainty in regulatory de-
9
cis ion rr.akina.
10
The second panel will be "!r. Bruce Jordan/ who
is Chief of the A.~\bient Standards Branch; Dr. Thomas Cur-
ran, who is a statistician with the Monitoring and Re-
13
ports branch; and Dr. Barnard Steigervald, who is Direc-
14
tor of bhc: Office of Reaional ?rc~rar^3. All in the Of-
15
itv Plar.riirvT an:;'Stinkards cf TPTv.'
16
Ccllocti-"-!'.-, the-.-- rill Discuss the incorpor-
17
::-j~ul:'.r.c-rv decision -aking.
18 ~ '
an rill address the is-
19
o- --.r --.;?.".if ;-:tar.;":-;i--'',r:. ".r. Curran, or Dr..Curran,
20 ' -
the inclusion cf statistical infcrr.ation in v;ork with
21
stzinflr. :"~ ".-..". re::u?.^. tlo-.-.s. ~.:\c*., Tinally, Dr. Steigorv-;ald
will tal.''. c.l:out the ;:se cf statistical tech»nicues in reg.-
23 ' J:
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
t-1 r (* »»«
-------
I
2 Mr. Jordan.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. H
102
all questions until all three panelists have spoken.
:S. JOP.DA": Thank you, Joe.
I was hoping that you were going to say, until
all the panelists have a chance to get out of the office,
or out of the rccm.
*»^
Good morning. It's certainly a pleasure to
t
be able to work with this group this morning. And to
share a few moments with you, where we can both illus-
trate some of the problems that we have in trying to
*
meet air quality goals. ^
You, as modelers, have problems. And those
of^us who are in the business of setting ambient stand-
ards, certainly, have problems. And collectively work-
ing togirV.sr, I thin?-:, we can baging to resolve some of
- _ - . _ 4
thj> 'm'ohl r."o thnh '.;il] face us in the rr.ear future.
".TV.: ir. tivs -- in c'p3.~.ir.7 my presentation, "thl
rr.oriiing, ~'2 like to rr.a';a a ccupla of general observations
for you. Tirs± of all, I think you can rest assured that
t:-.ors> -.ill c?r.ti:\v.a to h». nr.ticr.r.l amhient air quality
standard.! ^roorams v/ithin the U. S. for some time to
the major Act amendments that I have seen,
.vo -^.'v" I*::T%O~-.m:.>".." l'h~t t%: : .'s^biGnt air nual-
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
103
There are some fine-tuning, as people want to
call it, but, generally, I think ambient air quality stand-
ards v.'ill be around. And you will b>e required to model
them for some tine to come.
I also don't think that you will see any new
pollutants in the near future being added to those that
already exist, for which we have national ambient air
quality' standards.
And, third, the data.base, the scientific data
bass, for the basis of th^se standards,will be very, very
closely scrutinized during the review of each standard.
I That is, scientists, medical people will require a much
better data base upon which to make the decisions for
r
the national ambient quality standards.
And I don't think that you, as modelers, can
requirements will be any lees stringent.
c/ualitv nt'.ri-^'.^.j bo reviai'ed Xt least once ev
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
104
review only for one pollutant. And. that being ozone.
Carbon nonoxide is presently in the proposal state. And
v:e have v:ork underlay to revise TCP,. MO , and sulfur ox-
ide. T7e also have, a proposal cut to rescind the hydro-
carbon national anbient air quality standards.
!7ow, as we have undergone these reviews, there
have been so:r.e trends that have developed that I fehink
can significantly affect the way that we go about model-
ing to demonstrate attainment in the near future. 3
The first c.f these is that the medical cornnun-
ity 333"s to be prone to rely nuch more" on v/hat we call
~' *>
harp.an' clinical studies to set the levels of the ambient
standards. By human clinical studies, I mean where you
F
stic": a hurrian being into a char'ber, expose hip. to a level
'~:;~. rr-~ic;r.a.l'j; fcr c^L'vj T_'ii3 " is" tV.a!: "U'^sc- stud-
o;--.~il-_ '';-i.c.r.z.-..:, :-.:' it's r.-ucV. easier to i3.-la'J:'e cho
Th^ -.*.:.o=.l cc ;:uit_- '-.^5 ^l^o been prone to
s-j'--.ciry that !::ia "t.'.tus ; vcu,^^t against: the observed cf-
..':'.- !..".c.";! -;tu'"io^ hein-j, c,":r of, the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS .
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 200O5
-------
105
I these studies tend to be short duration, or done over a
2 short period of time.- T7e began to observe the health e'f-
3 fects from, short tern exposure. Arid there rseems to be -
a trend toward a much more shorter term standards, par-
5 ticularly for sonc of the pollutants.
. And a couole of examples, that we have looked
o ~
at extensively, are SO and nitrogen dioxide.
Q ' We in the standard setting business recognize
o
that this puts quite a "challenge to the modelers .in try-
ing to demonstrate attainment. The problem of asking
models to perform in an area where they perform, probably,
not as good as they could with" some of the longer aver-
aging time.
13
f
But, nevertheless, the medical community nots
14
so f.uch concerned about what the models can do, as the
lo
are for ~>rctactir.g the ~cneral mxblic.
16 " "
For this r£2sc-/;. I-.believe that vou v;ill r>oe-
17
nov: ?. poriod of tine ./har/ tho ir.oc.slsrs will b
18
o develop some: .-hat or surrogate to demonstrate attain-
19
r.ent cf the standards. And by surrogate T mean mavbe
20 " '
seme longer averaging time, modeling longer averaging
tirr.e, but, at the sare time, shovina that those model re-
22
suits five a reasonable assurance of nrotection aoainst
23 "
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
"?nr, TW-44^ WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
standards, over the old, what we called the determinis-
tic form of the standard. Realizing, many times, that
a major decision has to be made on a nodeling result where
you mcclel ancl you find the worst of all possible situa-
tions, and a decision to disrupt a new plant is made
yea and nay based on those modeling" results.
To some extend, we're trying- to do something
about this by looking at the form of the standards. And
working with the medical community, we have had some suc-
cess in getting the medical to think a little bit dif-
" ferently "than, just the pure statistic form I mean,
deterministic form of the standard.
I think you saw that when we revised the ozone
standard, we went to an expocted values. As we look at
the CO standard, we're also going to, I think, an ex-
pected vri/.ue-, ?.r.d, porsib.lv, even ore that will have
* ^ « . t _
r..:;-.c -.u_v:.j _> ;:::.;o?c--r.t:« tr it.
re.--- to '-;iv-2 you sore Illustrations-of the typ-
os of thiy.-;.^ that we face in t*'.-- standard setting busi-
]-^3S, I'c! li"-:;-. to run thrpagh several of the standards,
and ju-'it r:m::har,i? the kind of protection that the ned-
ic.-.l cr -"-.mit;;, ?.t lf:F?st, ss'2-%3 some pec-pie in the
medic."il cor~ur..:tv see*1) cc indicate to us that ere need-
202) 234-4433
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, O.C. 20005
-------
1
2
3
4
5
/. low concentration level, that nitrogen dioxide can cause
o
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
107
There is a faction within the nedical connunity that is
convinced that nitrogen dioxide, in relatively low levels,
has so~e adverse affects to hunan. "ow, what I can un-
derstand cf it, not being a r.edical doctor, myself, they
explain it to ne this way. They say that a relatively
some adverse effects in the body. And it starts off by
the body being exposed to it, and the effect occurring,
and, after a period of tine, the body just dismisses -the
nitrogen dicxic2e.
TTov, if, during this duration of exposure, there
13 a .se7C.-r.cI appearance, cr he gets the sane dosage all
over again, -the effects nay be additive, or they nay not
f
be" additive, but, anyhow, the nedical community cays that
the id-3~. th?.t it would! be desirable to /sort of, protect
=~rv^ ti-i;, the -;c:dical cor-.r\x; r.lty- indicat
s t>.r;t t":.-r2 is -. r3?.~on to ';^::t to protect against very
:i ;'/. c :ir.;>_A:.'crnt:i-.:'r. levels c" ".:"-, "or short poriocTs o.T tirr.c
y hij'-. , ~ r-.-3.;ai in the order, cf 1.5 ?.?'!, or naybe even
n r.o-.'.^Iin:;, if L::at 'or- of the standard
"-/ ^ : : :.~ ?. ~ c -.".*?::* that vlll t^ke on
ro
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
not so much interested in one hour, by itself, but in
2 back-to-back type of one hours. We're looking at the
3 possibility of having some form of a stochastic type stand
4 ard. Which I understand would be advantageous, from the
5 modeling standpoint.
6 Also for carbon monoxide we have a pollutant
7 for which the body tends to be an integrator. And it's
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
important how people move about through space. So, for
. D
that one, this time around, we've got some exposures mod-
els built, but they*re probably not totally acceptable
to the medical community, at this time, where they will
completely accept them. But, in order to account for
some of the spatial exposure, we're thinking, very strong-
ly, about the_possibility of going to a multiple type
of exceedance standard, where you have to have a concen-
a
tration level that will be an eight hour, as well as a
one hour., standard, but a possibility of allowing mutli-
ple exceedances of that standard.
For total suspended particulate matter, we are'
having some substantial changes in the ambient standards.
Instead of recommending a standard that takes into con-
sideration all the mash in the atmosphere, we're going
now we're thinking,very strongly, about going to a
size specific pollutant, where they will be dealing with,
for health standards, 10 microms, or less. And for this
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS '
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
r;.
109
one, it looks like the averaging times will still be in
the ballpart of annual and, probably, 24 hours.
We also expect that we may see a secondary TSP
standard which will require some long range transport
modeling. It will most likely -be based on something like
aesthetic effects, such as visibility.
And,finally, for sulfur dioxide, we see-some
evidence, at this point in time, that sulfur dioxide may,
in fact, cause some adverse health defects at "very, ve^y
short exposure concentration very, very short durations
of exposure.
So, we*re looking at that. And we see the pos-
sibility-that we may have to have like a one hour SO,
. r
standard.
All these are speculations, at this point. And
it's only thrown out to you to get you to think-about some
of the ways that modeling may have to be changed, in the^sE
future, to meet some of the ambient air quality standards.
Generally, I see that there is a trend toward
a shorter averaging time for many of the standards. And
I see that we will, probably, be thinking much stronger
toward going to a pure statistical form of standard. That
is, trying to get toward such things as a 99 percentile,
or a 95 percentile, or some form of that, and totally
away from the not to be exceeded more than once a year
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS .
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
type of standard.
Thank you.
2
MR. TIKVART: Thank you, Bruce.
3
Next we have Dr. Tomas Curran, who will be speak
4
ing on transition to statistically based standards.
5
Tom.
6
DR. CURRAN: I should preface my remarks by
7
indicating that my background is primarily in data analy-
8
sis, rather than in modeling, so that my perspective is
9
more from an ambient monitoring viewpoint.
10
Also, at the present time, the ozone-standard
11
is the only national ambient air quality standard that
12
13
14
15
16
17
18
19
20
21
22
23
24
25
has been revised to incorporate statistical concepts.
And models are not as routinely used for ozone,
in quite the same way they are used for sulfur dioxide.
So our experience with statistical standards and disper-
sion modeling, at the present time, is fairly limited.
What I plan to discuss, today, is, briefly,
background on expected exceedance standards, how they
-affect both monitoring and modeling,, a little bit on
what has come to be called the Ex Ex method, which was
used to incorporate sulfur variability into coal-fired
power plant impact assessment. I'll also mention some
further issues that, probably, have to be considered.
As far as background, I think most of you know
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
Ill
1 that in 1971 the initial national ambient air quality
2 standards incorporated short-term limits in terms of con-
3 centration values not to be exceeded more than one per
4 year.
5 Certain advantages to this type of approach.
6 The most-obvious one being simplicity. But there are
7 also certain disadvantages. Hissing data is probably
8 the most obvious.
9 What happens is there was no way, really, of
10 conveniently accounting for the effect of missing data
11 in the one per year format. So you got the type of thing
12 where one site might have very complete data for the year,
13 another one very incomplete data. And there was no way
14 of really adjusting for the effect of incomplete data.
15 One of the ways you can get around that is to
16 impose a minimum-data requirement. But it still leaves
17 you with the situation that the site that has complete
18 data is more likely to detect a violation than one that
19 just meets the bare minimum requirement, even if both
20 sites are monitoring essentially the same air quality.
21 The other point the other disadvantage is
22 a little more subtle, but it's still troublesome. And
23 I'll define I'll use the term exceedance, as kind of
24 a shorthand, to refer to a concentration value that is
25 above the level of the standard.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
,, nT, fr^r WASHINGTON, D.C. 20005
-------
a. A.-.
ill
1 And with a once per year standard, you're kind
2 of in the situation where the implication is that it's
3 acceptable to have a non-zero probability of having one
4 exceedance per year. But it's not acceptable to have a
5 non-zero probability of having two or more exceedances
6 during the year. Which, in reality, it's more likely,
7 that if you have a non-zero probability of having one
8 j exceedance, you're likely to have some year where you're
9 going to have two or more.
So, in a sense, the once per year standard al-
lows for the unusual event within a year, but not for
12 the unusual year.
Now, one way to get around this was to go to
r
14 an expected exceedance standard. And, basically, in 1978,
15
16
17
18
19
20
21
22
23
24
25
when EPA proposed revisions to the ozone standard, one
of the points -proposed was the introduction of the con-
cept of expected exceedances... And this was promulgated
in 1979.
Including in the ozone national ambient air
quality-standard was an adjustment for missing data.
.And the standard, itself, was worded in terms of the ex-
pected number of exceedances per year. Now, the expected
number of exceedances can really be thought of, intuitive-
ly, as a.long-term average. In other words, if you say
that the expected number of exceedances can't be greater
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
f2021 234-4433 . WASHINGTON, D.C. 20005
-------
1 - ».
113
than one, well, what you could have is site that some '
2 years has no exceedances, some years it has one exceed-
3 ance, other years it may have two or more. But the long-
4 term average number of exceedances won't be greater than
one.
Okay. That's the intuitive example. From a
6
practical viewpoint with monitoring data, we're never
really going to know what the true expected exceedance
o
rate is, because we're hot going to have an infinite num-
9
ber of years to average over.
So from a practical viewpoint:, we ended up
specifying that we'd use a three-year estimate, just to
estimate the expected number of exceedances in terms of
13
a three-year average of the observed number of exceedanc-
14
es, after adjusting for missing data.
15
It really is kind of a compromise. We recog-
16
nize that the more years that you have the more stabil-
17
ity you'd have, in terms of the estimate. But, on a prac-
18
tical basis, emissions are going to change over time, so
that using just the most recent data.may be more indica-
tive of the current status of an area. And, there's also
£* L
the practical viewpoint that using fewer years of data
ensures that timely action can be taken.
23
Next overhead.
24
The facts of going to an expected exceedance
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
->r,-»i >?/,-/;/,^ WASHINGTON ir Tinn«i
-------
114
I standard are kind of interesting. From a monitoring view-
2 point, probably the biggest effect that people saw was
the missing the adjustment for missing data. From
4 a modeling viewpoint, it's more likely to be the averag-
5 ing process. Certainly, with models, such as CRSTER,
some of these dispersion models, that give a predicted
7 concentration at a 100 receptor sites for every time per-
8 iod in the year, missing data is a relatively minor point
9 And*it's more likely to be the averaging pro-
cess. The idea that you need to address the probability
j of an exceedance, or the frequency of an exceedance.
12 With an expected exceedance standard, the high
values are not really ignored, but they're kind of weight-
lo
f
,A ed by how likely they are to occur. The obvious effect
14
of this type of situation on a screening model,which as-
15
sumes worst case condtions, is that with an expected ex-
16
_ ceedance standard, it no longer really suffices to show"
that under worst cases assumptions something can occur.
18
What you have to address, now, is how likely
it is to-occur. What is the probability of an exceedance.
In a sense, for screening models, I think this is a sit-
uation that people have recognized in the past.
If you assume worst case conditions, and you
A«U ;
don't have a problem, then fine. If you assume worst
case conditions, and you do have a problem, then it really
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
734-4433 : WASHINGTON, P.C. 20005 i
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
115
is kind of a grey area where you have to address how fre-
quently the problem will occur.
For other models, getting away from screening
models, it may turn out that the change to expected ex-
ceedances is really fairly easy to handle.
I have an illustration, I guess, of the Ex Ex
method. This sort of, briefly, indicates the type of
thing that can be done. Some of you have have seen it.
It's called the expected exceedance method.
Intuitively, all it is, is that you can run a
dispersion model, such as CRSTER, with unit emissions.,
Say,for a particular .time period, you get a predicted
concentration of .07. And what you really want to know
r
is what's the probability of being greater than .14 PPM.
Well, because CRSTER is linear in emission,
- i
the probability of being greater than .14 PPM is the
same as the probability of having double the emissions
So,- if you have a frequency distribution of
emissions, or some way of computing what was the proba-
bility that you would have double the emissions, during
that time period, then you can directly compute the prob-
ability of an exceedance for that particular time period.
And then you can just aggregate these over the year to
get an expected exceedance rate for the year.
The actual implementation of this procedure was
NEAL R. GROSS
' COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE, NW
f202) 234-4433 , WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
116
done .by SAI. And it was done in terms of a computer sim-
ulation to account for oiranissions .
The intuitive idea is the sane as presented
here, but from an implementation viewpoint, the simula-
tion allows you to handle more complex descriptions of
emissions , either as autocorrelated tine series. Also
enables you to get certain probabilities more convenient-
ly.
The point is that it does not change the model.
Basically, all we've added is to take the previous model
"that was used in a deterministic framework over to an
expected exceedance or probability violation framework.
It -was simply-a translation step that took the predicted
concentration and converted it to a probability of ex-
ceedance.
Next slide. ....._,
Okay. Up to this point, the potential effects
" ~" ~ ~*
of the expected exceedance, or statistical standard , may
be may appear kind of negative, either makes use of
screening models at little more difficult, or it's more
work for the translation step, in terms of other models.
But I think the net effect on modeling will
probably be positive.
And from *r ambient viewpoint, I think people
there's a tremendous intuitive appeal. People are
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17"
18
19
20
21
22
23
24
25
. ~ 117
familiar with concepts, such as 10-year flood, or 100-
year flood. And the idea that .we're weighting for high
values by how likely it is that they reoccur, just has
a certain intuitive appeal.
Also, from a more technical basis, with an ex-
pected exceedance standard there's no penalty for using
additional data.
One of the problems with the once per year for-
mat is it's second highest value for the year that de-
termines your status.
As you look at more years, you look at the high-
est of the second high values over the time period. So,
what you end up with is each time that you add an addi-
tional year of data, you cannot really lower the design
value. All you.can do is it increase.
And I think it is understandable that a source
may hesitate to use additional years of data, because
adding additional years has an inherent penalty, and there
:is no real gain for them.
Under an expected exceedance standard, it's
possible that adding additional years could either raise
or lower the design values. So, it does away with that
inherent penalty.
Also has advantages, I think, in terms of the
realism. A statistical framework, such as the problem
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 . WASHINGTON, D.C. 20005
-------
I
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
jf ' 118
that really led to the development of Ex Ex method, was
if you want to describe emissions statistically, which is
a more realistic description than, say, assuming constant
emissions for coal-fired plants, a deterministic frame-
work does not really accommodate statistical, or probabil-
istic, descriptions that conveniently.
So, the modeling may actually find greater flex-
ibility within the framework of a statistical, or expect-
ed exceedance,type of standard. In fact, not only in
terms of the actual descriptions of the processes involv-
ed, but even, maybe, the structure of the model.
For example, to compute expected exceedance as
a probability of violation, what is really required is
that you have the probability of an exceedance for each
particular time period. It may turn out that in some
*.
situations it's easier to develop a model that yields
a frequency distribution for a time period. ' ^
In other words, you can't pin it down * you
don't pin. it down to a specific concentration value that
occurred for that time period. But you have a frequency
distribution. You have an idea of, well, a probability
that's less than .12 in that time period is such and such.
The probability of .13 and from that it becomes pos-
sible to develope the probability of exceedances, or the
expected exceedances, probability of violation.
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS .
1330 VERMONT AVENUE, NW
(2021 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
119
In other words, model structures, that were v.
not allowed under a deterministic framework, may become
very convenient in terms of a statistical framework.
If you think about predicting frequency distri-
butions, for a particular time period, rather than a sing-
le concentration values, it's kind of interesting to see
what implications that would have for raodel validation.
I have just'a couple of final comments.
I guess, reitering the idea that our experience
to date with our statistical standards and dispersion
modeling is fairly limited.
There are some things that are worth consider-
ing. One-of the ones that I have up here is a CAP.
F
When I mentioned the ozone standard and the
idea that we use a three-year average of ambient data
to determine the expected exeeedance rate, one of the
things that's implicit in there is the idea that if I -
average the number of exceedances over a three-year per-
iod, and I say that can't be greater than one, no single
.
year can have more than three exceedances, or you cannot
attain the standard.
So the three-year averaging process for ambient
data, in a sense, puts a cap on the maximum number of ex-
ceedances that can occur in a single year.
From a modeling viewpoint, it may be conceivable
NEAL R. GROSS
. COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 -. WASHINGTON. D.C 20005
-------
1 to model a 100 years, or a 1,000 years. And one of the
2 questions that comes up is what does it mean from a health
3 viewpoint if the model predicts a 100 exceedances over
4 a 100 year period, but all the 100 of them occur within
5 a single year.
6 And, I think, perhaps, we should consider in
7 modeling analyses things like putting a cap on the number
\
8 of exceedances per year, or specify a higher'concentra-
9 tion value that can't be exceeded more than once in 10
10 years, or this type of stuff. Which leads to some inter-
H esting observations of the implications of probabilistic
12 interpretations for modeling results.
13 For instance, what does it even mean to say
f
14 that you've-modeled 100 years, or 1,000 years. Some of
15 these implications will be discussed by our next speaker.
16 MR. TIKVART: Thank you, Tom.
YJ With that introduction, next we have Dr. Ber-
18 nard Steigerwald, who is going to speak on the use of
19 statistical-techniques in regulatory analysis.
20 Bern. . . .
21 . MR. STEIGERWALD: Tom only thinks that I'm go-
22 ing to talk about some of those things.
23 I have a hard time talking to modelers anyway.
24 And I've never tried to talk to a hungry group of meteor-
25 ological modelers. So, I will make it very short.
NEAL R. GROSS
' . COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1: . - ' 121
'.
I'm going to be focussing mostly on some of
the regulatory implications of using statistical tech-
2
niques instead of deterministic techniques in modeling.
3
Basically, the message I have is I think we are
4
moving toward statistical techniques.* That it has a sur-
5
prising impact.
6
I'm had a couple of years using it, or attempt-
7
ing to look at it, and it has a surprising impact on the
8
regulatory end of the package.
9 .
And I think it, also, offers some opportunities
10 .
to the modelers.
11
First slide.
12
I think that we are moving toward statistical
13
techniques not only because of the changing form of the
14
ambient air quality standard, but for several other rea-
15
sons. I think because it's going to offer modelers and
16
regulators a better way of handling error and uncertain-
ly"
ty. A better format for expressing it.
18
I think that it has the potential for improv-
19
ing the reality "or the appropriateness of the the input
20
assumptions, of the non-meteorological input assumptions,
21
if you want.
22
And I will talk about the Ex Ex method a bit
23
later.
24
We are having an increasing difficulty
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
122
1 defending some of the input assumptions. Saying that
2 the sulfur content of coal does not vary, when everybody
3 knows that it does vary, is a difficult point to defend.
4 So that I believe that, even if we don't change
5 the form of the ambient air quality standards, that we
6 will be moving towards statistical techniques, because
7 it's going to help explain uncertainty and describe it.
3 It's going to improve some of the inpudt assumptions. And,
9 finally, because it may enable.us to handle a little bit
10 different kind of error than I've heard talked about to-
ll
12 Doug Fox did talk about the lack of representa-
tiveness. The problem of a regulatory making a decision
r
that should be over the lifetime of a new plant all based
on one year of meteorological record.
lo '
.. So, that I think that the statistical techniques
lb
18
25
are coming.
Next, I have been playing around with something
called the-expected exceedances methods, which, as far
1*7
as I know, from a regulatory point of view, is our first
attempt to move from a deterministic into a statistical
M 1
2_ type of modeling.
It seems simple to me. You just vary the emis-
sions over time, because the sulfur content of the coal
varied over time. Everything else stays the same. I
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
.WASHINGTON DC 20005
-------
..... ._-, . . 123
I didn't touch the model. I wouldn't dare.
2 We just changed the one input assumption. And
3 I was quite surprised at the implications that this hand.
4 And,in the next slide, I attempt to point out some of
5 these, quite triefly.
« I'm not sure if everybody can see that.
b
7 The first problem that came up is once you start
8 saying I'm not going to say never. I will accept a cer-
g tain amount of risk. It sounds better to say acceptance
of a degree of certainty. But, in fact, it's accepting
some quantitative finite risk that that emission limit
will.allow the air quality standard to be violated. Not
to be exceeded, once per year, but to be violated.
lo
-- That forced EPA to pick an acceptable viola-
14
tion frequency. Not much precedence for that. In the
15
Ex Ex method, we picked one in ten for the primary stand-
16 . ~ *
_ ard. Which means one year out of ten. You can have two
days, or two three-hour you can have two days above
18 '
the air quality standard number, and, if that occurs less
.1 y
frequently than one in ten years, it's acceptable.
For the secondary standard, we're in for the
PSD increments, where it did not seem necessary to have
the same degree of certainty, we'll be tossing out one
23
in five years, as an acceptable risk.
Not much basis for those numbers in the Air
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
">rm "'WI-4A1*! WASHINGTON, D.C. 20005
-------
4 - ' 12*
* "
1 Act or any place else. But it when you go to the sta-
2 tistical technique, and you throw out never, you have
3 to put something else in its place, as the basis for say-
4 ing yes or no to a permit, as the basis for saying yes
5 or no to an emission limit.
6 Secondly, the worst site doesn't become all
7 holy anymore. You're talking about the probability of
3 violating anywhere in the vicinity of that plant. A lot
9 of us made a mistake. We felt-that if the worst site
10 attained the one in ten year probability we were home free
1! In fact, you have to consider the probability
12 for all of the sites, because the real question is what
13 is the probability of that source causing a violation,
14 in the vicincity of the source, not just as the worst
15 site.
16 So, the whole concept we've had, we only worry
17 about .the worst site, goes out the window, and you worry
18 about about all sites.
19 I think that the worst year, as Tom said, as
2Q you added more years of data, you had the potential for
21 finding a year that had worse dispersion, a higher high-
22 est second high, and, therefore, a tougher regulation.
The Ex Ex method, since it is statistical, in-
^O
24 tegrates, or allows you to put a bad year into perspec-
tive with other years, because the answer you are looking
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
1 for is how often am I going to violate.
2 So that, I think, as Tom said, more meteorolog-
3 ical data, better data, means a better answer, not, neces-
4 sarily, a stricter regulation.
5 Fourth, we have the form of the emission limit.
6 Since the Ex Ex method, you are approving a source based.
7 on .the distribution of emissions. You, essentially, have
«
g to enforce the distribution of emission. Or you have to
9 pick enough points off of the distribution, so that you
10 have a fair shot at enforcing the distribution.
H In the old system, one number was used to de~
12 termine yes or no. And you only had to enforce that num-
13 ber, always. ; . ^ -
r
14 Now, we will have more complicated regulations,
15 because you will have to do what you can to set regula-
16 tions that, if enforced, limit the source to the distri-
.- bution that you modeled. - _ x
18 Five, determining the degree of emission re-
19 duction, it tends to be startling to regulators, when
20 they finally understand that if the criterion for accep-
. .tance or rejection is how often will that source allow
22 a violation, that the answer is not proportional to the
23 design value.
24 There is a great sense out there, now, that, if
on the highest second high day, or highest second high
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
126
three hour period, we double the allowable incrementf
regulators say we have to half the emission. If you have
double the allowable frequency of violations, it's not
an obvious answer how much reduction is needed to come
into a one in ten probability of violating.
And, I think, it is up to the modelers, now,
to give the regulators a range of answers in terms of
the acceptance-rejection criterion.
What does a little more control. What does
a little less control. And, finally, what kind of con-
trol do you need to get to an acceptable frequency of
violation.
r The information base has to be expanded. And
everything gets a little more complicated. If you toss
in autocorrelation, if you don't assume random coal lots
from one three-hour period to the next, or one day to
the next, you then have to toss autocorrelation in. While
we have that in the model, not many plants know much a-
bout the distribution of their emissions, much leas the
autocorrelation of that distribution.-
Finally, it forced us to decide what we were
going to allow to vary. Once you move from a determin-
istic, where everything except meteorology is aconstant,
over into we are going to allow something to vary,
we solve the problems of the acceptance criteria, you
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
127
then have to say why not let other things vary, why not
2
3
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
let load vary? Why not let BTU content of a coal vary?
Why not let sulfur retention in the ash vary? Why are
4 you only letting sulfur content of the coal, or the scub-
ber efficiency vary?
Those are questions that have to be answered,
and, I think, justified, once you move from a determin-
istic into a statistical technique.
3
So that you can see I didn't know that I was
getting into all of these, when we just simply were go-
ing to model with varied emissions. Bat it does nove ,
me,philosophically, into a whole new ballgame. And I
think that it's a good arena to be in. But I think it's
r
got its difficulties, too.
Finally, I think there are opportunities. And
>
if you can put up the final slide. I think it's got some
..
opportunities for the modeler. As I said before, intu-
itively, it just seems that it will facilitate the hand-
ling of air, and the uncertainty in the expression of
uncertain results.
And then, finally, a problem that has always
bothered me is the one of making a decision on a plant
that might be there for 60 years, and understanding, fin-
ally, that we use one year of meterological data to an-
alyze the air quality impact. And we assume that that
NEAL R. GROSS
. COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
. f
128
1 one year held for the 60 years that the plant was going
2 to be in existence. Not only that the year held, but that
3 every day in the year was the same. So that we had one
4 worst day and one best day that year, in terms of dis-
5 persion high, and that no day would ever be worse than
6 that in the 60 year life of the plant. And that no day
7 would ever have better dispersion than that. And each
t
8 would show up only once. Not twice. Not zero. So I
9 anticipate that we will begin looking into, possibly,
10 the use of dispersion as a variable. Curve fit the Chi
'n curve for each site for, say, 365 days. Or if you have
12 five years, for 1,800 days. Extend the tails of that
13 curve. And then pull a series of 365 day years out of
14 that.
15 . That can be coupled with variable emissions, ..
16 or any other variable that you want. And, possibly, move
17 us, a-little bit, towards better representativeness.
lg I anticipate that you could put climate changes
19 in and move the high distribution curve a bit. and get
at it even better, that way. I'm not sure that any of
21 those can be done. But I think going into the statisti-
22 cal technique gives us the opportunity to do that.
23 So, in summary, I think there isn't any ques- .
24 tion in_ my mind that we are moving towards statistical
treatment of modeling, and of modeling results, and as
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
129
1 the basis for regulations. And I think it's a funda-
2 mental movement in air quality management system.
3 -I would urge the regulators to understand what
4 it means, and to demand it of the modelers. And I would
5 urge the modelers to look at it as an opportunity to in-
6 fluence regulations and to get better regulations.
7 Thank you.
8 MR. TIKVART: Thank you, Bern.
9 Any questions from anybody on the panel here,
10 or clarifications, or further points?
11 No. Okay. With that I'll be happy to enter-
12 tain questions or observations from the .floor. If you'll
13 -quQue up at either one of the microphones.
14 While you're doing that, Shep Burton has asked
15 me to request that all of this afternoon's panelists meet
16 with him in the back at the conclusion of this session
17 and before you proceed to lunch. .
18 All of this afternoon's panelists meet with
19 him before you depart.
20 Yes, go ahead.
MR. BEAR: My name is Mitchell Bear. I work
22 for the U. S. Geological Survey.
23 I have two questions. ;
24 The first question is there a chance that we're
25 going to see a short-term NO standard in the next couple
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
fl
*-
1 of years?
2 MR. JORDAN: I wish I could give you a positive
3 answer on that.
4 The status is simply this. We have gone through
several attempts at putting together a data base from
which to either set or reject setting a short-term NO~
standard. And each time we think we've something to go
8 on, it tends to get shot apart. Very simply, that.
9 At the present time, we expect to go before
10 the Clean Air Scientific Advisory Council in October, or
11 around October the 7th, with a next proposed type of ap-
12 proach for handling the short-terra NO- standard.
13-1 suspect that there will be a short-term
14
15
16
17
18
19
20
21
22
23
24
25
standard. I suspect that it will not be a very stringent
short-term standard.
MR. BEAR: Okay.
My second point is I discern a transition in
EPA's thinking with respect to the standards that seems
to be very much in line with the Reagan's Administration
position on the standards, that we're going to move away
from a threshold .type of standard to more of a risk analy-
sis type of standard for all the pollutants.
Is this position being formally taken by EPA
during the upcoming reauthorization of the Clean Air Act
Amendments, or is it a kind of an informal position that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
1
2
3
4
5
. Don't be surprised. That's who we work for.
o
7
8
9 force in 1970. And we're learning that the old concept
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
131
EPA has taken?
MR. TIKVART: Somebody?
MR. RHOADS: Sure. Why not.
I expect a I detect a note of surprise that
EPA is going along with the Administration's philosophy.
There've been options within EPA for a long time.
We're learning things. We started in this business full-
of threshold, in many cases, is not medically sound.
We're learning that the deterministic approach to stand-
ard setting sometimes is overly stringent, sometimes is
not stringent enough.
. . p
So we have a combination of technical knowledge,
medical facts, and political philosophy, which just seems
to be coming together, basically, right now.
MR. BEAR: Thank you.
MR. NOCEMSON: My name is David Nocemson from
Los Alamos -National Laboratory.
What people have been saying, in terms of the
.actual processes going on, they're stochastic, meteorolog-
ic, and even the emission processes. So it seems to make
sense to go to some kind of statistical,or probabalistic,
approach.
It seems like you're going to get into a problem
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
132
1 in terms of how do you validate these methods. The meth-
2 ods are going to depend on the -dispersion models and also
3 on the statistical techniques. And, basically, the bot-
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
torn line which you're estimating is the expected number
of exceedances. How are you .going to validate your meth-
ods for expected number of exceedances, for these, looks
like, worst case conditions? It seems like you need
years and years of data to do that. It would be even
more difficult than trying to validate the dispersion
models.
MR. CURRAN: Actually, I think it's kind of
interesting, if you listen to the people earlier, who
were talking about model validation, the conclusion that
r
they seemed to be reaching was that the frequency
you know, the overall frequency distribution, if you ig^
nored the spatial and temporal breakdown, that that is
where the models were good.
I'm not entirely convinced that switching over
to an expected exceedance standard is going to make it
that much more difficult.
I think, in a way, that's almost what they're
looking at right now.
But I agree with you there are some implications
that will have to be looked into.
MR. NOCEMSON: But you're really talking about
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
'2021 234-4433 . WASHINGTON. D.C. 20005,
-------
133
the distribution well, looking at it from the statis-
tical point of view, you have to look at the distribution
2
of the expected number of exceedances.
3
MR. CDRRAN: Yes. And right now they're look-
4
ing at what the second max is* So both of them are kind
5
of one number a year.
6
MR. NOCEMSON: And there's kind of uncertain-
7
ty associated with that. You need you're- talking a-
bout extreme values. And, usually, you need a long year's
9
record, if you look at the work that's been done in terms
10
of the water area. You need they're dealing with rec-
ords of like 50 years or so. And those aren't always ade-
12
quate-to-talk-about the case of long-term, you know, a
13
10-year flood, or a 50-year flood.
14
MR. CURRAN: Oh, yes. There'll be problems in
15
terras of the length of the data. But on a practical bas-
16
is, I think that a lot of it won't different a whole lot
17
more than.it does right now, as far as looking at the
18
second max.
19
Yes. I agree with you. I think part of it
20
is going to be almost like a philosophical definition of
21
22
23
24
25
what it means when we use these models. Do we really
want to talk -about the expected exceedance rate over a
1,000. years, or 3 years, or what.
So there are almost some philosophical
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
f2021 234-4433 WASHINGTON, D.C. 20005
-------
4 134
" ' '.5 r.
I definitions that'll have to be made. And then, you know,
2 some practical considerations.
3 MR. HAYNES: My name is Eldewins Haynes. I'm
4 with the North Carolina Division of Environmental Manage-
5 ment.
6 First of all, is there any plan to account for
7 synergistic effects in the standards in the future?
MR. TIKVART: Bruce.
... ;,
9 MR. JORDAN: That will strongly depend upon
10 the capabilities of the medical community to demonstrate
the synergistic effects.
12 MR. HAYNES: Okay.
Secondly, about the Ex Ex method. I believe
r
there was some discussion about a autocorrelation, a co-
14
rrelation, I suppose, between various emission parameters
ID
. .._ i
.. and meteorology and time of day. Is that what was im-
lo
plied? ""^
MR. STEIGERWALD: Autocorrelation, as I used
18
it, was the correlation between adjoining lots of coal.
iy
Is it an appropriate technique to go into the distribu-
tion independently every day. Or, you know, every three
hour period. Or does what happened yesterday influence
where I should go into the distribution today.
So,when I use the term autocorrelation, I meant
the ability to alter the way we pick tomorrow's lot of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE. NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
135
coal to be burned. And it did not have to do with the
meteorological variables. We do not modify, at all, the
output of the meteorological dispersion model.
MR. HAYNES: No. I didn't mean the modifica-
| tion of the dispersion model, but incorporating the var-
iation of emissions to the meteorology, or to the time
of day. For instance, a power plant will have a varia-
tion of emissions depending upon those factors.
MR. STEIGERWALD: We have our Federal Reg-
ister notice, in which we will be, I hope, proposing the
use-of the-Ex~Ex method, allows only the sulfur content
of the fuel to vary, or the sulfur content of the fuel
and the efficiency of the control device. It does not
allow loading to vary.
The modeling is done with the current assumption
of worst case load held constant. And that was done,
partially, because we do believe that they aren't inde-
pendent variables. That loading and meteorology, both,
are some function of time of day, often. Or of season
of the year, often.
So, we have not allowed loading to be consid-
ered a variable in the Ex Ex method, as we will be pro-
posing it soon.
MR. HAYNES: Thank you.
MR. TIKVART: Jim.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. O.C. 20005
-------
136
MR. PRESTON:^ My name is James Preston^ 1 work
for Tenneco, Inc.
2
My question would be directed at the entire
3
panel.
Basically, as long as we're considering Gaussian
5
plume models, that do not modify the plume's trajectory
6
with meteorology, or time, or space, we're looking at
7 .
a problem that can fairly readily divorce the problem of
8
time from the analysis,- and treat it as a spatial sta-
9 :,
tistical problem, at a considerable savings in machine
10
resources, with no substantial modification results.
If we look at the deposition reaction mechan-
12
isms, and-the transport, long-range transport consider-
13
ations, has any consideration been given in terms of the
14
significance of the time or temporal considerations that
15
the future noIds in this area, in these considerations
16
that we're talking about.
17 " -
And the questions open to any one.
18
MR. TIKVART: Jim, that's kind of a complicated,
19
involved question. I think the answer is we're still
20
struggling with basic principles, and haven't gotten to
21
the specific technical issue, or implementation question
22
that you're asking.
23
MR.PRESTON: Okay. Let me put it in little
24
more down to earth. September the 1st, the State of Nortl1
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
WASHINGTON, D.C. 20005
-------
137
- Dakota's going to hold a hearing on long-range transport
models, for use in the regulatory matters. Inside EPA
4* ^^~-~M~~
held several interesting articles on what will EPA do,
3
what will they do.
4
_ So, what I'm getting at, the regulatory envir-
5
onment is going to walk into this problem very soon, re-
6
gardless.
The question is: are we looking at the time
8
aspects,like the presentations .today showed that on a
y
time correlated.basis we don't have much results that
10
look very favorable. And what I am indicating is the
transport time factor, so long as it does not impact the
*£
plume trajector, like, for instance. In any way, then
13 r
you can take the time out of it. Mostly, your point re-
14
ceptor correlations assume instanteous transport. There
are some modifications with distance away. But it does
16
not vary from hour to hour. If it takes two hours from
17
the time it's released to reach the receptor, you're not
changing the direction of the plume flight.
19
So, therefore, time is not a factor. But once
20
you get to long-range transport, this is no longer true.
21
So, what I am getting at, if we make decisions
today that divorce ourselves from the time and spatial
23
aspects, will we not be crippling ourselves in the near
term future when these become more significant factors?
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
- ^ 138
- Is any one looking at that? This is what I
am asking.
q
MR. TIK7ART: I think the answer is no one has
looked, yet. But, perhaps, Martin Shock would care to
5 comment. Or would he not?
6 .MR. SHOCK: In the context of the upcoming hear-
7 ing, no. No comment.
8 MR. TIKVART: In a technical context.
9 MR. SHOCK: That is correct.
10 MR. TIKVART: Jim, I don't think we've got a
11 good answer for you. You're asking a valid question.
12 But it's too far down the road for us right now to see
13 . the answer.
14 Next question.
15 MR._.ARBRECHSTADT: Marcel Arbrechstadt. Motor
16 Vehnicles Manufacturers Association.
17 It seems to me that'ulitmately the stochastic
nature of-the instantaneous concentration of a pollutant
19 has to be-considered in the setting of the national am-
20 bient air quality standards. And I'm thinking of the
21 inclusion of this factor in the risk analysis techniques
22 that the EPA's working on towards improvement of this
23 setting of national ambient air quality standards.
24 And I was wondering if Mr. Jordan would like
25 to say a few words about how that process is proceeding
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
139
That is, the inclusion of the risk analysis in
the standard setting process.
MR. JORDAN: It's still, you know, I'd just
like to say that we in the standard setting business got
ahold of the term risk a couple of years before the new
Administration came into being, so it*s not something
new with us.
We do have an extensive risk assessment program
under development now. And, hopefully, sometime about
the middle of next year that program should have its first
test in an environment to see if we can, in fact, set
ambient air quality standards based on risk. And have
medical people tell us what the risk of being exposed to
r
this level, or this pollutant, is, for these'kinds of
health effects.
I am hopeful that the work that Tom and Bern
spoke of here will give us some framework then, as we go
to try to implement the risk program in setting the stand-
ard, to go to the medical people, and talk to them in
terms of what does it mean to be- exposed to the pollutant
four, five, six, and seven times, in that sense.
And so that ultimately, if everything worked
out, you could probably get this standard set into a sto-
chastic concept also.
MR. TIKVART: If there are no further questions,
NEAL R. GROSS
* COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1 I'd like to bring this morning's session to a close.
2 I'd like to thank you for your patience with
3 us, because we had an awful lot of information to present
4 to you.
5 This afternoon we will deal with a summary and
6 a panel discussion of the Airlie House workshop. This
7 afternoon's session will begin, promptly, at two o'clock.
Panelists, please meet with Shep Burton in the
9 back now.
10 (Whereupon the morning's session was adjourned
at 12:28 p. m.)
12
13
14
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
* COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
234-4433 WASHINGTON, D.C. 20005
-------
!' .f
CERTIFICATE OF REPORTER '
I hereby certify that the foregoing transcript
represents the full and complete proceedings of the 8/10/81
aforementioned matter, as reported and reduced to type-
writing under my direct supervision.
NEAL R. GROSS
''
NEAL R. GROSS
1330 VERMONT AVENUE.
-------
GFC ~
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
v
17
18
19
20
21
22
23
24
25
GOVERNMENT OF THE UNITED STATES
ENVIRONMENTAL PROTECTION AGENCY
SECOND CONFERENCE ON AIR QUALITY MODELING
-MONDAY, AUGUST 10, 1981
. . AFTERNOON SESSION
The conference was held in the Thomas Jefferson
Auditorium, South Agriculture Building, 14th Street and
Independence Avenuey- S. W. >; Washington, D. C.-, - Joseph .- . ,, ,..
f ' ' - - _
Tikvart, .Chief,_Source Receptor Analysis Branch, Confer-
ence Chairman, presiding.
PRESENT: (First Panel)
(202) 234-4433
JOSEPH TIKVART, Chief
Source Receptor Analysis Branch
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
RICHARD RHOADS, Director
Monitoring & Data Analysis Division
U.S. Environmental Protection Agency
Research Triangle Park, N.V7. 27711
JAMES DICKE
G. THOMAS HELMS, Chief
Control Program Operations Branch
U.S. Environmental Protection Agency
Research Triangle Park, N.C. 27711
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
WASHINGTON, D.C. 20005
Chairman
-------
1
2
3
San Rafael,-Gal. 94903
4
MR. STEPHEN .CONNOLLY, President
5
6
7
8
Lexington, Mass. 02173
9
LEWIS KONTNIK, ESQUIRE ;
10-
11
12
13
14
15
16
17
US.
19
20
23
24
3Btes»53SEr
PRESENT (Continuectfr
DR. C. SHEPHERD BURTON, Vice President
Systems Applications, Inc.
101 Lucas Valley Road
Schwartz and Connolly -
1747 Pennsylvania Avenue, N.W.
Washington, D.C. 20006
DR. BRUCE EGAN, Vice President
Environmental Research & Technology, Inc.
3 Militia Drive ...... ' - - -
Kunton'Se Williams- .. : -
1919 Pennsylvania Avenue, N.W.
Washington, D.C. 20006 --- -
MR. JAMES SALVAGGIO, Chief
Air Quality Planning Section
Pennsylvania Department of Environmental Resources
"PV^O. Bo'Jf 2063 -":-- -- -
Harrisburg, Penn. 17120 - -- . . .
DR. MICHAEL D. WILLIAMS '
Los Alamos National Laboratory . ._ ' -
Group S3, Mail Drop 603 a
P. O. Box 1663
Los Alamos, N. M. 87505
- - -
DR. STEVEN WISE .
Mobil Research & Development Corporation
Research Department
Paulsboro, N. J. 08066
ALSO PRESENT:
21
VERN WALKER, Legal, Hawes & Symington
JOHN WOOD, Central and Southwest Services
LOWELL VAN VLICK, Tuscon Electric
KENNETH McGUIRE, Kentucky Air Pollution Control
LOU TOSIE
DAVID NOCHEMSON, Los Alamos National Laboratory
ELDEWINS HAYNES, North Carolina Division of Environmental
JAMES PISTON, TENNECO
«v
DOUG FOX
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 200O5
-------
T *T *"* E X
2
3
4
5
6
Introductory remarks by Joseph Tikvart 4
Statement of Doctor C. Shepherd Burton/
Vice President, Systems Applications,
Inc 8
Statement of Mr. Stephen Connolly,
President, Schwartz; and Connolly 14
7
8
9
10
.11
12
13
14
18
Statement of Doctor Bruce Egan,
Vice President, Environmental
Research & Technology, Inc. 27
Statement of Doctor Michael D. Williams,
Los Alamos National Laboratory 36
Statement of James Salvaggio, Chief,
Air Quality Planning Section,
Pennsylvania Department of Environ-
mental Resources . . 44
Statement of Doctor Steven Wise,
Mobil Research. &_.Development Cor-
poration, Research Department 51
15
Statement of Lewis-Kontnik, Esquire,
16 Hunton & Williams 56
1-7- Statement of G. Thomas Helms, Chief,
Control Program Operations Branch
(MD-15), U,S. Environmental Pro-
tection Agency. ............ ... . 64
19
Summation by Doctor C. Shepherd Burton 69
20
Questions and Comments 81
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
202} 234-4433 WASHINGTON, D.C. 20005
-------
i'l
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
C E E D'l N'G S
2:00 p.m.
MR. TIKVART: I would like to begin this after-
noon's session, if you all would get settled in, please.
This morning I read you a tentative list of
speakers for tomorrow. There has been a number of changes
and additions to that list, which I would like to document
for you right now. The modified list of speakers is
posted outside, but for the record I will read it Bright
now.
-For"the Government agencies, those who have re-
quested to speak will speak tomorrow morning starting at
9:00 o'clock a.m. in the following order: for the Federal
Aviation Administration, Mr. Sundataraman; for the Federal
Highway Administration, Doctor Jongedyk and Doctor Carpen-
ter; for-the National Oceanic and Atmospheric Administra-
tion, Mr. Heffter and Doctor Draxler; for the U.S.
Geological Survey, Mr. Goll; for the Nuclear Regulatory
Commission, Mr. Markett; for the National Park Service,
Mr. Henderson; Department of Energy, Doctor Shull; State
of Texas Department of Transportation, Doctor Moe;
Ontario Mystery of Environment, Mr. Mishra; U.S.E.P.A.,
Region Five, Mr. Trout; Department of Health, State of
Maryland, Mr. Banta; Louisiana Department of Natural Re- -
sources, Air Quality Division, Mr. Raol; and Kentucky
: NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
,, afternoon will be Alan Witten, Richard Hanson, Robert Kohm,
o
7
8
9
10
11
13
14
15
16
17
18.
19
20
21
22
23
24
25
Division of Air Pollution Control, -Mr. McGuire.
Individuals who have asked to speak and there
are some deletions which I will not mention; I will only
mention now those who have indicated a desire to speak.
Those include those who are going to speak tomorrow*
Donald Moon/ Jerry Pell, Ray Wright, Ralph Sklarew,
Richard Fine, Mitchell Wormbrand, Philip Youngblood.
Those are the list of speakers I have. With^ the
increased list of Government agencies, I doubt that any of
~the~individual-speakers will get on tomorrow until after
lunch..
One individual on that list has asked for permis-
sion to speak early so that he can make another commitment
If there is anybody else who needs to speak early tomorrow
afternoon, please inform Charlotte Hopper or Ann Asbeal of
that.
I would like to repeat what I said this morning
and that is, there is a verbatim transcript being main-
tained of these proceedings. If anybody would like a copy
of that verbatim transcript, they should talk to Miles
Anderson, the Reporter here, who is with Neal R. Gross and
Company, Incorporated. He will be happy to discuss with
you obtaining a copy of the transcript. \
Moving into this afternoon's proceedings, we
NEAL R. GROSS fc .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
18
19
20
21
22
23
24
25
---. ' ., MI; ,. 6
have '& number of presentations and "a "panel "'discussion' "
dealing with the workshop on model uncertainty that was
held at the Airlie House this last May. This workship was
held specifically as input to the Modeling Conference to
develop ideas and concepts that you, the attendees at
this Modeling Conference, could consider and comment on.
When we originally formulated the concept for
this Modeling Conference and what its emphasis should be,
we were concerned that it would be very difficult for you
to grapple with the still somewhat abstract ideas of deal-
ing with model uncertainty and decision-making; therefore,
we got together the group of recognized experts in this
general_.area,...if there is such a thing as an expert in
model uncertainty and decision-making, a group of invited
individuals last May to deal with this complex problem in
a relatively closed environment in groups that could deal
17
16
together and come up with some meaningful ideas and recom-
mendations.
That workshop was organized and chaired essentia
ly by Doctor Shep Burton of Systems Applications, Incor-
porated. Shep will be the main speaker for to summariz
the results of that conference this afternoon and will be
more or less moderator of the panel discussion.
I would like to introduce the individuals who '.,
are participating in the panel discussion this afternoon.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"Starting with my right is Doctor Shepherd "Burton, Vice -, 1
President, Systems Applications, Incorporated; next and
the seat is vacant is Mr. Stephen Connolly, President .
of Schwartz and Connolly; Doctor Bruce Egan, Vice Presi-
dent of Environmental Research and Technology, Incorpora-
ted; Doctor Michael Williams, Los Alamos National Labora-
tory; Mr. James Salvaggio, Chief, Air Quality Planning
Section, Pennsylvania Department of Environmental Resources
Doctor Steven Wise, Mobil Research and Development Corpor-
ation; Mr. Kontnik of Hunton and Williams; and Tom Helms,
who you-met-this morning, Chief of the Control Programs -
Operations Branch of Office of Air Quality Planning and
Standards.
r
Before I turn it over to Shep, there are any
number of other individuals I should not say "any num-
ber," because the total was 40 but there are a number
of other individuals in the,audience who participated in
the Airlie House workshop. I am sure that Shep would like
to share the credit, if I may say credit, for the workshop
results with you. Would these individuals stand because
I am sure if I tried to name everybody I would miss some-
body. Dick Londergan; Don Moon; Vern Walker; Lou Kontnik;
Dick Kerch; John Wooten; Doug Fox; Norm Bowne; and Martin
Shock there in back. Anybody else?
(No response.)
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 V/ASHINGTON. D.C. 20005
-------
1
r "
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
8
11
. J5
.... , MR.. .TIKVARTii: if. you. are,--you will not share the
recognition.
Okay, with that brief introduction, I will turn..
it over to Shep Burton, and more or less leave this after-
noon's proceedings up to him. I will be here as a traffic
cop, more or less, and to keep the thing flowing. But,
Shep, I would like to turn it over to you now.
DR, BURTON: Thank you, Joe.
Basically, where there is so,many of us and we
are trying to keep it to, on the average of about ten
minutes apiece, I am not going to have that much time, but
^
what I would like to do is to thank Joe and thank everyone
for coming and for having us here and indicate that I am
really going to serve as a moderator and then but
prior to that, try to set the tone, provide some back-
ground on why we met and what we dealt with there.*
As Joe indicated, we were really looking at the..^
role of models -in regulatory decision-making and we. sort
of there were actually 49 of us, Joe; some crept in or
more EPA people showed up than we really originally though:
However you want to look at it..
As you indicated, there are a lot of people
here today that were present and they are out in the audi-
ence. One person that you a couple that you did not
mention that I knov; were there and arc here today, but
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS* '
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
f
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
".probably took arrl-ong^ lunch,, B.ruce Jordan, Tom Curran and -
Bern Steigerwald, so let me mention those.
The as I mentioned, the workshop was held to
address perceived growing concerns associated with using
air quality dispersion models as an air quality management
tool. Now, in planning the workshop, it was recognized
that the practice and the technique of modeling has really
grown considerably and in the past ten years there has beer
an unprecedented increase in the skills associated with.
modeling. In part, because of what you heard today in-
just the overall feeling of the participants in the work-
shop, I think that everyone would agree that we could ex-
pect the increase in skills to continue.
, If the increase in modeling skill is to produce
an increase an not a diminution in the quality of the
clean air management decisions, there has to be a corres-
ponding increase in wisdom and parenthetically I might
add that the perfect model might still be misapplied in a
regulatory or policy-making setting with one consequence
being incorrect decisions.
But there has not been as great an increase of
j wisdom by either the modelers, the decision-makers or the
regulatory system they serve. The official rules by which
management decisions are made are still those that were
initiated in the early 70's.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
-ff.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
i. .-,.- 10
-.-.. «..»/- /<.*-*» -, , £ ) , .-,- * T-. - r-,V T. ..c.-r.-r- ^ '* < ~ " .«.»..-» ^,
' - ' ' ^ - - . ,.~ . " . ~._", ,'_ .. ,. ' { ... " . _ ->
Notwithstanding this morning's discussion regard-
ing future trends in modeling, modelers still dutifully
provide decision-makers with a single concentration esti-
mate located at some place for some brief period of time
in a year and regulators still base decisions solely on.
such estimates with only infrequent and informal consider-
ation of the uncertainties or the implications of such un-
certainties in those estimates.
So, skill without wisdom appears to contribute
to our problems. If they are to be cured, it will be not
- by a mere increase of skill, but by the growth of such
wisdoms as the time demands the times demand.
Thus, the workshop devoted little or no effort
r
to discussions directed at model improvements per se. The
efforts of the participants were devoted primarily to .
identifying from their individual and collective experi-
ences practical recommendations which, if adopted, should,
one, reduce the doubts and risks concerning modeling meth-
ods employed in and conclusions derived from air quality
impact assessments; and, two, should assure the wide ac-
ceptance of the use of modeling in air quality management
by interested parties and by the public.
Now, the structure of the workshop consisted of
three independently functioning groups of approximately
equal size who addressed the same four sets of questions.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20O05
-------
1
2
3
4
5
6
7
8
9
10.
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
113
They addressed those questions and tested their
answers or their approaches dealing with those questions
against three practical applications: a PSD permitting
problem in the oil shale area? a SIP revision problem in
the Chicago AQCR dealing with SO^ and ozone, changes in
the SIP; and the transport of pollutants across political
boundaries.
Each work group is represented on the panel to-
day and the panelists hold a variety of interests and
views about the recommendations, which individually they
will share with you. They will discuss the findings of
their respective groups, as well as comment to the extent
they wish-on-the workshop recommendations as a whole.
r
There has not been any strong attempt to really
orchestrate the presentations that will be made today.
You might read as much into the summary report itself in
that it was-really a collection of the ideas of the par- ..
ticipants and the recommendations- of the participants.
Before I do this before they do this, let me
outline in broad terms the recommendations offered by the
workshop. They are five:
First, the utilization of cooperative processes
whenever possible that provides for earlier and substan-
tive involvement of interested parties and it encourages \
the anticipation, definition and resolution of potential
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
12
areas of conflict.
Second, the utilization of air quality impact
assessment plans or protocols to identify and define
models, tasks, analysis steps, data bases, potential dis-
putes and the means for resolving them, and the schedule
for accomplishing the impact assessment.
Third, the utilization of advisory groups to pro-j
vide oversight, guidance and peer review.
Fourth, the explicit stipulation of uncertain-
ties through the best available means in all modeling-re-
lated-decisions;- encouraging modelers and decision-makers
to become familiar with methods of estimating uncertain-
ties and the use of those estimates in modeling-related
decisions; seeking the identification of measures other
than rare events for use in decision-making and identify-
ing other readily available modeling outputs to augment
the rare event measure currently used in decision-making.
Fifth, the selection and application of new or
modified modeling approaches rather than insisting that
existing guideline models be used for all circumstances.
Improving the methods used to convey changes in models,
methodology and process; consideration of the establishment]
of a modeling center concept to centralize certain model-
ing activities and insulate the technical modeling tasks'.
from the political decision-makino process.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20-
21
i
22
23
24
25
;, -- 13 ,t.
' -t , t
Now, all "of those may sound like a lot of apple'.
pie and motherhood and but, in fact, in a lot of in-
stances, or at least instances of which I am aware, they
have been practiced and they have worked. They may not
work in all circumstances, and I would hope that people
here could comment on that.
An additional recommendation that was not ad-
vanced by the workshop, but which did appear to me on read-
ing all of the recommendations, to be harmonious with tha
workshop recommendations involved establishing a modeling-
related quality assurance activity in air quality manage-
%
ment.
In the-summary report you can find, in general
r
terms", not too specifically though, an outline of the prin-
cipal elements of that air quality modeling modeling
quality assurance activity. ,
Now, as Joe indicated, or as I indicated at the
beginning and I indicated at the beginning, each panel-
ist will discuss the workshop. We have two broad cate-
gories of participants and I would like to be able to re-
frain from putting labels on people, but at this particu-
lar point I cannot those who can speak for a particular
work group as a whole. In other words, they do not really
represent interest industrial interest, environmental
interest or governmental agency interest and so on. They
NEAL .R. GROSS
COURT REPORTERS AND TRANSCRIBERS"
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
i
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.1.4 .
come from the consulting or the research7 community'.
The second group falls into the category of
those who do represent someone,. who.do have some interest
in addition to the broader research interest.
The' first two will speak Steve and Bruce will
speak on the first category and the last group will speak
from a special interest. They are all free to comment on
any aspect of the recommendation's.
Steve?
MR. CONNOLLY: Thank you.
STATEMENT OF STEPHEN CONNOLLY
\
MR. CONNOLLY: What I would like to do is pro-
ceed in the-following order: I will state for you what
r
our Work Group Number Three problem was and talk a
little bit about the membership of the work group and its
organization; how we proceeded, the approach and method~
ology? and relate some specific concerns and then finally
review briefly our, the group's generalization and spe-
cific findings.
Work Group Three's specific problem was the ex-
tent to which air quality modeling can assist in the devel-
opment of a defensible regulatory posture with respect to
the issues of the transport of pollutants across inter-
state political boundaries.
The group started by noting that it felt that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1 .
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.JL5
the problem was not to deal with the issue-of-trans- ;
boundary for its own sake, but rather to explore it only .
in terms of the principal workshop questions, and as we
j shall see in a minute, we ended up concluding that the
important issue was the long-range or mid-range transport
and the modeling with respect to those and not so much the
\
question of the intervention of boundary problems.
.In terms of the membership, we had, out of about
i 15 or 16 members, we had representatives from EPA, includ-
ing ORD, OAQPS and various regional offices, other depart-
ments in-the-Federal Government departmental agencies
in the Federal Government, from two state governments,
from industry, from the modeling community and from severa!
law firms.
The work group approach and methodology can be
characterized-first by giving you some background on the
context in which the group operated and then talk spe-
cifically about the approach and how we proceeded, step-
by-step.
The group began by deciding to limit the deliber-
ations to the question of utility of modeling and dealing
with long-range and mid-range pollutant transport issues.
The short-range transport problems were to be raised in
other sections having to do with SIP revision and perhaps
v.'ith PSD. Therefore, we felt that it would be batter to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
t330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
13
19
20
21
22
23
24
25
16
,t .
deal with the- mid- and -long-range and -leave *the- -short- V.
range to the other sections.
We also thought that the problems that would be
raised by short-range transport were essentially the same
modeling issues as would be raised in other contexts such
as the SIP modeling and the PSD modeling.
We also focused on the modeling aspect of the
long-distance transport problem and not, as I said, the
trans-boundary nature of the problem.
We defined mid-range transport as transport over
20 to 200 kilometers. We defined long-range as that ex-
tending beyond the 200 kilometers.
Then a set of distinctions were made between
modeling problems that the group felt were associated with
long-range transport and modeling problems arising out of
other scales of modeling.
Long-range must deal with multiple emission
sources, in some cases, treating a few sources as individ-
ual point sources and other sources as area sources.
Another distinction was that greater difficulty
was experienced in identifying and quantifying the con-
tribution of individual point sources than ambient concen-
trations at any point down wind.
There is greater difficulty defining appropriate
! source groupings and geographic boundaries for area sources
NEAL R. GROSS
. COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
17
treated" in" the" modefs'.. " Tt" was* agreedv* 'however/ that the
use of longer averaging times tend to reduce the relative
uncertainties in all model predictions, including those of
long-range models.
Because of the number of sources and the dis-
tances involved, the data needed for long-range transport
models that is, emissions, wind speeds and directions,
deposition rates and so forth are frequently both more
numerous and more difficult to acquire than it is for
shorter-range modeling. Few, if any, well-defined proced-
ures or precedents governing the use of long-range trans-
port models have been developed for the legal and regula-
tory arena.
Finally, the air quality problems associated
with sulfates, other fine particulates and acid rain pre-
cursors and the high stakes involved in regulatory deci-
sions on these problems led the group to explicitly recog-
nize the need for better understanding and further develop-
ment and application of long-range transport models. A
bow towards more research.
In terms of the approach of the group, we
basically went through four or five phases or steps.
First, we identified a list of issues and questions. We
developed these from these issues and questions from .
the perspective of the regulator or the policy-maker.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19.
20
21
22
23
24
25
18
t ,
These'were -questions- and "issues with respect to the long-1'
range or mid-range transport which modeling might help
eliminate. We basically asked ourselves the question:
"What do we need to do if we were regulators or decision-
makers? What do we need to do? What do we need to know
to do it?' What, if anything, can modeling do to help us
in knowing this and in doing this?"
. The list was designed to help guide the modeler
in defining, prioritizing, designing and conducting model-
ing activities and in presenting modeling results in the
-form-most-relevant for'.the regulatory decision-maker.
%
Step Two, the group .identified model capabili-
ties with respect to long-range and mid-range pollutant..
transport. What we did was to develop a matrix which dis-
played model capabilities by general model type. This
would be used to assist not only the modeler, which is ob-
vious, but also the policy-maker or the regulator in deter;
-*£%:
mining and choosing the model most capable of dealing with
the long-range, mid-range transport problems.
Step Three, the groups stepped back and attempted
to identify attributes which can help determine the legit-
imacy of the regulatory process and the use of models in ..
it. The work group determined that it would be useful to
generate a list of the general attributes which helped to.
determine whether a regulatory process or the choice and
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
-1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
-19
applica'tiori^of ~a 'p"articular"air qualiy"'nSclei is accorded"
.*
legitimacy by all interested parties. -
The group by legitimacy, the group meant is
it accepted as fair and is it likely to produce outputs
which all interested parties or most interested parties
| will accept. Such a list of "processed legitimacy attri-
butes" was then generated. It would be used by policy-
makers, regulators and modelers in attempting to choose
regulatory approaches and. tools which are most likely to
be accorded legitimacy.
Step Four and Step Five togehter really was the
generation of responses to the .principal workshop questions
and-then-generalizations and findings. To do this, we
split into two sub-groups. Each answered the four princi-
pal work group questions; one with respect to long-range
modeling and the other with respect to mid-range modeling.
Upon completion of their separate tasks, the
two sub-groups reconvened as one, developed a consolidated
set of responses to the four questions and then a set of
generalizations and findings regarding the role and use of
mid to long-range transport models within a regulatory
framework.
Before I review some of those findings, let me
just touch on the issue of legitimacy of current applica-
tion of modeling' and uses of model output in the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
regulatory setting.
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
ill
' * ',
20
Work Group Three determined that the policy-
maker, the regulator and the modeler each should be keenly
sensitive to the need to assure the wide acceptance by
interested parties and by the public and of the regulator
decisions and processes and the application of technical
tools in those decisions and processes.
It is this widespread acceptance, fhis legiti-
macy that assures that the process of government and the..
public policy decisions those processes produce is acceptec
by the public. The work group developed a list of attri-
butes which were not intended to be exhaustive, but more
as examples of process attributes which may legitimate
r
that process and which should be attempted to be maximized
i
by anyone who is attempting to operate such a process.
It was recognized to a certain extent some of
| these attributes may be opposite poles with the same, con-
tinuim. It was urged by the work group that policy-makers
regulators and modelers consciously consider and balance
each of these attributes and others which they may subse-
.quently develop in constructing a regulatory process or
i
selecting and using air quality i?.o'.;->ls.
Let me give you examples of some of these attri-
butes. They are not listed in any certain in any
necessary order of priority. One would be the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
18
19
20
21
22
23
24
25
_-'- ,; - 21.
minimization of resource use? two would be the maximiza- *"
tion of simplicity; three, assurance of timeliness; four,
consistency of process and results within a process and
across various processes; five, accessibility for inter-
5 ested parties; six, that they are understandable to partic
6
7
8
9
10
11
12
13
14
15
16
ipants and to observers alike; seven, that they are both
flexible and adaptable; eight, that they have the ability
to handle uncertainty and express it well to identify
it and express it well; nine, the ability to handle change
both in the process and in the inputs; ten, that it is a
defensible process producing defensible results; eleven,
that it can be publicly'documented; twelve, that it is
administratively easy or relatively so; thirteen, that
there is ample opportunity for public participation, but
without .undue delay; fourteen, that it is scientifically
valid.
17 ! - The work group also recognized that the use of
models and model output in a regulatory setting is more
likely to be accorded broad acceptance and legitimacy if
cooperative processes were utilized. Such processes will
be particularly important in assisting .the technical and
policy.analysts that make it clear to participants and the
public the capability and limits of models in assisting
the decision-maker in using the model results.
The process, whenever feasible therefore, should
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. MW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
--- 22
(r - :
incorporate adequate opportunity for early involvement of"
the interested parties, in a cooperative effort to antici-
pate and to define areas of potential conflict and areas
of potential agreement and to narrow disputes and focus
cooperative efforts on the early resolution of these dis-
putes.
Let me review for you some of the key generali-
zations and findings. Within fairly large areas the
example we cited was 50 _X 50 kilometers estimates de-
rived from long-range transport models by themselves are
.decreasingly-useful for allocating control levels to spe-
cific individual sources. However, it does not follow
that these models have no contribution to make to regula-
tory policy decisions. The group found that emissions
from one grid square may have a significantly different
impact on concentration levels in a given receptor area
than emissions from another grid square.
Furthermore, emissions released on a windy
winter day may contribute significantly less than emis-
sions on a stagnant summer day. Consequently, long-range
models can provide information on a regional scale and may
be expected to quantify the relative contribution of a
particular source's or area's emissions to pollutant con-
centrations in another area.
The ai£ quality model application and selection
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"~c,^ r- " 23
«=- « -* * 4&.J
li
criteria for the medium scaTe"and" long-range 'scale pollu-'
tant transport and transformation evaluation are not appre-
ciably different, the group found, except greater sources
greater source and receptor geographic resolutions is
obtained for the mid-range and for the long-range. Over-
all, uncertainties are less for mid-range models than for.
long-range models. These uncertainties typically decrease
as averaging times increase.
As we go through these findings, I think you .
will note that, not surprisingly, some of them will sound
fairly obvious; that, however, doss not detract from their
importance as a concensus finding of a working group of
very different experiences and interests.
i"
-Another finding was that the availability and
magnitude of data processing resources, both hardware
and staff, technical know-how of the user in time and fi-
nancial resources can significantly affect the resolution
and uncertainty of the modeling'predictions in any specif-
ic application. These factors must be taken into account
in the process of it, to be legitimate.
Exceptability/defensibility of long-range trans-
port models depends upon the level at which and the number
of times model evaluations have been performed. This is
true to a greater extent than for the nore traditional and
widely-used short-range model. The intended use of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
long-range transport models^shoulcL-deterraine whether and ' -
how such previous model validation is required as a selec-
tion criteria.
Regulatory activity should rely on those models
which have demonstrated that that is the model evaluation
results.
Appropriate procedures for the validation and
testing of MESO scale and long-range transport.models
should be developed and agreed upon by a reputable scien-
tific group with extensive peer review.
..Current.lack of data and lack of adequate tech-
niques make it difficult to identify and examine all
sources of uncertainty in long-range transport models.
However, sensitivity analysis, if explicitly developed and
documented, combined with knowledge of tha uncertainties
and input data, model parameters and model algorithms will
assist the decision-maker to identify, understand and
express for the regulator the associated uncertainties
arising in model results.
Dependence on single-number, numerical represen-
tations that do not express uncertainty should be avoided.
Attention to graphical representations of model results
may be particularly important for assisting regulatory
decision-makers in understanding long-range transport
i ir-odel results; thus, for expressing their associated
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
uncertainties and also to demonstrate to the regulator
what other information the modeling has generated other
than simply that this one high number.
Uncertainties should be recognized and expressed
at all levels of the decision process. Present regulatory
process and structure, in some respects, is not well con-
structed to deal well with the uncertainty that is inher-
ent in the use of models and model outputs. Therefore,
dealing adequately with this uncertainty may require revi-
sions of the regulatory process and structure.
It should be recognized that the use of uncer-
tainty may be subject to misunderstanding, to misuse or
to abuse as this information is passed up through the de-
cision process. In order to limit the potential for such
misunderstanding, misuse or abuse, decision-makers should
be required to explicitly state on the record how he in-
tends to take uncertainty into account and to identify,
describe and balance, among other, non-modeling considera-
tions that affect his decision.
One important response to dealing with uncer-
tainty in model results and in the entire regulatory pro-
cess should be the adoption where feasible of a coopera-
tive process which helps the modelers and the decision-
makers to find areas of agreement and disagreement and to
focus on.narrow issues of conflict.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
202) 234-4433 WASHINGTON, D.C. 20005
-------
1
8
10
11"
12
13
14
15
16
17
18.
19
.20
21
22
23
24
25
26
-...-.- . ...nt. :- --;--. --2.:c -:c ^u'l^'-rr::...:-.: ; ^ e,-e v.:..:. ...:.---
Decision-makers"should-have: the responsibility ~
to recognize, to take into account the various levels of
uncertainty associated with model results. They should
make reasonable efforts to understand the effects of model
uncertainty upon their capability, to determine the rela-
tive effectiveness of alternative control strategies for
regulatory policies.
The last two, because of the regional nature of
pollutant transport, statutory and regulatory frameworks,
that is, the Clean Air Act itself and the implementation
regulations, should more fully recognize the nature and im-
plications of these regional transport processes.
Long and mid-range transport models are expected
r *
to undergo rapid development and improvements in the fore-
seeable future. The process may result initially in more
complex models that are data-intensive.. These models may
require a high degree of technical skill to understand ...andr.
to effectively operate the models. The cost of using thess
models could be high, compared with using short-range
models. Such factors, obviously, may prohibit some users
from having access to the better models or to the latest
improvements.
One means of helping to reduce any potential prob
.lems would be to establish centralized air quality model-
ing center for the more important models and data bases
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
27,
1 " ~" ' * ---.--.> --
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
to be maintained and facilities-for-using the models to be
provided to all interested parties.
We will be able to accept questions, I think.
DR. BURTON: The next speaker is Bruce Egan, who
will describe the activities.of Work Group One that dealt
with the Oil Shale problem.
STATEMENT OF BRUCE EGAN
PR. EGAN: Thank you, Shep.
I am pleased to be here after seven hours on^ an
airplane this morning from Boston. I might have driven if
I had had--one-more hour, I suppose.
The Group One addressed the four workshop ques-
tions which were identified earlier in the context of the
new source review process in the context of the Oil Shale
region. 'Thus, problems associated with permitting incre-
ment sharing and permitting multiple sources located in
the region of complex terrain,'affecting both Class One
-*£§
and Class Two areas emerged from the discussions._
Much of the discussion within this group x-;as
oriented towards the basic need for physical and scienti-
fic realism in the application of air quality models and
their role in providing information to decision-makers.
Without a solid technical basis, decisions made
could not be defended and would be expected to have
biases of various forir.s.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
10
.11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
28
I would like to just briefly overview our de-
liberations and then go to some vu-graphs for specific
items.
The contexts and the feeling the understand-
ing was that the Clean Air Act, in effect, requires the
air quality management process to rely, in part, on models
or modeling techniques which are believed to represent or
simulate a sound in scientifically defensible understand-
ing of atmospheric dispersion processes.
It was expressed early in the workshop that be-
cause of the inherent turbulent nature of atmospheric
processes and because of the limitations of technology,
the air-quality manager has to deal with the fact that
r
numerical values predicted.by'any such model will not pre-
cisely mirror reality in detail and, therefore, compari-
sons of predictions with measurement will inevitably demon-
strate that any predictions v/ill have some uncertainty.
Sub-group One took the view that the presence of
uncertainty by itself should not deter the air quality
manager from using model results in the decision-making
process, provided the model represented a scientifically
sound simulation of tha relevant adversary processes.
The model, in essence, is asked to provide the best esti-
mate of what will occur based on current scientific know-
ledge. The decision-maker, in turn, nec-cJs to incorporate
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
29
-. C" >
these findings,, together,, .with Bother.-information, into the"
decision-making process.
This context, Sub-group One discussion focused
considerable attention on what should be done to provide
the decision-maker with the most appropriate information.
The models used and their input data should be
scientifically justified and technically appropriate.
This is basic to providing confidence to the decision-
maker on the scientific merit of the subsequent uses.
The group addressed what alternative output data
that is, alternative to that presently required, .might--
form a better basis for air management decisions, espe-"
cially alternatives which might reduce uncertainty*
Common practice in the making of decisions in
the presence of uncertainty is to call for additional,
sometimes peripheral, information to help frame the most
___ X
defensible decision.
"*£;^
'The Sub-group One report identifies additional
information presently available in the routine output of
computerized air quality models which could assist the
decision-ir.akers in this regard. The group also identified
other information which can be obtained with only minor
modifications to presently available models and which
would further provide useful supplemental information to
the decision-making process.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
.7
8
9
10
11
12
13
14
15
16
17
18
. 19
20
21
22
23
24
25
t-: 30-
The group' further "discussed how~in"formation' on
the uncertainty levels of predicted advice might be pro-
vided by modelists and endorses the concepts of the need
to have uncertainty information available.
However, two cautions were raisedi that, one,
it is difficult to do at the present time, that is, to
provide information on uncertainty levels; and, second,
that the process of decision-making may not necessarily be
improved until guidance is provided to decision-makers on
how this information should be used.
The report also draws attention to the upgrading
of models and the need for that and mechanisms for such
upgrading are outlined.
r
Let me now talk over a couple of the vu-graphs.
I have stated this viewpoint in my opening re-
marks, that basically the model, if it provides a scien-
tifically sound simulation of relevant atmospheric pro-
cesses, one would have uncertainty and one must learn to
live with it in sores way. Then, of course, the question
is what is the magnitude of that. But from that context,
the recommendations, general ones, I would summarize,
strive "for the most scientifically sound models; identify
model output of better use to decision-makers; and have
uncertainty in model predictions factored into the deci-
sion-making process.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1?
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
31
.? ii
We addressed .fouridifferent;guestipns basically.
--_-_
On the matter of the choice of models and this was in the
context of the Oil Shale PSD Program, that conservatism
in models presents problems and especially with respect to
increment tracking and v/hereas one can be conservative for
a number of reasons for a single source, for multiple
sources, the model would probably not only over-estimate
! the magnitude of impact of a single source, but would'mis-
j
locate that magnitude such that when you have multiple
sources, you have some trouble in the increment tracking
process...
More generally, appropriate models would have
this list of attributes not in necessary order of priority
availability, of course, because of the regulatory setting
and adequacy of documentation from the same point of view;
previous model-performance, in a positive sense, of course
that the input data requirements would match the type of
input data available for use? cost of computations be some-
what reasonable; simplicity by itself is an attribute in
the sense that because of the regulatory aspect and the
number of deliberations, it is much easier to deal with a
simple model if you can than a more complex one; scien-
tific validity and acceptability, very important that pre-
vious use of it would demonstrate validity and overall
accuracy and/or bias in outputs could be identified in
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
some way,
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17"
18
19
20
21
22
23
24
25
32
Now, at the end of the list we enter right into
the matter of how does one provide uncertainty.
In the use of models, I think all of the groups
came upon the conclusion that an agreed upon protocol was
very desirable for complicated regulatory settings and in
this particular case a protocol which selected the models
as best they could beforehand which defined the meteorolo-
gical and emission state or input requirements beforehand
and which defined what performance evaluations would be
would take place and what would be the measures of per-
formance presumably, if possible, would all be agreed upon
beforehand.
f
Overall, the theme was that scientific credi-
bility regarding the representativeness and quality of
the input data was important.
We rephrased this question slightly to what we-
really answered in our panel, which was how can uncer-
tainties be dealt with. This was going beyond simply how
do you quantify uncertainty, but the general notions of
the need to reduce the uncertainty by improving models -
and/or by using models to predict quantities with more
certainty. I will corns to that in a moment.
A general feeling in the context of the above
would be to strive generally for stating model predictions
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
V
18
19
20
21
22
23
24
25
in ;.a ".probabilistic 'sense. Some discussion, I missed this>
morning, but you have heard about "XX" of the methodology.
We are talking about basically having people think about
model output in a statistical sense rather than as a sim-
ple worse case does it or does it not excede a standard
type of thing in the absolute sense, supply additional in-
formation to decision-makers. Here, we were addressing the
fact that presently, a typical use, a model is used ?..-
through a computer to generate a large batch of informa-
tion, of which only really one number the highest,
second highest value at the worst receptor is used for the
decision-making process.
The output, we feel, is rich in other informa-
tion. We have some confidence that your model mirrors
what is going on and the decision-maker should be very
interested in, for example, the frequency of values which
are, say, 80 or 90 percent of the standard or the incre-
ment, or generally the frequency distribution of high
values.
The average of the top ten or, say, 20 highest
values at the worst receptors, to give you some idea of
how extreme the extremely high values are and how often
they occur.
The episodic nature of the highest values. Do
they occur randomly through the year or do they occur
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENU2. NW
(202) 234-4433 WASHINGTON, D.C. 200O5
-------
1.
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
34 ,
sequentially with different sort of health implications? '*
The location and extent of the geographic areas where
standards are where increments .are mostly likely threat-
ened. Is it a remote area, largely unpopulated or it is
an urbanized area?
Finally, we identified the fact that models can
compute dosages? that is, concentrations times population
density. We shied away from identifying this as a specif-
ic recommendation because it does imply a fundamental
change of the motivation behind the Clean Air Act to con-
siderlooking -at such matters, but as supplemental informa-
tion, we think decision-makers would be interested in that,
On the item of reducing uncertainty by improving
models and predicting quantities of more certainty, for
example, there was identified that the 95th percentile or
the 98th percentile is a more predictable event than the
highest, second highest. So, again, try to reduce uncer-
tainty by having models address .something which they can
do a better job of.
On Item Four, provide estimates of uncertainty
associated with model predictions and the caveats I men-
tioned earlier about the real difficulty of coming up with
numbers at the present time this docs imply more model
validation studies, more comparisons of models with data.
There Is a separate question of how the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
T. ' 35
infornfa'tiori^would' be v'used^Byvde~cisi'ori-mafce*rsT" We made' one
suggestion which was in terms of looking for a variance,
to compare a variance with, if you will, if you think of
a model as predicting a mean value of an expected rare
event and if one were to provide air bars about that value,
a measure is obviously the standard and another measure of
the air bars would be to look at the air bars or uncertain-
ty factors of safety, if you will, associated with the
standard that you are looking at.
This would be and you could scale this to be
-anything-you-would like, one-tenth of the factor of
safety or whatever. But it is obviously decision-
makers need to have measures to compare the uncertainty
with. . .;
And Five, encourage decision-makers to deal with
uncertainties in this type of argument I have been making.
Finally, our group looked at the process of
changing models. How do you incorporate changes of models
modeling methods to the decision-making process and
we identified two things I have summarized here; that EPA
should focus consistency needs on the process for selec-
tion of models rather than the models themselves. This
came up in the context of the Oil Shale permitting prob-
lem that basically what was needed was a protocol and
agreement of how-to proceed, given the complexity and the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
, V
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
V
18
19
20
21
22
23
24
25
known deficiencies of. existing -models in that .circumstance,
Secondly, that a review board, if you will,
which should be established within EPA staff, would review
new models and such a review would have an overview func-
tion an overview function would be provided for such
review by a standing review committee of both government
and non-government representatives.
erations.
I think that completes my summary of our delib-
I guess I will introduce Mike Williams, an
analyst in my group. I should mention that Doctor
Williams and Doctor Wise were both members of the Group
One Panel.
STATEMENT OF DOCTOR MICHAEL D. WILLIAMS
DR. WILLIAMS: First of all, I should clarify,
I am not representing Los Alamos National Laboratory for
the purpose of this conference. The opinions I will give
you are strictly my own. They are probably colored more
with my past experience with environmental groups than
they are with current activities with the Laboratory.
I have a number of concerns about the workshop.
I think we had some useful ideas, but there were some
potential dangers that I would like to go through a little
hit in discussing some of our recommendations.
On the broader set of material in the workshop
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS-
1330 VERMONT AVENUE. NW
(70?) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18 .
19
.20
21
22
23
24
25
37
"summary,- there were'a few other issues that I1 felt worth -
commenting on and I will add those in, too. For instance,
in Section III-2, it talks about principal issues emanat-
ing from the current regulatory use in models and talks
about how management decisions are made.
In many instances, air quality management decir
sions are based on a single number; however, in the case
of the Class I PSD considerations, the act is very expli-
cit that a decision on increment consumption is only one
of a part of a larger decision to permit facility siting.
This provision in the Iax7, to date, seems to have been
ignored, although it appears to be consistent with many of
the improvements in air quality management decisions which
critics have been advocating.
Perhaps more attention should be given di-
rected to implementing the air quality related values
test which are part of the act and less on altering the ___
---*£;
increments or the mod&ling approaches.
Second, I have not found that most environmen-
talists believe that all uncertainty should be given to
protecting the environment. In fact, I have observed en-
vironmentalists generally want to use the best unbiased
estimate for decision-making. In some areas where models
such as CRSTER have indicated under predictions and flat \
terrain, particularly the 24-hour average, for instance,
NEAL R- GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
i.
18
19
20
21
22
23
24
25
38
environmentalists have "advocated making .adjustments .in. mea-
sured results to reflect the apparent bias of the models
predictions, but the emphasis is usually on a good, un-
biased estimate.
The suggestion that in marginal situations all
national interests, not just environmental protection,
should be considered in the final decision has some diffi-
culties. First of all, who makes and based on what pro-
cess a. decision that a project is in the national inter-
est? Would the same test be applied in marginal compli-
ance, that is, the facility would appear to meet the incre'
ments, as well as in marginal non-compliance?
Currently, facility siting is examined in a few
limited contexts and there is no determination of the
national interest. Suggestion that the decision be based
on the national interest is a suggestion that we make a
major overhaul...of our decision-making apparatus with re-
spect to facility siting. It 'would seem to unavoidably
make the decision-making more cumbersome with a great deal
more government intervention. In many cases, I doubt that
groups such as major mining interests would be willing to
open their books to permit a full exploration of the
economics of~a particular siting decision.
Getting to the sections that relate more to some
of our work, Section 4.2, as long as the screening models
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
39
are -t-ruly- conservativeV'th^y^-should be* useful"-tod-Is" which-'"1
do not create undue confusion. These models merely allow
us to make inexpensive decisions now and deal with the
more difficult problems later.
As long as all plants relative to this decision
are considered with the appropriate model, when it is time
to make the difficult decision, no confusion should arise.
Also, in Section 4.2 I understand that some
regulatory agencies tend to consider best available con-
trol technology as new source.performance standards plus
-what-is.-needed to make the increments. I do not believe
this is the intent of the Clean Air Act. It seems to me
that that was intended to be independent of the dispersion
modeling. If this is the case, the screening model used
should have no impact on best available control technology
decisions.
Specifically, with respect to the Work Group One
report, Section 4.4 dealing with incorporating uncertain-
ties, I felt a lot of the material in there was quite use-
ful, but I am very concerned about a couple of aspects of
*
it.
One of my major concerns is with the suggestion
that we should change the short-term increment from a sec-
ond highest to a 95th percentile or to the average of the.
ten highest. I believe there are several things wrong
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
40
with "such" ari approach/ including, one, change will move
us further from the actual basis for the increments or .7
standards; two, the change will even will be even less
meaningful to the lay audience; three> the modeling will
probably be less readily tested against measurements; .
and, four, such a change will benefit isolated major
sources relative to complexes of sources.
In the case of ambient standards, the basis is
generally -- generally affects data which shows that above
i
certain levels of concentration, damages have to occur.
The use of the highest, second highest is already a step
away from the primary data. A change to the average of
the highest, of the ten highest or to the 95th percentile
r
would be an even greater change from the primary data.
In the context of ambient standards, I doubt if
there is any justification for such a change. In the
case of the Class II PSD increments, there is still a re-
lationship between levels and damages.
j I know several of you are thinking that secondr
ary standard is set up set at the lowest levels at
which damage can occur. However, if you examine the re-
vised SO- criteria document and look at the chapter on
Vegetation damage, you will find several cases in which
SO-, in combination with ozone or nitrogenoxide, damaged \
vegetation in three to four hour exposures at about 0.10
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
10
11
.12
13
14
15
16
1.7
IS
19
20
21
22
23
24
25
uor ten part per million of S0_ to two-tenths of a part per
million of SO..
2
As our principal reference suggested, damage can
occur at levels comparable to the Class II increments.
In the instance of the Class I increments, the
function in the increment is, again, merely to define who
carries the burden of proof. In view of this situation,
it seems that there is no reason to change the Class I
increment without changing the others. In other words,
you generally like them to bs consistent. .Class I already
has an appropriate decision-making mechanism built into
it anyway.
.It is currently difficult to explain the meaning
and the rationale for the highest, second highest basis.
It becomes even more difficult, that is to a lay audience,
it becomes even more difficult to explain the 95th
percentile or the highest average of the ten highest.
I believe the modelers should eliminate as many
barriers to public understanding as possible and.the pro-
posed change would instead, add yet another.
Using either the 95th percentile or the highest
ten will mean that we are predicting a lower concentration
which is readily compounded with instrument threshold or
poorly characterized background concentrations. I believe
that accurate measurements are the highest in the 95th
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
.- " -. .'. -_~>^ - - - 42 ' ' ;
of the ten highest in the 95th percentile will be more
difficult; consequently, there will be less accurate test-
ing of the predictions.
Finally, change to the 95th percentile or the
ten highest will change the current set of equities. In
the KCAQ 4-corners study we found that the ratio betx^een
the highest and second highest betxveen the highest,
second highest and the 95th percentile varied markedly
from one site and source configuration to another. Such
a change will benefit the isolated sources at the expenses
of complexes resources or sources with higher wind fre-^
quencies in critical directions.
That is, basically we are changing the rules
. r
somewhat if we change this kind of increment. It is impos-
sible to define a current an increment that is exactly
equivalent to the highest, second highest in terms of the
95th percentile or the average of the ten highest. "^^
In addition, we will have intended to make the
short-term restrictions more nearly redundant with the
annual restrictions and, thus, lost some of the control
over the entire frequency distribution.
It is not evident to me that the changes sugges-
ted here will produce greater accuracy. In NCAQ 4-corner
study, we found that the center line concentrations
and these were based on measured data were relatively
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 . WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
-7
8
9
10 .
11
12
13
14
15
16
17
18
19
20
21
22
23
24
r .-- 4.3
'insensitive to changes~in meteorological changes. You had
about with about the the restriction that we are
talking about, the same stability classes. Uithin that
range, you could change the wind direction and wind speed
very significantly with little change in the concentra-
tions.
While, of course, the concentrations were very
sensitive to the wind direction for any particular monitor
Changes well, the center line would not be. Changes
discussed here will move us away from the center line con-
centration, which appears to be relatively stable and
toward the off center line conditions, which are not
stable.
r
Another suggestion in Work Group One with which
I disagree is the use of exposure or dosage estimates.
Use of dosage estimates implicitly alters the right of
the individual to clean air. It assumes that people in
large communities have a greater right to clean air than
people in small communities.
This is a major change in the concept of the
Clean Air Act and I think it should not be made under the
guise of a technical change.
The use of dosage estimates also implies that
effects are linear with those seeking independent concen-
trations. In the case of NO , where concentrations arc
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
S
9
10
U
12
13
14
15
16
17 .
18
19
20
21
22
23
24
25
44
*
on the order of a part per million, this-is not the case.
I expect that other contaminants may have a similar ef-
fect.
It seems to me that major changes are being sug-
gested here which alter the equities between sources and
the rights of the individuals under the guise of a techni-
cal improvement. I do not believe that such changes
necessarily lead to significant verifiable improvements in
model prediction. I also do not believe that the basic
changes we are .discussing here should be the sole purview
of the air dispersion modeling community.
Many of the other suggestions we went through I
think are very.helpful and would be useful for decision-
makers and represent contributions to the air quality
management picture.
MR. BURTON: Jim Salvaggio.
STATEMENT OF JAMES SALVAGGIO
MR. SALVAGGIO: I would like to talk upon part
of a subject that we discussed at the conference. It came
up a number of times today already, and that is the use of
uncertainty once you get into the decision-making process.
Since the early 1970's when the Clean Air Act
was adopted, the entire decision-making process in air
quality-related matters has been dominated by air quality
modeling. Discussions on air quality modeling, in turn,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS '
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
45
has centered on the technical factors involved in model
selection and application. Such issues as worst case
scenario, terrain adjustment, mixing height and fumigation
analysis have been at the center of the debate. In the
past, these issues were often debated in the decision-
making process as if the underlying model uncertainty
could be resolved.
This situation occurs because the air quality
management process, as embodied in the Clean Air Act, is
based on modeling. It requires that models be used to
precisely predict relationships batween emissions and air
quality. It implies that if the existing models are un-
able to provide these precise relationships, then improved
r
models will have to be developed.
This philosophy has resulted in modeling being
the final determinant in many decision-making exercises.
For this reason, air quality management has tended to con-
centrate on the technical issues related to modeling.
Likewise, the decision-making process centers on academic
exercises in diffusion modeling. The perception of the
decision-maker and interested parties is narrowed. Bounds
are placed on the range of problems and solutions. These
bounds are intentional. Their purpose is.to force the
decision-maker to focus on meeting the goals of the Clean
Air Act rather than wandering too far toward judging other
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
_ 1.
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
societal issues such as employment/ cost,and energy con- -
sumption.
This focus on air quality modeling has led to
controversy. Over the years the air qua!5.ty management
community has come to recognize that modeling uncertainty
cannot be eliminated or reduced to insignificance. This
has created problems for both the modeler and the decision-
maker. It has frustrated achieving the Clean Air Act goals
by causing delays and no decisions. It has also confused
and frustrated attempts to efficiently administer the
Clean Air Act.
Recognition of modeling uncertainty in the deci-
sion-making process will deemphasize the importance of air
\
quality modeling. It will no longer be the dominant fac-
tor in the decision-making process.
This- is a first important step in air quality
management. I am not saying that formal recognition of
uncertainty in air. quality modeling results in controver-
sies and -improve the decision-making process; on the con-
trary, it will merely shift some of the controversy from
the technical aspects of modeling to societal issues.
The decision-making process will have to balance
these societal issues with modeling uncertainty. It will
have to more formally recognize and discuss these issues
in the public forum. At times, it will be forced to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
resolve conflicting societal needs. If the regulatory
decision-making process is to be improved, this shift
will have to be recognized. Policy, managerial and organi-
zational changes, will have to be made which recognize
modeling uncertainty.
There are a number of critical areas in which
these policy managerial and organizational changes will
be necessary. One is the requirement for demonstrations
of attainment of the National ambrent air quality standard:
and PSD increments.
In general, the current Clean Air Act and EPA
requirements fail to formally recognize that attainment
of air quality standards or PSD increments cannot be demon1
strated with absolute certainty. In fact, the permitting
of a new plant for promulgation of a control strategy
attainment of the National standards requires such an ab-
solute demonstration.
In general, these demonstrations are based on
past meteorological conditions and anticipated pollution
emission characteristics. In seme cases, the demonstra-
tions are further complicated by complex terrain or mete-
orological situations that are beyond the state of the art
of air. quality modeling.
Future events and situations beyond the state
of the art of modeling cannot be predicted with certainty.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
48
Nevertheless, the decision-maker must make an absolute -
decision. For example, he has to approve the construction
of a new plant or not approve it. He has to implement a
control strategy or not implement it.
Currently, decision-making process comes to a
halt until the modeler produces'a document that has an.
illusion of certainty. Formal incorporation of uncertaint;
in modeling results will force the decision-maker to make
decisions without this illusion of certainty. He will
have to formally balance the uncertainty inherent in the
air quality demonstration with issues such as the strin-
gency of emission controls, employment, energy and cost.
The current decision-making process is not
formally permitted to balance these issues. If the role
of air quality modeling in the decision-making process is
to ba improved, then the probability or statistical nature
of air quality management must be effectively incorporated
to the decision-making process and the decision-making
process permitted to formally consider societal factors.
A second closely related problem is open and
formal recognition that the decision-making process must
proceed in spite of the limitations imposed by state of
the art of air quality modeling. It should not surprise
anyone that the lack of universally accepted source recep-
tor relationships in many air quality problems have been
NEAl R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 V/ASHINGTON, D.C. 20005
-------
.1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
i.
18
19
20
21
22
23
24
25
- --.- ' 4 9
used as an excuse ^to delay_incurring^costs associated With1
.
their pollution control.
Considerable latitude exists in choosing the
assumptions and input parameters for models. The incorpor-
ation of uncertainty and the modeling results will add
another technical element to the list which experts de-
bate and lawyers litigate. Lack of scientific certainty
on source receptor relationships and the probability of.
violations should not be permitted to delay resolving en-
vironmental problems. Modeling results will have to be
recognized as a tool to aid the decision-maker; not the
final determinant. The decision-making process, will have
to be structured to focus on environmental problems and
societal issues.
The technical aspects of modeling will have to
take a secondary role. The final decision will have to
bs based on a balancing of issues related to public health
and welfare, employment, cost and energy.
The third area of concern is equity and standard
ization. The incorporation of modeling uncertainty in the
decision-making process will increase the need for equity
and standardization among EPA regions and among states.
Even if emissionu are considered to only affect immediate
vicinities of the emitting source, the issue of national.
equity and standradization is still important.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NV/
(202J 234-4433 WASHINGTON, D.C. 20005
-------
1
^ 2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
50
The PSD, and new source review programs, with
their reliance on air quality modeling, affect the compe-
tition for new plants and economic growth. It can have
major effects on the nature and location of economic
growth.
Emissions, however, do not only affect air qual-
ity in the immediate vicinity of the source.? emissions can
N
be transported across state boundaries and over long dis-
tances .
Consumption of air resources and degredation of
the environment in down-wind areas can ©ccur. How meaning-
fully can the decision-maker in one state balance socie-
tal issues-and modeling uncertainty as it affects a down-
wind state? How often will a decision-maker in the. emit-
ting state unilaterally impose additional cost and burdens
on his own state for benefits that will be realised in
another state? Although equity and standardization prohrsg?
lems already exist, they will be even more difficult to
daal with when uncertainty is formally incorporated into
modeling results.
Without national equity and standardization, the
decision-making process will regress to the lowest common
denominator. Further, confusion and controversy will re-
sult. Degradation of the decision-making process will
occur, rather than desired improvements-.
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
'I
*
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.., ......... . ............ ..... si
Incorporation of modeling uncertainty must be-
accompanied by equity and standardization improvements in
the decision-making process.
In closing, I want to reemphasize that formal
incorporation of uncertainty in a modeling results will
not eliminate the controversy surrounding the use of air
quality models j it will merely shift the center of the
x
controversy. This could, in fact, result in degradation
of the decision-making process rather than the desired im-
provements .
If this degradation is to be prevented in policy,
institutional and organizational elements of the decision-
making process, it must also be changed to accommodate the
f
shift in emphasis. These changes should take place at the
same time that modeling uncertainty is incorporated in the
process.
Thank you. " \
DR. BURTON: Steve Wise, from Mobil.
STATEMENT OF STEVEN WISE
DR. WISE: As a member .of Work Group One, as
ascribed by Bruce Egan, I participated in the discussions
surrounding the Oil Shale issue. However, many of the
indings and discussions surrounded issues that were
f
equally applicable to most industrial applications models
For the rv.ost part, we found that the summary
MEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
report that SAI has put together did reflect an accurate
representation of the deliberations and I am happy to say
that we in industry have generally agreed with most of
them.
Industrial use of models surrounds not only the
regulatory permitting process, but it also surrounds the
problem of design of new facilities. Therefore, not only
do the model outputs have to satisfy the regulator, they
have to provide some certainty to the builder or the de-
signer of new facilities that what he is designing is
going to. perform as it was designed to do.
This means that models have to we have to have
some confidence in model predictions; not just that they
are going to predict second high results at some point in
time and space, but we do have to know whare in time and
where in space that these predictions will be accurate.
For instance, the first slide indicates what the
regulator would call a perfect model prediction. This
represents a model study of the an industrial data
base as being increasingly used by model developers that
| was put together at tha Dow Midland Complex. It represent:
two power sources indicated by the wast and the south,
inside the plant boundaries, and nine monitors surrounding
the plant boundaries* '
There is also at the nuclear site, indicated at
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
53
- ....", i
the south boundary, an on^site MET Station. The diversity
of the data that is available within this source has been
described in several places. API was interested in lookinc
4 at two model comparisons that would, in effect might be
5 used in an industrial complex approximating a refinery.
6 This is probably the most complete data base
7 that exists anywhere of this nature.
8 We looked, in this case,'at the RAM Model which
9 was an approved model at the time the study started and the
10 1C Model which was on the scene, and I guess is still on
11 the scene,
12 The "off's" and the ."on's" refer to the method
13 by which atmospheric stability was determined* The "off's
14 refer to off-site stability determined in an airport ap-
15 proximately 14 miles to the south and east of the site;
16 the "on's" refer to stability being calculated through the
17 older method, using 10 and 16 year wind shear available
18 on the site.
*
19 The RAM Model used the urban coefficient and the
20 RAM-R Model used the rural coefficient and the ICS Model
.
21 i was run in the rural mode.
22
23
24
25
As you can see, only the highest prediction was
missed by the ICS Model, although the RAM Model managed
to include it. Both models used actual hourly emissions.
This is a "little bit different from the regulatory version
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS -
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
2-1
25
-^1- 54'
"of ISC because it really only uses average emissions.
Next, this vu-graph indicates for SO,,, of
j course, the three-hour and 24-hour values are what was im-
portant and even there, the ISC is doing a better job.
Here is of 24-hovir was actually hitting or exceeding
the predictions or exceeding the actual emissions.
This is what we would call the perfection, in
model application in the regulatory sense. However, from
an industrial point of view, we need to know where around
that site we might have problems should we want to do
further development- or should another developer come
within range of our site,
The next slide shows the problems that we can,
therefore, get into. This looks at only the bullock
monitor at the south and west corner of the site. Here
you notice that the "X's" which are lost in these legends
! at the bottom right-hand side of the slide are for the
most part over predicted by all three versions of the
model, although the RAM Model comes pretty close for the
highest two predictions.
If we look at the next slide, hoxvever, on the
24-hour and three-hour basis, we are in big trouble be-
cause net only are we out a prediction with the ISC Model,
we are -out greater than quite a ways, shall I say.
As we move around the plant site, however, in
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS *
J330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
'1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
the next slide, you-see-we-begin'to do;pretty well again.'
As we go west east of the plant, again, the models
seem to be predicting fairly well. The ISC Model and the
RAM Model, the Ram-Rural Model is falling in line again.
The next slide shows the 24-hour and three-hour averages
again, we would call this a good model representation
of the high values observed at those monitors.
The final slide, however, shows that we can get
in trouble, particularly no, that is I think we are
out of order. Keep going.
The final slide shows what can happen, however,
under certain conditions. Here the ISC Model is under-
predicting. This, we do not want either, particularly in
this instance, since .the Austin Monitor is in the city of
Midland and the monitor directly north of it also was
doing the .same-kind of thing.
We also have a problem, although the rural ver-
sions of "RAM seem to be doing okay. It is the urban ver-
sion I mean the rural version excuse me. The urban
versions of RAM seem to be doing okay. It is the ISC
Model that we are principally interested in in this in-
stance because it contains provisions specifically aimed
at a site like this which is the building down-wash out-
rythm, :'
Here, we have determined that it is not working
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
- 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
particularly v7Sll~7:inL-Wi:-s-s
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.-41'- 57
Policymakers a'nd de^isTdn-maTCefs are*" of ten "tempted "to over-
look this uncertainty, but that does not solve the dilemma;
it just disguises it.
UARG believes that air quality management pro-
grams must be designed and implemented in a manner that
uses models as the tools they are but does not demand the
impossible of them. Models should be used to help investi'
gate reality; they should not be used in a sqlipsistic man-
ner to create reality.
Models should be used to aid in the air quality
decision-making process? they should not become the de-
cision-maker by default, We believe that realism and in-
formed judgment, not computer printouts, should guide air
qua.lity "management decisions.
Looking first at the design of air quality man-
agement programs, the basic tenet of the air quality man-
agement-program is that ambient air quality, not merely
emissions, should be regulated -in order to protect the
public health, welfare or aesthetic values.
The Clean Air Act requires EPA to protect these
values with some limited margins of safety; it does not
authorize the Agency to reduce ambient air pollution belovr
these levels v/ithout legal justification,.
This implies, however, that any effective air .
quality management program must strive to represent the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
-"'- ^ ' 58
*-- "-" ~* '"' ." '-, ^^-^i-'jATxi^jp'-l. "* - - -, - - -..-v ,. ,
1 real world ambient pollution concentrations as accurately
2 as pos.sible.
3 The PSD Program is a classic example of a progran
4 which misuses models by demanding too much from them and
5 by making them too important. The PSD divorces models
6 from reality and creates an artificial system defined only
7 by models.
8 The program ignores the emissions of certain
9 sources, while modifying emissions from others. These3
10 features of the PSD frequently make real world comparisons
11 between modeled and monitored results impossible. More-
12
13
14
15
16
17
18
19
'20
21
22
23
2-1
25
over,'because there is no link bstween the calculated
concentrations and actual concentrations, any real world
T" .,-.,.
benefits from the program are largely fortuitous.
The PSD Program, with its tertiary standard
increment system further shuns reality by requiring ds-
cision-makers to account for ambient air quality concen-^cp
trations that are too small or too ill-defined to.be moni-
tored.
The Class I increments are below the threshold
of the most sophisticated monitoring equipment and visi-
bility cannot be monitored in a psychophysiologically
meaningful manner. Finally, the PSD increments themselves
have no firm epidemiological or welfare effects basis.
The utility industry has long opposed tho curren
NEAL R. GROSS - -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1'
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
;59
PSD Program for a-variety of reasons. And, while this .is
not an appropriate forum to debate the legislative alter-
natives to the present PSD Program, it is, I believe, the
proper forum in which to point out, as I just have, those
abuses of models which make it an unworkable program.
UARG believes that the agency and Congress shoule
recognize these defects and should not extend them into
new programs when the Clean Air Act amendments are dis-
cussed. Specifically, Congress should be cautious in the
current debate over long-range transport to avoid divorc-
ing -models "from the real world air quality conditions
which are the basis of the legislative concern.
Models should be asked to provide information
which they can provide; they should not be used to create
new, artificial parameters such as the PSD increment sys-
tem.
Now, I would like to turn to model use in the_
decision-making process under otir current regulatory pro-
grams. Decision-makers use models in different ways; too
often they are treated as predictions of immutable facts.
We believe that realism and rationality should
guide the decision-making process within our current air
quality management programs.
Decision-makers who administer our current pro--
grams must consider the inherent limitations of modeling
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. B.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Hi
17
18
19
20
21
22
23
2-1
25
60'
results together with tha social value of the goals being
protected and the potential burdens imposed by the deci-
sion.
Not to do so would be irrational and to claim
that it is not done would be naive. UARG believes that the
following examples illustrate a rational approach to using
modeling results in the decision-making process.
Thinking first about attainment area designa-
tions. Non-attainment designations should not be based on
| modeling information alone. If computer models predict
violations of a standard in a particular region, but no
such violations have been recorded by monitoring, the
agency should designate the region as unclassifiable and
f
should undertake monitoring in order to determine whether
the violations actually exist.
This area should be redesignated based on the
monitoring information according to a specified time
table.
Turning next to secondary standards and the
"tertiary" standards or PSD increments, because the PSD
increments and secondary standards protect values other
.than the public health, modeling calculations should be
interpreted in a flexible manner.
For example, if computer modeling shows that a'
PSD increment or secondary standard might marginally be
NEAL R. GROSS
COURT REPORTERS AND'TRANSCRIBERS
1330 VERMONT AVENUE. NW
(20?) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1?
18
19
20
21
22
23
24
25
-^1 61
i ^ i
"~ .-.-_.' . .v,!ri»7 v <-
exceeded, provisional permit approval should be granted..-
The approval should include an operational monitoring re-
quirement designed to determine, within some specified
time, whether the increment or standard is actually being
exceeded.
The provisional approval could also include a
condition specifying a mutually agreeable remedy designed
to correct any subsequently monitored violati6ns. This
approach would help to prevent the imposition of unneces-
sary restrictions at the outset and it would give all
j parties notice at the outset of what the remedies might
ultimately be. .
Turning to primary standards. In the case of
r
primary standards, designed to protect human health, the
best estimate of the future conditions should be used as
a guide in the decision. The agency should consider moni-
toring information, robust statistical measures, expected
exposures and other relevant information in arriving at
this best estimate.
We believe that such flexible approaches to the
interpretation of modeling results would introduce more
realism into the. decision-making process and would allow
decision-makers to exercise their informed judgment.
There is a tendency in air quality decision-making to use-
exclusively tho short and long-term not to be exceeded
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVCNUE. NW
(207) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
'62 ,;
basis of the standard "in interpreting the air quality '
modeling results.
This is an unduly narrow approach. The agency
has no obligation to use the highest-second-high value
calculated by a computer model merely because the standard
is written in that fashion. Rather, the agency has a duty
to conform the computer results upon which a decision is
based, as closely as possible with the real conditions
which would actually occur.
In order to do this, the agency should strive
to use whatever information is available from the modeling
calculations and from whatever other sources, such as
monitoring or physical modeling, to estimate the most
probable results.
We believe that the agency has an obligation to
use probabilistic methods in estimating ambient concentra-
tions, to use more robust statistics, and to employ any-
other relevant .information in interpreting modeling re-
sults.
V7e also believe that the decision-maker should
weigh the values being protected in the particular environ
mental program and the burdens being imposed on the permit
applicant and on society by the decision.
It would be irrational not to consider the tre-'
mendous costs involved in many air quality management
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
decisions, given the reversible nature of the decisions
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
and the great uncertainty in most models. Should I say
all"?
Likewise, it would be naive not to consider the
value being protected in the decision. As the motivating
value decreases in importance and as the burden increases
in severity, greater uncertainty in a model's accuracy
counsels for a stronger presumption that the permit should
... D
be granted. Operational monitoring would then be avail-
able to test the correctness of the decision in the areas
of concern.
This approach provies the needed balance which
emphasises realism and rational judgment.
f
This Conference comes at a propitious time.
Congress is considering amendments to the Clean Air Act.
»
We have had more than ten years experience with the use
of air quality models. And we have seen them elevated to""
the decision-making role. We are becoming more acutely
aware of their predictive limitations and we are at the
point where we should redefine their use in the design
and implementation of air quality management programs
precisely to harmonize their uses with their capabilities.
I would just like to conclude with one observa-
tion from this morning's session, which is that the infor-
mation concerning the ctccuracy of models presented today
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS .
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
*Lr~T -- - 64,.
.--- '- "*
is the kind of information which should be discussed in an
agency-sponsored proceeding and. it should be the kind of
thing that all parties have an opportunity to explore and
discuss before any results are accepted as being demon-
strated levels of accuracy for existing models.
Thank you.
DR. BURTON: The last speaker is Tom Helms.
STATEMENT OF G. THOMAS HELMS
MR. HELMS: A couple of observations I would
make real quickly. As I sit up here/ I had had I think a
large order of french fries, double hamburger, chocolate
milkshake and a piece of apple pie. When the lights went
out, I went-out awhile ago; so I thought I would do a sur-
r
vey and I counted 18 people out there asleep, plus or
minus five people for uncertainty. So, stand up for 30
seconds and stretch.
Okay, let us move on with it. I hated to talk
to 23 people or 13 people asleep out there, so maybe I
have gained five or six of you;
What I would like to do very quickly is talk
about the SIP process. I was in Work Group Two. We dealt
with tha SIP process, the state implementation plan, revi-
sion process. Most of the discussion, at least that I par
1 ticipated in, at the workshop dealt with point source
!
' models for sulfur dioxide and particulate matter.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
J330 VERMONT AVENUE. NW
(202} 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
t.
17
18
19
20
21
22
23
24
25
We talked about ozone-* but I really will not ,
address any remarks about ozone today.
Some of the comments I will make apply to new .
source review; PSD permitting, as well.
As I went into this workshop, I had a number of
questions in my mind, a number of ideas that I wanted to
explore. Some of these are just when do you model; when
is it necessary to model? When can you take existing data,
air quality data, emission data, and make a decision with-
out modeling? I will not go any further on that today,
but I think there probably is a role place and type to do
that.
How much confidence do you place in the results
f
of models? I was real curious about that. When do I go
to my boss and absolutely put my job on the line, that the
decision should be no, based on this model? When should
other things-be factored into-.a decision, an air quality
management decision besides the results of an air quality
modeling effort? When does the size of a plant, the age
of a plant, the existing fuel supply, when should this be
factored into a decision?
When can you just can you decide in a process
that you are just really arguing for the sake of arguing
and that the real world will not know the difference of '-.
the point that you are arguing about?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, HW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
6'fc
1 Finally-,-whenyou--decide to make a new decision,
2 how can you make this decision, implement new policy, come
3 up with a new program without undoing ten previous deci-
4 sions and disrupting all the ongoing projects?
5 So, I xtfent into the conference with these things
6 in mind. I should say "workshop" with these things in mind
7 There was a lot of discussion. I skimmed this morning over
8 the summary report and pulled out, I guess, the following
9 items that might somewhat address these points. ~D
10 First of all, I am going to assume that we have
11 decided to model. We have decided the correct way to make
12 a regulatory decision in the SIP process is to model.
13 The first thing I would recommend, the partici-
r
14 pants recommended, Bruce, I believe you recommended it
15 earlier on and that is to get together to set up some
16 ! kind of protocol. I am presuming there is going-t© be a
17* : potential for disagreement right now in the modeling ac-r-_ti
18 tivity, so get together early on, set up a protocol-:, figure
19 out which model is most appropriate, set up a process to
2.0 resolve technical disputes because they will happen, as
21 you all know. I have never seen two or three modelers
i
22 get in the same room and come out with the same answer.
23 They will usually compromise, but there is no black and
24 white in the program.
25 Finally, getting together, you - have got to be
NEAL. R. GROSS
COURT REPORTERS AND TRANSCRIBERS
U30 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
^. 67
to come up with a protocol, you have got to get a com-
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
mitment from all parties to accept the results. It does
not do any good to get 90 percent down the way and have
one party, either industry, environmental, regulatory, or
whatever, decide to back out; they do not like the way the
results are going.
I think the second point I would like to make is
that it is very important to provide the decision-makers
with all the information possible, since they put a real
world view on the decision. I heard confidence limits-
mentioned; I heard 95 percentile mentioned; sensitivity
analysis may or may not have been mentioned; all these
things are important. I think a decision-maker has got to
. f
divorce himself from the Sigma Y or Sigma Z and to try to
look at what this decision will mean in the real world.
At the conference at the "workshop", I should
say, there-was a lot of discussion about striking a
*
balance in modeling, a balance between standardization,
rigid ways to model in every case versus flexibility. It
is very difficult to strike that balance. In my present
job, I see a lot of lowest common denominatoring. VThat
do I mean by that? You take a model that is run one way
in one area; it may be run correctly. Let us presume it
is. A case conies along in another direction, another
location and I find people wanting to go use that model.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
rr
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
Maybe it is not the correct-model. It gives a better ans-
wer, either for the environment or against the environ-
ment, but people want to use that decision or that prece-
dent in a second decision. So, I think we have got to
work very hard to strike a balance between standardization
and consistency on one hand and flexibility. Flexibility,
meaning to allow you as modelers to select the most appro-
priate, the best model for a particular situation.
When you do make changes in models, when things
are done differently.. I think the participants at the work-
shop felt like there should be some way to grandfather
past decisions in and not to disrupt things. That is easy
to say and very, very hard to do.
There was some discussion at the workshop support
ing a mechanism to convey changes in models, methodology
and processes to all interested parties. There was some
discussion of a modeling center. That is a good idea. It
is going to be awfully hard to pull off in the future with
everyone facing resource juts. I would be interested in
some feedback on that.
Summing up, I think our challenge as modelers
and as administrators to try to get the best available
modeling information you can, temperate it with monitoring
! results, put it into proper perspective with all other
2.-, !' factors that impact on the decision and come out with a
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(702) 234-4433 WASHINGTON. D.C. 70005
-------
responsible, yet environmentally sound decision.
2 ' I will quit there, Shep.
3
4 agreed that now is the time to implement the 15-minute
5
6
7 I believe, kind of wrap up the panel discussion and then
8
9
10
11
12
13
14
15
16
17
18 '
19
20
21
22
23
24
25
69
MR. TIKVART: Shep and I have more or less
break that is on the agenda, so we will do that. When
you come back at let us say around five-to-four, Shep will,
we will open the floor up for questions and comments and
- D
I would like to see a good interface or good discussion
between you, the audience, and the panel members here.
Oh, one thing. The Department of Agricultures
asks your cooperation in not bringing food or drink into
the auditorium, so please bear with us on that. No food
or drink in the auditorium.
(Whereupon, at 3:36 p.m. a
break was taken.... piscussioi
resumed at 3:54 p.m.)
MR. TIKVART: If you will take your seats, those
of you who are still standing, we will continue with
back to Shep Burton and a summa'ry of the panel discussion
and then perhaps some additional discussion among the pane
members and then to you. We would like to see an exchange
.between the audience and the panel members to bring out
the ideas that h^ve been initiated here and to have to
perhaps foster r.orae new ideas.
DR. l-URTON: I had not actually heard what each
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS.
1330 VERMONT AVENUE. NW
'(207) 734-4433 WASHINGTON. D.C. 70005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
70
of the panelists was going to say until you did and I was
trying to write down a one-sentence summary and then it
got to two sentences, three sentences and then I got about
two panelists behind. So, I do not really have sort of
an overall summary, but a few questions did come to mind
that I thought I would ask the panelists and perhaps some
of the panelists have questions that they would like to
ask each other since they had not heard what each other
was going to say prior to today either.
The one comment or one thought that I had in
reading over the draft reports that came in from the vari-
ous work groups was the extent to which all of the recom-
mendations appeared to be a tremendous overlay of addi-
r
tional administrative burden if you will accept that
phrase for the moment and to what extent, although
offered in the with the best of intent, I am sure,
would these recommendations cause further delays, somothl'nc
that people are already concerned about.
Tom commented Tom "Helms, in his closing re-
marks, commented that in one respect the modeling center
concept really presented EPA with some considerable re-
source burdens or additional burdens in time when the re-
sources are being constrained. And there are probably a .
number -of. others which would do that.
So, there is one panel member'who has had
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBES*
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
71
experience, considerable experience/ outside the modeling
area and who has dealt with these kinds of process ques-
tions and that is Steve Connolly. I would like to ask
him the extent to which he sees the recommendations that
are raised that have been suggested would cause additional
administrative burdens and delays.
Steve?
MR. CONNOLLY: I think if one is talking about
systems of developing an advanced protocol for modeling
processes, choice of models, use of choice of data and
input assumptions, use of model outputs or if one is talk-
ing about cooperative processes or additional periods, it
has to -be-conceded that there is the possibility for
f
another layer of bureaucracy, another more opportunitie;
for delay.
I would say that this is simply a factor of
there is the same number of people to go through more
steps, more stages. It is also a.possibility, if you as-
sume that either there is a tremendous difference in
perception on, let us say, amongst the participants in
a modeling or let us say, a permit proposal and permit
review; There is a tremendous number of differences in
the perception of the issues or the values or perhaps of
the motives of the others or if there is bad faith on one-.
side or the other; that is, that for one reason or another
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
another opportunity for delay.
The way I would respond to that is to say that
the current adversary process can and is used to delay.
That is perhaps one of its strongest features: one can
never tell when it is being used for delay or when it is
... ;>
being used just as an adversary process.
The cooperative process, I think, has the advan-
tage early on of first, assuming that all parties are oper
72
one party to the issue would like to see the process
stopped or delayed. Then this presents an opportunity
it could be argued that a cooperative process presents
ating in good faith, and assuming that the other side is
operating in good faith. It has the opportunity in those
situations to cut through the trash quickly and get to the
I issues that are of importance to people, those that are
I likely to come up somewhere down the line, get then
;
fleshed out early, set the other issues aside and move on7
The cooperative process also has the advantage
of making it more difficult for one side or the other to
| stonewall it without becoming very clear early on what it
is they are doing. So, that if you do use a cooperative
.process, you are probably going to know what is important
to both sides and whether or not they are serious about .
! dealing with these issues; whether they want to see an
i
outcome rather than a stalled process. Earlier than if
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE, NW .
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
73-
you are in an adversary process.
The other thing about a cooperative process, in
terms of whether it will actually be another layer of
bureaucracy and delay is that the key issues in any, let
us say a modeling dispute, or let us back off and not call
it a let us call it, in any permitting process in which
there is the potential for disagreement and controversy
over the proposed facility or its site, each of the issues
that is going to arise in the modeling process is going to
have to be addressed at some point. The question is: who
addresses them and how and in what setting and when? Do
they address them up front at the beginning; do they ad-
dress them openly and together and attempt to decide which
of the key issues for each side and why or do they address
them separately in their own operations and, if they hap-
pen to come up at some point in the future, fine. Then
we will fight about it. .
If-it is true that the key issues are going to
come up at some point, then it seems to rr.e that it is
smarter to bring them up, get them hashed out, if it is
possible, before there is a major commitment of resources
of energy, and most importantly, of egos. And after there
has been a lot of modeling performed and the controversy
is in full bloom.
It seems to ma that all the cooperative process
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NY/
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
24
25
'* 74
73 --.'
does then is acknowledge that which is going to happen
inevitably, try to bring it out in the open, at an early
date; try to identify areas of agreement and disagreement;
try to focus on those areas of disagreement and come to an
agreement, and if not, very early on identify for the ulti-
mate decision-maker what he or she has to have; what kind
of information he or she has to have and what the ramifica-
tions, regulatory or litigatory what the ramifications
of a decision are. :>
In other words, it is just a way of saying let
us take this whole process if modeling is going to con-
-»
tinue 'to be used, if it is going to continue to be an im-
portant factor in the decision process, and let us take it
r
out of the-closet and let us take it out of the closet at
the beginning, rather than bring it out of the closet when
everybody has got enormous amounts of energy and time and
ego invested in a particular outcome. -^
I do not think it has to be delay. In fact, I
think it can expedite in most cases, the process; not de-
lay it.
DR. BURTON: V7ould any other panelist care to
22 ' "
23
21
comment?
One other question that you did not touch on,
but that could also entail delays and introduce a lot of '
inertia in the system is the* utilization of advisory
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
75
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"groups or external peer review groups and so on to approve
models to recommend changes. Some people perceive that
that could also add additional delays.
Do you have any comments on that?
MR. CONNOLLY: My only response is if I were
trying for a permit/ I sure would not want to wait for a
peer review to be
DR. BURTON: Well, I do not mean peer review; I
am talking about now introducing changes.
MR. CONNOLLY: Well, I was thinking of it if
I were introducing a new model which I thought was a fair
representation a fair and scientifically credible repre-
sentation of the real world, I would not I would not be
r
overly concerned about a process that was cooperative and
open which merely, at an early date, allowed me the oppor-
tunity to present that model and explain it and see where
people are going to have trouble with it and try to deal
with that, but if that were to degenerate into a process
where, let us say, you plugged it into the idea of a na-
tional modeling center or whatever we- are talking about,
it would have to go through some extensive peer review
there and would could, not be used in a regulatory pro-
cess until such time as that whatever peer review committee
acted. I would be tremendously concerned about the oppor-
tunities for inertia and delay and, therefore, expense.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
18
19
20
21
22
23
24
25
That is one area where you have to be careful.
MR. TIKVART: I would like to ask two questions:
one is how formal should the process of agreeing on models
and data bases be? How formal should that process be, the
protocol? Do we need a formal written document or does
everybody sit around a table and agree? The second ques-
tion is there are going to be occasions when there is dis-
agreement on the modeling approach to be used. Admittedly,
this will result in delays. How do you deal with those
situations? The last question deals with the answer you
jxist-gave-to the former question and that is, you said you
would hate to see delays associated with litigating or
arbitration of a modeling technique; but admittedly, there
r
are going to be cases like that. Is there any way to
minimize the impact on time and resources in resolving
those issues in the process mode?
MR. KONTHIK: I would just like to go back to
Shep's question for a second. Sorry. I think it relates,
I guess, to yours, too, Joe, about the delay point. One
perhaps promising, but also troublesome thing about the
summary of the workshop results was that it appeared that
there were going to be a proliferation of different organ-
izations, you know, of advisory committees, of modeling
centers, of dispute resolvers, etcetera, and I just think
i it is really important that that be a streamline process
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
*,
17
18
19
20
21
22
23
24
25
..--.--.--."...' . .- 77
involving as few nev; organizations as possible, maybe one
preferably one, and that it be adequately funded and
staffed so that you do not get tremendous delay and it
becomes an impediment rather than an actual benefit.
MR. TIKVART: Can you clarify exactly what you
are referring to?
MR. KONTNIK: In the workshop, the summary of
workshop recoinmendations, I think that there were at least
j 3
there were two or maybe three distinct or possibly dis-
tinct groups: one for resolving disputes in the develop-
ment of modeling protocols and another, which would be i;he
modeling center, which x^ould take on a variety of may
or may not take on a variety of different functions; and
the thind being some kind of advisory group empaneled to
look at including new methodologies in the guidelines and
»
so on and so forth. My point was just to say that it
seems like it is valuable probably to have some kind of "^
panel of experts to deal with those sorts of issues and
that that panel should be composed both of agency and non-
agency people. It would bs unfortunate if that group
wound up because of lack of funding, because of lack of
staff, whatever, that it became actually an impediment to
the decision-making and dispute-resolution than actually
facilitating that.
MR. TIKVART: Okay. Then back to Steve. Given
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
78 : ;f
that we have some group to do this, would you comment on
your concern with this as an intent, as an impediment to
the decision-making process, namely, given arbitration of
what modeling techniques or what data base you should use
for a given source application?
MR. CONNOLLY: I think my first assumption would
be that you ought to set up a system so only under the
most dire circumstances is there anything called arbitrcx-
tion. I think you have to set up the system so that it
encourages cooperation and negotiation between the parties
involved in the process; not bringing in outsiders, except
where there can absolutely be no other recourse.
Now, I am sensitive of the fact that if you hold
out some arbitration process and somebody wants to be a
stonewaller, then that is where you are going to get soonei
or later and so I have not thought enough about it to
figure out how to make it very hard to get to arbitration
or how to give people incentives to not to want to get to
arbitration.
I think it is important that those who are ths
participants in a process come to an agreement on what the
assumptions are, what the models of choice are, about what
the data to be used are, and the techniques of modeling to
be used.
So, I tend to think that you cannot have a
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, N\V
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
-13
14
15
16
17
18
19
20
21
22
23
24
79 ;
permitting process that is dependent upon except in the
most extreme circumstances, and perhaps you ought not to
have an arbitration permitting process; perhaps it ought tc
just go to court eventually but you ought not to have
a permitting process dependent upon some arbitration cen-
ter somewhere. You would be better off just flipping a
coin than getting to that situation.
DR. BURTON: Steve, let me interrupt because I
might be guilty of something in the report which I did not
intend to do. I do not recall anyone at the workshop or
in any work group actually proposing that a body of arbi-
trators exist who would be called on demand to resolve dis-
putes. What I do recall being discussed was that within
r
the context of each protocol, if issues are raised for
which the resolution of disputes are perceived to be diffi
cult in the future, that a scheme for resolving those
would ba identified in that protocol, which could ba one
one could be a project arbitrator, and that individual
could be designated in advance and that could be whomever
the group agrees to. It could be a member of one of the
groups who is perceived to be particularly trustworthy or
whatever. But it does not mean that there is on call a
group of people somewhere in the country who would be
brought out and it is possible that tha demand could ex-
cscd the supply, and in so doing, introduce delays.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
8Q :
I just wanted to clarify that. In terms of the
other two groups, I think there were two groups called
for this general advisory group and the or peer re-
view group, which would oversee the introduction of new
modeling concepts and another group serving in the role
of the modeling center, which sort of would dispurserthose
approved modeling
MR. CONNOLLY: My notion of the modeling center
was a technical assistance center.
DR. BURTON: Right.
MR. CONNOLLY: To move the new techniques, the
new models out into the community.
DR. BURTON: And at most, I think most people
r
that were 'present looked at existing groups within EPA.
»
Perhaps with some slight modification, perhaps with hardly
t
any modification, just the reorientation of priorities to
serve the role of the modeling, center or the center of
technical excellence, so that, in fact, there would only
be on group left and that is the peer review group which
would pass judgment on the introduction of new modeling
concepts, so I think that meets with Lew's concern about
not a proliferation of groups.
As you might suspect, one of the reasons for
this hasty ad hoc discussion up here was to try and give
people a tir.is to overcome what may bs information overload
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
and for you to dream up questions or to pose questions that
you have already dreamed up, so I would like to try and
start doing that now, if people do have questions.
Vern?
MR. V7ALKER: My name is Vern Walker. I am with
the law firm of Legal, Haves and Symington here in town anc
we are currently representing a group of 14 utilities in
the interstate pollution matter that is before the EPA.
If it was one thing we- were very good at at
Airlie House, it was coming up with long lists of
j especially of sources of uncertainty that in many cases
|
! were cumulative and attended model results.
I
r We also seemed to bs in agreement that a reason-
able decision-maker would attempt to take into account as
much as possible and appropriately what uncertainty there
was in the model results. I would like to pose for the
panel the following hypothetical: a dscision-maker who
has on the record before him or her modeling results, but
does not have a quantification, a reasoned quantification
of the uncertainty associated with those results. Would
you advise the decision-maker to make any use of those
results and, if so, why,.and in what circxamstances and
with what safeguards?
DR. BURTON: I get to just sort of dodge these
and reflect these over to people and I am sitting here
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(20?) 734-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
2-1
25
82
,V v' ;>
looking at Tom Helms, who does a particularly good job at
those kinds of questions,
MR. HELMS: Let me see if I can first of all,
I am probably going to have trouble recalling your com-
plete question, so I am ramble.. Stop me if I am not hit-
ting some of the points. Can we make decisions I am a
decision-maker; I am sitting there; someone shows me model-
ing results. They do not give me the information, the
uncertainty information associated with thatj can I make
a decision on it? People are making decisions on that in-
formation every day. What I think you need to do is at
least-what I would recommend is to put some real world
thought, into it. Again, go back to let us talk about
r
power plants, for instance. How old is a plant? Where is
it located? How much life is left in the plant? Is it a
mine mouth plant? What options do they have for eontrol?
Are they Washington thair coal now? Are they even burning?
coal or are they burning something else? Do they .have
access to gas? Look at all these types of things. How
much air quality problem is there associated with it? Kow
big is the plant? Take all that into consideration and
see if that helps you with the uncertainty.
I do not think you go out. and you say, yeah, the
mo-del shows real bad air quality problems; we have got to
do something; let us t>ut a scrubber on a 50-year-old plant
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(207) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
- 7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
, 83
You do not make decisions like that.
What we probably cannot do is make some of the
look at that one individual plant with the uncertainty
associated with it and project way down when 300 or 400
miles and talk about that one plant and its air quality
impact. But in the vicinity of the plant, you could
probably make a judgment; you could factor in some of the
real world constraints and I guess I submit people are mak'
ing judgments like that now, around the immediate vicinity
of plants
MR. WALKER: Well, without addressing, you know,
what actually happens, because I think we are all familiar
with that, I was trying to isolate the very important fac-
r
tor that the workshop came back to time and again and that
was the use of uncertainty or the fact of uncertainty.
There is no doubt that in that case, a decision-maker has
to make a decision and that he or she looks to many other
factors.
My question was given the hypothetical that I
posed, is the decision-maker justified in being more con-
fident in his or her decision by the fact that he or she
has these model results? Do the model results, absent
quantification of uncertainty associated with them, pro-
vide any basis for that decision at all?
MR. HELMS: Modeling results will tell you the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
84 :
order of magnitude of the problem from a very bad air
o
quality problem to no problem at all type and there are
3
the modeling results. I do not think I think sometimes
5
6
7
8
9 a result of what your model tells you. You do not have to
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
probably ranges of control that could be associated with
we get hung up too much on whether an emission limit
should be 3.51 pounds per million BTU or 3.0. There pro-
bably is not that much difference, frankly.
I think you could look for ranges of control as
be so absolute.
MR. WALKER: So you are entitled to rely on ths
model, even though you cannot quantify the uncertainty so
long as you do it on a sliding scale?
f
MR. HELMS: I do not think I said that. I think
I was trying to say use a little common sense with what
you have got. If you do not have the uncertainty asso-
ciated with it, you have to do the best you can.
MR.. WALKER: Thank you.
DR. BURTON: Is there anyone else hold on a
minute, Vern. There might be someone else who would want
to fell that one. Does anyone? Bruce?
DR. EGAN: Yes. I would just comment that I
think the hypothetical may be a little unreal in the
sense that you can get some other information. You have
got sonie predictions from, say, a single model and you
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
85
have got a very specific series of numbers that corr.e out,
but there are other things that you can look at to gain
some confidence about that value. Look at another at
somebody else's assumptions in the model, for example, if
there is a dispute about deposition rates or whatever.
I think you can look at the assumptions within
the model; try to trace that back in terms of a scientific
validity.
There are a number of things that I think you
could examine to gain some confidence about what sort of
*
certainty you would have with the predictions.
DR. BURTON: Another thing that there must be
some information available about whether the model has a
tendency for a bias or not. You can there are ways in
which in the workshop various work groups suggested
j
ways of doing this looking at sensitivity analyses, for
example. So, that you can get a handle on the quantifies.-
| tion and then, even if pardon me? and even i'f there
are if you do not choose to'do this, there are other
: ways that you can get a handle on ths uncertainty through
simulation of the model so that where you might not have
an explicit specification of the uncertainty for your hy-
pothetical plant, you could come up with the range on un-
certanties and whether or not there is the potential for
bias and then look at what the effect of or tha implicatio
MEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 . WASHINGTON. D.C. 20005
-------
of that range of uncertainties would be on a decision.
Once again, bringing to bear a fair amount of common sense
3
4 respond to it.
5 MR. WALKER: But you think that that collateral
6
23
24
25
on the consequences of that uncertainty and how you might
information should be on the record in a particular case?
Southwest Services. Was there any consensus at the Airlie
House meeting that a nev.' model need not necessarily give
the exact same results as a previously approved EPA model?
DR. BURTON: Would you repeat it again? I do
not know if you are
MR. WOOD: Was there any consensus at the Airlie
House meeting that a new model need not necessarily give
the exact same results as a previously approved EPA model?
DR. BURTON: Describing the same phenomena?
MR.- WOOD: Yes.
DR. BURTON: I do not think that question was
addressed, but if no one else it was?
99 DR. EGAN: I think that was implicit. There was
a. lot of discussion about developing new models. I think
people understood that they would produce different num-
bers, in that sense, yes.
NEAL R. GROSS
COURT REPORTERS AND .TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15-
16
17
18
19
20
21
22
23
24
25
.'87 ,
I'know we are not supposed to discuss the guide-
book, but
DR. BURTON: Oh, that is up to Joe.
MR. TIKVART: Okay. Let me ask you a question
now. Let us say because this ties into the process
question. Let us say that there is a standard EPA model
that is suggested-for an analysis and the source has a
model that it would prefer to use and they do not give the
same answer. Considering the process question of let us
sit down and discuss it and establish a protocol, etcetera
etcetera, how do you resolve the problem of having models
that give conflicting or estimates or estimates that are
not within a same reasonable range of values? How do you
r
deal with that?
MR..WOOD: Of course, I would not go to EPA if
I did not feel the range of tha values were reasonable,
and yours were not. So, I guess we would have to have "SOUK
arbitrator.
DR. BURTON: The context of your question then
i is one it is not abstract? In other words, it is not
the introduction of a new model which when applied to some
facility somewhere would produce probably a different re-
sult, but rather one source has a model, EPA has a model,
and they do give different results, and how do we recon-
i cile the differences?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005-
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
88
MR. WOOD: Yes.
DR. BURTON: That particular question was very
much addressed and contained in the recommendation for the
protocol and to get that up you know, we tried to get
that resolved at the beginning and without arbitration.
I mean those there are differences the differences
are real.
MR. TIKVART: I guess my question is given that
that needs to be taken care of as a process, how would you
approach EPA or some other regulatory agency, a state
agency>-with~that issue? And try and resolve it in the
process mode that Steve Connolly was talking about?
MR. WOOD: I have no response for that, sir.
MR. CONNOLLY: One of the things I think that yoi
can do is come to the regulatory agency and say here is the
model we propose to use. As bast as we can determine,
here are the major differences between the way we treat
certain data or certain assumptions in our model and the
way you treat them in yours. Here is why we think our way
is superior. Agency please respond oh the record why you
think your way of doing things is preferable or more rea-
sonable, or why ours is unreasonable. I think that the
more you can get into those processes early and force the
process to be explicit and force one side or the other to
demonstrate conclusively, if that is possible, why one
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
20
21
22
23
24
25
£9
approach is reasonable and one is unreasonable. I do not
think.you have any mechanism other than to come with a
model and shake it in somebody's face and say, hey, our
model is better than yours. Unless you get down to those
very explicit details early on,, and unless the source
assumes some of the responsibility for forcing those is-
sues, then I do not think that the source is in a .position
to come back, which I think it should be, come back to the
agency and say, okay, agency, here is ours. Why is yo\irs
better? You show us why yours is better.
I have not been directly involved in, but have
!
observed processes where neither side has adequately demon-
strated why its approach is preferable.
*»
DR. BURTON: That could be done through theo-
retical or performance evaluations and I think tha work-
shop definitely suggested that when nev; models were being
introduced in the regulatory process that tha new model ...^
bear a the burden of being evaluated and I know, that
not all models being used at this point have borne similar
costs or exercises, but it seems that to introduce new
models, that that is something that has to be done.
The next question?
MR. VAN VLICK: * wantcd tO takG iSSUS with
a statement that you made that you think that modeling
has come a long way in the- pa.
-------
'.*
5V
f
1 .
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
90
me we are at about the same place we were in about --63.
We still have Delta T or Sigma Theta and we still look up
in a cookbook what the dispersion rate is going to be and
there is plenty of evidence around that you cannot even
specify the dispersion rate from measured meteorologial
variables. We are talking here about we are going to find
new models; we are going to come up and develop these new
models and what we need is a meteorological variable that
will, in fact, type the turbulence and specify the diffu--
sion rate. We do not need a new model; we just need to .
find a way to tie these together better.
DR. BURTON: I want to clarify something. I
certainly-did not mean to imply that all of the recommenda-'
r
tions were based upon the existence and introduction of
new models. As a matter of fact, I intended to say that
regardless of whether we had a new modal or not, they
would still be misused or likely to be misused or mis-
applied.
I think in terms of whether or not there have
been advances or not depends on whether you are looking at
the fraction of the glass that is filled with water or the
fraction of the glass that, is still empty and I do think
that thex-e has still been no specification of the turbu-
lence, the parameters that you indicate, and that is a
need. We still will ru-.ke decisions v.cing models and the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
2
purpose of the workshop was to improve the quality of
those decisions in light of those kinds of uncertainties.
MR. VAN VLICK: Of the major meteorological
variables that are going to be nex* and different from what
we are using; is that right?
DR. BURTON: I think that there are plans to in-
! troduce new and improved ones and I am not certain that
those in fact, I am confident that those will not re-
| duce a lot of the uncertainties that you might be alluding
to or that others have alluded to.
MR. VAN VLICK: Another point well, this
morning they presented that the'frequency distribution
resembled each other between CRSTER and what was measured
in the field, but they failed to specify that did this
so-called highest values occur in the same meteorological
class? They may get a high value from Type B or Type A
or something that is wrongly typed and really ended up
! with strong, steady winds. I would hardly call that a
validation of a model.
DR. BURTON:. I know th
MR. TIKVART: Thank you,
I NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
*r & 92
DR. BURTON: !.Yes, sir?
2 MR. McGUIRE: I am Kenneth McGuire of Kentucky
3 Air Pollution Control. It is nice to hear all of this
4 reasoning and people willing to sit down together and try
5 to get together over these differences; however, we have a
more contentious situation and I was wondering if I could
7 get soms reflections from you all on that.
8 I can remember a few cases one of them is witl
9 an oil refinery in the eastern part of the state v/here^we
10 have 200 days a year of stagnations and no matter what we
11 could do, the oil company would never recognize modeling.
12 They put monitors at most any place we would suggest, in
13 addition to ...their own places. They would always get cora-
14 pliance and we never would.
15 (Laughter.)
16 MR. McGUIRE: Now, thio got settled -in Federal
17 Court. Let us put it that way. That is where it always^.^
IS
19
20
21
22
23
24
25
goes. ..:;
DR. BURTON: I would like to comment since I was
/
personally involved, I think, with that particular refin-
ery.
(Laughter.)
DR. BURTON: And with that particular circum-
stance and I believe the way it did coir.e out was that, yes
there WG.S no agreement on node ling and a monitoring progra:
MEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. HV/
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18.
19
.20
21
22
23
24
25
*-", !' .'.
p : i .<. 93
? (
was adopted and, yes, there were exceedances of every
standard that was on the books three-hours, 24-hours,
and annual averages and the refinery was just as sur-
prised as anyone else about those and quite embarrassed
about them, as a natter of fact, and did institute a fair
amount of control to assure that those standards the
exceedances of those standards would not be threatened,
but I still have to say that the controls that were imple-
mented and agreed upon would, I think at least from3 my
own point of view and I think the EPA and the state
people concurred protect against exceedance of the
>
standard.
The model still showed tremendously more addi-
i*
tional controls would be required. I think that was just
a way in which to deal with the uncertainty in that situa-
tion which was to spend a lot of money to go out and get
the observations because it was for an existing source;----S
MR. McGUIRE: Yes, that is right. But when we
have an adversarial type situation, it is very difficult
for us to get a situation of agreement on EPA models be-
cause they seem always to sho'v/ higher level of emissions
than can be-obtained by monitors. Our own monitors or
others. '
DR. BURTON: It does seem in this situation that
both sides were right,
NEA!. R. GROSS , .
COURT REPORTERS AND TRANSCRIBERS
. J330 VERMONT AVENUE. N\V
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
'11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
... y± I , 94
MR. McGUIRE: Thank you.
DR. BURTON: Lou?
MR. TOSIE: Hi, Shep. My name is Lou Tosief
not as Tikvart tried to turn me into Lew Kontnik back
there.
I was at the Airlio House Conference. I am with
the law firm of Fuller-Henry and in response to the ques-
tion I think it was the second question ago: what do
you do when you have competing 'models like dualing models,
dualing banjos? Our group and the only reason I am up
here is to pass on what our group talked about a little
bit and we had some ideas I do not think they were re-
duced to writing, so I will just report them as a bard;
if I had a banjo, I would sing it to you.
The point was that someone in our group felt tha
if EPA had one model and another group had another model,
the best way to look at those competing models was on some
type of equal comparison and if EPA had either an empirical
or theoretical base methodology to support its model, the
agency should be willing to accapt an alternative model
on an equally sound or better empirical base.
For instance, if the model they had had a set of
dispersion curves based on 20 tracer measurements, they
should be willing to accept another set of dispersion
curves based on a like kir.d study.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NV/
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
.11
12
13
14
15
16
17.
18
19
20
21
22
23
24
2f>
, I " -' - 95
«A j!|
(..'
One of the problems I have experienced person-
ally is that you have the agency developing guideline
models on one set of information and to come forward with
a competing model, you have to go through a continual
analysis and reanalysis to prove what the agency did not
prove in the first place.
So, in our working group, there was at least
that thought. If the agency had a model that was purely
theoretical, some physicist drew it up, and you had a
better physicist, your model should win. If the agency
had a model based on two measurements, empirical testing
and you had one based on three, and yours looked batter,
you should win. That was a very crass idea, but at heart,
i-
I think a rather important one of consistency.of criteria
and, of course, it is always good to have advance agree-
msnt because when you get into the courts on this, I
j maan half the people in this room do not understand what
we are saying and I do not think we do. In courts them-
selves, it is much harder.
I thought I would just pass that on.
DR. BURTON: I appreciate it, Lou. One comment.
There arc examples where that has actually been imple-
mented in practice in the PSD permitting arena and it seem
to work in introducing new models, for example.
Yes?
NEAL R. GROSS
. COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
15
18
20
21
22
23
24
25
O-f
-
Stev£ would >like to respond --to that. . !
MR. CONNOLLY: Lou, I would add to what you said
that to the extent possible, not only should the criteria
be consistent, but that they are to be made explicit and
in advance and as many of these as possible can be com-
mitted to paper as general guidelines for how like situa-
tions will ba treated and how unlike situations will be
treated. Then that facilitates cooperative processes,
I rather than having to invent the wheel every time you
start a new process.
DR. BURTON: And I think that the things that
I Doug Fox talked about this morning probably provides a
\ basis for reaching some of that agreement .
14 YeS?
MR. NOCHEMSON: David Nochemson, Los Alamos
National Laboratory. What measures would you recommend to
17
16
use for characterizing undartainty in models and how would
| you explain these measures to the lay public and, third,
how would you validate the methods for estimating these
1 measures?
DR. BURTON: Anyone? Bruce1, do you want to
I have an answer, but I do not want to answer.
DR. EGAN: Well, I minsed the morning session,
! but I assume Doug Fox may have described some of the re-
sults of an AMS-sponsored workshop held last a year ago
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
\
-------
1
2
3
4
5
6
' 7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
September, which described some measures in terms of at
least looking at airobers and s5 forth on models.
I am not sure if you are looking for a specific
list of statistics to be used or what. Maybe you could
clarify that.
MR. NOCHEMSON: The statistics that AMS recom-
mended in terms of characterizing uncertainty and things
like error estimates and correlation coefficients?
DR. EGAN: Yes. There is a table in this re-
port I do not have it with me, so I cannot read it off
to you but it talks about bias and gross error and othei
measures which then have some interpretation associated
with them so you can relate these to correlation coeffi-
cients and so forth. But more generally, your question
really does not have a single answer in the sense that
depending upon whether you are looking at one hourly or
24-hourly. You may want to quantify the uncertainty in
different ways.
The first approach would be to think of using
standard statistical means conference limits and normal
curves and so forth to describe the uncertainty. That
would be the way people would approach it first off. It
gets to be more complicated when you look at, in my view,
.when you "look at real meteorological events which may have
son\e persistence associated v.-ith then; and things, so you
NEAL R. GROSS
COURT REPORTERS AND .TRANSCRIBERS
1330 VERMONT AVENUE. N\V
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
_.;: . . . 1 98
~ ** "7"- - * Ji* * ..*.!». :
i
may want to adopt other statistical measures to describe
uncertainty.
DR. BURTON: I think there was one other com-
ponent to your question and the validation of those maa- .
sures with that report has not been written on how to
explain them to the lay public yet. But in terms of vali-
dation, I think that that is an issue that the Electric
Power Research Institute Study is dealing with,.' in particu-
lar, the influence of auto-correlation in various, measures
on assumptions about independence contained in those mea- .
sures and whether or not the influence of auto-correlation
on those measures is important and so on. So, I think thai
that is also-a report not yet written and the importance
t*
of it, although one can speculate on what the effect would
be.
Yes?
MR. HAYNES: Eldswins Haynes, North Carolina-
Division of Environmental Management. 1 hope I am asking
the right questions of the right people. I may should
have asked this before. It occurred, to me that the air
quality modeling that a modeler doss, if he uses statis-
tical techniques such as a percontile ranking or whatever,
you are forced to assume that accuracy of your output;
otherwise, your statistics that you develop arc more or
less meaningless.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
J330 VERMONT AVENUE. NW
(702) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7-
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
- ... , 99
Row, if the model employed always over predic-
ted -r its values were always shown to over predict, it
would be sensible to use the a 95th percentile or take
its tenth highest value or whatever, whatever you have to
do for your regulatory decision, but if the model may just
as often underpredict as it may overpredict, there is no
there may not be an inherent advantage replacing this
statistical technique to high, second high other than that
perhaps you will gst lower numbersf which would help in-.
dustry pretty good.
If we accept the possibility of model over-
"V
estimates in instances of model non-attainment, we as
scientists should also be willing to accept the possibil-
ity of model underestimates in instances of model attain-
ment. Let us say ten or twenty percent below a standard.
What do we do statistically or .is there1 any plan
to statistically adjust the decision-making or model im-=-.<-2
pacts if we accept the possibility of underestimates in
the modeling and what percentile rate might we apply to
the maximum concentration which may . that we get which
may itself be an underestimate?
DR. BURTON: Mike? Bruce?
DR. WILLIAMS: Well, my inclination right now
is to say that what you do is you go with your bast esti-
r.r.te. In oV.her words, v:c know something cibout tha
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS.
1330 VERMONT AVENUE. NW
(20?) 234-4433 WASHINGTON, D.C. 20005
-------
'"- '.;... 7 10-°
1 uncertainty, but I do not think it guides the decision in
2
3
4
5
6
- 7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
a more complete situation which- I would argue might happen
to be might require other people in the current set of
air quality managers. You might look at things like sup-
pose you have a decision that it is critical within the
range of confidence and you find that the control is not
much more expansive than any part of that range. Well,
then you might go to the higher end of the range in that
case. The naxt time you came back and found that you were
at a critical control level and costs jumped up dramatic-
ally, you might 'Stay you might accept that one as giv-
ing the benafit of doubt to the facility.
I think that is the kind of thing you would have
to do if^ you really were going to incorporate uncertainty
into your decision. Right now, I think what you do is you
make your best estimate and you live with it.
DR. EGAN: Let ma just comment on one thing you
mentioned, that by changing to a 'S5th percentile or some-
thing, you necessarily make it more lenient or whatever
towards industry. You do hot have to do it that way. If
you right now took a standard and then changed the fre-
quency recurrence so you allox-; it to occur more often, you
would be doing what you are saying, but you have the alter-
native of resetting the standard to another number if you
wanted to r»air.tain i_ho oc.ir.a cLrir.goncy, whatever that mean;;
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
8
9
10
VV* & '- 101
- - " K:
And the notion. JLs_ that^some...tighter, value could be ex-
ceeded more often. The advantage from a modelers point of
view is that that, if you allowed a value to occur more
often, you might be able to predict it with more certainty
and you do not necessarily have to give up environmental
., goals in so doing that.
b
Now, there is a difference. Mike Williams
j second highest value and it might even be cite specific
12
13
14
15
Ifi
17-
18
19
20
21
22
23
2-1
25
pointed out earlier that it may not be a single scale fac-
tor that you can apply to do that. It is not necessarily
a simple matter to define an equivalent to the highest,
for situations of complex terrain or whatever, but the
| notion would be that you would gain some certainty in
being able to predict the values, hopefully.
MR. HAYNES: Yes. I asked this question because
I am anticipating that whatever we decide in the future,
we will -probably have soma resistance from environmental
groups. They will need to be fully informed about what
this really means.
DR. BURTON: In the practical sense, I can tell
you what happened when I experienced in a model evalua-
tion activity, we were trying to introduce a new modal in
a PSD permitting situation and the evaluation of the model
was made separate from the application of the model to
detcrr.ine the impact of the proposed facility.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
ifii ; «' -.A : 102
? ' ' :
_ . - in the course of doing the model Devaluation-, .-the
high, second high value was, in fact, observed to be lower
the predicted value was observed to be lower than the
observed value and the protocol in this case stipulated
that if that happened, all values would be raised in pro-
portion to that value.
I would not I cannot defend that from a
scientific point of view, but that, in fact, 'is what was
agreed upon. Then but that was done in the context of
model evaluation; it had nothing to do with the applica-
tion to the particular facility. It was based upon tracer
data.
Then with that sort of fudge factor or that
r
factor for bias, in mind, the plant, the model was actual-
ly applied to the facility and then that factor applied
to all of the calculated concentrations of that facility
and then comparisons with the., standards ware made. It--is
a procedure, but it is which everyone could agree to
in advance, but, as I said, it is not one that you would
want to advance for all situations.
Doug?
MR. FOX: My name is Doug Fox. I would like to
ask a question of the panel and it is really addressed to
some comments that Mike Williams made with regard to the '-
Class I PSD increment and the air quality related values
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
-r-. ur/ i"
I'. 103
test and if I am quoting you or paraphrasing you correctly,
Mike, correct me if I am wro'ng you said to the ef-
fect that some of the reliance on the specific numerical
value was reduced and diffused because of the air quality
related values test that is implied and required in the
Clean Air Act for permitting in Class I areas. You were
the only one that made any mention of that. I would like
to address the panel in general to see if they agree that
such an added complexity to the problem reduces the pres-
sure on air quality models.
I am not sure it reduces it. In fact, I think
maybe it increases it because it increases certain degrees
of uncertainty associated with what you are talking about
*
in the whole realm of air quality related values.
So, the question really is would anybody else
on the panel like to address that business of the ACRV test
and whether that is a suitable surrogate for using spe-
cific numerical values.
DR. BURTON: Does anyone? Bruce?
DR. EGAN: Maybe' I could just comment on in
terms of som-3 of the discussion that our panel had in
fact, I think it ona day when Mike was absent, so that
maybe he will hear this for the first time also.
But we identified that fact, that there is a
>
j variance procedure with Cluss I a::d, in fuct, thought abou
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
it in terms of opening the door to adding to bringing
O
other information into the decision-making process. In
3 other words, here in the law already was a case where you
4 do not have to depend upon making a decision basically or
5
6
7
Q
0 thinking that perhaps that sort of thinking could be ex-
tended in other areas.
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
IK
104
have a decision usurped, in fact, by a value of the
highest, second highest value. There were soms ways to
go around that. We talked about, in that context, and
Now, most of all, I think, as you recall, our
group addressed the PSD issues rather than non-attainment
issues and we felt that there was at least some precedent
for- allowing other supplemental information to be provided
to decision-makers in the PSD program which might be less
acceptable, if"you will, for looking at compliance and
>
non-compliance questions.
MR. CONNOLLY: I would like to add that there "Is
an interesting history to the creation by Congress of the
air quality related values test. The time that that was
created, I was working for the House Committee on now,
Energy and Commerce on the Public Health and Environ-
ment Subcommittee and it is an interesting story of the
struggle between values: one between certainty and the
or between certainty on one hand and flexibility on
the other.
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
; . . 105 ,
The House said let us set up these increments".
They may be silly; they may be arbitrary, but at least if
at least you know what the game is: it is modeling..'
Up or down. Yea or nay. If you are over, you do not
build; if you are under, fine, you can build.
The Senate created the air quality related
values test and it was actually the creation of Senator
Howard Baker. What they were looking for was flexibility.
That they wanted a subjective case-by-case analysis in the
Class I areas to determine whether or not there was actu-
ally going to ba soma problem caused to something they
called air quality related values in a National park or
I National wilderness area and that that decision as to whe-
! ther the permit ought to be granted ought to be based upon
that, not upon some arbitrary measure of air quality or
modeling of air quality.
They thought wa were nuts creating this fantasy
world of modeling and we thought they were nuts creating
this fantasy world of flexibility where somebody would be
! making a decision somewhere on \%'hat basis God only knew
and perhaps he or she could not get involved enough in it
! to find out.
The deal that was made that we also had the
Bro Amendment along on the variance and we slapped the
vairanca on the air quality related vc-lues; test and that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
106
is the base that walked out of the room. But it was a
struggle again between flexibility and uncertainty. The
values the flexibility uncertainty, the values that
we are talking about here.
DR. WILLIAMS: I certainly do hot think it re-
duces the burden on modeling. It gives them a whole ser-
ies of, in many cases, more difficult questions: ozone
and other things.
On the other hand, it gets away from reliance
on a single number and probably a number that in many
cases is not very directly related to impacts and gets"
away from that kind of thing. So, I think it is in the
line that we are talking about here,, improving the deci-
sion-making.
I do not think it that that kind of improved
s
decision-making is any easier; it is more cumbersome, I
would say. Every time you gst away from that simple num-
ber, it is going to ba more cumbersome.
DR. BURTON: Jin, would you care to comment be-
cause I think you touched on this when you in your
presentation?
MR. SALVAGGIO: I agree with what Mike said.
Once you get away from the numbar, you are opening up the
ball game for a whole ncv; s/st of judgments and there is no
basis for those judgments. Any decision on that matter
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW -
(202) 234-4433 WASHINGTON DC 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
n . J.;.;. : 107
you are going to have questions of equity and consistency
between regions of the country, between Class I areas, be-
tween state agencies, between companies applying for per-
mits, what have you.
I think it makes the process much more difficult,
I think the.issue you mentioned initially about attainment
and non-attainment areas, that problem is going to come up
much more pointedly in the future as we get uncertainty
into the decision-making process. There is not this re-
lief valve of other related values that you can bring up
to answer questions when you are sitting right on the
border of a permit causing a violation or not causing a
violation and you are not sure where it comes.
r
DR. BURTON: There is one more question and I
think this.will ba the last one before we ask Joe how he
intends to implement the recommendations in the workshop
reports. - \
MR. PRESTON: Ky name is James Preston. I am
employed by TENNECO, Incorporated. My question is one of
the classical solutions in modeling to solving the prob-
.lems we are discussing is reducing the uncertainty limits
to the point they are arbitrarily small. On the basis of
the data that is available for us to use in modeling being
classically Weather Bureau data, etcetera, has any thought
been gix*en to the consideration of the cost benefit to the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
!{ - ... - { 108-
end consumer or the American public in general in terms of
funding an adequate data acquisition program nationwide
to reduce this thing to a better foundation in the model-
ing industry, which would probably make them the whole
business of modeling a more viable profession in the long-
run?
Then it would if everyone recognized it as a
tool that gives a reliable result, I would say it would
probably be a better recognized instrument for everyone to
use. What I am talking about is Congress said that the
EPA would do this and EPA officials said that is nice, but
they did not give us the money to accomplish it with. 1 Has
anybody been looking at the potential cost to do a good
job in this area?
DR. BURTON: Tom, you are the only one from EPA
next to Joe up here.
MR. HELMS: I certainly do not think any new
major program I would hate to 'calculate the cost bene-
fit of that. Nov:, I do not see us undertaking any major
. .
new program to collect that data.
DR. BURTON: I do not even I am not aware
it would be a hard study to do and it looks to me like the
EPRI Study offers that possibility because it has both
from which to make ths comparison.
!'K. PRESTO:-:: Okay. Let: ne clarify ny statement
NEAL R. GROSS
COURT REPORTERS AND . TRANSCRIBERS
1330 VERMONT AV£HUE. NSV
f202| 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
' - - 109
in this regard. I am a member of various organizations
that have been entering into contracts with environmental
consultants all over the United States to acquire neces-
sary data. We are spending millions and I am not talk-
ing about one and two millions'of dollars every day.
This data may or may not enter the public domain. In lieu
of that, is there any cost benefit nationwide to doing thii
in a public manner for accessible to everyone to save
individual organizations or groups of"organizations from
doing this on a regional or small basis here, there and
yonder? There might be some larger benefit out of it.
MR. TIKVART: Jim, I do not thin); that anybody
has done a cost benefit of the sort that you talk about.
I am not sure how you would do such a cost benefit, but
in response to your question.perhaps everybody here is not
aware that every EPA and DOE are actively discussing a
joint complex terrain study to take place in a time frams^
of a couple years from now.
I do not know that it is necessary to elaborate
I on thcit. Somebody here somebody else here Norm
Eowne may want to comment on it, but those discussions
are taking place.
DR. BURTON: We will fill the available time, I
can assure you.
MR. COX--:OLLY: V/hile I was on Capitol Hill
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
; " 110 :
working for the Committee staff, we heard week after week
an enormous amount of and numbers of criticisms about
modeling, air quality modeling and the use of air quality
modeling.
I was trying to recall as you asked your ques-
tion whether we ever heard about any suggestions on how
the basic data that you put into the models could be im-
proved and whether there was a national interest in doing
so and how you would go about doing that and how you would
fund it.
I do not recall anybody every having from
industry, from EPA and that was in two Administrations
..the late Ford and early Carter or the environmental-
ists ever proposing anything like that, or even proposing
that it ba looked at to determine whether it would be
cost-effective to do it.
I think that is the ~sort of thing that in terms
of research that is a sound suggestion and is one that
demonstrates good faith. We want to build a system that
I works and to build a system with 'good works, we have to
i
I have reliable, credible data that people are willing to
live with and we do not have that right now. We do not
have that at all.
DR. WILLIAMS: There have been a lot of sugges-
tions of various panels like the NCAQ, air dispersion
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
2-1
25
rv.- -£~- ZLi_ ~ -111
panel, about a list of what additional measurements ought
to be made and made suggestions to NOAA that things be
done, but as far as I know, nothing was done beyond that.
Lots of people are not aware and are not talking
about it. Maybe it did not get any audience.
DR. BURTON: Are there any other questions or
comments that people would have, especially the panel?
(No response.)
DR. BURTON: Mike, do you?
DR. WILLIAMS: Just one comment I would like to
make. I do not want to bring up it is obviously not
the time and the place and I. do not think it is the
time and the place at least to debate the 1977 Clean Air
*
Act Amendments and if anyone is interested in the basis of
the PSD, there is an extensive record in both Houses of
Congress that deal with it and ones that wants to deal
with Lewis' discussions, I think should refer to that
material.
MR. TIKVART: Okay. Thank you.
Shep, I would like to express my appreciation
to you for organizing and running this panel. I would lik
to thank all the panel members for their participation
here today and I would like to thank the audience for
bearing with us.
The ball is in your court now. Our goal today
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NV,'
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
'11
12
13
14
15
16
17
18
19
20
21
22
23
24
112
was to present you with a lot of information on the status
I of model accuracy and uncertainty in decision-making and
we wanted to present you with some of the concepts that
are running around and the necessary information. It is
i
now up to you to feed back to us your ideas on this diffi-
cult issue.
I think a lot of us are suffering from informa-
tion overload right now, but perhaps tonight after a couple
of drinks, that overload can bs short-circuited and some
ideas will spring out. In any case, I am looking forward
to be hearing from you tomorrow with your ideas, your
thoughts. We will start promptly at 9:00 o'clock a.m. wit!
presentations by the Governmental agencies I mentioned
earlier and wa will probably not get to the individual
presentations until after lunch.
I would like to close the mseting with that and
i thank you.
(Whereupon,
at 4:57 p.m. the conference was ad-
journed, to reconvene at .S:CO o'clock a.m. the folloxv*-
ing morning.)
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 .WASHINGTON. ,D.C. 20005
-------
r -
v'C
2
3
4
5
6
CERTIFICATE OF REPORTER
This is to certify that the attached proceedings
before the Government of the United States/ Environmental
Protection Agency/ Second Conference on Air Quality Model-
ing, Afternoon Session/ held on Monday, August 10, 1981,
in the Thomas Jefferson Auditorium, South Agriculture
7 Building, 14th Street and Independence Avenue, S.W.,
8 ! Washington, D.C., were held as herein appears and that
9
10
11
12 j
13
14
15
16
17
18
19
20
21
22
23
21
25
this is the original transcript thereof.
NEAL R. GROSS
*
Official Reporter
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
T
U
E
S
D
A
Y
-------
2
3
.4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
-1 ii ri nif 1
GOVERNMENT OF THE UNITED STATES
ENVIRONMENTAL PROTECTION AGENCY
SECOND CONFERENCE ON
AIR QUALITY MODELING
./TUESDAY, AUGUST.ll, 1981
The conference was held in the Thomas Jefferson
Auditorium, South Agriculture Building, 14th Street
r ' - - -
and Independence Avenue, S.W., Washington, D. C., Mr.
Joseph Tikvart, Chief, Source Receptor Analysis Branch,
Conference Chairman, Presiding. »
PRESENT:
Joseph Tikvart .
Richard Rhoads
G. Thomas Helms
James Dicke
Chief, Source Receptor Analysis Branch
EPA
Director, Mon-itoring & Data Analysis
Division, EPA
Chief, Control Program Operations
Branch, EPA
Chief, Techniques Evaluation Section
EPA
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
INDEX
Page
PRESENTATION OF;
Mr. N. Sundataraman 7
Dr. Howard Jongedyk .;... 17
Dr. William Carpenter . , 25
Dr. Roland Draxler .-44
Mr. Jerome Heffter 46
Mr. John Goll 50"
Mr. Earl Markee, Jr. 62
Dr. Roger Shull .~ 70
Dr. Rodney Moe 82
Mr. P. K. Misra . ................ 96
Mr. Dennis A. Trout 105
Mr. William K. Bonta 117
Mr. Kenneth U. Mequire '...".' ~.~ .'132
Mr. Jerry Pell .' . -.138
Mr. Richard Hanson , . . --.-,d50
Mr. Richard S. Fein 159
Mr. Ralph Sklarew 172
Mr. Alan Wittoa '. . . .182
Mr. Robert Kohn 194
Mr. Ray Wright 198
Mr. Donald Moon 208
Mr. Mitchell M. Wurmbrand 218
Mr. David Maxwell . . :..... .225
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW " """"
-------
1
2 (9:00 a.m.)
MR. TIKVART: Good morning, and welcome
4 to the second day of the Conference on Air Quality
5 Modeling, the second conference.
Since some of you came in late yesterday
or did not arrive until yesterday afternoon, I plan
s
to excerpt my opening remarks of yesterday to make
sure that all of you are aware of the ground rules
under which we're operating today.
I am Joseph Tikvart, Chief of the Source
12 Receptor Analysis Branch of the Office of Air Quality
Planning and Standards. I will be your chairman
for this conference.
Participating with me on the hearing panel
to solicit your views and take your comments are*
17 Richard Rhoads, starting from-my right, Richard
18 Rhoads, Director of the Monitoring and Data Analysis
19 Division; Tom Helms, Chief of the Control Programs
20 Operations Branch; and James Dicke, 'Chief of the
Techniques Evaluation Section.
22 I'd like to thank all of you for accepting
23 our invitation to participate in this conference.
24 The conference is being held in response to requirements
of Section 320 of the Clean Air Act. .A conference
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
,*,,*,« - « i' F r~"" *' ^"" '- ^ , *> * ^ r
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
on air quality modeling is required at three-year
intervals to help standard!zed-and improve modeling
practices within air pollution control, programs
such as prevention of significant deterioration.
The conference is designed to encourage
an information exchange* Topics to be addressed
are the use of models in regulatory processes and
the accuracy and reliability of those naodels.
We are interested in recommendations for improvements
both in modeling procedures and in regulatory processes
with the goal of insuring optimal use of air quality
models in all programs which require tbteir use.
As an aside, I'd like to mention that my
r
impression of our discussions and the various presentations
yesterday were there were a lot of shoulds we should
do this, decision makers should do that, modelers^should
.do the other things, et cetera.
There weren't very many hows, how to do those
things, so the individual presenters today, I will be
looking -for and asking questions about bows, so I'm
not badgering any one individual. I'm trying to get
as much information out of you as I cam, and I hope
that the other panelists will assist me in asking question
about hows.
I'd like to note that we have specifically
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
nr
-------
invited those governmental agencies identified in Section
320 of the Clean Air Act to participate in the conference.
Those agencies that responded to our invitation were
listed yesterday, and their representatives will speak
this morning.
If there are any governmental representatives
or members of the public who wish to make a presentation
and are not on the list outside the door, or your name
wasn't read yesterday, and you wish to speak, please
make arrangements with Ann Asbill in back or Charlotte:-
Hopper, and we will get you on the list of speakers.
. . ^
12 As required by the Clean Air Act, a verbatim
transcript of these proceedings is being maintained.
,. The recorder is Miles Anderson of Neal R. Gross and
14
Company, Incorporated. Speakers are encouraged to provide
.fi extra copies of their presentation for' the "convenience
of the recorder and the chairman. Interested per sons-*^S:
will be permitted to enter into the record any written
comments they do not present orally. The record will
2Q remain open-for written statements and comments until
21 September 14, 1981.
22 The transcript and all written statements
will be maintained in EPA Docket No. A-80-46. If you
24 would like a copy of the transcript of this conference, ''
please see Mr. Anderson, and he'll make arrangements
£O
NEAL R. GROSS ''
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
15
16
17
18
19
20
21
22
23
24
25
for you to obtain a copy of the transcript.
The comments and discussions during this con-
ference will be informal and nonadjudicatory. We will :
try to allow time for questions after each presentation
if those in the audience desire to ask questions, and,
as I've indicated, I and other members of the panel
will try to obtain clarifications and ask questions
to bring out your ideas as we go along.
Individual presentations should generally
be limited to ten to 15 minutes with time for questions,
so the total presentation by each individual should
be about 15 minutes. If somebody needs longer time,
we will try to accommodate you.
14 -When making a presentation, please give your
written statements to the recorder and summarize your
remarks if they are very lengthy, but, as I said, we
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
on any written statement. We plan to have a 15-minute-
break this morning approximately about 10:30, depending
upon how that time falls relative to individual speakers,
so we will have a 15-minute break this morning. We're
not going to run you like we did yesterday.
If at an appropriate time you have a question
or a brief observation, and we will try to allow for
such at the. end of each presentation, go to the nearest
microphone, either one -of- the two on the floor or here
at the podium and clearly state your name and affiliation
for the recorder. I have to emphasize that. It really
helps the recorder if you state your name and your affilia
tion before you proceed.
^
.Before I introduce the first speaker, someone
left a-yellow'tablet'with notes written in it in the
back of the auditorium yesterday. Ann Asbill has it
«*
outside. I don't know how .important the notes were,
but if you left a tablet, she has it for you.
Now I'd like to proceed and introduce the
* -
first of a series of presentations by various government
agencies. The first speaker speaking for the Federal
Aviation Administration is Mr. Sundataraman. I didn't
say that right, I'll try it again.
i
MR. SUNDATARAMAN: Thank you, Mr. Chairman.
Regarding my name, it is Sundataraman. Nobody pays
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
4
10
11
12
13
15
16
17
18
19
20
21
22
23
24
25
attention to that, so I would expect all of you to forget
it immediately. The nickname is Ram, spelled RAM.
I don't know how that happened. I assumed as.-1 landed .
on Ellis Island a few tens of years ago, the immigration
officer would have assigned that name to be.
I am the Chief of the Air Quality Division
in the FAA, and my main purpose here today is to tell
you something about what we have done with regard to
the uncertainties that "We see in the models that, are
used in calculating the aircraft related air quality
problems.
I begin by referring to the fourth and fifth
recommendations of the EPA-sponsored workshop on the
f
use of mathematical models as management tools held
during May 3-7 of this year.
To quote them partially, "(4) The explicit
stipulation of uncertainties -through the best available
means ..." and then it goes on I don't want to
read all_of- it and take up your time, n(5) The selection
* " .
and application of new or modified modeling approaches
rather than insisting that existing guideline models
be used for all circumstances ..." These are taken
verbatim from the workshop summary report Role of
Atmospheric Models in Regulatory Decision-Making; EPA
Contract No. 68-01-5845.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
4
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Even though we did not participate in that
workshop, the operating philosophy of the FAA in air
quality studies has been (1) to emphasize the need tc
identify and quantify, where possible, and you all know
that is almost impossible, the inherent uncertainties,
and (2) thereby or otherwise to continue to refine,
improve and verify aircraft-related air quality assessment
models.
One of the objections of the air quality studies
in FAA is to develop techniques to determine quantitatively
the contributions from aircraft engines and other airport-
related sources to ambient air .quality, so really two
different kinds of sources we are looking at, one is
f
aircraft itself, and the other one would be everything
that is non-aircraft, but airport related activities.
Such determinations are for purposes of (1)
**
assisting your agency, namely, the- EPA, Mr. Chairman,
and the International Civil Aviation Organization in
evaluating the need for and type and timing of aircraft
i
engine emissions standards, and (2) environmental assess-
ment of Federal actions at airports and/or airway' facili-
ities as required by the National Environmental Policy
Act of 1969.
The meaning of the term federal actions is
somewhat vague at the moment. May I be permitted to
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 2O003
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
10
insert at this juncture, Mr. Chairman, our expressions
of genuine satisfaction at the very excellent, continuing
and responsible cooperation and help that we have received
from the EPA in the general area of airport related
air quality assessments.
With regards to the uncertainties in the air
quality models we employ, we have begun by identifying
the sources of uncertainty, and today we can list about
six of those. These six, I am simply going to read
them all to you. They are not. unique particularly.
They will start fairly general.
The non-steady or the impulse nature of the
airplane~as"a source. The emissions take place maybe
for a minute or so while the plane is taxiing and taking
off. In terms of taxi, it will be a little bit longer
than that, but certainly take-off doesn't take very
long, so in a sense it is like an impulse source.
Number two, determination of background levels
because you-want to find out enhancement of the background
i
levels, how do we determine the background levels.
Number three, the four operational modes for
the aircraft. These are somewhat uniqne to the aircraft
because the emissions defer at different modes. For
example, the carbon monoxide emissions .are the highest,
for example, the idle on the taxi, on the queuing
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 2OOOS
-------
4
8
9
10
11 -
.12
13
14
15
16
17
18
19
20
21
22
23
24
25
11
-- -
- situation, while take-off, it is noxious which is import- ;
ant. The four modes are the engine start, taxi in a_nd.^_.
out, queuing and take-off.
The fourth source of uncertainty is the high
engine exhaust temperature and velocity which affect
fHeTvertical and horizontal extent of the plume. I
will have a little bit more to say about this later
on. . .
Five, substantial changes in idle emissions
rates, for carbon monoxide and hydrocarbond, for rela-
tively small changes in power setting.
Six, other non-aircraft sources associated
with airport operations.
* We have undertaken oftentime jointly with
other federal agencies, including yours, monitoring
studies as needed at selected airports across the country
*»
to address these uncertainties.
'
In regard to the non-steady or the impulse
nature of the source and the determination of background
/
levels, we have used' high time resolution measurements
of concentrations and associated meteorological-parameters
close to the taxiway and/or runway to study the emissions
from essentially each aircraft passage. To such moni-
toring, we have assigned the name single event monitoring^.
By co-locating noise monitors, there is one
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005 --- .--
-------
12
good "use of noised "We "can differentiate between airplane
and other sources. Such measurements give more objective
measurements of both .time average values and background
levels. I am not going to go into more details than
this, because it will take up much too much of the time.
Wi.th regard to the different operational modes,
we have developed specific models for taxi in and out,
8 queuing and take-off. These models lend themselves
more readily to verification. Also, we exercise these
models in a diagnostic sense in order to formulate and
refine the multi-mode models.
The hydrocarbon emission rate as a function
of power __sett ing is being studied in an engine test
1^ ""
14 bed in a cooperative effort with the EPA and the Air
Force. This effort will include an assessment of alterna-
15
16
17
18
19
20
.21
22
23
24
25
tive fuels as well.
Further, the emissions will be directly fed
into a portable smog chamber for -studies on photochemical
oxidant formation.
i
With regard to the plum problem. I feel the
FAA has made some contribution in this area. In the
other areas, we are making attempts and we have some
results, but it's a continuing process-
We did conduct at Dulles International Airport '-
a monitoring program in connection with the decision
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. 13
made by the Secretary of Transportation on the Concorde
landing, and this program was undertaken in 1976, and
we did make some observations at that time at the airport,
Dulles International Airport, which have clarified the
characteristics of aircraft plume rise.
I would like to show you two viewgraphs at
this stage, and I apologize for one of them. It i-s
you obviously cannot see it. I can't from here.
You may feel why does this idiot show such ^
viewgraphs, but I can assure it is not intentional.
The point I want to make here is you can see the taxi
>
path, -and there are three spikes sticking out. I don't
know that you can see the third, but certainly there
r
is a line joining those. They are the locations of
the three.towers.
The three towers were not varying "in h'efght.
They were all 84 feet in height actually, but this was -
to show simply the location with respect to the center-
line of the taxiway, the location of those towers.
i
We'had five stations on the first which is
about 200 feet from tho center taxi line, and the second
tower had also five stations which was another 300 feet
away, and then we had a third tower whicti was an equally
tall tower which had only three stations on it, and
that was about 200 feel away from the second tower,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
- . }
14
r
so in all it was about 600 feet, 200, 300 and some frac~;_
tion of whatever is left over.
2
And what we did was to measure really carbon '
3
. monoxide. We did measure the temperature gradient from
4
the vertical gradient of compressed air, and we had
5
concurrent wind observations.
6
Can I have the next viewgraph please? This
7
gives you the result of what we found. This is on one
8
occasion. Obviously the results depend upon the wind
«7
speed and direction and the other characteristics.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
As you can see, the first panel on the left
shows the measurements taken at the first tower. The
measurements were at 6 feet, 26 feet, 41 feet, 56 feet
and 80 feet, and the middle panel shows the measurements
in the second tower, and the measurements on the third
tower are shown on the third panel, and you can see
;. -*» ~"~
the plume rise visually here.~
What we have done is we have collected quite
a bit of this data, and going through certain analysis,
we did come up with values of plume rise that would.
be applicable to aircraft in a certain general sense,
and now we do use this plume rise characteristic routinely
in our models. ;'
And the inclusion of this plume rise render
a discrepancy between model calculations and observations
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
133O VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
,15
j. - - - f « - .-.. _ j_
much less severe than otherwise*
That concludes my remarks. I thank you for
the opportunity to tell you about the PAA program. My
associate, Mr. Segal, who is with me or I will be happy
to answer any questions you may have. Thank you.
MR. TIKVART: As promised, I would like to
ask you a question. That is, you've listed areas"of
uncertainty in the model estimates, and you've shown
what the measurement program showed. How do you plan
to take this uncertainty into account in judging what
the emission limits from the aircraft should be?
MR. SUNDATARAMAN: The best thing I can do
is to say that first of all, you use the measurements
in a sense to really check your model, and the two don't
agree. I'd be surprised if they agree. And then what
you do is to go back and see where you can really,make
improvements, and one area which we specifically has
is plume rise, and the plume rise, we can give you the
numbers. ( -
In fact, it appears in our report that was
published jointly with the EPA. It was a study, and s .
we do incorporate the plume rise in our; models these
days routinely, and from the same study, we also had
determination of the box size parameters which we also
include, and this gives you a technique of perhaps coming
NEAL R. GROSS "
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7 correct?
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
IS.,
closer to assessing the future impact of airport opera-
tions as against what we have today.
MR. TIKVART: Okay, so as I understand it,
basically you're talking about going through a model
evaluation and improvement exercise until the model
conforms acceptably well for making a decision, is that
MR. SUNDATARAMAN: Yes, maybe what you are
trying to get at is can we really quantify these uncer-
tainties. Is that the thrust of your question? ~"
MR. TIKVART: Well, okay, that certainly is
an aspect, can we quantify the uncertainty, but more
importantly perhaps the difficult question is once the
r
uncertainty is .quantified, what do you do with that?
How do you factor that into your decisionmaking process?
MR. SUNDATARAMAN: I don't think I have^answered
any of those questions, but .1'e.t me just throw some thoughts
out.
It is not impossible, it's very expensive,
but it's not impossible to quantify the uncertainties
.to the extent that we know the sources of uncertainties.
You can do an exercise, like a Monte Carlo model, for
example. We have done this in connection with aircraft
emissions in the stratosphere. That program is also
under me in the FAA, and you can come up with certain
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
2
if
^.- fcK.
IT 1*
results. You can come up with certain measures of uncer-
tainty, but what you do with those uncertainties especially
in a regulatory context is something that is not frankly
upto me at the moment.
4
MR. TIKVART: Thank you. Does anybody else
5
have a question? Any questions from the audience? Okay,
6
thank you very much.
0
o
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
7
MR. SUNDATARAMAN: Thank you.
MR. TIKVART: The second speaker this morning
there are two speakers. Is this a joint presentation?
No, okay. There will be two speakers for the Federal
Highway Administration. The first is B)r. Howard Jongedyk.
DR. JONGEDYK: To clarify this as we go along,
the speech about how this all fits together. The Federal
Highway Administration has a role play in air pollution
inasmuch as highway sources have 'been identified as
a major source of air quality problems.
In our investigations, we have tried and are
trying to identify and clarify .the problems, to see
how we may make measurements, models to more clearly
quantify particular situations.
As we identify these problems more thoroughly,
we have .inputs into other types of modeling efforts,
automobile emission control strategies,, transportation
contrcl~~measures and highway designs an
-------
8
9
10
11
12
13
14
15
16
17
19
20
21
22
23
24
25
in general. We are not alone in this regard. The Federal
Highway Administration really has a fourth full wave
of operations besides the overall cooperation of other
federal agencies.
There is actually the work done by the staff.
Then we have what's called administrative contracts,
and we work with the NCHRP program of the National Academy
of Science, and we have cooperative relationships with
various agencies who may do the work themselves or work
with state universities or other places in the state.
At various times, we've had work in many states,
New York, California, Colorado, Texas, Virginia and
several others which I won't name right now. Now, in
f
this regard, Dr. Carpenter who is with us this morning
has been at the University of Virginia, and he has parti-
cipated .heavily in work which Virginia has done in coopera
tion with us, and rather than saying anything more right
now about that, I'll let him talk about that.
Mr. Moe from Texas Department' of Highways
and Traffic will discuss some of the work they are doing,
so I will try to avoid saying anything about that as
well.
Now, our overall role in what we call our
federally coordinated program is to look at the role
of highways as sources and as places from which emissions
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
13
14
15
16
17
18
19
20
21
22
23
24
25
--«=«?
3 -;r_;. - 19
go from, in other words, the dispersion. In our efforts
0 to do this, we've tried to look at what happens to vehicle
' ' '
movements, the unsteady flow, the nonuniform flow .of
the traffic, and trying to do that, we look at the emis-
_ sions at sites specifically.
_ For example, currently we have a major study
-in cooperation with the Transportation Systems Center
a I] in Cambridge,, Massachusetts where we are trying to quantify
what actually happens as vehicles operate various modes.
Vehicles are put on the dynamometer, a test
,, similar to the ordinary test vehicle from the automobile
. manufacturers, but also they are tested under a wide
variety of modal operations and various applied external
loads, meantime collecting the emissions put out by
those vehicles under those various loads.
Then these vehicles are taken out in the field,
and we find out in the field why we have these loads.
Loads come from acceleration, lift, a vehicle going
up a grade,,aerodynamic drag and road resistance. We
are finding out that the road resistance is not a constant
factor. It varies with velocity of vehicle. It varies
with the extremities of a vehicle going around a curve,
it varies with the roughness of the pavement and so :
forth.
Interestingly enough, as you. look at the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
4
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
7:7^- : ; '' 20 '
variations of source, what we have is a highly variable
number for the emissions factor, per sc, of a given
vehicle, never mind what happens when you have a whole
big assemblage of vehicles going down a particular highway
or a highway network. . . . . . . . -. .
.«
For example, we have found out the emission
rates for a highway going up a hill at grade are much,
much higher than the stated FTP test cycle value. At
the same time, a vehicle operating in a cruise mode
at a moderate rate of speed had much lower emissions
than .highway Federal Test Procedure womld give.
Much of our work in the past !has been done,
however, looking at the quas steady uniform line sources.
The first.efforts were really done primarily in the
New York City area looking at how highway configurations
affect the-air pollution pattern around a given higway
site.
This is followed by a major effort with the
California Department of Transportation in Los Angeles.
A caline 2 model was evolved from that- That work in
California has continued. In most recent years, they.
have produced a caline 3 model. The caline 2 model
was a line source, the caline 3 source is assemblage .
of area sources near the point of consideration.
The caline 3 which is currently widely
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
' ~ 1330 VERMONT AVENUE. NW
-------
2
10
11
12
13
14
15
16
n
18
19
20
21
22
23
24
25
distributed considers such factors of dispersion in
the wide direction with the parallel wind and dispension
in the Z direction with perpendicular -wind, and both
factors are considerably greater than we had originally .
thought.
In the meantime we had a major effort done
by SRI international in which they tried to look at
some of the basic mechanics with the help of wind tunnel
measurements and field measurements of the highway dis-
persion from again a quasi-uniform steady line source.
We looked at numerous configuration, varying7
roughness parameters, we considered a highway on a side
of a hill, fill, viaduct and so forth.
Many of these evaluations have been incorporated
into what we call roadmap models. SRI efforts were
contributory to the work which was done by General Motors
in Michigan and work which was done in INew York and ~~"i*^^
work which was done in Texas, and then all these efforts
have formed a data base which is now being looked at
against the various models by SRI under an NCHRP program
to try to look at the validity of these various models.
One of the areas which we think we're weak
in is where we have a great deal of uncertainty or unsteadi
ness especially near complex sites, complex sites which
have large geometric variations due to the buildings,
NEAL R. GROSS * "
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
-------
_ - -^ 22
1 due to the highway cross sections themselves, inter-
2 changes, intersections;where the sources are harder
g to quantify, dispersion is harder to quantify.
We are looking at in. many ways the degrees
5 of uncertainty. We realize our models are not exact.
6 They give us just ballpark numbers frequently; however,
7 until we have something better, we do the best we can.
8 Now obviously what we have clone is interacted
g with working with what EPA has done with the highway
10 model and other work as well.
A major part of our work, for example", has '
12 been done in cooperation with Texas Department of High-
13 ways and Public Transportation, and Mr. Moe will talk
about that later. If there are any questions, I will
entertain_them-at this time.
16
17
18
19
20
21
22
23
24
25
>
MR. TIKVART: You haven't said too much" about
uncertainty. How do you deal with the problems that
you mentioned in the models and how you use those model
estimates? 'How do you deal with the uncertain factors?
DR. JCNGEDYK: Well, first off, I would like
to defer to Dr. Carpenter who will follow me where he
will address some of the probablistic aspects of modeling.
As far as the uncertainties are concerned, we have tried::
to look at the effect of stability on the models and
realizing how these factors can come into play, we have
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
133O VERMONT AVENUE, NW
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.- - -"" 23
___ * -"---.-. ^ . ,^' ^
tried to even incorporate hot and cold starts in some
models,, but that's just a big guess too.
I don't know. I think the answer to your ques>
tion in large part is largely in the users' hands to
have them have enough understanding of what the models
look like or what they consist of and treat the models
accordingly. ':
MR. TIKVART: Tom?
MR. HELMS: You mentioned these models give
you ballpark numbers. Could you maybe comment on the
role the models played in your decision to build or
not to build a highway?
We heard yesterday, they were talking about
point sources. It seems that models give a go/no-go
type approval to construction of say a new power plant.
Again, what is the role of these models that give^ the
ballpark numbers in your decision?
DR. JONGEDYK: Well, basically, we have usually
taken a very conservative aspect. Most all of the models,
when they are applied, consider worst case situations,
- and with some exceptions where we have sometimes insisted
for a higher degree of dispersion than sometimes the
models were earlier calling for, we usually go with
high volumes of traffic, traffic situations which are
conducive to higher rates of emission per unit of distance
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
'202^ 7.H4-4433 WASMIMGTON D.C, 200O5
-------
2
4
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- ?4
or per unit of time, and maybe on usually adverse meteor-
ological conditions, and then we hope everything's fine/
and then if it is underneath considering background,
underneath the standard, then we have no more problems.
Understand, we have already taken the worst
situation. Now we've more finely tuned this, and there
have sometimes been situations where I believe the highway
agencies had to do some tough negotiations or maybe
in a realistic way, we have said, all right, here is
what the situation is, and there's notching we can do
about it.
Let me just give you one quick example. Frankly
speaking, any tunnel portal with heavy traffic in the
mediary of that runnel portal has violated the ambient
air quality standard, but somehow or another, we're
not called on that one, whether it's a grandfather situa-
tion or whatever.
MR. TIKVART: Thank you. Why don't we proceed
with Dr. Carpenter? Go ahead, ask your question now.
MR. HALBERSTART: I'm Marcel Halberstart from
the Motor Vehicles Manufacturers Association. You men-
tioned rather serious departures during portions, of
your testing from the emission patterns shown in the ;
Federal Test Procedure, and I was wondering if you were
perhaps working on development of alternative test
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
procedures. These are serious implications, of course,
for fuel economy testing as well as emissions testing.
DR. JONGEDYK: Let me add that fuel economy
is being measured in parallel to this, and one of the
ways which we have tried to respond to air quality situa-
tions or violations in the past has been definitely
to show either a new facility being constructed or a
new traffic pattern. We will try to have more cruise
mode situations occurring because of timing of traffic
lights and so forth.
In answer to your question about the legal
implications about the federal test procedure, I think
the regulatory agencies and the vehicle manufacturers
p
both agree they do not want to change legal status of
that current federal test procedure, because they have
so much already invested in that particular" procedure,
to moot the requirements and so forth. """" ~"
MR. TIKVART: Okay. Why don't we proceed
with t)r. Carpenter then?
*
DR. CARPENTER: Good morning^ My name is
Bill farpenter. I am today representing the Federal
Highway Administration. We have a need to predict future
air quality impacts for either existing or proposed
facilities. There's a basic problem. The problem is
we cannot predict the certainty.
NEAL R. GROSS '
COURT REPORTERS AND TRANSCRIBERS
' 133O VERMONT AVENUE. NW
(202) 234.4433 WASHINGTON, O.C 20005
-------
i -- 2'6
k ' ~ ' *.
1 This uncertainty arises from two places. There
2 are imperfect air quality prediction models, and there's
3 also the random behavior of the inputs to these models .
4 which is meteorology, source emissions, background con-
5 centrations.
6 Two obvious methods for dealing with uncertainty
7 one is averaging. When you take a bunch of.: numbers
8 and average them together, you reduce errors' or uncer-
9 tainty resulting from unbiased errors. You cannot remove
JQ bias, but the averaging process will reduce unbiased
jj uncertainties.
12 Another approach is simply a probablistic
jo approach which is simply to quantify the erros, try
to measure them.
Looking at air quality standards with respect
to uncertainty, you could take air quality standards
and write a standard in terms^of say a long-time average
pollution level, such as a yearly average level. The
uncertainty present in this long-term average will be
i
significantly less than the uncertainty in any one hourly
.pollution level.
For instance, if there is .no bias and if,
assuming for a moment, that the errors were somewhat
normally distributed, the error in an annual aveorage
would beon the order of one percent of the error in
14
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
27
a one-hour value. Another approach would be to look
at average pollution levels and a probability of exceeding
some long term average as an annual average.
You can also look at pollution standards as
a lot of them are written today which specify maximum
allowable pollution levels not to be exceeded more than
say once per year, and you can look at the probability
of violating that standard.
The two different kind of standards, the long-
term average and the maximum level, have different kinds
of applicability. Annual average type standards are
applicable to chronic exposure health effects. The
annual maximum are applicable to acute exposure, and
the two do not overlap.
With annual maximum type standards, you really
cannot address problems due to. chronic exposure, and
with annual average type standards, you cannot address
problems due to acute exposure.
'There are three basic approach to the annual
maximum type standards. I'm going to address the maximum
"standards, because those are the ones where the uncertainty
plays the biggest role.
We have the worst case type approaches which
Howard Jongedyk addressed a little earlier. In the
worst case type approach, subjectively, someone says
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
j||- we will use a given emissions factor, we will use a
ii
given background concentration, we will use a given
2
set of meteorological conditions, input these into some .'
3 ]
model and come up with a pollution level, a single number.
4 \
If that single number exceeds a value specified
5
by a standard, you can have a no-go situation. If it
6
does not exceed, you have a go situation. There are
7
the lognormal type models which are also worst case
8
in nature where a lognormal probability function is
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
fitted to a set of data. An order statistic from this
lognormal_probability function is obtained to estimate
an expected annual maximum concentration. This expected
annual maximum concentration is treated as a worst case
again". If it exceeds the value specified in the standard,
again you have a no-go situation.
A last approach is a probabilistic approach
where one might try to estimate-.the probability of vio-
lating the standard. On to the next slide please.
PThe worst case type approach, first off, com-
ounds uncertainty by its subjective choice of the para-
eters to be input to a model. Also, it's a deterministic
pproach to what is basically a stochastic phenomenon,
random phenomenon.
The lognormal approach does eliminate sub-
ectivity. It assumes universal lognorraality which
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(2021 23-W433 WASHINGTON, D.C 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
^~~ "~ ... - 29 f
.--, . _,. * *
- is really a pretty hefty assumption that in itself intro-
duces some uncertainty. The lognormal type approach
fa
does employ averaging. The parameters of the lognormal
3
distribution are obtained by an averaging type process,
4
so they turn out to be relatively stable, given the
5
assumption of lognormality.
6
The lognormal type approach does ignore the
sequential dependence of hourly pollution levels, and
they do depend on each other sequentially.
Given the lognormal assumption and given the
assumption of independence, the logncurmal approach can
be used to estimate a probability of violating a stan-
dard. In general, however, the lognormal type models
are used to estimate worst case.
The probablistic approach aJLso eliminates
subjectivity. It employs averaging and the determination
of parameters. It does address sequential dependence.
It does not assume any give distribution. There are
no uncertainties introduced by assumption of a type
of probabilistic model, and the probabilistic method
obvious does yield a direct estimate of the probability
of violation.
We've done some simulation studies, and what
we do here is we input a meteorological history, generally
something like a ten-year history. We input deterministic
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
30.
emissions patterns. These are the predictable patterns
from hour to hour, day to day, weekday to weekend, month
to month, whatnot, of emissions patterns. These are
deterministic.
Then you general random perterbations around
these emissions levels. You also input deterministic
background patterns. Again, these are hourly fluctuations,
and you generate random perterbations around the backbone
level. °
You employ some sort of an air quality model,
such as, for instance, highway. You could, we haven't
done this, generate perterbations around these predicted
values, in-other words, enter in the uncertainty in'
the" predicting model itself.
You have to be able to quantify that uncertainty.
That's why we haven't done it. From the result"ydu
estimate .the probability of violation. We did some
work with N02, and the example that I have on the next
slide I will show you the results, but let me describe
the example.
We had a highway with six four-meter lanes,
a 12-meter median, the receptor was 15 meters from the
nearest lane, the roadway extended about two kilometers
in either direction away from the receptor. The ADT
was 100,000 vehicles per day, speed limit was 55 miles
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
: 31
0
an hour. The average emission rate was 3.6 grams of '
NOx per vehicle per mile, and that was at the standard
19.6 miles per hour. :
The average annual background concentrations
were 2/100ths of a part per million of NO2 and 3/100ths
of a part per million of N03. Trie did a ten-year simula-
tion and we addressed three possible standards. At
the time this was done, there was no standard for N02,
and we looked at three different possibilities, The
next slide please.
Looking at these results, we looked at standards
not to be exceeded more than once per year of .25, .35
and .45 ppm of NO2. At a .25 standard, all of the methods
f
gave a 100 percent probability of violation. Let me
just quickly describe these methods.
.The P(V) method is a,method that we've been
IS.
working on for the Federal -Highway Administration, and
it attempts to address the true probability of violating
a- standard.
* .
What we call the simply binomial probability
.of violation is you look at, we had a ten year simulation,
you look at each of the ten years, and each year either
violated or did not violate a standard, so you're going
to end up with one to ten over ten, or 10, 20, 30, 40
possibles of violation. The one above P (V) is actually
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
. _...._' ,- 32
a combination of the simple binomial overall possible
2 one year periods.
3 The lognormal is, as I've discussed before,
4 the lognormal worst case is either a go/no-go depending
5 on whether or not the worst case predicted by the log-
g normal exceeds the value in the standard, and the worst
7 case is either a go/no-go. It will be zero or 100 percent
g depending on whether or not it would exceed the value
9 given by the standard.
At .35 part per million standard, again, we
have virtually 100 percent for all the methods. At >
12 a standard of .45 parts per million, you'll notice what
13 x ±JL call the true probability of violation was at about
40 percent. The simple binomial does give a very good
15 approximation to this and is quite a bit cheaper to
actually implement, and that gives about 30 percent.
The others though, the lognormal probability
of violation, the worst case lognormal and the true
19 worst case .all still give a true 100 percent probability
20
21
22
23
24
25
of violation.
The last three methods don't seem to agree
as well with the data as the first two methods do, and
I'll go on now to an example using carbon monoxide.
The scenario here is very similar. We had
again six four-meter lanes, a 12-meter media, the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
.-.-_.--- -v. :33
receptor was 15 meters from the nearest lane, roadway
2 extend two kilometers in either direction, and the ADT
3 was 100 vehicles per day.
4 Here the vehicle speed was actually 55 miles
an hour, not the speed limit. The vehicle speed and
the previous problem varied around the 55 miles. The
average emission rate was a little bit over 30 grams
of CO per vehicle per mile at 55 miles per hour. The
9 annual average background was 3.2 PPM of carbon monoxide
1Q and we did six ten-year simulations.
11 The first one used only deterministic source
12 and background terms, no random perterbations. The
12 next four used ten percent random perterbations around
the initial deterministic source of background concen-
15 trations, and the last one was deterministic. It was
the same as the first, but it was a five percent-overall
increase in all source and a 1-1 background terms.
Next slide please. For reference, the CO
19 standard was 35 parts per million. That's one hour
20 not to be exceeded more than oncre p&r year.
We have across the top the six different runs
22 that we did and the same probability terms we had on
no . the previous example. We see that, for instance, look
24 at run number one, the true probability of violation
was about 26 percent. The binomial again is a good
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
34-
approximation to the nearest ten percent~which is all
1
it will give.
2
Using the lognormal probability of violation
3
and the lognormal worst case, both of these yielded
4
a zero percent probability of violation. In other words/
5
the lognormal ended up with a much shorter tail than
6
the true distribution of the data.
7
The worst case method, however, came out to
8
be again 100 percent probability of violation. The
9
worst case method is really not stable because of its
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
subjective-.inputs. Two different people can do a worst
case and get to drastically different results.
-The next four runs two through five again
involved random perterbations around the data that was
used to run number one. Actually if you'll look at
the averages over two through five they approximate
4»
what happened_in run number one, . and the results are
basically the same, same source of patterns as you see
in run number one.
/ - ' .
In run number six, we had the true five percent
increase in all background and source, and we notice
a significant jump in the true probability of violation.
The binomial matches it again to the nearest ten percent.
The lognormal probability of violation in
lognormal worst case again, the tail on the lognormal
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
_ 100 percent violation.
D
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
thing to hear, the last slide. Just to briefly sum-
7
8
marize the probabilistic method, it does require no
subjective input. That's a true benefit for the user.
,. -.- X , . . 35
did not match the true data, and it comes up showing
a zero percent probability of violation.
The true worst case/ again it's subjective.
It does not match the data, consistently doesn't match
the data, and it shows an overestimate probability of
The last slide please. Itfs always a pleasant
He doesn't have to sit down and scratch his head trying
to figure out what sorts of numbers he's going to put
in..
It is distribution independent as opposed
to say the lognormal, and there are other approaches
other than the lognormal. I've used that simply"because
it's the most common, but this probabilistic method
we've been working on does not assume any one probability
distribution function.
The probabilistic method does quantify uncer-
tainty. It gives a measure. That's a lot of what we've
been looking for. It has the potential for addressing
model uncertainty. If you can quantify the uncertainty.
in a prediction, then that can be entered into the pro-
babilistic method and will, in turn, affect probability
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
' -"-. : . 36-,.
_ ""* - -£_ _ ~
that you would violate a given standard.
The probabilistic method is more realistic.
It does match the data better than the other methods
considered. As a final point, it is really quite cost
effective. There was a total of around $300 in computer
time spent, to do almost a million simulation, 700,000
7 or so simulations for the two examples that I showed
8 .vou-
That's all I have. If there are any questions,
I'd be glad to answer them.
MR. TIKVART: Thank you. That was a very
12 interesting presentation. One question, how do you
13 define the probability of a violation? Does that mean
two excursions .in one year of the series of years that
15 you looked at?
-fi I'm not sure what you mean by that. ^
17 DR. CARPENTER: With reference to the standards
18 that we were looking at, like, for instance the CO stan-
dard, the standard says not to be exceeded more than
20
25
once per year, we looked at all possible one year periods,
and you can start on any hour of the year.
22 So you take all possible one-year periods,
23 and you- look at the one-year period, and you ask the
question, was the standard exceeded more than once.
MR. TIKVART: When you say all possible years,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
.C 20005
-------
what does that mean?
2 DR. CARPENTER: For instance, a year can start
3 on January 1 at midnight, it can start on January 1
at 1 a.m., 2 a.m., 3 a.m. There are 8760 possible startinc
5 points in every year for a one-year period.
6 The standards do not specify calendar year,
7 fiscal year, a year starting on April 30 at 3 in the
morning. Standards only specify a one-year period.
By implication, therefore, thestandards must address
all one-year periods. That's what the probabilistic
method that we've employed does, as contrasted with
what was called the simple binomial which simply looked
at ten calendar year periods.
f
.. There are many, many more possible one-year
periods, and what you do is you average them altogether.
.., MR. TIKVART: One other question. Do I-under-
lo
17 stand from your last slide that you did not consider
model uncertainty in your results, but were only looking
19 at the uncertainty of a violation probability?
20 DR. CARPENTER: Yes, what we were looking
at was the uncertainty in source terms, source emissions,
background concentrations and meteorology. We did not
specifically consider model uncertainty. The reason
we didn't was we weren't able to quantify it. The tech-
nique is able to address model uncertainty if someone
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
38
»
can sit down and say what the bounds are or how much
variability there is around any one prediction. That
could then be entered into the technique and which show
up in the final result which says the probability of
violation is X percent.
.
MR. TIKVART: Have you written some reports
that we could see and study?
DR. CARPENTER: Yes, I'm not sure when they're
going to be out. Do you know?.
DR. JONGEDYK; The draft copies have been
fairly well distributed, including your office.
MR. TIKVART: Okay, I'll have to go back and
look.
r
DR. JONGEDYK: But I might introduce one little
topic here, and that is the basic probabilistic approach
that he .is dealing with, the probabilities of the^ various
inputs,could obviously be applied to any type of a highway
or any other type of model. It's "sort of a conceptual
idea involved here.
/
DR. CARPENTER: That is a good point. The
concept is appropriate for any type of source. We've
simply applied it to highway sources, because Federal
Highways was paying for the research.
/
-MR. TIKVART: Go ahead.
MR. ARTICOLA: I'm Bill Articola of REOTEC.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
39
You appear to be using a Monte Carlo method
in this kind of approach that'5 assuming that the model
is basically okay and that you're going to come up with
an area distribution for your various inputs, your back-
ground concentration or your emission levels.
What kind of thought have you given to the
type of area distribution these various inputs would
have? For instance, would they be flat, Gaussian distri-
buted, and I guess the corollary to that question is
10 why in particular did you choose a 10 percent index
ffcr your random perterbations? Why not 15 percent,
12 for instance, or five percent?
13 DR. CARPENTER: That was not to reflect reality.
f
That was simply to make an example, something that because
we would know exactly what we were putting in, and then
know what we were getting back out again.
<*
- The ways that we address the uncertainties
18
19
20
21
22
23
24
25
are as follows. The meteorological uncertainty, we
assume is taken care of, by an"actual ten-year history
of meteorology, so we don't perturb the meteorology.
We assume that the ten-year history is sufficient to
represent its population.
We assume again that the FTP emission factors
are appropriate, largely because they are virtually
legislated, so we don't touch them. We then take for
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
14
15
16
17
18
19.
20
21
22
23
24
25
40
*
every hour of the year, we have an expected traffic
volume, expected. From traffic theory, you can go to
the Poisson distribution which can be represented for
large numbers by a normal distribution, and what we
say is that expected value, there's a distribution about
that expected value, with a variance equal to the expected
value.
So we calculate a true traffic volume. Then
we look at the speed traffic volume relationship to D
calculate a true speed from the initial speed limit
of 55 miles-per hour. This then gives us a final emission
>
factor based on traffic volume and the traffic speed
for the different vehicle types.
. 1. . As far as background is concerned, what we
do I can't remember what it's called, SAROAD thanks,
somebody knows what I mean has data on. vario.us»sites
around the country giving means and variances.of pollutiojv,
levels, and we simply use these in the model, so the
plus or minus ten percent uniform ddstribution that
I used in the examples was simply for the purpose of
example only.
MR. ARTICOLA: So essentially you're finding
your nearest meteorological station, taking ten years
of history, compress it into one year, seme up with
your distribution from there, and you're using whatever
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 ' WASHINGTON, D.C 20005
-------
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
data you've got for CO monitoring stations where your
particular site is a background.
DR. CARPENTER: There is no compressing into
one year unless I misunderstood you.
MR. ARTICOLA: I have the impression that
you're taking a ten-year history, and you're looking
at the probability of say Class Ff one meter per second,
et cetera, et cetera, over that ten-year history and
then come up with a distribution that should be applied
for one year?
DR. CARPENTER: No, there is tio distribution.
. y
It's a straight simulation, and what you do is you end
up with in-the neighborhood of 100,000 one year one-
f
hour observations. It's a string of zeroes and ones
and you look at this string of zeroes and ones, and
you look at each set of 8760 of them, and you" mea'sure
your probabilities that way. ' --~
MR. ARTICOLA: Thank you. ~"
MR. TIKVART: Pleaseidentify yourself.
MR. WITTEN: My name is Alan Witten. I'm
from Oak Ridge National Laboratory. Now, to generate
your random source terms, you're taking source terms
that are randomly but uniformly distributed within plus
or minus ten percent of the mean?
DR. CARPENTER: No, this has apparently led
NEAL R. GROSS "
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18.
19
.20
21
22
23
24
25
42
to some confusion."" in the example that I presented
for CO, that was done simply for the sake of simplicity.
In practice, what you do is you have an for any hour
of the year, the traffic engineers in the Highway Depart-
ment should be able to give you the expected traffic
volume on a link. What you do is you make what's the
standard transportation Poisson assumption, and that
is that the distribution of the number of vehicles on
the road during that one-hour period will have the average
given by the numbers that they gave you, and the standard
deviation equal to the square root of that average.
The Poisson can be approximately by the normal
distribution.,for large values, hundreds, two hundreds,
a £housand vehicles or so, and so what you do is use
the normal approximations to the Poisson distribution
to come up with a true a predicted traffic.volume
for that length for that time. . ^.^
Then you take the traffic volume speed relation-
ship to calculate the true speed relative to the posted
speed limit,- you take that true speed, your true traffic
volume and go and find an emission factors.
MR. WITTEN: Okay, so. that's not really distri-
bution independent. It's dependent on the distribution
you assume back at the beginning?
DR. CARPENTER: That is true.,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3 Are the emission uncorrelated from one instance to the
4 next?
_
o
6 out to be highly correlated. That auto correlation
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
24
25
43
MR. WITTEN: Now, have you looked at auto
correlation in emissions? Is there any assumption there?
DR. CARPENTER: Oh, no, in fact, they turn
is taken into account by the deterministic pattern that
the traffic engineer has given you which shows the traffic
pattern or the expected traffic pattern on an hour by 3
hour basis. They're highly correlated.
MR. RHOADS: When you evaluate a highway segment,
you consider not only its impact on air quality, but
also on safety, traffic flew, the effect on the entire
r
network. . .
Have you attempted or do you intend to apply
these concepts for evaluation of some of those-other
factors, or do these concepts at the moment apply only
to air quality impacts? .,:
DR. CARPENTER: I haven't, and I haven't found
somebody whti would like to sponsor the research.
MR. RHOADS: Understood.
MR. TIKVART: Okay, thank you. Next speaking
for the National Oceanic and Atmospheric Administration,
we have two speakers, Roland Draxler and Jerome Heffter.
Dr. Draxler?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 2000S
-------
1
2
4
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
44
DR. DRAXLER: My name is Roland Draxler. I
work for the Air Resources Laboratories which is part
of the Environmental Research Laboratories which is
part of National Oceanic Atmospheric Administration.
5 It used to be called the Weather Bureau.
Our research is funded by the Department of
Energy, at least our group. Our group in the Air "Resource.-
Laboratories in Silver Spring, Maryland is conducting
research in long-range transport of atmospheric pollutants
Research efforts are divided between the develop
ment of numerical models and the design* and execution
of atmospheric transport and dispersion experiments
to verify the modeling studies. The transport distances
over which these models are applied range from the hundred;
to thousands of kilometers.
A long-range transport and dispersion model
that was developed by ARL and is frequently used by ^=
various researchers, calculates the pollutant transport
from archived meterological data. The pollutant is
assumed to be uniformly mixed in the layer above the
ground that may vary from hundreds to thousands of meters.
The pollutant is transported In this layer
following the average wind flow. The pollutant grows
in horizontal extent with increasing distance from the
source.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
*
..-'"' . .45
*
From some limited experimental data collected
during the past years, we have found that this model
may overpredict pollutant air concentrations. One parti-
cular experiment where an inert gas was released from
Idaho and sampled 2000 kilometers downwind in the mid-
western United States, the dispersion model calculated
concentrations five times greater than what was measured.
These model overpredictions are attributed
to much larger horizontal growth of the pollutant due
to wind shear and mixing of the pollutant out of the
layer near the ground into the upper troposphere.
wind speed and direction change rapidly with height
near the ground and the assumption that the pollutant
transport can be represented by a single average wind
in this layer may not be appropriate.
Model calculations with the pollutant distributee
through several layers reduced the excess calculated
concentrations to only twice the measured.
Further reductions in air concentrations are
expected-from pollutant mixing futher aloft into layers
not simulated in the model. Both of these processes
are probably enhanced over areas of.mountainous terrain.
These problems are currently under investigation and
a revised versions of the long-range dispersion model
will account for the effects of enhanced mixing by
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW .
-------
1
,T_ 46
*
permitting multilayer transport and dispersion calcula-
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
tions. Thank you.
**
MR. TIKVART: Thank you. Why don't we proceed
O
with Mr. Heffter before we take any questions, and then
4
we can ask questions of both of you?
5
MR. HEFFTER: I'm Nick Heffter from the Air
6
Resources Laboratories, NOAA, and Il!ll be speaking as
7
the second half of this presentation on tracer experiments
for verifying long-range air pollution models.
Concern over air pollution on regional and
international-scales has led to the development of long-
^
range atmospheric transport and dispersion models such
as those just discussed by Dr. Draxler.
'-- -Experimental verification of these models
is essential to determine the accuracy of assessments
based on the model, calculations. *
To meet this need, several long-range model--
verification experiments have been completed or aore
now being planned. A 2% year experiment was conducted
t
at the Savannah River Plan, South Carolina, starting
In march 1975.
The experiment was a joint Air REsources Lab-
Savannah River Lab project in which weekly and twice
daily concentrations of a noble gas enitited during normal
operations were measxired at 13 sampling sites (the red
NEAL R. GROSS "
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
. 47
dots on the diagram) from 30 to 140 kilometers surround-
2 ing the Savannah River Plant.
3 A model verification workshop sponsored by
the Department of Energy was held in November 1980
in which models developed by several national labora-
tories and other government and private agencies were
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
tested against these data.
A continuation of this experiment is now being
planned which will extend the range of sampling to 600
kilometers, ACURATE (an acronym for Atlantic Coast
Unique Atmospheric Tracer Experiment) will take place
over a 12-month period starting January of 1982.
Four sampling sites on an arc 600 kilometers
r
northeast of the Savannah River Plant will take twice
daily samples. During the experiment period, routine
and special meteorological data over the entire ACURATE
area will be archived, and" about 3000 twice-daily samples
will be collected and analyzed.
Npw the Air Resources Lab has recently developed
a new system for long-range verification.-)studies using
perfluorocarbon tracers and utilizing automatic sequential
samplers which provide rapid, inexpensive tracer analyses
down to -parts per ten to the 15th. ;
The capabilities of the systems were success-
fully demonstratoa in a 600 kilometer -experiment during
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 wt*M'vr-*^v "ir
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. I f 4.8
July 1980 in which two perfluorocarbon tracers were
released simultaneously in Oklahoma. Samplers were
located along arcs at 100 and 600 kilometers to the
north and northeast of the release point at Norman.
. Deployment of many samplers over a large area
in a long-range experiment can be very costly and present
difficult logistics problems. One of the objectives
of this experiment was to test the feasibility of using
/
the National Weather Service substation network numbering
over 12,000 sites.
Samplers were operated at 38 selected sites
on a 600 kilometer arc by cooperative observers who
take routine temperature and precipitation measurements.
9>
The observers carried out their assigned role with compe-
tence and enthusiasm. With the cooperation of the Nationa
Weather Service, future long-range tracer experiments
.will take advantage of the sampling capability inherent
in this substation network;. "
We.have also begun planning for a major long-
range dispersion study called CAPTEX (Bross-Appalachian
Perfluorocarbon Tracer Experiment) involving six separate
tracer releases in the Ohio Valley during August and
September of 1982.
This location was chosen because it1 s a major
pollutant source area affecting air quality in the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
OC 700O5
-------
Northeast. It is also suspected that effluents from
the Ohio Valley contribute to acid rain affecting many
lakes in the U.S. and Canada.
4 Tracer concentrations will be measured from
Ohio to the East Coast at 80 sites located from 300
to 1200 kilometers from the release point. The influence
of the Appalachians on transport and dispersion of pollu-
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
tants is of particular interest.
In conclusiori~a~ll of the experiments just
described should provide tracer data to improve and
verify long-range pollution models currently used in
environmental studies.
MR.'TIKVART: Do either you or Dr. Draxler
r
have any comments on what sort of. statement of the accu-
racy of models is most appropriate. It seems like at
this time most researchers have their own preferred
method of stating accuracy/ a'nd there is not a great
deal of consistency in doing that.
Do you have any recommendations on how to
1 .
go about stating how accurate a model is?
MR. HEFFTER: I don't personally. I think
the workshop, the DOE workshop that took place in November
will go into the answer to that. The paper is in draft
form now,, and I really don't think that I'd want to
say anything more until that comes out. It does deal
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
2,14-4413 Mf»«w«vr»r>v OC
-------
50
exactly with this problem of how to make statements
2 about accuracy.
3 MR. TIKVART: Okay, the workshop you're referring
to is that that was conducted at Hilton Head in November
5 of 1980?
6 MR. HEFFTER: Yes, that's right.
7 MR. TIKVART: Anybody else, any questions
of Dr. Draxler or Mr. Heffter? Thank you
The next speak is John Goll for the U.S. Geo-
logical Survey.
MR. GOLL: Good morning. My name is John >
Goll, and I'm with the U.S. Geological Survey's Conservation
Division, National Center Mailstop, 640 Reston, Virginia,
22092.
15 -Before I begin, I just want to mention, we
1C have a small hand-out that we left on tiie table in the
ID
lobby that has references to a few of the sites that
*>
I'll be giving, and also names, addresses and phone
19 numbers if any of you are interested in contacting us.
To begin with, the U.S. Geological Survey
, is within the Department of Interior and is responsible
22
23
24
25
for reglating the mineral activities on the outer conti-
nental shelves of the United States. Primarily in the
past, this has been oil and gas development, and the
Survey has been doing this for several decades, but
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS _
1330 VERMONT AVENUE. NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
fairly recently we received a new responsibility to
regulate area missions from OCS facilities. Our rules
on this became effective just a year ago, June 1980,
and, again, are referenced in the hand-out.
For those of you who are not familiar with
the OCS, the outer continental shelf, it's essentially
\
the area of the oceans three miles and beyond the coast-
lines of the U.S.
How far this goes out depends upon who you
talk to. Roughly, you might call it the zone of influence
of the United States out to 200 miles.
Briefly the rules we developed for our air
quality program are roughly patterned after those of
f
EPA. We were not trying to reinvent the wheel or to
go off in a totally different direction, and our sort
of mandate from Congress implied that we should contact
EPA and again not totally go'off in different directions,
so roughly we have screening procedures to separate
large and small sources and then, of course, controls.
The .-screening procedures do differ a little
bit from those of EPA in that we have a first screen
which is an adntssions screen based on the facilities
emissions, versus its distance from shore. Roughly,
*
the further you go out from shore, the more you are
allowed to enu : .
NEAL R. GROSS
\ COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
12031 234-4433 W»
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
>- . 52
Due to the location and size of most facilities,
most OCS sources over the first year of regulation have
been exempt from controls. Again, we're talking things
out beyond 20-30 miles from shore in most cases.
The second screen in the program is through
use air quality modeling to compare on-shore impacts
to what we term the significant on-shore concentrations.
The significant concentrations, we adopted
3
the significance levels of EPA's emission offset policy.
In the intent, Congress implied that we should not control
a facility until it caused some significant impact on >
shore.
v
Finally, if a facility is not exempt by either
r
the first or the second screen, then it would require
controls.,. again similar to those required by EPA.
, ».
The first reference on the sheet, again, fully
explains our regulatory program for those of you who
* " »»'
are more interested in looking into it a little further.
The second .screen as we noted 'required air quality model
ing, and as we were developing the rules, and as we
were implementing the rules, of course, we were faced
with two major problems.
One, there really is no suitable air quality
model for over water applications, that is in the
regulatory context, and, second of all, where many of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW -
(202) 234-4433 WASHINGTON, D.C 20005
-------
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
these facilities are data, meterological data can be
2 I! a problem to collect, especially existing data to find.
So to solve this problem, the Department took a three-
step approach. First of all, for an interim period
only, until better models could be developed, we allowed
f
the use of crster with some minor modifications to the
program.
Now the second site on our hand-out explains
the modifications and our views on the limitations of
using crster for overwater flow. We also encouraged
in this notice that any parties that were interested
in developing models to contact, us and also to submit
any overwater models for future consideration, because
r
we realized, of course, that crster was not the answer.
Second of all, to get better data and to find
out really what was going on for'over-water dispersion
and transport, the Department through the Bureau of
Land Management has been sponsoring over-water tracer
tests, primarily now off the coast of California.
The results of this"first series of tests
should be available later this year. In addition, inde-
dependently, the American Petroleum Institute has also
been conducting tests in the Gulf of Mexico, so hopefully
these and other tests that have been done beforehand
will give a little more light on how good or bad we
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
4
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
54
*-
are with regards to modeling, and also to provide data-
for any future modeling efforts.
The third point of the Department's program
was that the Geological Survey would help to develop
a suitable over-water model, again, for OCS facility
applications. Now, first of all, we wanted to make sure
that we were not duplicating anything that EPA was doing,
so we have been in contact with a Dr. Hank Cole of EPA
who has been working on" coastal problems and also is
developing a coastal version of crster.
The second point is that later this year we
hope to issue a contract to begin development of an
acceptable overwater model. Our first intent, if we
r
are able to do it, depending upon the data from the
tracer studies and also the availability of data on
the OCS would be to try to stay with something similar
**
to crster. .i -^
Now whether we take Dr. Cole's coastline version
of crster or have to go to another version, this would
i
be decided. Some of the main points, in considering
on the final approach would be well, through the
contractor we would hope to summarize available overwater
tracer data and the type of meteorological and oceano-
graphic data available today as input into the model.
Second of all, to determine the meteorological
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
_ 1330 VERMONT AVENUE. NW
(202) 234-4433 ~ " WASHINGTON, D.C 20005
-------
55
parameters of prime importance to describe on-shore
concentrations. Again, our facilities are three miles
and beyond from shore, and we're supposed to measure
impacts at the coastline, not over water.
Third, we would be providing guidance to this
contract on using the available data to describe the
meteorological parameters that we do cleem to be important
to describe these on-shore concentrations, and, finally,
we hope to provide guidance on the meteorological and
oceanographic data that should be collected off-shore
11 for use as input into air quality models.
12 Now as part of the model development through
13 this contract, we do plan to subject a. draft of the
,« model to peer review, so one purpose of my being up
here today is to ask any of you who are interested in
taking part in this review or wishing to receive'copies
of the draft to please contact us, and -we'll try to
18
19
20
21
22
23
24
25
maintain a list and send copies of the model out when
it .is available, the draft of the model..
Likely, this will not be until 1982. Again,
so that we're not duplicating EPA efforts, we would
hope that this model, if it is acceptable and of good
quality, will be included in EPA's guidelines and air
quality models in the future.-
That's my formal presentation, if there are
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
133O VERMONT AVENUE. NW -
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
13
14
15
16
17
18
19
20
21
22
23
24
25
56
any questions?
MR. TIKVART: Yes, does USGS have any plans
or programs to consider uncertainty in the model esti- .'
mates?
MR. GOLL: Through the contractor, we would
hope, right now for the crster use, we looked at different
conditions that you might get comparing what crster
would produce versus situations where you may be overprer-
dicting or underpredicting. Again, it was explained
in the Federal Register notice that I referenced to,
that certain conditions you may be overpredicting than
others.
However, we would probably be playing a game
that we would, wait until we received some of the tracer
data and that through the contractors really take a
look at where really we are. That would be considered
in the contract. ". "X
MR. TIKVART: Any questions?
MR. GOODIN: Bill Goodin, Deems and Moore
in Los Angeles. A photochemical-oxidant, I suspect,
may be of importance, considering the hydrocarbon emission
from some of the sources. Will you address that, or
strictly the pollutants that EPA is concerned with?
MR. GOLL: For this model at this time, we
were just going to be considering non-reactive pollutants.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
4
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
" '- - -v 57
We hope to again follow EPA's lead as far
as reactive modeling.
MR. GOCDIN: As far as what model you mean
you choose or .
MR. GOLL: Well, there are several approaches
that we can take to reactive modeling. One, the Bureau
of Land Management in their environmental impact state-
ments do include reactive modeling, and through that,
3
we can get an idea of where problem areas may exist
in the future.
We can depending on the scenario or the
data they've considered in that modeling, we can watch
in the future as new facilities go in. If that scenario
^
appears to be coming true, then we could take action.
Right now, for specific facility by facility
_ . . _ - i
review, again, the regulatory programs considers .VOC1s
in a certain way, but it does not consider air qualit
modeling for them at this time.
MR. SKLAREW: Ralph Sklarew, Form and Substance.
We're doing the photochemical modeling for the PLM,
and one thing really strikes me is that the photochemical
modeling is very worst case oriented, very trajectory
oriented. By its nature, it has to be a gridded model, .
and the results and methodology is very inconsistent
with the crster methodology. I've been wrestling with
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW .
'202' 234-4433 WASHINGTON D.C 20005
-------
1
2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
58.
that for a number of years, and I'm wondering if there's
any recognition and comments from the USGS point of
view?
MR. GOLL: There is a provision in the rules
for cumulative impacts, and by cumulative impacts, what
.
is the effect of multiple facilities on. a shoreline,
especially with regards to ozone.
The words in the regulations are not that
specific on what the GS would o'r would not do. We would
consider information, again, for example,, from the BLM
studies plus information from the affected state and
from the affected leasees, the oil facility operators.
I guess what I'm saying is that there is not a clearcut
yes or no way we would deal with that. It would have
I to.depend on the. information that was available at that
time. *
Again, for example, if we took locking at
the BLM scenarios, looking at that, knowing that photo-
chemical/modeling only gives indications, it is not
an exact science, that it indicates things are getting
worse or indicates that things are not so bad. We would
compare what went into that modeling with the reality
at the time, with how many facilities are actually out
there, and versus, also, what the situation is on-shore,
but we would get a relative, are our facilities causing
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW .
1-202) 234-4433 WASHINGTON. O.C 2OO05
-------
IA
13
14
15
'
lo
i*J
. .. "59
a problem or not causing a problem?
MR. SKLAREW: Speaking of multiple impacts,
are you planning to go on to an MPTER type approach
for looking at the number of sites at one time?
MR. GOLL: For this modeling?
o
. MR. SKLAREW: Yeah.
6
MR. GOLL: For the purpose of this contract/
we would be looking at a single source; however, we
are asking that the final model come out in modules
that hopefully could be applied to other models also.
MR. TIKVART: Anybody else?
MR. DESCAMPS: Val DesCamps, Region I, EPA.
Could you describe the sources? I had no idea they
r
were that kind of a problem.
MR. GOLL: The sources?
__ . _ . _ _.. i
MR. DESCAMPS: Yes.
MR. GOLL: Primarily you're talking about
power generators from most facilities. Gas turbines,
diesels, in, some. cases.
MR. DESCAMPS: That are on oil rigs?
MR. GOLL: Yes, they are, yes, perform explora
tion activities, but primarily from production and de-
velopment activities .
\
Depending upon the size of oil or whatever
you're looking at, ycu're looking for, you run from
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
714-443.1 WASHINGTON D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- , , 60
very small sources, very small facilities with a few
wells to some very large facilities with upto perhaps
50 wells or so on, with quite a lot of equipment.
If you are not pipelining the production on
shore, you might run into tanker emissions which could
cause a problem again with the VOC and the ozone problem.
Areas of sour gas, that is gas with a high sulphur con-
tent, you might run into SO2 problems, but for the most
part you're talking about NOx.'
MR. DESCAMPS: So you're not concerned with
say plumes? Are you considering the plume model or
anything of that nature? You don't figure that your
emissions will cause any sort of a plume reaching shore?
MR. GOLL: Yes, we would. Essentially I would
think the models would be applied mainly to facilities
within 20 miles of shore. It depends on again your
meteorological conditions. Some areas of the Atlantic
perhaps, a plume may never get on shore. Other areas
of the OCS,. it may go a majority of the time. It really
would depend on where you are.
MR. DESCAMPS: USGS is also involved in acid
rain and monitoring and so forth, so wfriat sort of an
organization is it developing to take care of all these ,-
atmospheric problems?
MR. GOLL: Okay. The acid rain is being done
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBES^
1330 VERMONT AVENUE, NW
'202' 234-4433 WASHINGTON D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.61
'
at a different division, in the Water Resources Division,
and to my understanding, they're more working on the
deposition side as part of the interagency cooperative
workgroup.
It1s independent of the group that we are
with. Our division is sort of a wierd part of the Geo-
logical Survey in that we're a regulatory division whereas
the rest of the survey is more research oriented.
MR. TIKVART: Thank you. One more question,
and then we'll take a break.
MR. VIERATH: Doug Vierath, General Electric
Company, Schenectady, New Yock. A follow-up to the
question-concerning perhaps a major installation off-
r
shore, the plume impact upon shore. To what extent
would terrain impacts be part of your model? Perhaps
you might elaborate a little bit -on the coastal model
as well.
MR. GOLL: Okay. Primaxily the terrain should
not be too much of a problem for what we have looked
i
at so far. Most releases are roughly 50 meters to 60
to 70 meters, somewhere in that area. You may get plume
rise under varying conditions of 50 to maybe 200 meters
or so on.
Again, remember, we're three miles off-shore
and beyond. Most of the plume will be down to ground
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
62 -
** i
level by the time we reach the shore line. There will
not be we're also concerned about, you know, as soon
3 as we hit the coastline.
4 Now most areas of the U.S., well, in the eastern
U.S., of course, do not have, you know, high topography.
Areas on the west coast and Alaska would.
MR. VIERATH: May I interject perhaps also
some Class I areas in the west? Does that get a little
sticky? . _ -
MR. GOLL: Again, the Class I area would have
to be-along-the coastline.
12 MR. TIKVART: Okay, thank you. We'll take
a break.now and begin at 11:45 with Earl Markee from
r .
the Nuclear Regulatory Commission.
15
16
17
18
19
20
21
22
23
24
25
(Whereupon, a brief recess was called.)
"MR; TIKVART: The next speaker will be Earl
Markee speaking for the Nuclear Regulatory Commission.-
Earl, would you come on up, and while Earl is coming
up, I would- also like to mention that on the list outside
t ' .
we had indicated Don Henderson would be speaking for
the Park Service. That is an 'error. Don will not speak.
So the first speaker now will be Earl Markee.
MR. MARKEE: I'm Earl Markee. I'm principal
meteorologist with the Nuclear Regulatory Commission
in the area of licensing. Today, I will present some
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
r I would like to focus on one round which I feel we have
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
63
'
NEC licensing experience which may be of some use to
the EPA in meeting their air quality modeling objectives,
Since EPA regulated source characteristics
are usually quite different from NRC regulated sources,
in common, and this is the acquisition of .meteorological
data to provide inputs to the models.
The models and the data inputted to the models
in my opinion are of equal importance in providing assess-
ments. If a model demands data which are not readily.
obtainable, the model is of little value in the regulatory
- -- _ "*
process.
In research, the data requirements are stringent
f
so "that fundamental processes can be documented and
defined. In the regulatory process, the capability
to paramaterize these fundamental processes-by data
*%
is needed to be feasible and cost effective. "
Basically, what is needed to drive air quality
models are parameters which describe atmospheric trans-
f
port direction, atmopsheric dilution and stmospheric
diffusion rates.
The basic questions which must be answered
on data requirements are what kind of measurements should
jbe made, whre should I make these measurements, and
how many measurements should I make.
NEAL R. GROSS "
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- 64
».
The kind of measurements to be made usually
boils down to measurements of wind direction, wind speed-
and an estimator of atmospheric stability.
Sometimes there is a measure of turbulence
or atmospheric fluctuations as in the case of the lateral
fluctuations of wind direction; however, the other ques-
tions are not answered as easily, namely where and how
many measurements should be made.
The where and how many questions are functions
of things such as source characteristics, complexity
of terrain and the models. For example, if we wish
to describe flow around an obstacle like was presented
here yesterday, in a presentation, is there any way that
r
we can parameterize from single point measurements where
the air will flow over a hill or around the hill.
Another example might be coastal terrain.
*a
Can we make-a reasonable assessment of the penetration
of sea breeze or lake breeze effects by a single measure-
ment source, or do we have to have several sources for
/
these measurements.
There is a reasonable concensus that on-site
meteorological measurements neart the source are needed
as was concluded in the workshop conducted by EPA in
January 1980 at Raleigh, and also is described in NRC
regulatory guide 1.23.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
' 24
25
65
i
If meteorological conditions are very radical -
temporally and spacially, we may have a problem. If
the site is located in homogeneous terrain where we
don't expect these variations, a simple model, and a
single set of onsite measurements may staff ice.
Therefore, we must consider tlaat we must
question ourselves as to whether temporal and spacial
and homogeneities affect the licensing and the regulatory
process.
This can be accomplished either by increasing
the data requirements or increasing the.modeling require-
ments. In some instances that we have in the NRC, which
are for annual average dispersion models, and so forth,
r
the relative importance of these homogeneities is much
less than the consideration in emergency situation where
we have an accidental release of effluent from a nuclear
facility and are trying to protect the public against
the conditions existing at that point in. time.
Another factor to be considered is as the
modeling arid/cr data requirements become more complex,
the cost goes up. Now cost effectiveness becomes a
consideration in the regulatory process.. For example,
a typical nuclear power reactor with a rrseteorological ;
program which complies with regulatory gruide 1.23 requires
a capital outlay somewhere between $300 and $500,000
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
'202^ 234-4433 WASHINGTON D.C 2000%
-------
5
power plant. The money spent on the meteorological
program amounts to much less than a tenth of a percent
8
9
10
11
12
13,
14
15
16
17
18
19
20
21
22
23
24
25
66
with annual costs for maintaining the equipment in the
range of $50 to $100,000.
However, the total cost for a single nuclear
power plant is approximately $1 billion, and there can
be severe consequences although remote from a nuclear
of the entire cost of the facility,and, therefore, has
been considered as diminimus.
However, if we take the other extreme in the
nuclear area which is a pharmaceutical firm which handles
radioisotopes, the cost of a regulatory guide 1.23 program
wc-uld represent more than 10 percent of the operating
costs of the plant.
However; with a pharmaceutical firm, the con-
>
sequences are much less, and the requirement for the
-
rigorousness of meteorological definition can be less.
»
Therefore, .in,the Nuclear Regulatory Commission, we
have taken 'the position that compliance with regulatory
guide 1.23 is only expected where it is cost effective
and needed to insure safety.
Other procedures can be used to insure safety
at the smaller facilities. For example, taking the
pharmaceutical firm again, we have made estimates of
what might be the worst relative dilution factor that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
5- * 67-
would occur anywhere outside of that facility and have
used a screening technique whereby we fiave found that
well over 90 percent of these facilities can meet those
criteria described by the Commission.
5 It's only in this very small fraction that
6 we would need to do additional studies to make that
definition and to make the finding of acceptability.
g I'd just like to close with these thoughts.
Data requirements should be optimized so that they require
»Q only the minimum data needed to perform the regulatory
jj evaluation. Direct measurements rather than parameter!-
12 zation of atmospheric conditions should be used with
jo caption since they are the cost of measurement is
., high in relation to ne effectiveness of the measurements.
15 . And, finally, there should be a .balance between
16
17
18
19
20
21
22
23
24
25
the modeling and data requirements. Thank you for allowinc
me to present these comments.
MR. RHOADS: Sir, we share another thing in
common. We' both represent regulatory agencies. We've
been talking about uncertainty in air quality analysis;
however, ou regulate for a lot of other criteria, tremors,
accidental release of radioactive material, et cetera.
Under the other criteria which you used for regulatory
purposes, have your people found any better way of handlinc
uncertainty than we have in EPA? That is, your data
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS _
1330 VERMONT AVENUE. NW ~
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
68
for earth tremors, for example, for accidental release
of radioactive material must have uncertainty in it
also. From that standpoint, do you have any advice
to us?
MR. MARKEE: . .Well,)'in the NRC, if the conse-
quences are very severe, we usually take the what
we call the conservative approach. If we were, for
instance, to consider, say, that we would have an incident
at a facility, we would take the measurements of the D
entire radioactivity in the containment at that time,
and predict the consequences of that, although we know
>
that only a very small fraction will become airborne
and get into the ambient atmosphere.
f
So this is the procedure that we use. I might
add that the 95 percentile criterion was mentioned.
We use this for analysis of what we call design- basis
o
accidents for a nuclear facility, which are the more .-.
severe class of accidents, and using that criterion,
this has a reasonable basis in.regulatory policy.
i
If we compare to those objectives that are
in our regulations a value based on the 95 percentile
level of meteorological conditions, we have conservatism
built into the source condition. As I mentioned before,
we've considered the entire build-up of inventory and
so forth.
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
/c
?. 6:9
* ' ':'
V However, the thing that we do have in our ' *
back pocket is that we know that if we use the 95 percent
g tile criterion, it is deminimis to assume that concentra-
tions in the air would be greater than an order of magni-
tude greater than the predicted value, and the reason
we are using the order of magnitude is because our regula-
tions are not based on they are based on obvious
injury to the public. They are not based on fatalities.
If you were to increase those by an order
of magnitude, you would approach the LD50 levels for
11
12
13
14
16
18
20
21
22
23
24
25
one criterion which is the only body do'se, so, therefore,
we have a safety factor, and we also have the conserva-
tism built-into the analysis, and this is how we arrived
r
at"that as being the reasonable factor by which to design
the plan.
MR. RHOADS: I see. You had mentioned *the
control of the pharmaceutical industry and that approxj
mately 90 perciir.t of the industry could comply wit'h, I
believe you said regulatory guideline 1.23. The remaining
10 percent became questionable, and you evaluate those
on a case by case basis as a 'criteria?
MR. MARKEE: lie, we use a maximum value, an
arbitrary maximum value based on experimental data.
The maximum relative concentration that could exist
anywhere outside of the facility. We did an evaluation
NEAL R. GROSS ''
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
to see whether the doses met the applicable guidelines.
We have short-term guidelines and./we have annual average
guidelines.
v_
If they could meet that, then we felt it was
acceptable, because this was the worst case consideration..
Howeverr, those 10 percent which failed, the screening
test would be r«2~ examined and incrementally, we would
have to apply more meteorological information, more
site specific meteorological information to that evalua-
tion to see if the maximum value was, indesd, reasonable
or whether the maximum value was not reasonable.
M.R. RHOADS: So essentially instead of explicitly
considering-uncertainty in those remaining 10 percent,
V
you're trying to reduce the uncertainty?
MR. MARKEE: That's correct.
. _ 3b
MR. TI'XVART: Anyone have questions for.Mr.
Markee? No? Thank you, Earl.
The next speaker will be Dr. Shull for the
Department oil Energy.
DR. SHULL: Thank you. Good morning. Roger
Shull, Office of Environmental Programs. U.S. Department
of Energy, Washington, D.C.
First of all, I'd like to compliment Joe here
on a well planned and well run program. I know it's
required by law, but you people acted like you really
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
-1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
1
-_wanted tcx^ do-it ..-^"I" think it1 s a useful" meet ing. I'
concurs with nearly everything that's gone on so far.
I'll give you a little thumbnail sketch of 20,000 people
called the Department of Energy and some of the concerns
we have with modeling and uncertainty.
DOE has several different functions, one of
which is general energy policy development. An example
of that is the recent energy policy plan that was released
Regulation, if you've been reading the papers, is a ^
declining function at DOE, because the current adminis-
tration believes that the market can handled many of
>
the problems that we've been trying to handle in a regula-
tory mode beforehand.
r - We also do a large amount of energy research
and development at the national labs across the country.
We have representatives from those sites here, .andiyou've
heard mention of some of those facilities from. some ______ ^
of the earlier representatives. . ..^
What some people may not realize and what
accounts for a lot of the number of employees in DOE
is that we are a very large manufacturing organization.
All of the fissionable uranium that's burned in civilian
nuclear reactors is concentrated in DOE gaseous diffusion
plants in Tennessee and Kentucky.
All of the plutonium that's used in military
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
72 t
I weapons is all produced at DOE facilities at the Savannah
2 River and Hanford, Washington.
3 Finally, we are the nation*s largest producer
4 of hydropower. We supervise five power marketing adminis-
5 trations around the country and actually generate several
6 billion dollars worth of federal revenue by selling
7 that hydropower.
Why is an agency like that interested in air
quality modeling? I guess it becomes fairly obviously.
First of all, we have these several large facilities
11 around the country, several dealing with nuclear facili-
12 ties, but they all have their own power facilities,
13 and other- types of activities that release air pollutants,
^
so we are, in fact, subject to EPA and state regulations,
and any kind of a tighter regulation translates, of
course, into a higher capital and operating budget for
our people which is a higher federal budget and on and
on.
Secondly, in our environmental policy role,
we'd like to be assured that if there are air quality
constraints on developing the nation's energy supplies
that those are, indeed, valid constraints and not just
an artifact of a mathematical model, conservatism.
Why are we interested in uncertainty? The
basic reason that even the best mathematical models
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
133O VERMONT AVENUE, NW .
1(202) 234-4433 WASHINGTON. D.C 20005
-------
73
that we have now have ranges of uncertainty that translate
into millions of dollars of control and compliance costs
2
variations.
3
These variations can be retrofits on existing
4
facilities, suboptimal siting patterns, requirements
5
for changing fuel sources, decreased ^reliability of
6
an energy system because of delays irn building new faci-
7
lities, forcing facilities to go to tine very highly
8
efficient air cleaning methods, and dlispersal of the
9
facilities which has a lot of diseconomies from other
10
points of view.
11
Some examples, we have been experimenting
12
with the~use of long-range transport nsodels in possibly
13
developing national control policies for long-range
14
transport effects, and we coupled some cost estimating
15
algorithms with some of these long-range transport models
16
and looked for optimal control cost solutions at using -
17
different values of transport, and we used the matrix
18
method for this.
19 ' . .
If we simply use July meteorology as opposed
20
to January meteorology, we can see tremendous shifts
21
for responsibility for the burden of controlling such
22
things .as fine particulate sulphates or SO2 precursors,
23
great shifts between regions of the country depending
24
on which particular set of meteorology we use.
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
74
So we would argue that before any kind of
13
14
15
16
17
18
19
20
21
22
23
24
25
thought be given to using models for long-range transport
control policies that we increase the range of certainty
o
on those.
5 Other reasons for concern here is that there
_ are many other factors affecting the siting of energy
facilities, besides just air quality. Any of you who
have written EIS's know that the regulations'you have
to consider just go on "arid on, dozens and dozens, water
quality, waste disposal, water resource laws, land use,
wildlife, archeological, those sorts of thiogs.
12 A simple example, the Intermountain Power
Project in-TJtah, back in the mid-seven ties, spent over
$2 million checking out groundwater sources in a likely
site area because there wasn't enough surface water
there, and they did, indeed, find adequate ground water,
but that .site was subsequently ruled out because of
a PSD judgment.
And, I don't believe.the company considered
i .
it necessary to go back and challenge that judgment,
but let me just say, supposing that the shift 30 miles
down the mountain to a different site was a result of
the model overpredicting on concentration. Just consider
the amount of increased cost that was placed on that
utility, and, of course, that comes out in your and
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
---^=. ^---^- ' 75 li
ray electric bill sooner or later. Another thing we
2 have "to consider is that there are many other factors
« besides clean air that are of interest to states, indus-
tries and citizens, and, of course, those are economic
growth, jobs, moderate energy costs. We used to think
of low energy costs. I think we're convinced that those
days are gone forever, but we still would like to hold
those down where possible, and, of course, economies
of scale in putting togeter energy supply facilities.
An erroneous forecast or overprediction of
an air quality concentration which would result in either
increased costs or preclusion of establishing a certain
13 facility there could easily deny these other benefits
to the citizens for no measurable benefit that would
15
16
17
18
19
20
21
22
23
24
25
14
have occurred otherwise, so we want to be aware of these.
. -. ^
I heard talk yesterday that there were high
stakes involved in these modeling activities. I hadn't
heard to date too many examples of that.
We believe if the air quality modeler properly
expresses the uncertainty involved in his forecast that
the state or regional policy makers can appropriately
consider these other state goals in the political context
which is his prerogative. .
For example, we may say that our prediction
is accurate within a factor of two. That means times
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
76
are divided-by two,-and-that can be a fairly large range.
For example, if a state administrator is in
2
a state that really has a bad unemployment problem,
3
he might assume, well, let me guess that the model was
4
really overpredicting, that it really did plus two of
5
what the real value is, so I'll allow the facility to
6
site there.
7
On the other hand, if he1 s in a state that
8
advertises its clean air-and long vistas and is into
9
tourism, he may assume I'll apply my ov/n safety factor,
10
I'll assume the model underpredicted so we won't let
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
the facility site there.
In that way, the burden and the judgment are
*
on his shoulders as a representative of the people,
not on the burden of the air quality modeler who is
far removed from responsibility and accountability to
»
the people. . -
So that's another reason why I'm glad that
we're addressing this topic here. That gets us into
safety factors, and uncertainty is one of the things
*
that brings us in the engineering profession to use
safety factors.
The Clean Air Act actually specifies, and
we said the ambient air quality standards that we have
an adequate margin of safety. I'm not intimately
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- »<»-«--*'«"--i"'..' * - -- * - *-- - "-» ,- .
familiar with how the PSD increments were set, but I
would assume that there were some either explicit or
implicit safety margins there also.
The question is how many safety factors do
we need, especially if they're all compounding and acting
in the same direction, and the next question is, who
should set those safety factors. Should it be the state
administrator, the Congress or the air quality modeler.
I won't answer that question.
Finally, how can uncertainty be reduced in
air quality modeling. The simple answer is money, just
more money will giveyou lots of improved certainty.
But what do we do with the money.
r
Based upon my experience both in modeling
and in dealing with policy matters, I would argue we
spend more money on site specific data rather than on
extensive improvements in the modeling mathematics,
because we alwrys have to demonstrate these things to
people who aren't air quality modelers, and I've heard
other people argue for simplicity, but we need to have
easily recognizable factors that people can help to
educate themselves.
So calibration of the model with the tracer
studies that our NOAA colleagues have talked about,
I think is something we should do a lot more of.
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
6
8
10
11
12
13
17
18
19
20
21
22
23
24
25
__ -"- 78
_ . . .^Something, else:-we!, need to do is utilize along ;
.*
with the air quality modelers who may be mathematicians
or computer specialists or whatever actual'field trained
meteorologists or climatologists. If the air quality
modeler says, well, I have a model that pretends that
the plume will do this when it hits the mountain, but
I'm not sure .that it really does that, we need to find
somebody who knows what happens to wind patterns in
that part of the country, and he can say, well, it does
do this, it doesn't do that, and those kind of people
working along with the modelers always tend to add credi-
bility to the results.
That concludes the DOE remarks. I m open
14 . to" quest ions.
MR. TIKVART: No questions from the panel.
Any from .the floor?
15
16
MR; SKLAREW: Ralph Sklarew, Form and Substance.
....'.-
One of the big arguments agains-t control is the long-
range transport, and since you. people would perhaps
t
be interested in the cost effectiveness of control,
perhaps you've looked at the trade-off of control cost
effectiveness versus long-range transport, cumulative
impacts.
DR. SHULL: I can't speak in detail on that. :.
That is one strategy that we have begun looking at again
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
3
4
5
6
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- f 79
in the past couple of years, but I don't know the details
2 of that.
MR. SKLAREW: Is there anyone in your organizatio
» would know.
that may be the individual?
DR. SHULL: Yes, if you'll see me after the
.«
meeting, I'll put you in contact with somebody that
MR. GOLL: John Goll, Geological Survey. Your
statement about leaving it upto the political types,
you know, when you have a situation where you might
be plus or minus two from what the modeling shows, that
leaves me a little uncomfortable, in that I would think
that we would loose credibility as modelers and as the
r
person giving that information if we didn't along in
these cases where you might be approaching a standard
or you might be approaching a situation where it w,as
a tough decision for the political person in charge.
I think in our position" one thing we should
do is at least give that decisionmaker pur best guess
of whether, in this circumstance the model is overpredicting
or under predicting rather than letting him really be
faced with the option, well, it could be, you know,
a factor of two under or a factor of two over.
Did I misunderstand what you had said?
DR. SHULL: No, I think I agree. My position
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
W-
20
21
22
23
24
25
80
-*
-was that the-modelers-shauld-explain-the-uncertainty 5
and his best guess. I assume the number, if you do
present a number as your best guess, that'you present
that and you present the uncertainty in your professional
judgment as best you can, but what I'm arguing is that
you don't argue that your best guess is the exact answer
and thereby playing a stronger role in a decision than
you should be.
MR. GOLL: As long as. we let him know what - 3
went into the decision, the numbers that we gave him.
DR. SHULL: Right, right, and, you may also
>
present him with what the likely effects would be at
the upper and lower bounds, but the comprehensive evalua-
tion of what that means in a societal sense is his decisior
and not yours.
MR. GOLL: I was just worried about situations
where you're very close to a cut-off rather than, well, .
- --i-»
if you can call anything a normal modeling situation.
^DR. SHULL: Right. Well, the ones that are
i
way above-and way below are never a problem. They're
sort of, a priori, and the model works, but you hardly
even need the model. It's the ones that are close to
the line that are always going to be controversial no
matter -how good the model is.
MR. STRAITIFF: Dan Straitiffr. Standard Oil
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
2
3
4
5
6
7
8
10
11
12
13
14 -
15
16
.17
18
19
20
21
22
23
24
25
-^-^ ~" - 81
of Ohitq.: D.QJE .i.s^a.majiQ^^participant in-.the...shale oil
development in Region 8 which is an area- of very complex
terrain. Given the undertainties of complex terrain
modeling, can you offer any suggestions to Region 8
as to how they should overcome these uncertainties in
allowing these projects to go through.
Your suggestion of perhaps more money is not
maybe an immediate enough solution. Permits are going
forward now. .
DR. SHULL: I guess I would have to think
on that, before I make a pronouncement too much, but
I think the tracer studies, of course, that would take
time. Perhaps tetroons, the radiosondes, those kinds
of'things so that we actually learn more about the wind
patterns in those areas, or perhaps taught some local
hunters to- know how the winds blow around there.
I mean get the real facts rather than trying
to build a better model of air -to flow around those
mountains.
MR. STRAITIFF: But other factors, socio-
economic, you're going to weigh those as important or
not important. I know this isn't your decision. You're
kind of on the other side of the fence, but I'm just
looking for information. =
DR. SHULL: Well, I believe those factors
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.-.= -- 82
'are be-ing Iweighed. in-Lthjeiidecjisions;, and :the.y' re brought
up in the public hearings, and if the uncertainty in
the models are brought out, and the uncertainties are
quite high there, the decisions' may fall in the correct
place, wherever that is.
MR. TIKVART: Thank you very much. Next we
have Mr. Moe for the State of Texas, State Department
of Highways.
DR. MOE: Thank you. I'm Rod Moe, State Depart-
ment of Highways and Public Transportation D-8, Austin,
Texas, 78701.
Well, after sitting here for a day and a half
and listening to everybody say that wind speed and wind
directions, stochastic process, it's random in nature,
and that's all that ever occurs, I'm going to try to
show that it's not always random.
. - 1
»»
-Would\ you turn on the slide projector please,
-----*«ajS
someone, and get the lights please? Now, for about
*
seven years, we've been working on model validation
studies on line- source models at the Texas A&M University
in College Station, Texas, and we've established a data
.base with about 400 hours of data, screened data, and
we have monitored at six different sites, in four major
cities in Texas. ;
When we examined this data base, we noticed
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
:l£
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
' - 83 ,
. t,:
thatany, -timefrwa, fradVnegative; wind "shear*, :-we: had high" ^ '
CO levels. The models predicted well about what our -
carbon dioxide levels would be, at least reasonably
well, ballpark estimates, except when we had negative
wind shear. By negative wind shear, I mean a decrease
of wind with height somewhere within our tower structure,
so we-have a tall tower here at 30 meters, and we have
four different levels of weather instruments on the
way up.
We found that if there was a decrease in wind
between any of these sets of instruments that we had
fairly high carbon monoxide levels. When we tried to
plot these carbon monoxide, wind speed and wind direction,
we'll show you what kind of graphs we came up with.
May I have the next slide please?
Thos 'shows our monitoring tracer. Next slide
v>
please, and this is the view, toward the east, next slide,
a view toward the west. It gives you an idea of what
kind of obstructions.
'
This is an at grade site about one mile west
of 1-45 on north loop 610 in Houston, Texas. Traffic
volume about 60,000. Next slide please. This is an
overall view of the experimental site. It shows a tall
tower, and we have instruments mountedon telephone poles'
and some of the instruments up the sidestreeet there,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
4
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
84
_ _ .... .* - .
andI we^"ra"isb^have7""iTrstrlOTentsToh"both"s±des' of the'street.
for monitoring carbon monoxide at 12 different locations.
Next slide please. We plotted this this
is an example of several carbon monoxide tracers. This
is five-minute average data in the afternoon. Notice,
we have almost 8 parts per million of CO under fairly
high windspeed conditions. Next slide please.
What we did, we plotted a five-minute average
windspeed data, and you-notice that only at one of these
particular data sets where we have it cross hatched
do we have -a -negative windshear.
Next slide please. This shows the traffic,
traffic volume and speed. You notice that at about
the same time-we had the negative windshear we also
had a cessation of traffic in the westbound direction.
Next slide please.
This overlays several of these at the same .._
time, and you'll note that the high concentration of
carbon monoxide occurs at the same time as a negative
windshear and low volume of traffic. Next slide please.
This is a plot of the carbon monoxide on 30
second time*-increments. It's a plot of several of the
carbon monoxide monitors. Notice that the peak CO con-
centration was 16 parts per million, a modeling meteoro-
logically the same conditions, you get about 1.5 parts
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
85
- 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
per mi±I^io^-usi:iig.j:.s^^^KS^MP<^&l5t^so,jiff:get an, prder.. . ..
i
of magnitude increase in peak CO for a short time period
as compared with what the model would say for one hour.'
Next slide please.
Now this is a plot of the windspeed, so what
I've done here is shade a negative windshear, so the
red areas are areas of negative wind§hear. You notice
the periodic pattern. Also notice that it's 16:15.
There seems to be an interruption of the pattern. Next
slide please.
This shows a temperature trace for the same
time period. You'll notice the yellow areas are inver-.;
sions. The green vertical lines are strong downdrafts.
By strong downdrafts, I mean say more than about .4
meters per second.
Note that the downdraft at a 9 meter height
*»
at 16:14:34 is five miles an hour which is a pretty
strong vertical wind.. Next slide please.
Okay. This matches .the windspeed with the
*
temperature trace. You notice the wind inversion plume
occurs at 16:14^ or so, well, about 16:15, shortly after
the strongest downdraft in this period.
Now this is a short period. We're talking
about five or ten minutes here, but notice periodic :.
pattern and also the updrafts and the change in
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
1.9
20
21
22
23
24
25
? -"^.wwte 86
temperature... ^-Nex.t s 1-ide^._pj.-ease.; .This...matches the carbon i
monoxide with the temperature in the wind. You notice
that where the negative shear is heaviest is where the
carbon monoxide is highest, and that at the time you
get the downdraft and the warm temperature plume, you
also get a sudden decrease in carbon monoxide. Next
slide please.
Now we looked at what happened to the wind
direction at the tower, the same time as we got the .,
strong downdraft, and these are the changes we noted.
What you've got is a stronger change, greater change
>
in the" wind direction towards the bottom of the-tower,
the smaller change as you go up. But this is a very
sudden change and could possibly be caused by a plume
that came off "the roadway, and this was air that was
filling in behind the plume. _
**
.Remember, this tower is about 70 feet to the
side of the roadway. Next slide please.
Okay. This shows acoustic sounder trace with
i
some strong shear echoes on the lefthand side there.
Next slide please. These are from aero environment
when they're catalogued, but you notice the shear echoes
and their closeness to the ground on the lefthand side
of the top trace and the ways on the lower trace there
on the lefthand side. These are early morning areas
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
-l!i- that^wel'rr^^taikangBral/biitX'^Next -slide-please. This""~' " ->''
2_ shows the gradual increase in the inversion height, ' -
2 and the ways forming along that inversion with top of
4 fog, shear echoes on the lefthand side.
5 Next slide please. This shows way forms aloft
6 or representations of that on an acoustics sounder.
7 Next slide. .
8 Now this is an acoustics sounds trace super-
g impsoed microbarograph _traces. Microbarograph trace
1Q superimposed on the righthand side of the diagram is
11 the dark line, and then there are a couple of other
12 traces of microbarograph complex that show different
13 time periods.
14 r - Notice that evidently at the same time these
waves are occurring, there's a big fluctuation in baro-
metric pressure. Next slide please. Now this is an
interesting slide. It's the only slide I have of radar
traces, of gravity waves and Kelvin Helmholtz instabili-
ties.
You'll notice the top trace up there with
f r
two sets of regular waves. They're identical as far
as period frequency goes. The middle slide shows the
same picture at a lower gain setting. You notice that
the lower gain setting you see there, pattern of Kelvin- -.
Helmholtz instabilities that are obscured in the top
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
1(202) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
photo..c,^The_ bottom-traee^hpws a JK-H: instability with '.
breaking waves. The vertical streak could be what they
call turnover, you know, in a gravity wave or K-H insta-
bility.
Now, these waves can have the cusps up or
the cusps down, and this may be a matter of what kind
of shear you have at the time with a positive or negative
shear. What you can get for that bottom trace is possibly
what they call turnover or intermittent turbulence.
You get wave formation, very stable conditions, form,
and then every few minutes, you get turnover and you
get intermittent turbulence, and under these peculiar ' .
conditions, this is the sort of thing that seems to
occur.
Now, a lot of these is not well understood.
The energetics exactly what.happens when you have turnover
»
«»
or whether it really is turnover, it's very difficult
-
to measure- these phenomena. You can take pictures of
them pretty well with radar and things like that, but
try to understand some of the relationships is something
the meteorologists haven't been able to fathom too well
so far, but these may be features that may need to be
taken into consideration. Next slide please.
Okay. Here's the second case. Now this is
an early morning case. I think this shows the pattern
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
89
that" you 'ftave .:high 'con-
centrations of CO in the early period there, and gradual
decrease, very slow decrease in carbon dioxide with
time. Next slide please.
Okay. This is the windspeed again showing
the shear. The shaded areas are shear. Notice that
until 8:00 you've got heavy shear. From about 8 to
8:06, you have what you might call moderate shear start-
ing to break up, and then after 8:06, the shear virtually
disappears.
The strongest downdraft is 2.7 miles per hour
at 8:00, and the downdrafts seem to be related to the
weakness in the shear. Next slide please.
* This shows the thermal pattern for the same
time period. Notice that temperatures changes at three
or four levels" at the same time. Notice that the down-
drafts seem to be related to the colder temperatures.
Next slide please.
This is an overlay of the two, windspeed and
temperature. Notice you get a thermal plume at 8:00.
Next slide please. Now this is an attempt to integrate
the negative shear, to try to quantify it in terms of
each minute of negative shear, and that's the solid
line. The dashed line is the 3 point moving average :.
of that. Next slide please.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
14
15
16
17
18
19
20
21
22
23
24
25
90
i °
>
1 .( - - r -.".^.-r*:=.. I^hi'sv-p^erlay^t^he^'CjaTbon^monoxixle' levels -on .
the integrated negative shear, and there does seem to
be some relationship. If you average the CO levels,
it would pretty close to the integrated shear. Next
slide please. Okay. You may say what does five minutes,
ten minutes or 20 minutes even make in terms of modeling
ifyou're trying to model for aone-hour average or some-
g thing like that.
9 Okay. Here's a case where the same sort of
IQ thing happened for about an.hour and a half, early morning
11 case. It lasted most of the morning. The highest con-
12 centration of carbon monoxide if about 26 parts, per
13 million which is about 10 times the predicted caline
2 concentration for one hour.
Order of magnitude increase. Notice that the
highest concentration of carbon monoxide occurred at
_ . i
three different CO levels at the same instant of time
almost, reven at the 30 meter height, and notice the
relationship. The two are plotted there all the way
through. It's a pretty close relationship in terms
of peaks. Next slide please.
Okay. This shows a relationship of carbon
dioxide at the different locations. You take 4H, that's
the top of the 30 meter tower. You compare the solid
line with the dashed line in the 4-H, and I think
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 . WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
91
^--^-^r-!.'~-~: ;- rr^r?::* }
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
" 92
is ttie."third "day'' "af teY-cbrd'Tfr'orit "passage?"when the high ,
cell was over the Appalachians, and we were in the south-
west corner of the high, you might say, at an east wind
blowing parallel to the roadway, and you start out with
freezing temperatures and increase quite a bit as you
can see here.
Now what we did, we related the carbon monoxide,
the wind shear, weakness in the wind shear, the down-
drafts and the cold temperature stucture to eight sets
of ways all within a 15-minute period.
Now, lights please. I think that study of
*
conditions like this admittedly we don't know how certain
they are, how often they happen, because we haven't
observed that many cases.
In our data base, you've probably got 200
short-term cases, and this is probably the only long-
term case. We monitored this sort of thing about one
percent of the time, and we monitored about one percent
of the total year, you might say 100th of one percent.
We did no-monitoring at night. We monitored only in
the daytim. We don't know how often this thing happens
at night.
There's been a lot of work with radars, sonars,
lidars, weather balloons in an effort to try to measure :.
these things, but a great deal more needs to be done.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 ' WASHINGTON. D.C 20005
-------
2
3
4
5
6
7
8
9
10
12
13
14
15
16
. .- *
93
18
1.9
20
21
22
23
24
25
jy than what our measurements show.
-I think -there: -needs %o~be-^a "lot- more -research. I think *
it's important to consider in terms of modeling as a
factor, to my knowledge, it's not been incorporated
into the models. I think it's important for monitoring,
that when we monitoring, we don't just monitor at ten
meters as far as windspeed goes.
If we need to check our windspeed at at least
two levels, we need to incorporate vertical windspeed
measurements, and I think AMS is considering a Bulk, 3
Richardson modulus where you monitor at 1.5 and ten
.. meters.
%
I think if you're going to do a model validation
study, you need to monitor windspeed at higher levels
as^.well, and you should measure temperature probably
more often than once a minute, because I think the temp-
erature structure is probably a good bit more details
I think we need, like Ned Meyers said, to
-.
study the fine scale data. I think it's extremely import-
ant. Now we save all our raw data, and I think in studies
like this it's important not to average our the data
to maintain that, but try to make enough detailed mea-
surements, so you can go back and reconstruct these things
for your data base.
Any questions?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(207) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
._ ' 94"
'MR'. TiKVARTT' Questions from "t"Ke audience? ".
MR. HAYNES: Eldewins Haynes, North Carolina' -
Division of Environmental Management. Was there any
relation, did you check these stability classes during
these periods?
DR. MOE: The stability classes were C and
D.
. MR. HAYNES: C and D?
DR. MOE: C and-D, so we're indicating neutral
or unstable, slightly unstable classes where the actual
stability-was-about F, you know, in terms of modeling.
MR. HAYNES: It .was F,
DR. MOE: The Pasquill Gifford class about
C or D.
MR. HAYNES: Okay, and this was also during
the periods of the negative wind shear?
DR. .MOE: Right, so.the Pasquiill Gifford
stability class doesn't work for this situation. In
fact, a similarity theory didn't work, the wave structure
and all that.
MR. HAYNES: Sort of -going through all the
detailed measurements that you had to go through, do
you have any suggestion of how we could make this opera-
tional?
DR. MOE: I don't know for sure on the negative
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
95
- ' .. -
. . ..- - --"-'"'",-". " '"*'
-~windshear>-*~ YouHcnowy-'that1 s pretty easy "to check' for.
What we did, we took a raw data, and we averaged five
minutes, 15 minute and 60 minute averages. If you look
at five minute averages, and you see you have negative
windshear by having a plot that shows what the windspeed
is for the different heights, then you can narrow this
thing down.
I look, of course, mainly for areas of high
carbon monoxide, the same time as you're looking for- D
that negative windshear. That narrows it down, and
-then once you narrow it down, you can use a fairly
>
versatile plotter, like say versatex or something like
that and try to plot your raw data, instead of doing
it-by hand like I did. That one graph took several
thousand points, and it's tedious, you know, but you
can do it with computers too. _ >
n.
The trouble with computers is you crunch so ~,
* «v--
many numbers, you crunch all the good ; information ..out
of it sometimes.
/
MR. TIKVART: One more question, Fred, and
then we'll have to move on.
MR. HAMBURG: My name is Fred Hamburg. I'm
with Radiation Management Corporation. I just wanted
to get straight in my mind the purpose behind this pre-
sentation.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
"**
"""""""ft Ts °oBvi6us"ly in ~the nature of a professional
paper, and as member of the TT3 committee of APCA, I ' -
would welcome a paper of this sort at our forthcoming
meeting whether it be in New Orleans next year or the
one precedigint that at San Antonio*.
Nevertheless, the question that came up in
my mind is why are we seeing this, and I thought about
it, and perhaps I have the answer, and I'd like you
to comment on this. _ ._
DR._MOEr Okay. I wanted to show if you think
you know all about what's going on out in the atmosphere,
you're wrong. I also wanted to.show that you can't just
summarize your data by applying statistics to a mass
of'numbers and come up with all the answers. I think
these are critical points. I think it's important to
go back and look at the detailed data.
MR. TIKVART: Thank you very much. Next we
have Mr. Misra representing the" Ontario Ministry of
Environment.
MR. MISRA: Thank you, Mr. Chairman. My name
is P. K. Misra, and I'm representing the Ontario Ministry
of Environment in Canada. My talk will deal with what
air quality models predicts and what we measure, what
the observations are, and why there are uncertainties
in models.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2 some of these uncertainties that exist in models. Dr.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
;< ' .97
*i^-
I1 ll-also-ta-lk-briefdy about how to quantify
Venkatrara a quarter of this paper that I am pre-
senting. May I have the first viewgraph please?
Some of the things that I want'to describe
right now are probably familiar to you, but nonetheless
I want to work through it just for completeness. What
I have .here is to show you the difference between per-
forming an experiment on laminar flow and a program
flow.
If you have two flow experiments under identical
conditions, and the laminar flow or variable flow will
be the same for both experiments. You do not expect
them to they will be different. Next slide.
Therefore, in doing an experiment in program
flow, we do not really work with just a single measurement
«*
but rather an ensemble of measurements, An ensemble
is defined by values of a valuable, say, for instance, .
concentrations of any gas or pollutants, of all possible
experiments under given conditions.
Now, the ensemble is defined by probability
mean of this ensemble, and I've given you a formala
there like if N stands for a number of experiments,
when N is very large, you expect a steady (UnintelligibleO
Now in the atmosphere, and keep in mind that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
5
6
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
98
' ' ' ., *
the experiment must be repeated under given external
conditions, and this is not possible in the atmosphere,
and it's also (unintelligible) to use time averaging;
however, time averages approach ensemble average only
for a restricted state of conditions.
Now if you do assume that these conditions
are (unintelligible) we can perhaps replace ensemble
8 || average by time averaging, and I have given you a formal
for the time average where we now have defined the average
(unintelligible) to infinity. May I have the next slide
please?
Now if you can define a time scale, for instance
(unintelligible) then if the sampling is sufficiently
r
larger than the time scale we can in principal
(unintelligible). However, in the atmosphere, the way
i
we perform the experiments, I do.not.believe that we
can get this time scale, because the time scale by defini-
tions is defined by that, the spectrum at zero frequency
i
and when you do an averaging, you take a finite averaging .
here and then you don't (unintelligible) zero frequency
scales. j
And so the time scale is not finite to atmospher<:-
and, therefore, in principle, we do not get an ensemble
average in atmospheric experiments regardless of the
averaging period or the sampling period. May I have
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
4
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
the Jiext slide please? i/Now, the best we.can do in any i
mortal is to predict the ensemble average whereas an
observation is just a member of this ensemble chosen
at random. Therefore/ the model prediction is not expected
to exactly equal (unintelligible), and this is a sort
of (unintelligible) atmosphere.
Therefore, uncertainties in predictions are
inevitable. I do not believe regardless of how much
you try that we can get through this uncertainty. May
I have the next slide please?
The next question is given this, how do we
quantify the uncertainty models? The one way of doing ,
it is to compare the model results with the experimental
data. Now, keep in mind that when you do an experiment,
your ensemble is defined by the input parameters. In
other words, if you have the first measurements, like
. >
we get one data point, that belongs to one ensemble,
the second measurements belongs to a different ensemble.
-»
It's a very important point, and, therefore, there's
no actually . (unintelligible) ensembles that are parallel
Now if the model predicts the ensemble mean,
then the data points that is of (unintelligible)
.atmosphere are distributed about, the model predictions
in some distributions which (unintelligible) of time.
Then if we assume how these data points are distributed
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
100
about -model pred-ictidirs^we" can perhaps "predict 'the
probability with which an observation should exceed
a given value, say CM which may be a standard. It can
be, like I said, (unintelligible) if we know what the
distribution is. Can I have the next slide please?
I'll show you as an example for normal distri-
butions if you want to predict the set of probability,
(unintelligible) , you can in principle, and you know
what the (unintelligible) are. Like I said before,
it's very difficult to obtain these parameters from
atmospheric data, because (unintelligible) are different
for each data point.
But you can't make an assumption like perhaps
tKese symbols are close to each other; therefore, you
can (unintelligible) . If you assume that the concentra-
tions are lognormally distributed, it can express your
observed concentrations as the predicted concentrations
(unintelligible) and ideally these should be an average
of (unintelligible) .
.*
Therefore, the average of "log absolute square
is the measure of uncertainty in models. Incidentally
this is the geometric standard deviation. Could I have
the next slide please?
With that introduction, I shall show you how '.
we obtained the uncertainties of the modeling study
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1 101
carried, put in OntaricLEnvironment.- - -Now- this -is a model
for the prediction of ground level concentrations of
(unintelligible) particles inside a ^unintelligible).
The model is based on the idea that most of the dis-
persion is controlled by the large scale eddies, the
updrafts and downdrafts in the (unintelligible), and
where the entire plume is divided into an updraft plume
or a downdraft plume, and the (unintelligible) to give
the final results of the ground level. 3
10 The observations, the model predictions are
11 compared with observations of Willis and Deardorff in
12 their Watertown experiments. . Now when you do those
13 statistics, assuming a lognorraal distribution on the
14 cdhcentrations, you get a.geometric of the predicted
15 (unintelligible) is 1.12 (unintelligible)which is not
too bad, and the geometric standard deviations_i§ 1.25
17 So this givens an idea of what uncertainties
are. May I -have the next slide please? This is a model
1.9
20
21
developed by Venkatram for dispersion of pollutants
for particles in the (unintelligible) atmosphere. It
gives you the ratio of the predicted observations, and,
22 I again, if you do the statistics, you'vO got a geometric
23 mean of 1.14, where value is 1, and a geometric
24 (unintelligible) of 1.69 (unintelligible). This is
25 II a large number of data points. Next slide?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
1(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
102
And this is a model for longrange transfer .
to air pollutants, a statistical model where statistical,
long-term, long-range plume is assumed for each source .
and applied to all the sources or most of the sources
(unintelligible) In the United States and Canda, and
the model also assumes the scavenging in a statistical
manner by assuming that (unintelligible). You can see
the rightmost column gives the ratio of observed predicted
concentrations, I'm sorry", the (unintelligible) of sulphur
in rain, and (unintelligible), the geometry mean and
(unintelligible) deviations are 1.16 and 1.64. Again,
the value of the (unintelligible) reasonably well.
Okay. Given the fact that (unintelligible)
f
are close together, and we can get a handle on the model
uncertainties, how to use this in the decisionmaking.
Generally, (unintelligible) controls are based on impact
*»
on the environment of sources-.. Its effects by
(unintelligible) averaging over several years, then
our decision on the control of.the source can be based
on model .prediction for long-term averages, and I say
.that with reservations, because (unintelligible) but
probably now the uncertainties are smaller, so it enables
you to make a decision on the long-term average.
However, (unintelligible) short-term
concentrations, the model predictions can not only be
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. -., i«3
used in the absolute sense, uncertainties would have
to be associated with model predictions to affect our
inability to predict the short-term concentrations exactly.
Even with uncertainty and with suitable assumptions
on the distribution of those (unintelligible) about
the predicted mean, we can only determine the probability
of which a given concentration of (unintelligible) mea-
surements.
Decisions would then have to be made on the
basis of this probability along with economic constraints
if the may have I have the last slide please? The
last view graph shows if you are not comfortable with
a normal .distribution or lognormal distribution or any
distributions, you can use these expressions which is
colored (unintelligible) regardless of the distribution
which says that absolutely (unintellgibile) the probability
that it will be greater than K- is less than sigma square
of the variance, so given sigma square, we can get an
(unintelligible) of this probability.
Thank you.
MR. TIKVART: Are there any questions from
the floor for Mr. Misra? Okay. It doesn't appear that
there are any questions; however, can you supply the .
recorder with a copy you have. Mr. Moe, have you
also supplied a copy of slides? Can you send those
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
104
to the recorder please? It's right after 12:00 so I
2 think it's appropriate to break for lunch now, so let
me give you some additional information before we do
break.
g There are four more governmental agencies
who wish to speak, and they'll be first up this afternoon.
o
Those are Mssrs. Trout, Bonta, Raol and Meguire. After
then, we'll move to the individual presentations and
the following individuals are those I have listed, Sklerew,
Witten excuse me, and they'll speak in this order, .
Sklerew, Witten Hanson, Kohm, Moon, Pell, Wright, Fein
and Wurmbrand. I hope I've pronounced those names cor-
rectly.
It will be tight, but we protoably will be
able to get through all those this aftesrnoon. Those
are all we have as of now with one exception. I've
been informed that the National Academy of Sciences
wishes to make a statement tomorrow morning, so whatever
happens this afternoon, we will be here tomorrow morning
*
to continue to take comments as appropriate.
We will reconvene this afternoon at 1:30.
(Whereupon, the hearing adjourned for luncheon
to reconvene at 1:30 p.m., this same day.)
12
13
14
15
16
17
18
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, O.C 20005
-------
105
_ *
. . AFTERNOON SESSION
. (1:28 p.m.)
6
MR. TIKVART: After further reconsideration
3
of the list of speakers for this afternoon and negotia-
4
tions with them, et cetera, I'd like to quickly just
5
revise the list of speakers and their order of presenta-
tion.
7
One of the government agency speakers, Mr.
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
8
Raol, somebody scratched his name off the list in bacR,
I assume he no longer wants to speak. If that's incorrect,
please let me know.
. -- -»
That leaves three agency speakers, Trout,
Bonta and-Mequire. The individual presenters following
r
them will appear in the following order, Sklarew, Pell,
Hanson, Fein, Witten, Moon, Wright, Maxwell,Kohm and
Wurmbrand. The last couple of names are in jeopardy
^t
as far as completing today's schedule. Ifyou want to-
speak today, we will accommodate you,, the last several
names, if you could appear tomorrow morning, that might
be helpful all the way around. We'll see where we are
after the break at about 3:50.
The first speaker then will be Dennis Trout,
the third last of the agency speakers.
MR. TROUT: Thank you. My name is Dennis
Trout. I'm the regional meteorologist'with USEPA in
NEAL R. GROSS '
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
2 60604.
4
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
lp'6
rp . The address is 230 South Dearborn, Chicago, Illinois,
First, I'd thank all the participants here,
and I'm particularly glad to see some of the state and
local agencies represented. I think we'd be well advised
to hear their comments and suggestions as well.
The last day and a half, we've heard numerous
talks oh motherhood and apple pie/ and I took a look
at what I had prepared and found it to be very similar. .
I had originally intended to present a summary of compari-
son criterias suggested for evaluating the adequacy
of air quality models for regulatory purposes.
I propose particularly after luncheon, my
^
attention span, if your attention.span is anything like
mine, maybe some of you won't be listening after three
minutes, so I'll try to hold my comments to as bri^ef
as possible. -- \
We've heard very much on the should1s as to
where you should go, but haven't heard too much on how
/ " .
we can get.there. In order to-improve model estimates
and account for and, where possible, reduce modeling
uncertainty, we have been constantly and consistently
advised in meetings such as these by statements such
as, one, models should be developed and revised following"
scientific methodology and adequate representative air
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20OO5
-------
8
10
11
12
13
14
15
16
17
18
19
' 20
21
22
23
24
25
0 107
!U
quality, meteorological and source monitoring data should
be developed and relied upon.
Two, models and analyses should make use of
proper input data. Three, model analyses should-toe
judged by their conformance to -accepted scientific princi-
plaes.
Four, potential investigators meaning industry
and their consultants, for purposes here, should meet
with the agency personnel to develop a protocol to be 3
followed, and, five, frequent communications should
be maintained between the investigators*and the agency.
»
I would like to note that I wholeheartedly
agree with,the-objectives of these statements. However,
r
what is routinely preached and generally accepted by
all of us attending these meetings today, this does
not appear to be what is routinely practiced. '" *
^
Many of the reasons given for departures
these proposed objectives are well understood. For
example, costs and time constraints which are probably
t
two of the most important concerns. Other reasons that
may bear on the situation include the objectives of
those funding the study.
For example, if the study is to develop an
alternative model that would in hope allow for greater
emissions by estimating lesser impacts than a reference
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1 1
10
11
12
13
14
15
16
ir
18
19
20
21
22
23
. 24
25
model, those funding the investigation may not be desirous
to begin funding or continue to fund a study that might
actually result in the development of a model that would
estimate greater impacts than the reference model.
Considering the facts that, one, most of the
specific recommendations of how to account for uncertainty
do not diminish resource requirements, either on the
part of industry or the agency, but actually require
greater amounts of time and money to accomplish these
specifics, and, two, that the historically listed con-
straints, time, money and objectives may be expected
to continue or remain constraints that we find ourselves
faced with in the future.
f
In view of these, I would like to solicit
the specific recommendations from industry, consultant
community and the state and local agencies for what
you might view as minimum proposed requirements for
the agency to follow regarding the five objectives pre-
viously stated.
The specifics on the how's or the means of
accomplishing the five should1s previously indiciated, if
these are not resolved or provided, we should not expect
that the problems associated with uncertainty will be
readily resolved'or that we will likely account for
the historically known resource constraints, so at
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
109
*'
this point, I'll either let you comment to me on what
I have said or I would suggest .as this is opened up
for public comment that you feel free to elaborate on
those concerns, because I think they are of particular
relevance and importance to us, in exepditing this process.
Thank you.
MR. TIKVART: Any takers on Dennis' question?
8 '|| I think it was a good question. Why don't you go ahead
and state the five should's and then your question about
it. Give them a chance to think about it.
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
MR. TROUT: Okay. One, models should be devel-
oped and revised following the scientific methodology
and adequate representative, air quality, meteorological
p
and source monitoring data should be developed and relief
upon.
Two,models and analyses- should make use of
proper input data. Three, model analyses should be
judged by their conformance to accepted scientific princi-
ples. Four, potential investigators, meaning industry
and consultants particularly for this meeting, should
meet with agency personnel to develop a protocol to
be followed in the course of their investigations. And,
five, frequent communications should be maintain betwen
the investigators and the agency, and I*11 just note
that we standarly hear these. In fact, comments have
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
-^ . 1.1,0
-k
been made before, the Senate Committee recently towards
these ends, but we find that they are very frequently
due to alleged time and resource constraints unable
to be complied with.
If they can't be complied with, when you come
in to talk to the region, and that1s my particular problem,
then we're making excuses from the start. We have to
8 build in the caveats, let's build them in now, and not
wait for later which is going to cause you and all of -3
10 us greater delays and dissatisfaction
MR. ._TIKVART: And what was your one or two
12 sentence question?
13 MR. TROUT: Okay. My question in regard to
the", specifics that have been suggested during this con-
15 ference are obviously complex, involved and cause greater
16 resource requirements, and, two, I don't think .that
17
18
19.
20
21
22
23
24
25
what we see historically is that greater funding is
going to be available or we're going to have greater
amounts of time or objectives in providing these studies
are going to.change.
In light of those given's, we would like to,
or I would like to solicit specific recommendations,
particularly from industry, their consultants and state
and local regulatory agencies for any specifics on the
how's as opposed to the should1s in accomplishing these
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
objectives wHoch' are he motherhood and apple pie state-
ments that we typically hear.
MR. TIKVART: Good, we have a taker.
DR. MOE: Rod Moe, Texas Highway Department.
There's a highway research project called NCHRP 2018,
and this project is well funded by the Transportation
Research Board, funded actually by a lot of highway
departments all over the nation.
The idea behind- the project is to develop
a data base of line source modeling/ and what they've
developed actually is a five data base. They haven't
developed these, but adapted five data bases so they
can be.-used.to-test models.
f
They're right now in the throws of publishing
a final report. They've also developed what they call
a figure of merit. The figure of merit has five sets
a-
of statistical tests to test different aspects statistic-
ally, the different test methods, different aspects,
like one might look at say how well it predicts all
;
kinds of values, you know, for any model you pick.
How well it predicts -say the extreme values,
you know, the highest to the second highest and so forth,
that may be another set of tets, but with the figure
of merit, the standard data base, you've got a testing
ground based on this research project where you can
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
112
-?. '
actually go out and if you have a new model, test it" :
against the data base and against different data bases
is what you should do. You shouldn't ever test it against
the data base you develop it with, and use the standard
set of statistical methods that handles different aspects
of the thing.
When you wind up, you can get actually something
that people can look at as kind of an objective judgment
to how well that model does. Now I think that same
sort of thing might be done for point sources, or you
might even use an entry data base.
That's not a question; that's a statement
of .one. approach.
r
MR. TROUT: There are five different data
bases?
DR. MOE: There are very few data bases that
a
are good enought to ^ -^ ' .
MR. TROUT: Did i misunderstand you'that you
said that there were five different data bases developed
*
for this study?
DR. MOE: Right, they'weren't developed for
that particular study. They're data bases that were
adapted for this research study. They're existing data
bases.
Data bases with enough detail, you know, that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
3,13
"*"
they had good enough measurements to be able to be used
2 to validate models.
MR. TROUT: I would suggest if you feel that
these have utility towards these objectives that you
forward a copy of this report or those proposals or
whatever you may have that you feel appropriate towards
the subject of this conference.
DR. MOE: The research study hasn't been pub-
lished yet. I'm working on a panel that chose a subject
and chose a research and will be published shortly.
When it is published, it might be made a part of the
minutes of this meeting possibly, but it1s not been
published yet.
r
MR. TIKVART: Thank you very much. Any other
takers? I think Dennis has asked a very challenging
question, and I would encourage members of any consultant
groups to take that question to heart, and perhaps after
you have some time to think about it", submit written
comments to us, because we're essentially asking you,
i
how would you prefer the agency to conduct business
in this area. That's the question we've asked you,
so we'd like some response on that. Thank you Dennis.
MR. TROUT: I'd just like to say that it would
be much more efficient for you as industry or consultants,
when you meet with your regional offices, if you take
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
the .time and effort now to save yourself the additional
cost and time.
MR. ANSIS: I have one question. I'm Dick Ansis
from ITT. You asked a good question. Would it be possible
that you'd have a written statement? It was hard to
write everything you had down that we could take with
us on that?
MR. TROUT: I'd be glad to give you a list
of what I have. Otherwise, there will be copies of
this in the record which is available from the gentleman
sitting up front here.
>
MR. TIKVART: Speak in the microphone please.
MR. ANSIS: That would be after you have a
f
chance to make a comment on it. .£'
MR. TROUT: After all the comments had been
received, you'rse saying? ~ *
*s
MR. ANSIS: If we got a list of your" question
now, we could have a chance to make a comment before
the report was closed. If we wait to get the transcript,
;
we may not.
MR. TROUT: If anybo'dy has particular, I'll
be glad to type them out if you give me a business card,
I'll have my secretary send you a copy of my version.
MR. TIKVART: Yes?
MR. VAN VLECK: Lowell Van Vleck from Tuscon
NEAL R. GROSS '
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10.
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
^ .... - -« 114 ,
Electric. A comment on your five criteria here, one
of which was the model should be scientifically defensible
or something of that sort, and presumably this means
something like taking into the curvature of the wind
height and maybe a large scale oscillations or some
other such things as this, and you want an adequate,
you say, meteorological base, and you want measured
source terms and measured meteorology.
MR. TROUT: Let me correct you a second. It's
the community that wants. I am repeating what has stan-
dardly been requested of us. You can hand be a black
box, and that's what you propose for what's necessary
for regulatory purposes, and that will meet with the
IP
public comment and scrutiny, I'm just repeating at this
point what we standardly hear as the motherhood and
apple pie statements.
That's why I said I wouldn't give my paper,
because it follows in some of these natures, very resource
constraining.
;
MR. VAN VLECK: Sut I think though and at
least my experience is that this is beyond the abilities
of the" meteorological community. They have been unable
to find those variables through many, many, many field
trials that have really any proven ability to predict
dispersion. The first thing we had was from Mr. Moe
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
t tt
115
today where he talks^aboiit'^the Helmholtz and gravity
2 waves being significant factors in perterbating our
n general conceptions of controlling a dispersion process.
His was all short time, and it's hard to see
- how we can apply this for 240hpur periods, but it's
fi the first thing that's come forward in about 20 years
now that may offer some chance for improvement, but
o I bet you you- can spend $140 million and take ten years
to make that work. ....... °
._ It may work, I don't know, but it will take
,. that kind -of an~ effort to prove it. First, you've got
*
12 to go. with his process. He's got windshear at every
13 level-
*- MR. TROUT: This is exactly the point. It
._ is a complex situation. Scientists recognize this.
,,, However, this is standardly what is recommended-.; » We
lb ^
. all recognize that, and you said as an excuse- why we .......
fail to do our jobs later, and if you can provide -^an
objective or a scenario that we can accomplish our objec
tives in the regulatory sense of protecting ambient
air quality and not being overly restrictive at the
same time, I would like to hear it as much as you would,
I feel.
MR. VAN VLECK: I think you just have to rear
range a track that will bypass this modeling business
18
19
.20
21
22
23
24
25
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 200O5
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. ^ 116
and yet provide the~atmosphere with substantially the
same amount of protection and illuminate it. It's an
administrative or procedure technique, but it could
be done, and you don't have to go through all this, modeling
and meteorological measurements and tall towers and
all these things.
MR. TROUT: I'm not disagreeing with you.
That is one alternative.
MR. VAN VLECK: But I don't think we're going
to get there, from what I know about all the things
that have been tried to predict dispersion.
MR. TROUT: I would suggest, and I would welcome
any scenario that you can provide to us, if you want
to "provide, the .detail that can stand up so that we can
try to implement this type of policy so that we can
evaluate. - -
MR. VAN VLECK: You mean the second- approach -
I mentioned? -^
MR. TROUT: Any approach. That second approach
is, I think,' will be evaluated in the same context as
any other suggestion that is -made, and I would open
this up to some free thinking, because I think it hasn't
been stressed sufficiently before, that the cost of
such studies are exorbitant.
We're talking millions of dollars for a source
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
i?
18
19
20
21
22
23
24
25
117
study".
MR. TIKVART: Excuse me. I have a feeling
we're going around in circles. What I"d like to request
is the following. Dennis, if you and Mr. Van Vleck
could clarify between you the issue and then if Mr.
Van Vleck could submit a written suggestion as to how
to approach the process of dealing with what model,
what sort of data to use the specific application, I'd
appreciate that very much. Okay? Thank you. Thanks,
Dennis. I'd like to move on. .
.
The next speaker is William Bonta, speaking
for the State of Maryland Department of Health.
! MR. BONTA: Thank you. It looks like you've
I f
| closed nup your ranks there. Maybe if I turn this this
way, I can hide behind it. Okay.
Good afternoon, ladies and gentlemen, and
panel "members. My name is William Bonta, and I; am the
Director of the Technical Services Programs in the Mary-
land Air Management Administration. I would like to
f . -
first thank you, Joe, for.this opportunity to take a
few potshots at EPA. It isn't very often that we have
that chance at the state level. I hope you guys remain
calm in your seats until the end of my talk, because
;
my aim isn't so good sometimes, and you've got to be
still.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Seriously, however, I find the theme of this
second conference very refreshing, and I would like
to congratulate whoever it is that1s responsible within
EPA for regrouping the effort around a more macroscopic
look toward the problem of modeling and especially its
interrelationship with air quality program management.
I've been sort of distressed in the past about
what seemed to be an inexorable slide toward uniformity
with the federal government establishing all of the
rules. After all, the objective in modeling has always
been to mode accurately w enever it's used. Uniformity
is a laudable goal, but it's not the primary goal.
For example, I don't see any reason why a
r
model can't be tuned specifically to a local condition
and applied .only there and nowhere else. Why, after
all, does a model have to work in Cleveland, Ohio and
Nome, Alaska, if it was empirically tuned to work on
the coastal plain next to the Chesapeake Bay.
We all realize that there are many parameters
that are obtained empirically. Theoretical diffusion
only goes so far in this game, and after that point
in time, there are many black box coefficients and vari-
ables that are built into the programs that you and
we both developed.
If we include explicit clima.tological terms
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(2O2) 234-4433 WASHINGTON. D.C 20005
-------
9
10
12
13
15
17
18
19
20
21
22
23
24
25
__. . 120 -
in the theoretical dif-fusion equations and then say
that these equations can be utilized throughout the
rest of the country, in other words, pick out these
as independent variables, then I think that's a large
mistake, because it's quite possible that some of these
black box coefficients that you've tuned into the models
that you've generated also contain clxmatologically
related .factors.
It might vary between say Nome, Alaska and" 3
Cleveland, Ohio and the coastal plain next to the Chesa-
peake Bay. I've had a number of conversations with
>
Al Cimorelli about this difference of opinion, and some-
times quite frankly, I wonder if I'm the only one that
r
has the difference, and everyone else thinks the same
way you guys do, but that point aside, I'd like to mention
several things about the background documen't whlcfti was
sent to me to prepare for this meeting. " <-,~=»@g}:
My first comments that I would like to talk
would be about my dislikes, I guess, and then I'll follow
those items with the things that impressed me favorably,
and that wa maybe you'll be'my friend again when I
sit down. Okay.
My first problem with the statements made,
they assume that everybody realizes that modeling is
necessary to all situations. Well, I don't agree with
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
that necessarily." : I'm"an 'administrator now a'fter many/
many years, and not a modeler any more, but I just note
v
with cryptic interests that it was a conference of modelers
that sort of made that determination, and I'm going
to let that point rest where it sits.
There are many situations where I suppose
you could say analogies could be drawn from operating
plants with adequate surveillance in the area surrounding
the plant over many, many" years would easily substitute .
for modeling analysis in the event the plant proposed
to change- emission rates.
The second point I noticed was that when taken
in context of a widespread urban area with many sources,
f
it seems to me as relationships between far reaching
program design and air quality levels can quite often
be drawn accurately utilizing simple proportioning tech-
niques rather, than diffusion modeling especially when ~~
ambient air quality levels are directly proportional
to emission -levels, and I know that may sound a little
bit reactionary, and I'm on the right side of the stage
here, but I still think that from the state level, some
program opinions should properly be drawn that way,
especially long range estimates of what's going to happen
with air quality.
The third point, and this is the one that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
t330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
. 20
21
22
23
24
25
122
I would like-to-dweil-^n-a-little J-bit more than the
rest is that I think I'm strongly opposed to the conferees
recommendations about introducing uncertainty, explicitly
into the air management decisionmaking process. I'm
uncertain why I even came here, today to fight with you
guys but that's beside the point.
Most of the problems dealt with in a regulatory
agency deal with continuous variables, and we know that.
We do not very often relate to exact situations excep
for maybe when the light switches on or off in the men's
-room when-you walk in the door.
>
Why dhould uncertainty be explicitly considered
only in the modeling aspects of the program when it
f
is"not considered in such a fashion anywhere else. That
doesn't make sense to me.
For example, why should not uncertainty* be
included in an engineering evaluation of a piece of
control equipment? That was mentioned yesterday. " After
all, we know that equipment does fail and these failures
t
sometimes in a stochastic manner. Also, the failure
rates and variability could probably be predicted during
the design of a facility.
Such a failure, if it occurred on the worst
possible dispersion day could just as easily cause an
ambient air quality standard violation as the situation
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
are considered implicitly in.the permitting section's
o
review of a plant in question.
In a similar fashion, the uncertainty over
modeling is also known and is considered in the delibera-
6
7
9
1G
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
193
4.A.J
'where a M^h'"s\rip'huri'Ii:iBve*l~in--'a^"load"of-coal is burned
on the worst possible day. In general, these factors
tions over permitting new facilities, at least in Maryland
I can say that with safe assurance that I'm correct.
It is just as unrealistic to say that equipment
never breaks down, and it is unrealistic to say that
the worstcase- day is the worst that we shall ever see.
Why, therefore, is EPA focused on ExEx methodology and
other such discussions for uncertainty in meteorological
calculations and not emphasize the same sort of uncer-
tainty in plant operational problems.
If a Monte Carlo sensitivity analysis is to
be performed, on a source that is proposing to build,
why shouldn't the analysis include all of the factors
that could lead to high ambient air quality levels?
Indeed, I find it curious that ork Group
I at the Airlie House Conference and EPA, in general,
have been sniffing about this uncertainty issue for
the past several years. I see it in the discussions
of the CASAC meetings with their discussions of risk
assessment, and I see it in EPA's proposal to use
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBED
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
'; ..'."" _ . " -. 124
' "
expected exceedences^ke^-eva-luate--remote-power plants,;
and I notice it in the Work Group recommendations to '
consider factors like, quote, "location and extent of .
the geographic area where standards/increments are most
likely threatened", unquote, and also where they recom-
mended looking at, quote, "exposure or dosage estimate,"
unquote.
I think that what I.'m seeing here1 is what
I saw 15 years ago in-the radiological health area where
certain groups within the Atomic Energy Commission,
.,then the Atomic Energy Commission, were basing decisions
for aerospace hazard evaluation on total risk assessment,
and other groups, the groups that were dealing with
r
reactor safety problems were dealing with maximum credibl
accidents and basing the design of plants on those kinds
of criteria.
I don't think it will be too long before some-
body in EPA gets the general idea that total expectations
can be used as the measure of a new plant being built.
/
That is to say that from the time you sign the permit
to the time the thing has been operating for a number
of years, how many cases of emphysema do you expect
to see, or whatever, and if the total expectation is
less than one, somebody would say, well, we don't expect
to see anybody get it, and that's all right, and design
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
10.
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
the -plant so ife1 S"doWft--to^ OT* to' give^a r"actbr"of a" '
hundred uncertainty measure.
This is what was done in the aerospace nuclear
field. The situation was different, but it's analagous.
I don't want to get into any detailed discussion of
these issues in the small time available, but I would
like to point out that the single value index is in
my opinion the number that modelers should be calculating -
and that explicit expressions of uncertainty will serve
no other purpose than to generate further uncertainty
over those procedures and numbers.
I know from talking with program decision-
makers that- they get very upset with me when I try to
r
express uncertainty about a calculated result from a
model or even a measured result from an ambient instru-
ment.
There are many instances where regulatory
standards at the state and local level are based on
arbitrarily fixed standards, highway speed limit, for
example, a frostline for building a water line, a color
form content in food.
The law recognizes that in dealing with continu-
ous variables, the standard is always determined arbi-
trarily; however, as long as there are sound and suffi-
cient reasons for choosing a particular number, the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
126
»
^^ What EPA .-
seems to be suggesting is that not only is that number
an uncertainty to be documented in the original standard
setting process/ but all of the subsequent uses of that
number must consider similar uncertainties as well.
Imagine what problems would unfold if the
state police were to treat a speed limit the same way.
It's been my experience that the pressure to form finer1
and more technically sophisticated standards, regula-
tions, et cetera, have stemmed from enforcement personnel
who oft times-get-mixed up in discussions where they
do not belong.
For example, enforcement personnel try to
r
answer questions posed to them by industry regarding
program design. The answer delivered in cases like
that should be, "That's not my concern, I don't write
the laws, I just enforce them."
Many times, however, the complications that
I've seen dreamed up arise through enforcement people,
/ .
particularly the lawyers who imagine that they can't
possibly defend themselves in court against the myriad
of questions that they are being asked by industrial
people.
Many times the lawyers don't have the courage
or the courtroom experience to make these judgments.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(702) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
?*-
r assure you "that-by-^e'sigiiing a more complicated system"
to answer all of these questions will in no way reduce "
the numbers of attempts that environmental lawyers will-.
make to circumvent compliance.
Bernie Steigerwald mentioned some of the prob-a
lems over uncertainty issues yesterday, like auto-
correlation and plant failures, for example. I have
the f eeling that he has just uncovered the tip of an
iceberg. --- . . . -
When one starts looking at the beginnings
of new program element, the element normally increases
its complexity very rapidly as' the program is imple-
mented .
f
I need only cite the significant deteriora-
tion program as an example that proves my point.
And yet here/ with the uncertainty proposal
the EPA is proposing to complicate thing's in yet another
matter. I know when a Program Administrator is forced
to make a decision in Maryland, he doesn't want to hear
i f
about the uncertainty. All he wants to hear about is
whether or not his best people have done the best job
that they know how to provide an accurate number and
whether or not that accurate number violates the law.
Adding an error band about a modeling result could very
well increase the legal problems rather than decrease
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, O.C. 2000S
-------
128
I 'them. On the other hand, if there is an well, environ-
2 mental groups, you know, the pressure from industry
3 to say plus or minus, you could be high by 30 percent,
and you say, well, yes, we could be high by 30 percent.
If you take that error band, what are you going to do
with it then.
7 If you say yes, indeed, the calculated value
8 could be lower by 30 percent, and the guy says, ah-
9 hah, I'm in compliance, and then he puts a little bit
10 of pressure on the politicians, and they twist your
arm and you give them the permit, and then the environ-r
12 mental group looking over your shoulder goes to court,
13
14
15
16
17
18
19
'20
21
22
23
24
25
because he says no, it wasn't 365 down to 330, it was
365 to 390, because it's plus four minus.
So I think what you're really looking at is
. 2
the mean value anyway, and I think we're all going to
. ..
target into that, and that's the best thing we ought
*
to be looking at.
Just maybe a cryptic comment, my impression
in the modelers dilemma was expressed in the sentence
from the support document, "Furthermore, for air quality
management and regulatory process to ignore modeling
uncertainty and to continue to base decisions on past
estimates, single value measure, such as the high,: second
high concentration, places an unduly heavy burden on
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW .
(202) 234-4433 WASHINGTON. O.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
129
*s -1-'
modeler's," unquote,""'an3"fiiyxs'overail"sentiment is. Aw,
gee, that's tough, fellas.
It seems to me that the fellows in the programs
at the firing level, and I'm sure you're there in many
situations who have been looking at hardware are doing
this all the time. The registered professional engineers
see equipment design and they know there's uncertainty
with that. If they approve a lemon, boy, they're going
to eat it, because it's going to be out there putting
out the smoke,, and the thing doesn't work right.
We all know that, and yet they aren't talking
-, ^
about uncertainty, so I think you ought to consider
this all around.
r
Let me get to the positive reactions. Now
that I've dug you enough, I might as well get a little
kinder. On the matters of those positive reactions,
I was very impressed with the protocol approach. I ~ "
like that. It sounds good, I hope it will work. "
I think that the establishing of protocols
i
will protect both the company and the sibate from a great
deal of delay due to misunderstandings which routinely
occur. Guideline documents should contain examples
of pitfalls and traps, such that the public agency can
have ready guidance on other things to look for in cir-
cumstances where maybe a streetwise consultant knows
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17"
18
19
20
21
22
23
24
25
: ^ . 130
more" about" tfie "mode lilfgTtzhan "the "individual and the
public agencies dealing with it, so it can work both
ways. It can protect industry and the control agency
from a lot of circumstances that you know of exist,
and we may not know it in the state agency.
Okay. With regard to the central modeling
operation, I guess Joe is the central modeling operation.
The question is are they well enough insulated from
a political standpoint. I don'.t know that question.
Perhaps it's good enough for the present time.
I think that I agree that a technical oversight
committee should be established, but I believe it should
be advisory in capacity and, therefore, not have any
approval rights, or else the advisory council may be
tainted somehow.
Oh, another thing about the recommendation,
the modeling center should be a suitable group to provide
protocol guidance. I also notice'd that in Egan's group
there was a recommendation that, quote, "a suitable
form . (up sup.)" whatever that means, "should be identi-
fied and established to resolve anticipated or unantici-
pated issues," close quote.
This wording appeared kind of odd to me since
it seemed to defeat the whole purpose of what you tried
to do in coming up with a protocol document. The wording
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 .WASHINGTON, D.C. 20005
-------
3
4
5
6
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
131
"
seems to be so broad as to allow anyone to escape anything
agreed to-/in-»the protocol document, and a smart environ-
mental lawyer, consultant could play a tune on that
one, so maybe you need a lot more thought on that concept.
Overall, I'm very impressed, and this is my
summary here, I'm very impressed by EPA's approach toward
this modeling conference. I like the open minded atti-
tude exhibited by the turn around from the first proposed
guideline/revision, and" I hope that going with the proto-
col and the modeling center, technical oversight committee
that the modeling chore can be simplified.
Perhaps the pendulum that we see that is about
.ready.to .dismember the PSD program will not also chop
r
and seriously affect our modeling activities as well.
State of the art modeling has progressed so well in
the 12 years I've been in the air pollution programs.
I'd hate to lose it at this time, because the trend
is towards simplicity in government, as you know, and
I hope that you can successfully cast the die towards
t
that end. Thank you.
MR. TIKVART: Any questions from the audience?
No? Bill, thank you, as always, for a thought-provoking
presentation.
The last government agency presentation I
have is Kenneth Mequire for the Kentucky Division of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
132
Air Pollution'Con tiro IT
MR. MEQUIRE: Good afternoon, ladies and gen-
tlemen, and gentlemen of the panel. I've had to rewrite
half of this in the past few minutes. I hope it won't
suffer too badly.
Actually it fits in pretty much with what
Mr. Bonta has just said in that the matter of statistical
scatter has not only administrative problems, but it
also is only part of the"picture and it's the remaining .
part doesn't seem to have been dealt with very much
here so far, and that is the fundamental accuracy of
Gaussian equation and modeling which is contained in
practically all of EPA models.
v
There are problems with that model, becuase
it requires a finite velocity of the wind on the one
hand and on the other, there are three fairly important
areas in the United States where there's an awful lot""
of no wind at all, areas that the wind power people
will never .make any progress with, and this has to do
;
with the area of long calms.
The problem there is all of our weather data
that we use in Kentucky comes from the EPA. It has
no velocities reported under one meter a second, and
usually everything we get is where you have a period
like that. They'll report one meter a second for each
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. O.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
4- -^-F-~ ' . 133
"."~ «
hour which is wfiat we^use hourly, 'and you know good
and well if you get six or 12 hours in a row reporting
one meter a second, and that's the lowest it'll go,
you know that it went below that, but you don't know
by how much, so that's what goes into the computer model-
ing, and we do get an artificial result as to our model-
xng.
What this tends to do in an area of calms,
it tends to understate the actual concentration, because
under a condition of no wind you're going to have over
a duration of time, and that's also som'ething that's
^
not directly taken into account in the EPA models, you
have.a build-up, and usually the heaviest calms occur
r
when you have a period lasting through the night and
into the second day, this type of thing. It can go on
for days. .... >
So these cause understatements of the actual
concentration in areas where you have maybe 200 periods
of calm, 200 days with these calms out of 365. Those
are in the southern Appalachians, in the Salt Lake Basin,
in coastal California, according to the EPA's own publi-
cation on the subject.
There's another problem we have, and that
is since we do have only EPA weather data, that comes
to us with the input as to the velocity with its
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
, 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. 134
_ *" "
limitation, and also "there-1 s~a problem of the stability.
It comes to us that way.
Now if we try to do any monitoring or get
weather data for modeling from a local area in the Appa-
lachia Valley, we are up against it, because we don't
have a system for converting various meteorological
factors into one figure, the Pacquill stability, and
we could very well need that for operating on our own,
if we were going to use the Gaussian model.
So my feeling is that this type of weakness
needs to have a better priority than a consideration
of the scatter, statistical scatter that prevails where
you're just~using a standard procedure to actually do
r
the computing in the computer.
So that was my comment, and I feel that it"s
an addition to the last statement, and it also represents
a comment on the first presentation asking --"Mr. Mr.
Dennis Trout asked for some comments on what's needed
in order to improve the meteorological data that's used
in modeling. Thank you.
MR. TIKVART: Are there any questions?
MR. HAMBURG: Fred Hamburg speaking, from
RMC. I don't understand the statement about the non-
reporting of calms. It seems to me if you get your
weather data from the National Weather Service, I presume
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
J
i
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
:?-.- J " /.. . 135 :-:-
d -''* - « . »""
you meant EPA? r ~
MR. MEQUIRE: Absolutely. I've called down
to NOAA in Asheville to try to get some weather data
for some additional pounds that I know has weather sta-
tions, but we don't have it, and'when I started asking
about stability numbers, they never heard of them. They
wondered what on eart I was talking about.
MR. TIKVART: For the record, I should clarify
that the data you're referring to is undoubtedly National
Weather Service data processed through a special program
by EPA-to generate stability classes.
>
MR. MEQUIRE: Right, that's what I found out
subsequently, it is like that, so the way we operate
now, please excuse me if this sounds too harsh, but
we are completely at the mercy of EPA when it comes
to getting basic weather data to put in our"computers.
MR. HAMBURG: I'm just at a loss to understand^2
this lack of meteorological data because it seems "to
be available everywhere. If you want to get synoptic
data, if you want to get timely data, you can get it
from Louisville Airport. Well, the only thing I'm trying
to bring -out as an old-time weather bureau hand myself,
I don't know of any situation where we were forced to
write one mile per hour or one meter per second when
it was a dead calm. There is circular uncertainty
NEAL R. GROSS " '
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
136:
provides for zero direction with zero wind. Now, of
course, modeling has a problem with that, and particularly
when you get into CRSTER modeling where you make the
assumption that when you have that, you have to look
way, way back to get the last one that happened, assume
it's continuing to happen, and that, of course, gives
you three hours in a row with extremely high concentraric:
tion. Now I understand your point, but with the criticism
of the meteorological data, per se, I think it is unwar-
ranted .
It makes the computers work, although in a
certain way, and I'm not sure exactly. The only thing
I w,ill say is this. We have never, ever been able to
get a concordance between monitored data and modeling
data. Now, the modeling data comes from some previous
year or years, hourly, 8760 hours worth for a year or
for the average or 64 to 70, that's six years or seven
years, whatever.
"That's the way it comes to us, and that's
the way it's processed. It comes* to us through the
EPA. We right now don't have any facilities that I
know of for handling information that would come directly
from the MWS. .<
Let me^ cite you a little problem we had recently
down in a certain industrial area known as Calvert City,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
, . 137
Kentucky by the Tennessee River near its mouth.
please?
MR. TIKVART: Can you keep your example short
MR. HAMBURG: We found out that if we used
Paduka weather, Salem, Indiana weather or Nashville,
Tennessee weather, each one of us gave absolutely differen
reports on the compliance or non-compliance of this
plant there, and I forget the choice that was finally
made, but we had a hot old time there. We were all
laughing the whole time.
MR. TIKVART: Thank you very much. Are there
any government representatives of government agencies
in the audience who wanted to speak but did not have
f
the opportunity to do so? No? Okay.
I will remind you that a representative of
the National Academy of Sciences has expressed a desire
to speak tomorrow morning. That will happen right at
9:00 tomorrow morning. . -
I'd like then to move into the next phase
/
of the conference whereby individuals have expressed
the desire to make presentations. The "first of these
will be Ralph Sklarew from Form and Substance. Where
did Ralph go? He was here. Charlotte Hopper, is Ralph
Sklarew out in back?
Out to lunch, did you say? Well, we'll give
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 2000S
-------
138
*
Ralph a chance later on. Next we have Jerry Pell. Jerry,
as I -understand it will,be speaking for the Meteorology
3 Committee, TT-3 of APCA.
4 MR. PELL: Thank you. Mr. Chairman, Members
5 of the Panel, I am pleased to have the opportunity to
participate in this conference on air quality modeling
7 I am Dr. Jerry Pell, immediate past chairman of the
8 TT-3 meteorology committee of the Air Pollution Control
9 Association.
I am also a certified consulting meteorologist
of the American Meteorological Society,* and I'm here
12 today on behalf of the APCA meteorology committee. There
13 are 41 members subject to revision, because we have
14 people joining us in leaps and bounds presently on the
15 committee all of whom are meteorologists and members
lg of APCA, and of the 41 at last county, 21 were also
17 certified consulting meteorologists.
18 Together, we represent the views of governnent,
19 utilities, industry, consultants and academia. All
'20 of us are concerned with air pollution problems and
2i with air quality modeling. The following statement
22 is the statement of the committee.
23 Undoubtedly air quality prediction modeling
24
25
is an important tool used by regulatory agencies in
the permitting process. However, a model must remain
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS.
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20O05
-------
2
3
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
139
* "
a tool and not itself become a decision making. Air
quality models are currently being used as tools in
two distinct types of regulatory applications. The
difference between these two types of applications and
their relationship to the history of air quality models
are not commonly recognized.. Nevertheless, they have
a strong bearing on the cost of regulatory policy.
Several years ago, the only air quality stan-
dards were health related. These standards were mandated
by the Clean Air Act and promulgated by EPA. The stan-
dards were held as a goal to be achieved at virtually
any price. It was felt that protection of the public
health made any violation of the standards impermissible.
p
Air quality models were developed and approved
by EPA eo ensure compliance with these health-related
air quality standards, and, as a result, these models
necessarily .acquired a built-in conservatism to reflect
their historical purpose of protecting public health.
Now, however, the Congress and EPA have set
t
"arbitrary" standards arbitrary, as we've heard from
.previous speakers, in the sense that they do not have .
a clear relationship to a known deleterious effect such
as mortality or morbidity. Included among these stand- .
dards are de minimis levels and Prevention of Significant
Deterioration increments; these are arbitrary numbers
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
1
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
* 140
set by regulatory fiat. Nevertheless,, we are still
using the same conservative "thbu shalt never" set of
air quality models for assessing cmopliance with these
arbitrary standards.
Models, as we all well know, are simply mathe-
matical constructs, of environmental processes. Early
in the history of model development, model performance
was related to reality on an average basis. Some pre-
dictions were higher and some predictions were lower
than actual field measurements. The average of the
predictions matched the average of the measurements.
To meet is mandate for health protection,
EPA conservatively chose, as one of the criteria for
success in model performance, to emphasize model predictio
of higher concentrations than observed, thereby ensuring
protection of public health.
Purely on the basis of this tradition, we
have begun using these historically conservative models
for other well intentioned regulatory purposes, but
without a full and clear understanding of the cost impact
to society.
Requiring the use of these highly conservative,
health-based, models for assessing compliance with arbi- .
trary regulatory standards can be very expensive and
probably is not cost effective.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
141
Since the numerical values involved in these
arbitrary standards are relatively small to begin with,
the use of a model with a strong conservative bias imposes
a very onerous compliance burden. The Committee, there-
fore, urges that, for these arbitrary standards, models
with more realistic performance' criteria be used.
Despite these reservations, modeling is the
only tool available to estimate the potential impact
of proposed new sources or of proposed major modification:
to existing sources on ambient air quality. It is also
the only available way to estimate the degree to which
emission standards for existing facilities, and hence
the costs .of air pollution controls, could be reduced
and still comply with the applicable air quality standard:
This is becoming increasingly important as we seek ways
to achieve air quality standards in'.the most economical
manner.
Nevertheless, air quality prediction modeling
is not an exact science and there are inherent uncertain-
t
ties in the predicted maximum concentrations. The result:
of an air quality modeling study and the regulatory
consequences of these results depends very importantly
on the particularly model chose, the input data provided
to the model, and how the results from this model ultimat
are used to develop air quality control regulations.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
.-..
- f;- _ . | 142
1 This assessment of the importance of; air quality
2 models is consistent with the conclusions of the National
3 Commission on Air Quality, and for those of you who
4 have the written statement, the recommendations of the
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
NCAQ panel on atmospheric dispersion models is provided
as an attachment to this written statement.
Accordingly, in recognize of the importance
of air quality models, the APCA Meteorology Committee
fully supports the continued reliance on models as basic
planning and evaluation tools, but the Committee hopes
that the uncertainties of these planning tools will
be recognize within the context of specific applications.
Turning to the discussion of the accuracy
of these models, the accuracy of an air quality model
depend on the purpose for which the model is being applied
A model can be expected to be more accurate in predicting
the maximum pollution concentration to be found in the
area surrounding the source than it is in predicting
the exact location of that maximum or the exact time
of that occurrence.
Studies comparing predictions with measurements
repeatedly have shown this to be the case. This is
probably due to the fact that the pattern of the wind
is not known sufficiently precisely. For example, a
two percent change in wind direction can lead to a 1000
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
I I I" i C1
~.~ '' 1.43
: - ' t ' » ' "
percent increase in the predicted concentration at a
given fixed location using Pasquill Gifford dispersion
coefficients for E type stability, or moderately stable .'
atmopshere.
By carrying out predictions over a large number
of observations and for a large number of locations,
the inaccuracies due to the uncertainties of the wind
field are partially compensated for.
In most regulatory applications, it is not
necessary for a predicted model to be able to determine
the exact time at which the highest concentration will
occur. In many applications, it is also not necessary
for the model to be able to determine the exact location
f
of highest concentration.
Thus, in the evaluation of how accurate a
prediction model is, one must take into account the
purpose for which the model is- being used, and on the
subject of purpose, we heard some earlier discussion
of the kind5 of purposes used in the nuclear industry
as opposed to more conventional air quality control
considerations.
Several issues arise in taking model accuracy
into account in regulatory decision making. These include
one, should generic ranges of uncertainty for models
be provided, and used explicitly in the regulatory
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
r.144 ,.
-; » '
process? Two, should site specific ranges of uncertainty
2 be developed and used in the regulatory process, and,
3 three, should uncertainty, once it is quantified, be
4 used explicity, quantitatively or qualitatively in regu-
5 latqry decision making.
Concerning the first issue, the accuracy of
7 a prediction model is highly site-specific depending
8 on the source characteristics, meteorology, and topo-
9 graphy of the surrounding area. Thus it is very doubtful
10 that a generic range of uncertainty can be specified
11 for a given model. The Committee therefore recommends
12 that generic ranges of uncertainty not be developed
13 or used- explicitly in regulatory decision making.
f
14 On the other hand, the accuracy of a particular
15
16
17
18
19
20
21
22
23
24
25
prediction model at a given site can be determined by
careful comparisons of predictions with measurements
at that location, such that a- reasonable estimate of
the accuracy of the model in predicting the maximum
concentration can be made.
If such an estimate is- known, it can play
a useful role in regulatory decision making.
The third issue, whether to quantitatively
incorporate uncertainty in model predictions into the
regulatory decision making process, has been considered
by EPA. The written text quotes an internal piece of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
'.? U
._ ic-* ..| H . - 145
correspondence at EPA. The evaluation concluded that
it is redundant and improper to* include modeling uncer-
tainty as an explicit variable in the Expected Exceedances
method. This conclusion was based on the fact that
modeling errors, will almost always result in an over-
prediction of maximum concentrations.
We support that conclusion. There are a number
of issues that arise in regulatory decision making that
cannot be quantified to the point of leaving no discre-
tion to the decision maker.
The Committee recommends that estimates of
modeling uncertainty be accounted for qualitatively
in decision making by regulatory authorities rather
than explicitly in the results of the prediction modeling
analyses.
In conclusion, I would like to note that the
TT-3 meteorology committee of the Air Pollution Control
Association would be pleased to assist EPA with establish-
ing model uncertainties by drawing on the experience
of our large and diverse membership. We furthermore
offer to serve as members of a peer review board if
such a board were to be formed to assess model accuracy
and uncertainty.
-Now I'll be pleased to answer any questions
you or members of the panel may have, Joe.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202} 234-4433 WASHINGTON, D.C. 20005
-------
146
MR. TIKVART: Yes, I have an observation and
a question. The observation deals with you're referring
to the SteigerwaIt-Burton memo. It should be noted
for the record that that report discussed conclusions
based on interim results.
I find that people are too anxious to grab
anything that comes out of EPA as interim progress report
3 and run with it. As a matter of fact, that work is
9 still ongoing, not completed, and I'm not sure that
10 when we get through with this, we will find that models
as they have been used in the past have been overly
12 conservative.
13 I don't know. That's an iffy proposition,
r
14 but to draw the conclusion that the way we have used
15 models in the past definitely is conservative is only
lg an interim conclusion, and the work is not done on that,
and from what I've seen it could well go the other way.
18
19
'20
21
22
23
24
25
That's the observation.
The question is, in the early part of your
presentation, yo.u seemed to indicate that the models
used have been overly conservative. That doesn't seem
to square with the presentations that were made yesterday
morning whereby the standardly used EPA model was shown .
to be unbiased and in some cases underestimated concen-
trations. Would you care to comment on that?
NEAl R. GROSS
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
147
MR. PELL: Well, on the question of the inter-
nal memorandum from Bernie Steigerwalt, I take note
of your comment, and I appreciate the input. The testi-
mony properly should have included the notation of the
status of that. work.
With regard to the conservatism of models,
the concept has been, and I think it's been reiterated
in some of the remarks that I heard today that when
the public health is in question, if you're going to
err, you err on the side of conservatism and if a model
design is to be approved for regulatory purposes, then
the best option would be when uncertainty is present
is to opt for a model which tends to overestimate the
f
concentrations rather than under.
So as a matter of procedural policy, my impres-
sion and the impression of the committee has been that
the working concept has been-when you're not sure, go
for the higher end of the scale rather than the lower
end, and if the plant turns out cleaner than you had
intended, so much the better. I-f it were to turn out
dirtier, you'd have a problem on your hands.
MR. TIKVART: Thank you. Any other questions
or comments? Dennis gets there first. ,
MR. TROUT: I would like to make some comments
on the word conservatism because we -r
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.-=rr . ;-; 148
MR. TIKVART:"^"Excuse me. I don't know that
you identified yourself.
MR. TROUT: Dennis Trout, EPA Region V. And
we're frequently accused of putting in all conservative
assumptions, and this is something that we1re particularly
sensitive to, and, as a result, we put together lists
of conservative, liberal and case by case dependent
factors,, and at a recent meeting of the APCA in St.
Louis back I think in April, I went through this list..,
Of that 25, 11 were classified as liberal,
seven as conservative, and seven case by case dependent.
>
At that time, I also solicited from that conference
which was heavily attended by industry lists of any
additional factors that they might consider liberal
or conservative.
To this day, I have not received one additional
addition to that list, so I think the statement that
our models err intentionally or on the side of cpn.serva-
tism has not been founded, and from studies that have
been shown including some of the EPRI studies and all
the others, I don't think there is technical support
to arrive at the conclusion that our models are conserva-
tive.
MR. PELL: Perhaps we can chat about that :.
over a drink some evening, Dennis.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 ' WASHINGTON, O.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
-MR."TROUT: Okay. _ - i , . - . , .
MR. TIKVART: Doug?
MR. FOX: My name is Doug Fox. Jerry, I'm
pleased to see you've made mention of the National Com-
mission on Air Quality, Dispersion Modeling Panel policy
statement, and my comment is just simply for the record
to point out that actually the statements of that panel
are contained in about a ten-page article that was. pub- ,
lished recently in a 200=-page report.
I didn't want anybody to have the impression
that this was, in fact, the only output of the NCAQ,
and I'd like to recommend to Mr. Tikvart that somehow
or other that NCAQ report be made part of the record
r
for this proceeding.
MR. TIKVART: Would you submit it, Doug? Thank
you.
MR. PELL: The codicilion point in the admit-
tedly brief excerpt which is used in my statement from
the NCAQ panel on atmospheric dispersion models is pro-;.
/
bably the sentence that that particular NCAQ panel recog-
nized, quote, "that models are useful technical tools
for air management decisions, but models alone cannot
be used as a sole determinant in such decisions," and
* I
I believe it was the sense of the committee that that
was the cogent point worth bringing to your attention
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
150
- today, but-^certainly-'there was never an *intent to dimi- !
nish the magnitude both qualitatively and quantitatively
of the work of either the Commission or of that panel.
MR. TIKVART: Any other questions or observa- i.i
tions? Thank you.
MR. PELL: Thank you .very much.
MR. TIKVART: Next we have Richard Hanson
who represents ITT^ Rayonier.
MR.. HANSON: Thank you. My name is Richard
Hanson. I'm the Air Quality Group Leader, Olympic Re-= :
--search Division of ITT Rayonier in Shelton, Washington.
My comments deal with the use of dispersion models in
the regulatory decision making process and on the accu-
rac,y of these models.
They relate primarily to modeling in complex
terrain. However, they are applicable to modeling in
any complex situation whether it is terrain, source
or meteorological.
Dispersion modeling in complex situations
is difficult'. No model currently exists that is uni-
formly adequate for complex terrain. For the most part,
existing models in complex terrain are conservative.
This may be satisfactory for a PSD analysis if enough
increment is available and no other new source is likely
until more advanced models are developed.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
v " ' 151
^ '«
, ,.' ,f f -/ -JU.J-T-Y ;^ <*?*- -r^"a ' '"**: /**«** * f ..--_.._ - -
However, regulatory agencies have considered
. designating non-attainment areas based on modeling results
and they have not adequately considered the conservative
nature of the models.
My first recommendation is that is non-
attainment decisions should not be made only on modeling
results. They should be made using actual monitoring
data. The modeling results should only be used in support
of the monitoring data. 3
If a model predicts a non-attainment .area,
the problem should be documented with actual monitoring
%
data.1 Yes, this would be expensive,- but not nearly
as expensive as unnecessary mission reductions.
r
You may have to monitor a long time before
the worst case or second worst occurs. Perhaps the
worst case does not exist in the real world. J
My second comment deals with the validation
or calibration of models applied in complex terrain.
There should be some physical .justification for changing
the measured meteorological data other than just that
the model doesn't work without it.
I'm talking mainly about the measured wind
direction; however, it also applies to the other meteoro-
logical parameters. Other likely causes that should
be considered include unquantified sources may have
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. O.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
152
a major 'impact';""'Fugitive'"sources' can cause high ambient
concentrations^ Their mission rates are many times '
not known.
The measured meteorological data used may
not be representative. If the wind direction shows
that a receptor is not impacted during an averaging
period, all the other meteorological parameters may
also be wrong.
The model may not allow for the proper plume
interaction with terrain. The model used may not be
-suitable for the situation. If the meteorological data
is adjusted it is possible that the model would show
that the emission sources which in the real world do
not impact the receptor during the averaging period,
are the cause of the ambient concentrations.
Under this situation any modeling is totally
meaningless. \ .
My third comment is about representative meteor
ological data. On site meteorological data should be
/ .
mandatory when modeling for regulatory purposes. At
least wind speed, wind direction and ambient temperature
should be measured near the emission point or in the
receptor area.
On site data should be used even if meteoro-
logical data is"available for a longer period of time
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10.
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
153
from anearby site I t>ata from -a few kilometers away
may be totally unrepresentative. If sufficient on-
site data is not available, it should be obtained before
decisions are made.
These points are not based on a hypothetical
case. My company was faced with a non-attainment design-
nation based primarily on modeling results. Ambient
S02 standards had been exceeded several times near a
pulp mill. We had determined that the source of the
ambiet SO2 was a storage lagoon.
A temporary control for the SO2 emissions
had been completed with plans for permanent control;
however,., two-dispersion model studies had shown that
»
the pulp mill excluding the lagoon could cause a.non-
attainment area.
Nonrepresentative meteorological data had
been used in both validation studies and attainment
analysis. These data were available for a longer period
of time than the representative data. The emissions
from the lagoon had not been quantified or properly
considered. The wind direction was adjusted in the
validation.
These studies resulted in totally erroneous
conclusions, incorrect sources were identified as the
problem. As a result, reduced emission limits were
,, NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
12
13
15
17
18
19
'20
21
22
23
24
25
7i ' - 154
proposed that would have had little impact on the ambient
air quality; however, they would have had an adverse
impact on the continued economic viability of the pulp
mill.
Fortunately we were able to show that the
major cause of the ambient S02 was not from the incor-
rectly assumed sources in the pulp mill. Enough on
site meteorological data was available to challenge
the attainment analusis. However, <'.this should not have
been necessary.
This is one case that shows tne need for some
. -. >
functional control over the use of dispersion modeling
in the regulatory decision making process. That's the
end of the statement that I filed with the recorder,
. but I have two comments that I'd like to make based
on previous discussion.
We're talking about uncertainty in modeling.""
Uncertainty doesn't mean anything unless representative
input data is available. Many times this is not the
case.
Secondly, I agree with the recommendation
about joint discussions between all parties before model-
ing is begun. That would resolve some of the problems
that we had; however, much of the other work was done
for a PSD application for another source.1
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS.
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
".', ..-.. , '..:. ' x '-55;'
They asked for a received a confidentiality '
agreement that restricted access to the data. Thus,
we were not able to participate before the modeling
was performed.
That' st. the end of my comments, and I appreciate
the opportunity for making them at this meeting. I'm
open to any questions.
MR. RHOADS: You were giving a real life example
to make your point. Itrs~not relevant to identify that
particular plant; however, as I understand it, some
modeling was done, and then you were able to point out
through monitoring that the modeling that had been done
was inappropriate.
MR. HANSON: No, that is not correct. We
were able to use representative meteorological data.
Enough of it was available during the key period of
time to challenge the modeling done with nonrepresenta^-
tive data.
MR. RHOADS: Okay, I understand. Did you
consider putting out a monitoring network. Do you feel
.that your company when faced with that situation should
put out a monitoring network to attempt to validate
the model, or do you believe that should be more pro-
perly the government's responsibility?
MR. HANSON: In a situation like that where
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
^ ."..-' ,- 156
^ #"
it's an adversary~role, I believe it's probably a joint
participation so everyone agrees with what you're doing.
MR. RHOADS: Okay, thank you.
4 MR. HELMS: You spoke of making a non-attain-
5 ment designation based only on modeling. Were there
6 measured violations?
7 MR. HANSON: Yes, there were. The source
8 was already controlled by the time the decision was
9 made to look at non-attainment, so it would have served
no purpose to declare non-attainment, because at that
fixed date, it was no longer non-attainment.
12 MR. RHOADS: You had corrected the problem?
13 MR. HANSON: We had corrected the problem.
14 MR. TIKVART: Any questions or observations
15 from the audience? Why don't you come all the way up,
16 Ralph?
17 ' MR. SKLAREW: Ralph Sklarew, Form and Sub-
18 stance. This points out perhaps" a big hole in the model-
19 ing that has. been done heretofore in the guidelines,
20
21
22
23
24
25
that is meteorological modeling, modeling to take meteor-
ological measurements from a distinct point and simulate
what the effects are at the local site.
I think that should be something EPA could
spend a little more attention to and provide techniques
for general use. That might get you out of the problem
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
.24
25
157 f
with non-representative data.
MR. HANSON: I would -agree with that. This
is a very difficult place. I would think it's one of
the most difficult anywhere we got in the country, but
I don't believe that payment on non-attainment decisions
should be made basically on modeling results.
The models just haven't been shown to be accu-
rate enough in situations like that. This model that
had been used, we had shown it .probably now, it predicts
by about a factor of two which is not really too bad
considering the difficult situation.
MR. BONTA: Bill Bonta, from the Maryland
Air Management Administration.
F
I heard this suggestion before from the men
from Hunton and Williams yesterday, and it seems to
be that you're inferring that you-would in the failure
of agreement on modeling terms set out a monitoring
program to prove the point.
Are .you extending that to evaluation by a
control agency such as ours of a brand new grassroots
facility?
MR. HANSON: Would you repeat the question?
I didn't get the last part of it.
MR. BONTA: You seem to be suggesting that
in the event that both parties do not agree on the use
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
21
22
23
24
25
""" ^ "-'.- 158
-r-
of a model and its" outcome that ambient monitoring mea-
2 surements take a precedent.
3 MR. HANSON: Yes, I would say that. I would
4 not imply that we did not agree on the model. We agreed
5 which model to use. We did not agree, although I think
the agreement is there now on which meteorological data
7 to use, so it wasn't the question of model as much as
g the input data to the model.
9 MR. BONTA: Well, are you suggesting that one
10 utilize the same technique in evaluating a grassroots
plant, one that is not now there?
>
12 MR. HANSON: It is possible to do that> but
13 that is a PSD type or an impact type of analysis, and
r
14 if you have enough available, it's not hurting you,
15 where attainment, non-attainment decision is a real
16 world number that you're looking at with a monitor,
17 and that is going to be more accurate than a model.
18 With the new source, you have no choiceV
19 MR. BONTA: Okay, you're saying with the new
20 source then you have no choice other than to use the
model and make your determination based on that model
alone?
MR. HANSON: That's correct.
MR. BONTA: Okay, thank you.
MR. YINGST: John Yingst with PG&E. Another
NEAL R. GROSS '
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
; . . 1591
'_
I way you might have gotten around your problem was you
2 had a lagoon that caused your observed concentration.
3 Now your model predicts an expected value.
4 Your observed concentrations were extreme values. What
5 you could have done is a rollback model type approach
whereas if you reduce your emission concentration by
7 X, your maximum observed concentration will not be linear
g but will be reduced by more than that.
9 So you might have gotten out of doing the
modeling approach.
11 MR. HANSON: At the time we started with this,
12 we did not know, nor did the regulatory agency know
13 the size or magnitude of the emissions from this lagoon
r
14 or their impact. It was only during the study that
15 that came about.
16 MR. TIKVART: Thank you. We'll go to the
17 next speaker,. Ralph Sklarew/- are you ready? Okay.
18 We'd better get those in and move to the next speaker
19 first while, those get mounted, if that's okay with you.
20 The next speaker is Richard Fein who is with
I '
the American Petroleum Institute. Mr. Fein?
21
22
23
24
25
MR. FEIN: Thank you. I'm Richard S. Fein,
a senior research associate with Texaco Incorporated,
and, as was. indicated, I'm speaking on behalf of the
American Petroleum Institute this afternoon.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
--'-: iU1? . 160
The API comments for this conference present
a brief overview of the findings to date of two model
evaluation programs. They also discuss the implications
of those findings for current regulatory applications
and offer recommendations for improvement of models
and modeling practice and for. review and revision of
the regulatory applications.
API's model evaluation studies has focued
on the standard Gaussian models recommended by EPA for
assessing air quality impact for non-reactive pollutants.
One program has utilized the results from short-term
tracer dispersion experiments, and utilized the CRSTER
model. The second program was described by Steve Wise
yeserday, and it compared observed unpredicted sulphur
dioxide concentrations around an industrial source complex
using one year of continuous monitoring data collected
by a nine station network.
Predictions were made using the industrial
source complex short-term model adjusted for hourly
emissions which is not a normal capability of that model.
Results from two of the tracer experiments
. are illustrated in Table 1. Note that 50 percent or
median predictions, well, we see a ratio predicted to
observe values for the maximum observed for the maximum
on each sampling arc, so these have been adjusted for
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
'202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
161 ,
.:.., v'." " . .-'-r- i
wind direction, but they are coincident in time.
Note that the 50 percent or median predictions
are baised towards over-prediction, but that ten percent
of the predictions are at least about a factor of two
too low. In other words, the predicted is only about
half the observed at ten percent cut-off point on the
frequency distribution of errors.
Also note that ten percent of the .predictions
are at least several-fold too high, much greater than
a factor of two, as a matter of fact. In fact, in terms
of the.bias_here-for these two programs, approximately
three out of four of the predictions were overpredicted.
The next table illustrates results from the
industrial source complex for what we've termed three
model use categories, and we've done this for three
different averaging times.
The first category, areawide cumulative fre-:_
quency presents a comparison of predicted and observed
regardless of time and/or location. The second category
sites specific cumulative frequency, presents a comparison
based on the top ten observed .and predicted concentration
values carried by a station, and the third category,
site and time specific concentrations, present a compari-
son based on the ten highest observed values with the '
predicted concentrations at the same time and station.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 2COOS
-------
_v . - 162.
-' - .? « .«
The ten percent median and^90 percent cut-
2 off levels are indicated. Results of these studies ' -
3 indicate that the median agreement decreases and the
range of uncertainty of the agreement increases as the
model is required to predict the location and time of
the impact.
Note that the highest observed concentrations,
the righthand column there, are badly underpredicted
g and that ten percent of these high observed values are
10 actually predicted to be zero.
11 The next figure, and I'll make a further point
12 with respect to this, it shows, the point that Steve
13 Wise made-yesterday, that predictions for some locations
14
15
16
17
18
19
20
21
22
23
24
25
are biased towards overprediction while others are biased
toward .underprediction.
Also, as we just got through seeing here,
most of the time, the highest observed concentrations
are underpredicted whereas, although we didn't show
it on the previous slide, most of the highest predicted
concentrations are overpredicted.
*
This seems to be a generalization that occurs
at monitoring sites most of the time. It's apparent
from the results of these two studies that the Gaussian
models, even in these rather ideal situations, and these-.
were rather ideal, are not capable of predicting
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(207) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
concentrations at a given time or location with any
substantial degree of certainty or consistency.
As time or location constraints are relaxed,
model performance improves until in the areawide fre-
quency case as was pointed out yesterday, the models
perform quite well.
In light of this finding, it is worthwhile
to examine the demands which common regulatory applica- -
tions place upon a short"term model1sipredictive capacity
The next table presents a schematic attempt to depict
the model use type corresponding to various regulatory
applications.
The first applications group, that is the
top group there, represents cases for which the model
is used to predict the magnitude of expected impacts,
but the regulatory decision is relatively insensitive
to the location of the impact. Hence, we talk about '~
areawide.
For an isolated remote source or for the case
/
where, impacts from one source ar.e predominant, the model
. is required only to estimate the concentration magnitude
independent of time or location for an appropriate avera-
ging time. For situations where multiple sources at
different locations or a complex of sources at one
location contribute substantial impact, the predicted
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
;-r
:.,.,,_ 'M
combined impact is highly sensitive to the location -
of impacts for each source for the same time interval
and when we say same time interval here, we mean for
a given set of meteorological conditions.
Hence for those cases where you -need to use
site and time specific concentrations. The second appli^
1 cations groups represents cases where the impact -at
8 a given location, such as a PSD Class 1 area or a non-
g attainment hotspot must be estimated. 3
10 These applications clearly require the model
to predict a site specific concentration and pose a
12 more'stringent requirement for model performance. If
13 the combined impacts of a number of sources must be
r
14 estimated for such a region, the worst case combination
15 scenario arises in terms of model prediction.
16 I think from what we've seen that it's*obvious
17 that modeling areas are too large to adequately
18 many, if not most, of the purposes of these regulatory
19 applications.
20 Thus, for example, modeling.for PSD baseline
21
22
23
24
25
determination and increment'tracking often requires
accuracy and precision, that is far smaller than the
approximately three to at least tenfold 80 percent confi-<
dence bands for predicted value shown in previous tables.
Use of inaccurate models for these PSD
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(207) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
purposes reduces the regulatory process to an arbitrary
" * ~ , :
paper exercise which, as was stated yesterday, only .
creates an illusion of certainty.
In this illusory exercise, the model predictions
take on a public relations and a legal significant,
but bear no demonstrable connection to actual air quality.
In view of the fact that modeling is the only method
available for estimating the impact of a nonrexistent
source, that is a proposed source or one that existed
in the past, it becomes imperative to try to bring mode14
ing applications for regulatory purposes into better
agreement with modeling capabilities.
This can be accomplished by improving models
anfl modeling practices and by modifying the regulatory
requirements for modeling.
We'd like to make the following recommendations
to the EPA, and these are summarized on the next slide^
We would recommend requiring that monitoring
data be utilized to assist decision making involving
/
modeling whenever such data is available. To do this
it's necessary to develop methodologies for adjusting
models with monitoring data and for using models for
interpolating and extrapolating monitoring data.
Thirdly, we believe that the EPA should very
carefully examine each of the regulatory applications
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
- 14
15
16
17-
18
19
20
21
22
.23
24
25
of modeling to determine how to bring these applications' "
into better congruity with modeling capabilities, and
then based on these careful examinations to either modify
the regulatory requirements for the modeling applica-
tions or to recommend legislative changes necessary
to modify these requirements.
Obvious modifications that are needed include
one that was discussed yesterday, the necessity to re-
express ambient air quality standards to better average
out the uncertainties of air quality monitoring and
model.
And, also we believe that in view of the errors
in modeling'that we should recommend elimination of
the PSD increment system since one cannot calculate
baselines or in any realistic fashion or increment or
track the influence.
To protect Class 1 areas, the API believes
that monitorable maximum concentration limitations should
be established for each Class 1 area to protect the
air quality values that are specific for that area.
I thank you and would be happy to answer questions.
MR. TIKVART: I have a question relative to
the previous slide, if we could go back. What do you
mean by the multi-source example? The reason I ask
that question is, in my experience, many times when
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
. . '-.. ....... ,£ . -..- -..- ---, - - 167 .
1 you have large sources which might be considered multi-
2 source, they are still located too far away to have
3 a joint impact at the same a joint maximum impact
4 at the same point, the reason being that most maximum
5 impacts occur within a few kilometers of the source.
6 Can you give some examples of what you mean
7 by the multi-source case?
8 MR. FEIN: A typical large oil refinery, for
9 instance, would be not would be a complex source
10 which contained afmultiple of individual sources.
11 MR. TIKVART: Okay, so in that case you're
12 talking about multiple point sources that are really
13 part of the same source complex?
14 MR. FEIN: Right, or other examples where
15 one could get into this sort of problem would be in
16 say the golden triangle area of Texas where there are
17 many oil and petrochemical installations down there
18 which jointly impact the same area and impact them signi-
19 ficantly.
20 MR. TIKVART: Okay. One observation on the
21 lower lefthand block. I think you make a correct obser-
22 vation, but it is probably also a function of the proxi-
23 mity of the source to the receptors you're interested ;
24 in, and yet the proximity is relatively close, it probably
25 is closer to the upper lefthand group if the distance
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
168
is large and the angles where there can be an impact, ''
the wind directions where there can be an impact is
a relatively small angle, your point is probably very .'
valid.
MR. FEIN: I would agree with that.
MR. TIKVART: Any other comments?
MR. RHOADS: Would you elaborate on your recom-
mendation for protection of Class 1 areas? You said
monitorable increments? >
MR. FEIN: Monitorable concentrations. For
each class 1 area, there are air quality related values.
Visibility is an example where perhaps you find particu-
lates which you need to measure.
r
As for the visibility that's needed to protect
the scenic values of that class 1 area, you might esta-
blish a concentration limitation for fine particulate -
in the atmosphere, and establish it at a level where
you could measure it, because you certainly can't calcu-
late with any degree of accuracy.
And then with monitors-, you could determine
if you were protecting that class 1 area.
MR. RHOADS: I understand.
MR. TIKVART: Any questions or comments from .
the floor?
MR. SKLAREW: Ralph Sklarew,. Form and Substance.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
16--9
I'd like to disagree with the agreement. If I understood
what was just aid/ then it goes against how things are
used presently. If we have an increment in the PSD
area, and it's partly used up by one plant in this whole
area, it impacts, let's say, in the northeast, and now
we build a plant that's going to impact in the southwest,
it still goes against the same increment, what's left,
and so the multiple source case turns out everybody
uses up the same increment whether they impact at a
coincident location or time which is a very, very conserva
way of understanding it or again air quality standards
where we're going to be exceeding standards.
If my interpretation of how things are working
r
is not correct, I'd like to be enlightened.
. , MR. .EEIN:. These are the experts on how things
work.
MR. TIKVART: The EPA doesn't have a position
that I'm aware of as to how to go about tracking incre-
ments. I think that's left pretty much to the discretion
of individual states. ' " '
MR. RHOADS: That's correct; however, if you're
looking at a receptor point, a hypothetical point, and
you have a source that is 20 kilometers north of that
and a source 20 kilometers south of that, it is unlikely
that those two would be additive in consuming increment.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
tive
-------
170
I That was not the intent of the increments. They were
2 to - even though we cannot monitor the values contribut-
ing to the increments because they are so small, they
4 are nevertheless intended to represent the real world
5 deterioation, not a hypothetical deterioration
MR. SKLAREW: But the problemcomes about in
7 application. If we realize that models in a sense can
g only give frequency distributions, areawide frequency
D
9 distributions, they can^t tell us what time impact is
10 or where it really is, then all the increments in worst
11 case assumptions are additive, and we get ourselves
12 into a tremendously overconservative snaffu.
13 MR. RHOADS: It doesn't necessarily happen
f
14 that way, but it could, you are right.
15 MR. SKALREW: Unfortunately a_lot.of the things
>
lg that we've got involved in, it seems like that's the
17 interpretation.
18 MR. HAYNES: Eldewins Haynes, North Carolina
19 Division of. Environmental Management. Could I see the
20 graph that was a comparison between predicted and ob-
21 served?
22 Now you made the statement that the at
23 least correlation wise, it's obvious that there's pretty
24
25
bad correlation between the predicted and observed,
and I'm personally surprised at the number of
HEAL R. GROSS ''
COURT REPORTERS AND TRANSCRIBERS.
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
overpredictions in this case. However, if we examine
2 this for regulatory purposes arid compare our maximum
3 predicted versus maximum observed, and even the second
highest predicted versus the maximum observed, that's
pretty close.
.
MR. FEIN: You're absolutely correct on that.
On the other hand, if you were supposing that the maximum
8 observed was too high, and you needed to take some correct
action, this particularly complex source consists of
two sets of four sources which are separated in space,
and if you want to take corrective action based.on condi-
12 tions:' for the maximum predicted concentrations you would
13 probably be making the wrong corrective action so that
14 there's a real need here, I think, to have the predictions
15 and the observed coinciding much better .than they are,
16
11
18
19
20
21
22
23
24
25
if we're going to use these models for the kinds of
decisions that we are forced to take.
This could be done if we dug deeper into the
distribution and averaged out some of these errors, .
I believe.
MR. TIKVART: One last question.
MR. CHASE: Ed Chase, Fairfax County Air Pollu-
tion Control, it seems to me that your proposal on moni- ;
toring in class 1 areas seems incomplete to me. You
established some criteria concentration, and you set
NEAL R. GROSS
COURT REPORTERS AND -TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
ve
-------
,_,,., !;. v _ . 172
up monitoring for it and then what do you do? Suppose
you reach it. What do you then know in terms of what
you can do from a control strategy standpoint?
MR. FEIN: Then you have to rely on models
to try £o take corrective actions and you'll have to
consider the various modeling uncertainties in doing
that.
MR. TIKVART: Thank you. That was a very
informative presentation. Ralph, are you ready? Let's
10 go.
MR. SKLAREW: Thank you, and thank you for
12 the opportunity to speak. It's nice to be back after
13 three years.
14 " First I want to apologize for not having my
15 slides. I left them in my hotel room and had to run
16
17
18
19
20
21
22
23
24
25
back and get them. From this point of view 'though4,
Doug Fox's comments are exactly opposite. The podium -
hasn't moved to the right, it's moved to the left,-"and
maybe some of the things that are happening now are
.
a little bit more to the left.
I'd like to speak about decision, making with .»
inaccurate models. Inaccuracies in the standards EPA
guideline models have been acknowledged for sometime.
EPA sponsorship of this conference demonstrates this
recognition of these modeling errors. This conference,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
. 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
'' *7-3 K
however, seems -to -be -implicitly institutionalizing these"
inacc.uracies. Instead, the modeling community should
concentrate on reducing them.
This paper focuses on four concerns about
decision making with inaccurate models: the reasons
for inaccuracies, the need for quantifying the errors
actually I don't have a button here, can I have the
first slide? I have slides just to illustrate the whole
thing. Next slide. D
The reasons for inaccuracies, the needs for
quantifying the errors, the cost effectiveness in modeling
v
and the use of the models in emergency response situations
Certaintly there are valid reasons for inherent
and irreducible inaccuracies in modeling the atmosphere,
but the standard models are far from this level of accu-
racy. ~
Modeling improvements rather than acceptance
of gross levels of inaccuracy would have been a more
appropriate focus. The highest priority in dealing
with inaccurate models is to quantify the errors,:.-not
for a handful of cases, but for many, most or even all
of the major applications.
Error assessment for all major regions of
the country, especially urban areas should be conducted
I routinely. Since the standard models are grossly
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 2000S
-------
-rv - - - -a; *
.-is;. . -. r-.. . 174
'inaccurately, cost effectiveness of modeling should
2 be considered. Substantial amounts of money are now
3 being spent on modeling itself, and decisions now based
4 on model results are monumental in their costs.
5 Emergencies occur in which toxic substances
are released into the atmosphere, and the result is
serious harm to the public. Accurate models are needed
tp respond to these emergencies, to allocate vital re- u
sources and to actually save lives. D
These four concerns are being presented because
they are ones in which we at Form and Substance have
>
12 been able to make, we feel, significant constributions.
13
14
15
16
17
18
19
20
21
22
23
24
25
The primary causes of modeling inaccuracies
r
are~~the dual nature of the atmosphere, deterministic
and stochastic, the inconsistencies in the models and
the use of models that are inapplicable. " " *
The atmosphere of levels of resolution used -^
in dispersion models has both a deterministic and a
stochastic component. The winds are divided into a
mean wind causing pollutant transport, and fluctuations
about the mean wind causing dispersion. The division
is arbitrary.and usually ignores both influences that
produce variations in the mean wind of things like surfac
roughness, terrain, vertical wind shear, et cetera,
and the organization of some of the assumed stochastic
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
.1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
_ -._
components, due to thermals, surface roughness, wind ""
2 shear, terrain, et cetera.
3 Some are amenable to treatment. For example,
4 considerable progress has been made in the simulation
5 of winds in rough terrain, the effects of change in
6 roughness and remote measurements of the winds aloft.
7 Slide 3. The ultimate stochastic component
g will always lead to a minimum irreducible level of error.
9 By their nature =, atmospheric measurements correspond
10 to specific realizations whereas models by their nature
11 correspond at best to ensemble averages over the unmea-
"*
12 surables. These errors must be accepted, but they may
13 not be large. There are certainly vast improvements
14 possible over where we are presently
15 A second cause for errors in the standard
16 models is the inconsistencies in the model .itself. A
17 Gaussian model, and all the EPA recommended models are
18 Gaussian, assumes a split between mean wind and disper-
19 sions. The mean wind is held constant in the model
20
21
22
23
24
25
and the dispersive portion grows with downwind distance.
For example, in ths slide I show two puffs.
One would be at perhaps one kilometer, one at three
kilometers, and they're growing at a different rate.
Instead of being Sigma growing at X to the one-half,
Sigma grows as a different power of X, and so it's
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 ' WASHINGTON, D.C. 20005
-------
_ 176
I not proportional downwind distance. This inconsistency
2 results in the Gaussian model producing incorrect results
3 from multiple sources. The model is nonlinear and super-
4 position does not hold.
5 This problem is most easily seen in modeling
two adjacent area sources; using the virtual point source
7 method used, for example, in ISC, two adjacent ar-ea
8 sources disperse faster when modeled as one larger source
9 rather than two separate sources. Next slide. There31 s
10 the S one half compared to the Sigma Y variations.
Another example if provided by a simple gedanker
12 experiment. Next slide. Consider a plume point source
13 at a distance downwind. If the plume's concentration
r
14 pofile can be replaced by Gaussian source distribution
15 such that the two concentration profiles are equal at '
16 this distance where there's a cross-section slice, further
17 j downwind the two concentration profiles diverge.
18 These errors due to the Gaussian model -incon-
19 sistencies may dominant multiple source and urban model
20 application's, models such as ISC, RAM.MPTER.
2i Another major cause for dispersion model error
22 I is the use of models that are inapplicable to the case
i
23 being modeled. Examples abound. Single wind Gaussian
24 models .used in areas of complex terrain, high wind shearb
25 j| of shorelines, single mixing height Gaussian models
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
XV '- I'tJ .:
1 ; -' '& . 1!77
in "areas of shoreline diffusion. Next slide. This
is an example of some of the errors in, plume rise. Urban
heat island or mountain upslope winds or plume rise
formulas when strong stratification^ wind shear or downwas
is present.
Most real world applications have complexities
such as these that limit the standard model's accuracy
t t
and if dominant, make the model totally inapplicable.
Models have been developed for specific complexities,
but more efforts must be encouraged, and their results
evaluated and implemented.
Evaluation or validation of models should
befthe highest priority for contributing to their correct
use in decision making. Next slide. Only if the error
bounds of a model are known can correct interpretation
of the model results be made. Models can and have been
evaluated in application to a "few intensively monitored
cases. However, another complementary approach is possi-
ble, routine model validation, using the routine moni-
toring data taken in numerous regions throughout the
nation. Next slide.
Monitoring networks have been or are being
established in almost all the major regions of the country
both urban and industrial/mining areas. Next slide.
Many or most of these networks are automated with a
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. O.C. 20005
-------
; i:/:V"~-"_ .A 7 8
central computer for control and data collection. Modern
micro and minicomputers performing as central computers
have excess capability with which to perform concurrent
4 model calculations, but they're underutilized. Next
5 slide please.
Thus, for over a hundred regions, model testing
7 can be routinely conducted at a small marginal cost.
8 Next slide. Results of such tests can be compiled into
9 error bounds, directions for model improvements and
10 ultimately an accurate region specific model.
We can only guess and plan how well decision
12 making can improve if the accuracy of the models are
13 known for the specific regions and even for the actual
- r <
14 worst case conditions. Next slide. Next slide
15 And the next slide. The costs of applying
lg inaccurate models include the modeling itself and those
17 j associated with the decision's, based on model results.
18 Applying models typically involves an in-house modeling
19 specialist-or consultant to compile input data, make
20 | computer runs, analyze the results and generate reports
21 as well as agency personnel to review the reports, and,
22
23
24
perhaps, verify the results by applying models inde-
pedently.
Modeling specialist charges can run man-weeks
25 i to man-year per application. Input data compilation
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
l' "" ':-'- .-. - 17'9
may require local field monitoring. The computer charges
2 alone can be costly. A recent conference it was reported
3 that the ISC modeling requiring 32 hours on the EPA
4 Univac 1100/82 was conducted routinely, typically an
5 hour of computer time would cost $1000 commercially,
so we're talking about $32,000 for routine applications.
7 Some costs of model based decisions are well
8 known to everyone in the field. The costs, in human
9 terms, were brought home to me in the case of a youth
denied the funds to travel to an interstate competition.
*
The sponsoring local businesses were suffering from
12 closing of a major industry in town. It was closed
13 due to the high cost of complying with air quality con-
14 trols determined by inaccurate dispersion modeling.
15 Next slide. P rhaps the most demanding applica
16 tion of air dispersion models is in the planning of
17 an adequate response to emergencies such as a toxic
*
18 release or nuclear venting. Response time is critical,
19 and inaccuracy may mean severe injuries or deaths. Next
20 slide.
21 A response plan is based on a model. It may
22 be as simple as an evacuation of everyone downwind or
23 as detailed as multi-fluid hydrodynamics and exposure
24 calculations, next slide.
25 With limited resources available to respond
NEAL R. GROSS
COURT REPORTERS AMD TRANSCRIBERS
1330 VERMONT AVENUE. NW .
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
, - 180
- ,. - _-« _
to an emergency, correct focus, of efforts is vital.
Evacuation of peoples not endangered soay leave others
with a critical exposure. Next slide.
The best possible models are needed, and any
redicible inaccuracy cannot be tolerated. Next slide.
Furthermore, the models must be in a rapid response
easy to use package. Next slide.
Next slide. There is a modeling application
that is not entering into decision making supplemental
or intermitten control. In a supplementary control
system, emissions are reduced when there is potential
for deleterious impacts.
The needed reductions as based on model results
f
are produced by load reduction, fuel switching or engaging
effluent controls, for example precipitators.
It is not being used in the United States.
.In other countries, however, this methodology has proved
to be the most cost effective, in fact, we are developin'
a supplementary control system for a power plant in
Melbourne, Australia to control even photochemical ozone.
In keeping with the administration's push
toward industrial productivity and cost effectiveness
and with its decision'jmaking topic of this conference,
it is strongly recommended that supplementary control
be considered even with the present model inaccuracies.
NEAL R. GROSS
COURT REPORTERS AND. TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
J
--- -» .
Next slide. The concerns about decision making'
with inaccurate models were mentioned, model erros,
their causes and measurement, model costs and emergency
use. All have been the subject of recent advances in
the private sector. These advances should be reviewed
and applied in government decision making. Thank you.
MR. TIKVART: Thank you, Ralph. Time is grow-
ing short, and there are a number of people who I know
wnat to speak this afternoon. What I'd like to do is
take about a ten minute break.
I'm sorry. Were there any questions for Ralph?
What I'd like to do is take a ten minute break and start
promptly at 3:30 with Mr. Witten. Don't wander too
r
far please.
(Whereupon, a brief recess was called.)
MR. TIKVART: I hate to keep juggling the
schedule on you, but there are several people who have
indicated that they really need to leave this afternoon,
so I have six speakers left, and I would propose to
/
take the three who have indicated they need to finish
up in the following order, Witten, Konm and Wright.
The other three are Moon, Maxwell and Wurmbrand. Do
any of those latter three need to speak this afternoon
and leave .by a certain time? Do Moon, Maxwell or
Wurmbrand need to leave by a certain time? No, okay.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
182
' -
We will proceed with Witten, Kohn and Wright in that
2 order.
3 The next speaker then is Alan Witten from
4 the Oak Ridge National Laboratory.
5 MR. WITTEN: I would like to summarize the
6 stochastic air quality analysis effort at Oak Ridge
7 and discuss the implications of this work on dealing
8 with uncertainty in regulatory decision making.
9 Our work to date involves three aspects
10 single source model development, multiple source model
11 development and data analysis.
12
13
14
15
16
IT
18
19
20
21
22
23
24
25
The scope of this work is, however, expanding
to consider economics and health effects. Most of the
p
work was performed as a part of a regional assessment
for the Office of Fuels Conversion, Economic Regulatory
Administration of the Department of Energy.
The single source model is based on an analytic
technique utilizing Chi over Q values from an air quality
model in which the ground level concentrations are a
linear function of source strength.
The model used for the case presented here
is CRSTER. The Chi over Q values were generated by
running CRSTER with a unit emission rate. The model
requires the specification of a source strength cumula-
tive distribution function either as an analytic
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(2O2) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
18
19
20
21
22
23
24
25
183
- - .
expression or as a function constructed .from emissions
data. The model assumed perfectly autocorrelated emis--
sions. Autocorrelation data is not typically available"
so that some assumption is required.
The perfect autocorrelation limit was selected
as being most representative of large coal fired power
plants operating from a large coal pile and utilizing
coal preparation.-.prior to boiler feed.
The case presented here uses a longnormal
distribution function with a mean emission rate of 2000
grams per second, and a standard deviation of 600 grams
per second. Could I have the first slide please?
f . This slide shows contours of constant probabilit
of one and only one exceedence of the 24. hour PSD standard
for SO2.
You can see contours of ten percent, 20 percent
and 30 percent exceedence probability with the highest
probability occurring about four kilometers southeast
of the source. The .source is located at the coordinate
origin in the center of the figure. Next slide please.
This'shows contours of constant probability
of violating the 24 hour PSD standard for SO2. You
see five, ten and 15 percent contours. Once again,
the highest violation probability occurs approximately
four kilometers southeast of the source.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(2021 234-4433 WASHINGTON DC 20001
-------
;.''_. f.84
'* *" / c /
i
I The model also predicts violation the pro
2 bability of a violation anywhere in the region, and
3 for this case, the violation probability was 19.3 percent
4 The model can also be used to construct emission maps,
5 that is contours of constant violation probability where
6 the violation probability is defined as being anywhere
7 in the model region as a function of emission mean and
8 emission standard deviation.
9 Could I have "the next slide please? This
10 shows such an emission map. The horizontal axis mean
11 emission rate, the vertical axis is standard deviation
12 in the S02 emission rate, and this tells you the combina-
13 tions of these two parameters that can produce 5, 10,
^
14 15 or 20 percent violation probabilities.
15 If regulations were modified to allow, for
16 example, upto a 10 percent violation probability, then
17
18
19
20
21
22
23
24
25
sources would have a range, of^parameters in which to
operate while being in compliance. This is shown in
the next slide.
i
As in the previous figure,, the horizontal
axis is mean emission rate. The vertical axis is standard
deviation in emission rate. Staying in the white area
or the compliance side with these parameters will result
in violation probabilities less than 10 percent, so
that a source could operate anywhere in the white range
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
and" satisfy"this hypothetical^regulation. Before "going
on to the multisource results,. I'll briefly present
some preliminary data. Next slide please. We looked
at S02 data from a power plant that scrubbed. We have
inlet S02 data which is inlet to the scrubber and outlet
data which is the SO2 concentrations coming out of the
scrubber.
This is smooth inlet SO2 data. It's smoothed -
by taking a running 240hour average. The data was col-
lected at ISOminute intervals, so it represents a running
average of 96 data points.
The data is not does not show a great deal
of variability. The standard deviation here is approxi-
" f '-
mately ten percent of the mean. Next slide please.
Here is.a similar smooth plot time series of the outlet
S02, and here/you see that the variability has increased
greatly as a result of passage through the scrubber.
Standard deviation in this case is approximately
50 percent of the mean. This 'increase in variability
/
is an expected result. It comes about because of the
non-linear behavior of scrubbers. Next slide please.
This shows, the X's on this figure are the
actual cumulative distribution function for the inlet
SO2, and the solid line is the curve fit to a lognormal
distribution, and you see here that a lognormal
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
" " ----- fi ; 186
distribution in this case is a very good assumption.
Next slide.
Here is the same type plot, but now it's curve
fitting the outlet SO2 to a lognormal, and the curve
fit is not so good. In fact, that would almost defy
fitting to an analytic function. Next slide.
This is the autocorrelation in SO2 emissions
as a function of lag time for the inlet data. You111
see that it shows perfect autocorrelation. That's the
horizontal line at the top going right across at one.
Next slide.
The autocorrelation from the outlet data is
not does not show the perfect correlation; however,
r
the correlation time here is quite large. If you extra-
polate, you'll find that the auto correlation time
is on the order of 75 days.
That's all the data I have to present. I'll
go on now and talk about the multisource model. The
model is an extension of the single source method. It
utilizes similar assumptions; however, in this case,
the individual sources can be any correlation among
them can be specified. For the case shown here, the
sources, the individual sources were taken to be statis-
\
tically independent.
The model was used to assess the impacts of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, OX. 20005
-------
1
2
3
4
5
6
7
8
9
10.
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. . __ . r . v_ 187 !
«. * *
converting power plants-in and around the New York City
area to coal.
The 24 hour PSD standard for SO2 was used
as an indicator of these impacts. Next slide.
That's fairly dim. Can we dim the lights
any more? What this shows, whether or not you can see
it, is a base -map showing Long Island, Manhattan, portions
of upstate New York, Connecticut and New Jersey.
You may not be able to read the numbers but
you can probably see where they are. They represent
the source locations that were included in the study.
There are 30 sources. We took all of these 30 plants
to be unscrubbed. Next slide.
r
We used 107 receptor locations, and these
are shown in this slide. We used for the single source,
we used CRSTER to generate the Chi over Q values, in
this case we used RAM. Bear in mind that the region
we modeled is certainly beyond the limits of the RAM
model, but it's useful in demonstrating the method.
The model can allow for a PSD increment to
be independently specified at each receptor. This allows
us to calculate violation probabilities at each receptor
that represent violation probabilities based on the
locally available increment, and that's what we did
here. We broke the region up into areas of full increment
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
I two-thirds of the" increment" arid "half off "the increment.'
2 We ran the model for these 107 receptors, and constructed
3 contours of constant violation probability. We then
4 used that as a tool to identify sensitive areas or areas
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
showing high violation probability for further study.
Based on this, we selected 17 receptors for a more de-
tailed analysis.
Next slide. This figure shows the 17 receptors
selected. These sets of triple bars. Now in some cases
you only see one bar or two bars, but there actually
are three bars there. The location of these sets of
triple bars show the receptor locations.
The height of the first bar reflects the pro-
F
bability of no exceedences. That's drawn in green if
you can make out the colors. The height of the second
bar represents, the probability of one and only one ex-
ceedence. That's in blue,"and the height of the third""
bar which is red reflects the probability of the violation
of the 24-hour PSD standard for SO2.
.
The peripheral bar graphs,- what we call source
break-out for each receptor, you can see an arrow from
each receptor drawn to each bar graph, the height of
the bars in the individual graphs represent the relative.
contribution of the dominant sources of that receptor.
We can use this source break-out as a tool
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
t
8
9
10
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
...i. 189 »
* ' *
for identifying" sources that contribute to a high viola-
tion probability. For example, there's a receptor about
in the center, in the center of Long Island that's full
red. There's an arrow going to a bar graph, and that
shows, I believe its source is 7, 10 and 11 which are
contributing to that large violation probability. We
can then use this information as a tool for evaluating
controls to be imposed on various sources.
Next figure please.. This is another set of
triple bars. Once again, the locations of these triple
bars are at the 17 receptor locations on the previous
slide. The first bar represents thewiolatioh probabi-
lities as shown on the previous figure.
r
Now the colors don't show up too well. The
way these things read is it's the non-green portion
that represents the violation probability, so on the
first bar, it's the orange color, the height of the
orange color that represents the" violation probability,
and you see there are a significant number of receptors
with the very large orange region.
So based on the source break-out, we selected
three plants to scrub. The violation probability with
three plants scrubbed is the second bar of the set.
Now that's the blue area which isn't very distinguishable
from the green, but it turned out that only one receptor,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
190
»
-after scrubbing three plants, there was only one receptor
2 that had a non-zero violation probability, and I "11
3 go point out that receptor.
4 That receptor, in fact, had 100 percent viola-
5 tion probability. We went back/ looked at the source
6 break-out again and selected, two more plants to scrub.
7 That gave us a total of five plants scrubbed. The vio-
8 lation probability in that case is the third bat in
9 the set and at every receptor those bars are all green
10 indicating that with the five plants scrubbed, you would
achieve zero violation probabiliy at each receptor.
'
12 We were trying to get the violation probability
13 below ten percent, and we ended up getting it to zero.
r
14 The implication of this is that we satisfied a hypotheti-
15 cal ten percent violation probability standard by imposin
16
17
18.
19
20
21
22
23
24
25
controls on only five of 30 plants. - »
Excluding all other considerations "then, this-«~sa
would indicate just based on the 24-hour PSD standard
for SO2, any additional controls would be excessive.
Next slide please.
Okay. This is a repeat of the emission-'map
that I showed before based on a 10 percent violation
probability. The shaded area, the area shaded with
dots in the lower part of the curve shows the compliance'
region under the existing regulations. You see that
NEAL R. GROSS ,
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
r
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
it' s a very narrow region as compared""lib the 10 percent '
violation probability, and, in fact, most sources operate
in the end of that curve, in the end of that region
that's down towards the higher means.
This shows that under a probabilistic standard
that there is much greater freedom in how a plant is
operated. One way this method could be used in changing
regulations is to simply specify something like a 10
percent violation probability, anything upto 10 percent
is allowable.
Constructing a fuel map such as this or an
emissions map such as this and then allowing the source
to operate anywhere it likes in that region, this would
r
tend we feel that this would tend to decrease a utili-
ty's capital and operating costs by allowing greater
flexibility in both plant design and plant operations.
Such a standard would also likely reduce en-
vironmental effects. It would do so by encouraging
.emitters -to reduce the mean emission rate in order to
i
take advantage of the large standard deviation that1s
. available for the lower means.
I would also like to address a comment that
was raised before when someone made reference to Bernie
Steigerwalt1s discussion of the ExEx method and criti-
cized it for not accounting for plant reliability or
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. B.C. 20005
-------
18
19
20
21
22
23
24
25
equipment reliability. This method which is quite simi-
2 lar 'to the ExEx method, both the ExEx method and this
3 method have the capability of implicitly allowing for
4 component reliability.
5 If an emissions map such as this is used in
6 regulations then that woudl be the only restriction
7 placed on the source, The designer could factor in
8 any reliability information he would like and then any
3
9 upset conditions that occur at the plant are just another
10 data point that must fall in that compliance region.
11 ... I would like to conclude by briefly stating-
12 where we're going from here. We're about to embark
13 on a very ambitious study where we will look at some
14 of the implications of going to a probabilistic, regula-
15 . tory format.
. >
16 We will assume something like an allowable
17 10 percent violation probability, look at how a plant
would be designed under existing regulations, and then
also look at modifications to the plant design that
would be allowed under a hypothetical 10 percent limit.
We will investigate the economics of these
other scenarios to see if they will, indeed, provide
any cost savings to industry and utilities. We'll also.
look at the other side of the picture and look at the
environmental effects. We will be constructing a
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW .
(202) 234-4433 WASHINGTON. D.C. 20005
-------
3
10
11
12
13
14
15
16
17
18
19
'20
21
22
23
24
25
hypothetical city using demographic data. For each
scenario ..-.identified, we will be superimposing contours
of constant violation or exceedence probability. We
will be providing the health scientists any data that
they would like to see that they can use in making their
assessment such as specific information at particular
points, for example, probability of exceeding a certain
value, probability of exceeding a certain value three
consecutive days in a year, et cetera.
This will be done for a single source. Then
we'll look at how a probabilistic regulation will affect
development in an area, and we'll go through the same
kind of analysis trying to develop our hypothetical
r
city under the existing regs and under a hypothetical
probabilistic reg, and see what the differences are,
. i
and that's all I have to say.
MR. TIKVART: Thank you. Do you have a paper~*^'
that's available, or do you have copies of your slides?
MR. WITTEN: I've given copies of the slides
with the reporter. I'll leave him my notes for what
they're worth, and then I'll also say, the single source
method has been published. It appeared in the July
journal of the air pollution control association. That,
as far as I know, is the most recent. It's undergoing
internal review now, and will probably be submitted
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
probably somewhere in the next month. '
MR. TIKVART: Is Mr. Kohm still here? Do
you turn into a pumpkin in five minutes? Okay, can
we take some questions first? Any questions or comments
5 of Mr. Witten? No?
6 Thank you very much. The next speaker is
Robert Kohm representing Alcoa.
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
MR. KOHM: My name is Robert Kohm, and I'm
the manager of Environmental Planning and Analysis of
the Aluminum Company of America, Pittsburgh, 15219.
Under agreement with the chairman, I'm not going to
read my prepared results. You 'can all cheer now.
The my have been appropriate first thing this
r
morning, but they're a little bit. esoteric for this
time of the day. What I will do is jus/t summarize a
little bit what, in fact, was in there.
The first point tha't I wanted to make was
the fact that.really when we looked at the dispersion
models, .1 don't think that there's anybody in this room
:
who would be willing to stand up .now.and say that they
believe the dispersion models are accurate all the time.
I'd like to see who that would be, you know,
waste time trying to convince you at this point that
dispersion models accuracy is subject to debate, but
I also throw out a question that if everybody can put
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
10
11
12
13
14
15
16
17.
18
19
20
21
22
23
24
25
.* ' 195
- * -
their- biases aside and forget about any particular ego
trips they have to suffer because they're a model develop-
er, just think in terms of what is the likelihood of
making any quantum leaps towards modeling accuracy.
I think the probability of making a quantum
leap toward the improvement of'modeling accuracy is
quite small, so what we have to do is face reality,
in fact, that although the system may not be perfect,
we can improve upon models incrementally. We1re not
going to tackle in any meaningful way the discrepancies
and agreement with data we've seen presented here today
and yesterday.
So with that in mind, I think that realisticall
r
Alcoa's position is that we need to move rapidly towards
a probabilistic compliance approach. It seems to be
that EPA1s approach has been one. of if it can't be done
exactly, we're not going to do it at all, because we've
caught enough flack in the past for doing things prema-
ture .
/
We're not going, to catch it this time, so
we're going to let the status quo fly. There are some
differences. The ExEx model is a refreshing change
in view from the standpoint that the oftly way you can
meaningfully interpret those results in through a probab-
ilistic compliance attitude, but there's another more
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
196
" *. -
pressing problem, and that is that even if we were to
2 today have the authority to make the changes, to say
3 we're going to move towards a probabilistic compliance
4 approach, there would be a great deal of consternation
5 by state agencies and people who have to interpret the
modeling results because it's a new area for them. It's
7 an area that's beyond their level of expertise, and
8 true understanding of dispersion modeling results re-
9 quires literally a graduate degree in statistics, and
10 there probably aren't very many around, so when you
11 get to that point, the people who have to interpret
12 those results are going to be dealing with things that
13 are not in their bailiwick and that's going to cause
r
14 some more problems.
15 So what we'd like to recommend is that EPA
16 make some administrative changes'directly and expedi-
17" tiously to allow for probablistic compliance arguments
18
19
20
21
22
23
24
25
in the dispersion modeling field, and really not make
any direct recommendations as to thou fehalt do it this
way as has taken place in the past and let people take
a creative scientific approach to the interpretation
of data as exhibited to. some degree by the last speaker
and evaluate what comes in and through that experience
develop an understanding of what people can come up
with.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
*- 197
^
You can sit in ivory towers and try to formulate
the best life, but regardless of what happens, somebody's
going to come up with a better way that we haven't heard
up. So I'm just saying open the door and see what the
cats drag in, and review that with a realistic viewpoint.
Consider it honestly and whether or not it
makes sense. At this point, I'm willing to answer
any questions you might have.
MR. HELMS: Does the issue of consistency
bother you, going that approach?
MR. WITTEN: Oh, the issue of consistency
bothers me immensely, and what bothers me is that we
try to be consistent. Again, I'll take anybody in
this room who feels that we should take the models
as currently defined by the guidelines and apply them
universally everywhere that you're going to get the
same level of accuracy.
I don't think there's anybody here that believes
that, so the issue of consistency is just really a
way of presenting your own, you know, it's easing the
administrative burden, and one way to ease the administra-
tive burden is to stop trying to enforce your way of
doing things and listen to what is a meaningful way
of interpreting the data and whether it makes sense.
Open mindedness is a virtue.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
'2021 234-443.) Mf»rv'vrl-nv "it"
-------
1 MR. RHOADS: It doesn't bother you from a
2 I'll rephrase it. Does it cause you great concern
3 to feel that if you went into North Dakota, you would .
4 enter a different regulatory climate, different models
5 and.different interpretation of those models than if,
6 for example, you went into South Carolina?
7 MR. WITTEN: We have that today from region
8 to region. Don't we? I mean, that's the status today.
9 We're just accepting that fact rather than trying to
10 gloss over it.
11 MR. TIKVART: Any questions from the floor?
12 Okay. Thank you very much. The next speaker is Ray
13 Wright speaking for the utilities air regulatory group.
'
14 MR. WRIGHT: Good afternoon. My name is
15 Ray Wright. I'm .Director of Environmental Affairs of
16 the Ohio Power Company, a subsidiary of the American
17
18
19
20
21
22
23
24
25
Electric Power Corporation); and I am speaking today
on behalf of the Uility Air Regulatory Group commonly
referred to, as UARG.
UARG is an ad hoc organization made up of
88 electric utilities, the Edison Electric Institute
and the National Rural Electric Association. UARG's
members produce most of this country1s electricity.
UARG is pleased to attend the Second EPA
Air Quality Modeling Conference. I will outline UARG's
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10.
11
12
13
14
15
16
17
18
19
20
21
22
23
24
.*v- , 5- -- 199
f II ^
iJi
views on the use of atmospheric models and the Airtie
House recommendations. We will submit: more complete
written comments on these issues and others raised
during the Conference before the close of the docket.
We support the Agency's efforts to re-examine
the proper role of modeling in regulatory programs
as exemplified by the Airlie House Workshop and this
Conference; however, we are concerned with the inc»;easing
rigidity with which models are being used in an ail-
regulatory management programs.
Rather than being used to provide informq^ion
to decision makers over the past several years, uniested
models have been pushed beyond their limits. They
. r
have been used to impose emission limits on the bacjis
of single value rare events, even though most expei \ s
concede they perform most poorly when predicting t|(,.,se
events.
In many instances, modeling decisions have
caused .or have threatened to cause serious economic
consequences with no prior investigation of the accu
of the model.
Many times monitored data has been set au|
in favor of modeled estimates with emission limits
being based more and more on hypothesis in an obsei --.iLi
25 In the past models have been par:t of a decision mat- i m.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBES
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
t >.:
s?r , ' . T.Z, ;; 200
process which has failed to give adequate consideration
to the uncertainties associated with model predictions.
UARG believes that realism must be the central
goal in all modeling efforts. Realism is the goal,
the inclusion of the uncertainty information and.=a ^
regulatory process will help focus on this goal.
We concure with efforts such as the EPRI
Plume Model Validation Project, the Woods Hole Workshop
and AMS-EPA Cooperative Agreement which indicated that
model performance needs to be further evaluated and
that model uncertainty should be quanitified and reported
routinely as part of the decision making process.
Also, we endorse generally the efforts and
many of the recommendations from the Airlie House Work-
shop, and we hope that many of the Airlie House recommenr
dations will be elaborated upon and ultimately be re-
flected in EPA's air quality management program.
Now, as to UARG1s specific recommendations,
we support the Airlie House recommendation or recommenda-
tions that model selection should be a flexible and
cooperative process with participation by all interested
parties.
That process should insure the use of the .
most realistic model in all instances. With realism
as the goal, all models used in regulatory programs
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVENUE. NW .
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
, "', , : 2u§ 'f r . : »"--' .; 201
should be tested against available data, to establish
levels of skill and performance. No model should be
recommended by EPA, thereby establishing it as the
model of choice until it has been evaluated.
The skill of the evaluated, recommended model
should establish the minimum acceptable performance
level for other models of that class. If no tested
model exists, any theoretically sound computational
correct model should be acceptable.
Apparently from yesterday's presentations,
EPA has begun evaluation of some models. The information
presented yesterday morning is the type that should
be. made public for peer review and discussion prior
to the use in a regulatory context.
We stress that this review process must be
open to the public with full disclosure and ample
time for comments. . -.«.
For' instance, the .CRSTR evaluation reviewed
yesterday for the Muskingum and Stuart Stations offers
an example of the need for careful public participation
in modeling evaluations.
Although we have not had time to review and
prepare comments on the conclusions presented, we are
concerned that EPA's analysis improperly averaged pre-
dictions and observations. The evaluation results
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
- "" ' --ic i !!: ".::. »:'-' ,. 202
were reported for the~99th percentile, not the statis-
tically less stable but normally used highest second
high concentration which is currently used to set emis-
sion limitations.
The other results referred to as demonstrating
the skill of the CRSTR, RAM and other models misleads
one to believe that these models are accurate to within
10 to 40 percent. These data should be open to public
scrutiny and peer review before they are accepted for
use in regulation.
UARG will submit additional comments, on these
data. I would like to request today a copy of those
studies and data that were cited in yesterday's pre-
esentation. I could give you this afterwards. It's
in the footnote. .
In order to minimize disputes in technical
modeling, UARG endorses modeling protocols developed
in open and cooperative meetings prior-to the modeling
exercise. .The protocol should be detailed and should
bind the regulatory body to the same extent that it
binds the applicant.
Reasonable yet firm deadlines should be set
for both parties. UARG recommends briefing the responsi-
ble decision maker on the measured technical recommenda-
tions contained in the protocol and on major areas
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
; - -sr- n- ' ' ' --, :;: .203
i. "j
I of dispute which arose during this negotiation. Discus-
2 sion of these key technical.(.points iprior to the modeling
3 exercise would afford the decision maker a greater
understanding of the uncertainties involved in the
5 specific areas.
6 There is a need for some technical review
7 group composed of agency and non-agency modeling experts
8 to review numerous types of mod ling issues. Such
9 a group should be available to render technical advise
10 on disputes which arise in the development of modeling
11 protocols and in the selection of models for recommended
12 status.
13 f Such a group could oversee many of the func-
14 tions suggested in the Airlie House Summary for a model-
15 ing center. If such a group is established, EPA must
16
17
18
19
20
21
22
23
24
25
assure that it is adequately staffed and funded so
that it does not become an impediment to modeling analy-
sis.
UARG recommends that the modeling community
should seek to reduce the sensitivity of regulatory
decisions to all uncertainties by identifying and quanti-
fying the uncertainties.
Currently the perspective of the decision ;
maker is narrowed by the fact that reliance is placed
on a single prediction out of the multiplicity of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 . WASHINGTON, D.C. 20005
-------
1
2
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
204
of available model outputs. The use of one extreme
value based on freak, sometimes never observed, condi-r
tions often adds to the arbitrariness of decisions.
UARG recommends that uncertainties be made
explicit thorugh all available means, in all modeling
related decisions. It is important to develop more
effective methods for conveying modeling uncertainty
in a meaningful way.
Model sensitivity analysis and Monte Carlo
simulations may produce useful technical data on the
short-term. Field evaluation will provide more meaning-
ful information in the more distant future. The use
of the protocol briefings to inform the decision maker
of the uncertainties at an early point may also be
useful. - -
UARG expressed strong support for continuing
the efforts started with the ExEx methods designed
to incorporate probabilistic concepts into the modeling
framework in.place of unrealistic worst case modeling
approaches.
In the interim, we urge the use of alternate
model statistics in a decision making process. UARG
will submit additional comments concerning the replace- .
ment of the highest,, second highest concentration esti-
mate by other statistical measures of the type
NEAL R. GROSS
COURT REPORTERS AND . TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.'-.'-;' . 205
recommended by the Airlie House workshop. Use of such
alternate statistics might serve immediately to reduce
the impact of model uncertainty on final decisions.
UARG feels that additional flexibility is
needed in PSD decision making.' Since PSD air quality
increments are not strictly health-or welfare related,
the decision maker should not be constrained along
a rigid pathway.
.... 3
Generally, air quality management decisions
should reflect the uncertainties in models, the regula-
*
tory goal to be protected and the burdens imposed on
the regulated industry.
The decision making process should encourage
r
thorough consideration of available monitored data
especially where modeling results are highly uncertain
. >
or only poorly reflecting reality.
. .
Since monitoring data are in fact reality,
they can be used both as a check on model performance
and to temper the interpretation of available model
output.
Because realism should be the primary goal
of air quality modeling, improved and more realistic
procedures must be incorporated into the Air Quality
j Modeling Guidelines.
Since this will necessitate frequent revisions,
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
J330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
IT
18
19
20
21
22
23
24
25
UARQ" recommends that the Guildeline be updated at least
every 18 months. Under this schedule, every second
revision would coincide with the triennial EPA Modeling
Conference. However, as major changes in modeling
guidance occur, such as the incorporation of newly
recommended models, the Guideline should also be updated
In all cases, proper comment and review proce-
dures must be followed. As a matter of policy, however,
advances in modeling techniques should not be the cause
for continual review of established SIPs or PSD permits.
As the Airlie House participants noted, emission limits
based on past model demonstrations performed in good
faith should be grandfathered.
r
Even if past modeling results are drawn into
question through improved models, the inequities which
would result from a reanalysis of the existing emission
limits counsel strongly for grandfathering. Only when
monitoring establishes that a primary standard is being
violated should mandatory reanalysis be permitted.
In conclusions, while models are a necessary
elements in our current air quality management programs,
they must be applied with their inexact nature in mind.
I stated previously that UARG will be for-
warding additional comments in more detail.
MR. TIKVART: Two remarks. One is most of
., NEAL R. GROSS
COURT REPORTERS AND . TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
1
2
3
4
5
6
7
i
8
9
10
11
12
13
14
15 .
16
17
18
19
20
21
22
23
24
25
207--- .
the report referred to-and Bill Cox's presentation
are already available as contractor reports. I say
that with minor qualification. There may be one of
those reports that have not been released yet, and
I will have to check that once I see your list, but
the majority of those reports with one exception are
already available as contractor reports.
MR. WRIGHT: Yeah, I think one that was foot- ,-
noted number 4 was still in draft form.
MR. TIKVART: Okay. That's probably the
Teknekron report. That's the one I'm referring to,
so all but one are alreadl available.
The second point, with regard to the 99th
percentile, we also wanted to show the second highest,
but because these are reports that are already completed1/L
the individual values are lost, and we could only use
the values that were available to use from the report.
MR. WRIGHT: Well, I think that was the original
Mills work .that might have been done in '75.
MR. TIKVART: That's correct.
MR. WRIGHT: And as to the Muskingum situation,
there were four monitors. One was background. Three
were in-line downwind, and it showed overprediction
close in and underprediction far out, and about 10
to 40 percent.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
r-.--- . 208
' '"*
MR. TIKVART: Whatever the figures were were
reported.
Any other questions or comments? Any from
the floor? Okay, thank you, Ray. I'm sorry. Val?
MR. DESCAMPS: Val Descamps, Region I, What
frequency were you urging for the guideline update?
MR. WRIGHT: Well, we're not urging any fre-
i
quency at this time. I only pointed out that if we're
going to use like a 99 percentile to evaluate a model,
that doesn't seem to be quite squaring: with the fact
that we use second max1s to regulate. There should
be some consistency to make sure evaluation does re-
flect the regulatory process.
MR. TIKVART: Thank you. The next speak
on. the list would be- Don Moon. Don, if you care to
make your presentation now, go ahead.
MR. MOON: I'd better go while I still have
somebody left out there. Good afternoon. My name
is Donald W; Moon. I'm supervisor of Salt River Pro-
ject's Air Quality Division. Salt River Project, a
political subdivision of the State of Arizona, provides
electric power and energy to more than 300,000 residential,
industrial, commercial and agricultural consumers in ;
and around Phoenix and other portions of central Arizona.
SRP is a member of the Utility Air Regulatory
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
"021 234-4433 WASHINGTON D.C. 20005
-------
I'l . I >
:ri.i :;: " . ,f 209
I Group and endorses the major conclusions' given by them
2 at this conference. We're also a contributing member
3 of the Electric Power Research Institute and fully
4 support their plume model validation efforts.
5 SRP is the operating agent for the Navajo
g Generating Station/ a 2250 megawatt coal-fired generating
7 station located near Page, Arizona and the Coronado
8 Gneerating Station near St. Johns, Arizona, comprised
9 of two operating 350 megawatt units, and an additional
10 350 megawatt __unit under construction.
11 We are also part owner of the Mohave," Four
12 Corners, Craig and Hayden coal-fired generation stations.
13 Needless to say, we are vitally concerned about the
14 appropriateness and the accuracy of atmospheric dispersion
15 modeling as it affects our operation and the uility
15 bills of our consumers.
17 Today I am going to address only one aspect
18 of the problem the incorporation of model uncertainty
19 in regulatory decision making.
20 First, I believe we must recognize that uncer-
21 tainties exist to a degree that serious problems are
22 being encountered routinely in applying model results
23 in the regulatory arena. It's clear from discussion
24 presented in this conference and in the scientific
25 literature that such uncertainties do exist. The
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
210
I regulatory decirion makers need to fully acknowledge
2 that existence and effect on s.iting and abatement deci-
3 sions
4 Secondly, we must recognize that modeling
5 uncertainties are not going to vanish or decrease in
magnitude appreciably in the foreseeable future. The
7 record clearly shows that improvements in modeling
8 accuracy have been painstakingly slow. Regulatory
9 decision makers need to recognize that the solution
to the on-going uncertainty problems will not come
through pie in the sky basic model improvements.
12 Thirdly, and perhaps most importantly, we
13 must recognize that many of the regulatory problems
r
and uncertainties that are posed can be solved by regula-
15
16
17-
18
19
20
21
22
23
24
25
tbry or statutory change. Regulators needs to look
at alternative solutions to the dilemma generated by
the use of modeling.
They need to examine those standards for
which modeling is virtually the only technique available
/
by design to derive related emission limits, and where
possible use more straighforward empirical approaches.
We must remember that in nearly all cases
the use of modeling versus some other analysis technique
or combination of techniques is an administration deci-
sion.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. OX. 20005
-------
10
11
12
13
14
15
16
IT
18
19
20
21
22
23
24
25
In examining" the" incorporatibri of model uncer-
tainty in regulatory decision making, we must not simply
resolve ourselves to living with such uncertainties,
but rather take positive regulatory and administrative
steps to minimize or eliminate such uncertainties.
Let's examine the Federal statute concerning
the use of modeling. The Clean Air Act requires that
analysis be performed so as to assure that certain
pollutant sources do not cause.or contribute to air
pollution in excess of National Ambient Air Quality
Standards, Prevention of Significant Deterioration
ambient increments, visibility impairment criteria
and air quality related values.
f
The Act implies, but does not specify, that
stmospheric dispersion modeling will be the principal
tool for such analysis. Such use is as the discretion
of the Administrator.
Section 165(a)(2) concerning preconstruction
requirements limits issuance of permits to those facilitie
for which, quote, "the required analysis has been con-
cuted with regulations promulgated by the Administrator,"
unquote. Thus, if the Administrator chooses modeling
as the only means of analysis, regardless of suitability
or reliability, that section becomes law.
Section 165(a)(3)(d) also concerning
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
'.--. ~212
preconstructioh requirements states that the administra-
2 tor, quote, "shall specify with reasonable particularity
3 each air quality model or models to be used under speci-
4 fied sets of conditions," unquote. Thus, modeling
5 is implied to be preferred analytic tool and although
suitability of models to specified sets of conditions
7 is addressed to some extent, the issue of reliability
8 or accuracy is not.
9 Section 169(a) (3)(b) concerning EPA's report
to Congress on visibility impairment requires recommenda-
11 tions for, quote, "modeling techniques (or other methods)
12 for determining the extent to which man-made pollution
13 may reasonable be anticipated to cause or contribute
14 to such impairment," No mention is made of suitability,
15 reliability, accuracy requirements expected or desired.
Section 171(2) concerning nonattainment areas
17
18
19
20
21
22
23
24
25
states, "The term 'nonattainment area1 means, for any
air pollutant an area which is shown by monitored data
or which is calculated by air quality modeling (or
other methods determined by the Administrator to be
reliable) to exceed any national ambient air quality
standard for such pollutant."
Thus modeling is inferred to be reliable,
and all other methods must be determined to be reliable
before they can be used.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.-7-;- :: , 213
Section 320 provides guidance to the Adminis-
trator cm the conduct of air quality modeling conferences
such as (.his with special attention to be given to
appropriate modeling necessary for carrying out part
C of title I relating to prevention of significant
deterioration of air quality.
No guidance is given concerning the suitability
reliability or accuracy requirements expected or desired.
Further, no guidance is given concerning their use
if uncertainties are found to be unmanageable.
It is clear then that analysis, not necessarily
modeling, is needed to set emission limits so as to
assure that pollutant sources do not cause or contribute
r
to air Pollution in excess of National Ambient Air
Quality :«laiulards, PSD ambient increments, visibility
impairment tlnd air quality related values.
It is also clear that considerable latitude
is given i h,.« administrator in establishing how that
analysis wiijht be accomplished, or even if separate
analysis U^yond that accomplished in the setting of
new sourv<> performance standards is necessary.
i"'t me repeat that. Or even if separate
analysis Ul like to address some specific areas of
f.( NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
' 1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
concern where modeling uncertainties pose serious problems
in the regulatory decision process and suggest some
possible administrative or regulatory solutions.
Perhaps the greatest modeling uncertainty
is encountered in the use of non site specific and
unverified models in complex terrain and seashore applica-
tions. The degree of uncertainty has never been quanti-
fied to everybody's satisfaction which by itself leads
to serious conflict.
It's been our experience that uncertainties
of upto a factor of ten may be encountered usually
biased towards overprediction. Such uncertainty creates
havoc for planners when tolerances are narrow as for
PSD permitting and potential costs of abatement are
out of sight.
We have no overall solution to that problem,
but would like to offer the following suggestions to
decision makers.
One, recognize that modeling.results are
only fir approximations. Two, examine all other analysis
results including these using tracer and comparative
measure teciniques. Use representative ambient measure-
ments wherever possible, particularly where multiple
units are involved, multiple follow-on units, that
is, anci, four, verify model estimates with on-site
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
22
23
24
25
-:-;- ' * 215
1 ambient measurements and modify source emission limits
2 accordingly. ;
3 Uncertainty is encountered also through the
4 use of multiple worst case input parameters. To exa-
5 cerbate things, the probability that concurrent worst
case events might occur is usually ignored by the modeler.
7 To eliminate that uncertainty, v;e suggest that all
8 inputs be expressed in probabilistic terrnr>, as in the
9 expected exceedance methodology.
10 All inputs, not just fuel sulphur content
was described by Dr. Steigerwalt yesterday. Such a
12 requirement would necessitate that all models be modified
13 to^accept inputs in probabilistic input, a rather small
14 cost considering the potential costs involved.
15 The results, of course, would also appear
16 in probabilistic terms, which would be far more amenable
17 to standard statistical treatment than present outputs.
18 Conversion of inputs to probabilistic terms should
19 be done regardless of whether standards are ever re-.
20 stated in probabilistic terms.
2i Another modeling uncertainty is encountered
in the modeling of shorter term concentrations as comparec
to the modeling of longer term concentrations. The
uncertainty is undoubtedly greater in complex terrain
than if appropriate peak to mean ratios were applied. .
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005.
-------
I We would suggest that where short term, that'
2 is 24 hours or less, dispersion estimates cannot be
demonstrated to be reasonably suitable, reliable and
4
5
6
7
8
9
10
11
12
13
- 14
15
16
17
18
19
20
21
22
23
24
25
accurate, that is at least within a factor of two,
that only the annual average.ambient concentration
estimates be used for .emission setting purposes.
If shorter term comparisons need to be made
for other purposes, we would recommend the use of annual
to shorter term ratios as derived from actual mean
measured values in similar terrain circumstances.
Such an approach would not only help minimize
the uncertainties involved, but would also considerably
streamline the permitting process.
p (
The use of the standard steady state Gaussian
modeling techniques has been extended in time and space
beyond reasonable bounds forcing ever-increasing uncer-
tainties in the process. This, is basically the problem"
in attempting to estimate shorter term concentrations.
It is also certainly true in attempting esti-
i
mates beyond the 50 kilometer range from a source.
We suggest that no steady state technique be attempted
beyond 50 kilometers. Only dynamic technique should
be used.
We recognize that statutory change beyond
the scope of the Administrator in this conference may
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
;/ . .{; ',.- 2i7
- - ,*< -i - *
, < - -
be necessary to provide additional and sufficient relief
from modeling uncertainties.
This is true specifically of the PSD increment
system where the basic problem of atmospheric loading
from distant sources is not necessarily reflected in
ambient concentrations which we are attempting to model.
It is also true to some extent of the dispersio'n
module of visibility modeling attempts, of long distant
transport associated with possible acid precipitation
and of stack height modeling attempts.
In summary, the Administrator has many avenues
open^-within the guidance given in the Clean Air Act
to reduce modeling uncertainties and associated permitting
r .
conflicts and costs.
She should not try to live with the uncertain-.
ties but rather take positive steps to minimize or
eliminate those causes of uncertainty within her juris-
diction. There is need to eliminate shorter time period
modeling, to eliminate the use of multiple worst case
input parameters, to limit the spatial extent of steady
state modeling, to use measured ambiet values and extra-
polation techniques wherever possible, and, in general,
to not place such absolute reliance on dispersion modeling
results as has been done in the past.
We thank you for the opportunity to participate
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
218
in this conference. If there are any questions, I'd
be pleased to replay.
- MR. TIKVART: Any questions from the floor?
No? Thank you, Don. Next I have Mitchell Wurmbrand
representing Western Energy Company. 1 hope I've pro-~
nounced -your-name not--top badly,_pn ^several -occasions-.,
MR. WURMBRAND: Good afternoon. My name
is Mitchell Wurmbrand. I am employed by TRC Environ-
mental Consultants of Weathersfield, Connecticut. I
am a certified consulting meteorologist, and part of
my work involves assessing the accuracy of air quality
dispersion models. - -
I am speaking today on behalf of Western
F
Energy Company of Billings, Montana. Western Energy
Company is the owner and operator of the Rosebud Mine
near Coal Strip, Montana, and Western Energy Company's
ability to secure necessary permits to operate is strongly
influenced by the accuracy of existing air quality
dispersion models.
Western Energy Company appreciates the oppor-
tunity to address this conference. Today, I will talk
about two deficiencies in the existing air quality
models and in the way the models are applied by decision
makers.
These two deficiencies introduce a systematic
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
5
6
7
8
9
10
-Hr -
12
13
14
15
16
17
18
19
20
21
22
23
24
25
^.- '. rtr^~~_^*_. __ ' *oi n
_.-.-i^r_r^::; «"- ' 219
error in the prediction of total suspended particulate
matter downwind of open pit and area surface mines,
such that TSP concentrations computed by the models
are significantly larger than those that actually occur.
"The consequence of these deficiencies is
that some- surface mines that should otherwise be granted
air quality permits may instead be denied permits,
and that overly conservative modeling studies may predict
total consumption of PSD increment when, in fact, some
"PSD increment remains. ' "
. - -., _,-.^. ...WhiLe-,the consequences .of: the-modeling def IT --.--,
ciencies are most often felt at the state level, parti-
cularly in states whose enabling legislation requires
that fugitive dust consumes PSD increment, the remedy
can be effected most easily by EPA in its congressionally
. »
mandated role of fostering greater standardization
,»
in accuracy and modeling.
The first deficiency is the absence of a
deposition.algorithm in the type of dispersion model
that is frequently used to simulate surface mines.
Every meteorologist familiar with the intricacies of
modeling surface mines knows that correct simulation
of deposition of particulate matter is essential to
computing realistic concentrations, yet surprisingly
there are some western states in which the regulatory
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW .
(202) 234-4433 WASHINGTON. D.C 20005
-------
" .- \"K-^-Ov£U~~- . *- 220.
agency responsible for granting permits routinely simu-
lates the air quality impact of a mine by using models
that totall ignore deposition.
The use of these simplistic models introduces
a systematic error, because concentrations predicted -
6 by a model- that ignores deposition will always be larger
7 than a model that properly simulates deposition.
Let me give you an idea of the magnitude
of this systematic error. An EPA publication titled
survey of fugitive dust from coal mines presents a
curve of source depletion factor versus downwind distance
12 for a windspeed of five meters per second and a particle
13 settling velocity of five centimeters per second.
This combination of wind speed and settling
15, . velocity is one that is representative of conditions
at a typical surface coal mine. At one kilometer from
17 the source under neutral stability conditions the value
18
19
20
21
22
23
24
25
of the source depletion factor is about four tenths.
Assuming that the source depletion factor
correctly describes particle settling behavior, this
EPA curve says that a model which totally neglects
deposition will compute particulate concentrations
two and a half times larger than actually occur.
Of course, we all know that an aravis magnitude
is not uncommon in simulating atmospheric dispersion.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
" " " " «"~ . ,1 _ _. _ ,
- ----- r:^ - r 22a
But, we generally tolerate errors of a factor of two
only when we do not know what causes them. The conse- .
quence of ignoring deposition is that the modeling
simulations overestimate particulate concentrations,
and if feed ^simulations are used as a basis of checking
whether a given surface" mine will comply with ambient
air standards, then it is conceivable that a model
be wrongly denied a permit.
In regions of the west where several surface
mines are clustered together or in states whose regula-
tions allow fugitive dust to consume PSD increment,
the consequence of ignoring deposition in modeling
simulations is that the available PSD increment for
the total suspended particulate may be erroneously
consumed. .
While there are some who will argue that
deliberate neglect of particle deposition is desirable
because it makes the regulatory.analysis more conserva-
tive, I would point.out that there is already a great
deal of conservatism built into the emission factors,
models and the ambient air standards.
If more conservatism is desired, then it
should be added knowledgeably, not haphazardly. The
failure to properly account for deposition is not a
fault of the available models. There are a host of
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
--,. ->T f f'rt Hf ' rv'"hf"rf*i*f "IT
-------
-'--'--.."- *. - 222
models that simulate particle and aerosol deposition,
and I would refer conference participants to Dr. Ray
Hosker's excellent paper titled, "Practical Application
of Air Pollutant Deposition Models: Current Status,
Data Requirements and Research Needs."
Furthermore, EPA has funded numerous deposition
model development projects, one of which culminated
in the industrial source complex model, now part of
the unimap series.
10 The failure t6 properly include the effects
of particle deposition is actually a failure to.achieve
12 consistency in applying available models. To remedy
13 this situation, I ask that EPA require that regulatory
14 agencies empowered with the authority to review surface
15
16
17
18
19
20
21
22
23
24
25
mines, stimulate particle deposition in a manner con-
sistent with the existing technology. This requirement
can be enacted through the'state implementation plan
and attendant delegation of authority.
The second deficiency in the.existing modeling
review of surface mines concerns'the'manner in which
the available models simulate transport of particulate
matter from inside the pit to the mine boundary. Many
different fugitive dust producing operations occur
inside ..the pit, including overburden removal to some
extent, coal drilling and blasting and. coal removal.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
3 ""*" ^. "t V * " " " ** 1 O
~' f a *" 2 2,3
.*
Depending upon the mine configuration, there
may be a large portion of the mine's total haul road
emissions generated in the pit as well. In the western
states, these pits are sometimes as deep as 800 feet,
although I would guess that an average depth is more
likely to be 200 feet.
It is intuitively obvious, and recent measure-
ments support this observation that only a fraction
of the fugitive dust generated within the pit escapes
to the mine boundaries. Yet, when it comes time to
11 model a surface mine, the regulators generally employ
12 a flat terrain model that assumes that pit emissions
13 ocpur at the same height as the surrounding terrain.
14 More complex models capable of handling area
15 sources> such'as ISC, do not adequately simulate the
16 mine either, because they cannot accommodate emission
17 heights that are below the receptors. In fact, in
18 the ISC code, if the source is below the receptor,
19 the ISC model prints an error message and automatically
20 terminates execution.
2i The emission factors associated with surface
22 mining activities result from measurements taken inside
23 the pit, generally very close to the pit activity in :
24 questions. So what we have here is a situation where
25 a surface mine is modeled as if its pit emissions occur
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
224
at ground level, and these emissions are then transported
to the mine boundary along a horizontal plume center-
line, when, in reality, the pit emissions occur well
below grade, and only a fraction of this dust escapes
from the pit.
The phrase that is used to describe this
phenomenon is pit dust retention, and I suggest that
currently used dispersion models do not adequately
address pit dust retention. The failure to simulate
this phenomenon in existing models causes the models
to predict ambient particulate concentrations higher
than those that actually occur. As with the previous
problem I just discussed, this deficiency could also
prevent the permitting of some mines or could prompt
unnecessary control measures in order to demonstrate
on paper that a mine meets standards.
I ask that the TPA "as the driving agency
for research methods and air quality and meteorological
modeling address the issue of pit dust retention. Addi-
tionally the decision makers and" the regulatory process
and the scientific community involved with modeling
should be made aware of this modeling deficiency. Thank
you.
MR. TIKVART: Any questions from the floor?
Thank you very much. Is David Maxwell still here?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
225
Do you wish to give a presentation now? Okay.
MR. MAXWELL: I appreciate the opportunity
to be last, but I'm going to follow up briefly on what
Mitch just said. It's fortunate that the last two
dealt with mining, because I was surprised that in
the entire two-day session very little of it concerned
mining, and that is a big problem out west.
MR. TIKVART: Can you state your name and
affiliation?
MR. MAXWELL: Yes, I'm about to do that.
I was about to say I'm Dave Maxwell. I'm an air resource
coordinator with ARCO Coal Company in Denver, and I've
had experience in state and federal government including
EPA Region, 8, consulting companies, and now industry.
I just .want -tp start out by saying, that-the National.
Research Council in its recent study entitled, "Control-
ling Airborne Particles" concludes that improving modeling]
capabilities will require as a first step, expanding
emission inventories to include the size distribution
and chemical composition of conventional ducted emissions,
fugitive process enussions, anthropogenic and natural
fugitive dusts and othor natural sources.
Currently, the most dependable and accurate .
models are those which <>:;! imate local air quality impacts
of a single point smn ov, such as from a smokestack.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C 20005 ..,-
-------
226
Estimation of air quality impacts from miming
operations, however, is much irtore complex for several
reasons. First of all, mining operations are composed
4 of multiple individual sources, such as area sources,
5 such as emissions from conveyor belts/ storage piles,
and loadout areas among other things. Secondly, mining
7 operations are often located in complex terrain as
8 | they are in the western United States, surface and
9 underground mining, and this reduces the reliability
of modeling results.
11 So this reliability is compounded not only
12 by the different sources but also by the terrain. Third-
13 ly, the mining operations have high variable particulate
14 emissions over time due to the nature of individual
15. , operations, and because- of variations in meteorological
10 conditions, and, fourthly, the fugitive emissions ori-
17* ginating deep within strip mines or open pits do not
18 get totally emitted out of the pits, as Mitch said.
19 The in-plume particle removal rates or plume
2o depletion for low level fugitive sources is probably
2i much greater than current dispersion models allow.
22
23
24
25
The combination of source complexity, variable emission
output and high dependence upon variable meteorological
conditions together with conservatism in modeling
predictions will make mining operations prone to predicte<
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
. - _ -y-: ' -227
violations of PSD increments as well as national ambient
air quality standards. Furthermore, the wide variation
in the use of emission factors from such studies as
PEDCo and EPA studies contain a significant degree
of uncertaintly for model emissions input.
Now we may ask what are mining companies
doing to resolve model uncertainties. Well, ARCO among
other companies has sponsored a fugitive emissions
dust study which has taken about three years to complete,
and it has involved up in northeast Wyoming which is
the center for mining activity in the western United
States for surface mining, and it's taken three years
to do a detailed and rather laborious study to determine
the effects of fugitive emissions on the impact of
mining operations, and the re-suits, even though they-1 re
not printed yet, they will be published soon, indicate
preliminarily that the fall-out or deposition rate
is much shorter, a much shorter distance than what
has been given credit for in any models, and among
other things, it includes a look at sedimentation rates
and also in addition to the deposition, it looks at
the variable meteorology conditions that tests were
made in, so I would recommend that as part of this
record, if I can, submit to you a copy of this study
that was performed for a consortium of energy companies
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
228
in the western United States if we can get it before
the September deadline that you mentioned.
We really think that this is a way where
industry is trying to look at the uncertainties in
modeling of emissions from mines, and particularly
surface mines, and we hope that not only the federal
agencies will look upon this, but the state agencies
as well, because they're the ones in the west that
will have to work with the federal government, and
some states in the western United States have PSD authorit
to make the decisions -themselves. - "
So we feel that this is one way that the
mining industry has been taking a good look at the
impact of their own mining emissions on the impacting
of the environment in the-western United- States, and
I may say that the mines in the western United States
are pretty far away from any major metropolitant areas
also.
Only a few of the available dispersion models
have the ability to vary fugitive emissions with wind
.speed, and none have the ability to explicitly conserve
the effect of precipitation on emissions. This variation
could be significant since precipitation can reduce
emission to virtually zero.
Although several theoretical and empirical
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
.:.; :-> I1- .. ' - ^c- 229
~ s ;_ ' '
techniques simulate gravitational fall-out and turbulent
deposition, none of the settling and deposition technique:
have been adequately validated by field studies, and
we realize this is a very difficult process, and we're
working on that, so we're not trying to look at the
regulatory agency as being hard on the mining industry.
We're just saying, let's look at all the facts that
we have, and together let's evaluate some of these
models that have been previously developed and also
that may be developed in the future.
So we want to work with the regulatory agencies
in developing the best way to determine an impact on
the mining operation on the environment. I want to
conclude that although significant process has recently
bee made in the development..of. models.which .can predict ,,
regional impacts of sources of sulphate and other fine
particles, similar process has not been made toward
the development of regional models to estimate regional
air quality, impacts from inhalable particulate sources,
and this is where the breakdown comes that we discussed
yesterday.
We're looking at a certain breakdown, so
the mining industry feels that if you're going to have
a particulate standard, why not look at the health
effects, namely the ones that deal with inhalable
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
is- ' ' 230
. . i '--" i ; _.- '»""
particulates, 15 microns or less or whatever standard
o is set.
.
3 Further study of sedimentation rates of various
4 particles by size and composition is needed so that
5 atmospheric residence times may be more realistically
incorporated into dispersion models.
7 The ISC model which was mentioned, it's the
8 only dispersion model that addresses deposition, but
9 does accurately simulate the dispersion of fugitive
10 dust in mines..
11 The dispersion model that would incorporate
12 factors unique to the mining industry would eliminate
13 much of the guesswork in the modeling of mines and
r
14 thus enable regulatory agencies to more accurately
15 determine the impacts of the mine on PSD increments
16
17.
18
19
20
21
22
23
24
25
and the national ambient air quality standards.
So I would conclude that we've started some-
thing. We've gotten the regulatory agencies to work
with the mining industry, and -we need to work a little
.
harder. We need to deal with model uncertainties,
work with the mining industry, the regulatory agencies,
and any research and consulting companies and studies
that need to be developed to more accurately determine
what effects dispersion modeling has on the impacts
of emissions from mines, and out in the western United
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
{202} 234-4433 WASHINGTON, D.C. 20005
-------
6
7
8
9
10.
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
n; .. ' 231
States fugitive' dust is the big issue out there, parti-
cularly in the less populated states, like Wyoming,
Utah and Colorado where mining is becoming a more evident
source of energy development, and an important environ-
mental issue out in those states.
Thank you for your attention, and I'm willing
to address any questions.
MR. RHOADS: You mentioned that ARCO had
been participating in a rather- large study of fugitive
emissions for the last three years. Have any of the
regulatory agencies been participating with ARCO on
that study?
MR. MAXWELL: To my knowledge, the study
f
I'm referring to was basically a study that was done
just by the mining industry; however, any time something
like this is done, they would like, once they get the
results and the internal agreement among all the com-
panies, they would like the comments from the state
and the federal agencies on that.
At this point, that study is not available
for publication, but hopefully will be very soon. I'm
not the one to say when it will be, but we would like
to send a copy to the federal and state regulatory
agencies involv d.
MR. DICKE: I would just like to comment
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2?''
,-._ *' >n » 232
* *' - -
that while certainly we would like to have a model
which accomplishes a number of -these depositions, sedi-
mentation, wash-out, rain-out processes, no model that
was essentially no model has been submitted to EPA
for review in the context of the solicitation of models
last year, so none of the models that we have received
so far from private developers have accomplished or
incorporated these mechanisms which you're suggesting.
So we would be just as happy to receive one
for review as you would, I'm sure.
11 MR. MAXWELL: Right, and I think one thing
12 that has to be looked at, I think mining operations
13 can certainly improve their environmental programs
p
14 simply by even watering their haul roads on a regular
15 basis, and ARCO has done some studies on that that
indicate that watering haul roads' maybe twice an hour
17' depending on climatic conditions is just about as good
18 a control method as using chemical dust suppressants
19 on a more frequent basis.
20
21
22
23
24
25
So I just think a simple conclusion like
that would be enough to maybe help in the development
of an environmental program for a mining operation,
but maybe in giving a little more credit as far as
the modeling goes of the particular impact that the
mine would have.
NEAL R. GROSS
COURT REPORTERS AND .TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
I believe that most of the modeling that
involves fugitive dust really doesn't look too much
into the control methods involved, so I think that
we have to work with the regulatory agencies on giving
you information upon what we determine as good control
measures on mining operations to keep down dust.
A simple thing like watering haul road, EPA
used to give 50 percent credit for. Our studies indicate
higher than that, so I think that we can maybe work
on this together, not only ARCO but other mines to
work on a better way of improving overall emission
factors.
MR. TIKVART: Any questions from the floor,
p
observations?
MR. HENDERSON: I'm Donald Henderson from
the National Park Service, and I wanted to ask Dave
if the mining industry has decreased any of the uncer-~
tainties in the emissions for mining operations suffi-
cient that it would warrant applying models that have
deposition in scavenging.
MR. MAXWELL: At this point, that's a difficult
question to answer. I can't speak for the other mines.
I know that one of the things that the dust studies
that they've been looking at and some of the work on
the fugitive emissions, the mining consoritum has been
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, O.C. 20005
-------
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
looking at has involved some of this work. I'm afraid-
it's going to take quite a while, just like it has
in model development to actually come up to EPA or
the state agency and say look, we've got better data
thatn you have, so, first of all, we're in the process
of gathering the data, publishing it, and then submitting
it for comment.
When you get a consortum of companies, each
one is going to have different pet peeves, and we've
found that out in this case, and the difficulty is
to get everybody to agree.
Just like innany group of people, you're
going to get people who are more in favor of something
r
than another, so the first step was to get everybody
on the industry side to get together and agree on a
particular format and the way things were done as far
as the methodology. - \
Now we're in a stage of publishing it, and
now we're going to have to submit it for peer review,
so I can't say right now that that'll be any startling
.revelations or not, but at least it's a start that
a series of industries have gotten together and tried
to improve the situation and make it easier for, I
think, everybody to evaluate the impact of fugitive
emissions in a mine.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 2000S
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
' 20
21
22
23
24
25
-;"' ~V :-' ,, ' .~ ; 235
MR. HAMBURG: Pardon me if I can't help saying,
but there seems to be a meeting of mines.
Anyway, I would just like to make a suggestion.
I've been away from this deposition business for many
years, but in reviewing some ' sorry, Fred Hamburg,
RMC, et cetera.
In reviewing some of the current literature
on models that include deposition terms, I find that
they're extremely primitive compared to models that
were put out about 20 years ago in an entirely different
area, namely fall-out prediction.
We kind of got out of that, because the nuclear
money ran out with the end of the Nevada testing, but
f
the approach there, I think, should be reviewed. Much
of that literature had been confidential. They were
. ..- i
never into the secret category, but they were classified
as confidential, and they are surely available in the
open literature today.
I.think you will see'some explicit terms
used that could be incorporated into prediction models
today. I doubt very much, however, if the gaussian
approach is going to work with these. These were all
numerical type of puff type models, and it's the only
way they can get it to work.
One paper that I remember because I had
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW .
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
occasion to write a criticism of it back in the days" '
of the Journal of Meteorology before it became the
Journal of Atmospheric Sciences, was a paper on the
D model put out by Al Anderson of Navy Radiological
Defense Lab.
He had incorporated as much as he could in
the open literature of some of the concepts that were
being used in deposition type models, and I would just
suggest that you might"want to look into those things.
MR. MAXWELL: Right. That's a start. You
know, we have a long way to go, and I think it's up
to industry to work with the regulatory people and
even consultants as a third party for peer review not
^
only with visibility. I think visibility and the mining
models are just about in the same boat, except visibJlity,
I think, is ahead, because they've had some models
that have been printed up, and that has as least been ~
printed.
I don't think anything has been printed as
.
an official document and guidelines .for the mining,
.but I think we have to work together, because it* s
a very important issue, and mining emissions are related
to visibility, particularly in the western United States,
so I look forward to working with a lot of different
people on this, because it's such a wide open project.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
' :.v '' '":'-.- ' 237
i" - "
-It's-going to become a very important issue, and I
thank you.
MR. TIKVART: Bill?
MR. BONTA: Bill Bonta from the Maryland
Air Management Administration: I was a little bit
surprised to hear, Jim, that you said you didn't get
any inputs. Actually although we didn't submit Our
exponentially tilted plume model that we have developed
D
specifically for use in fugitive dust situations and
have used a number of times on coal export terminal
operations, similar, but not exactly the same, it wasn't
submitted to you in a formal fashion, but it was sub-
mitted, and I think you remember that.
r
We did show that with the glass bead experi-.
ments that were used to validate the ISC particle set-
tling subroutine that, in fact, this model of ours
seemed to produce slightly better correlation with "
the measured field data than did the ISO's tilted plume.
What's more, the exponentially tilted plume
does show that in far range deposition results which
we don't have too much data in the glass bead experi-
ments to begin with, and especially with the heavy
particles that there is a very much greater deposition
velocities and far lower concentrations.
But that you can have that, we can give
NEAl R. GROSS -
COURT REPORTERS AND TRANSCRIBERS.
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2 comment is on these emission factors.
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
238
you that pHper~any~time~~you want it, but one other
We find in evaluating these terminals going
into Baltimore that just one operation, a hundred ton
coal car dumpers, rotary car dumps, dumps it out flat
on the ground, we get estimates' of emission factors
that range over three orders of magnitude, just for
that one operation alone, from looking through the
literature, and I don't see how you can stand around
here, and everybody says just routinely that Tikvart's
models are inaccurate and need to be improved when
you look at the garbage in, what are you going to get
on the other side.
MR. TIKVART: Okay, thank you very much.
That completes the list of speakers we have for this
afternoon. Was there anybody that wanted to speak
that didn't get the opportunity to do so?
Any further questions or observations? As
announced in the Federal Register, the conference will
continue tomorrow. We know of at least one speaker
that wants to make a statement tomorrow morning. If
there are no further comments, we'll close for now
to be opened tomorrow morning at 9 a.m. Thank you
for your attention and contributions.
(The meeting adjourned at 4:54 p.m.)
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
239
!!
h. -
CERTIFICATE OF REPORTER
I hereby certify that the foregoing transcript
represents the full and complete proceedings of the 3-11-81
aforementioned matter, as renor.ted and reduced to type-
writing under my direct supervision.
NEAL R. GROSS
>
3
NEAL R. GROSS .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
w
E
D
N
E
S
D
A
Y
A
U
12
-------
1
2
3
4
5
6
7
8
g
10
11
12
; i ^
* """" ~JL
GOVERNMENT OF THE UNITED STATES
ENVIRONMENTAL PROTECTION AGENCY
2nd CONFERENCE ON AIR QUALITY
MODELING.
.Wednesday, August 12, 1981
The Conference was held in the Thomas Jefferson
Auditorium, South Agriculture Building, 14th Street &
»o
°- --Independence Avenue, S. W., Washington, D. C., Joseph
Tikvart, Chief, Source Receptor Analysis Branch,
Conference Chairman, Presiding.
14
15
16
*>
17
1R
PRESENT:
19
20
21
22
23
24
25
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
Joseph Tikvart
Richard Rhoads
James Dicke
-------
19
,20
21
22
23
24
25
MR. TIKVARTT^Good morning. I'd like to
welcome you back to the third day of the Modeling Con-
ference. This morning we do have two scheduled speakers,
so I'd like to begin with that right away and perhaps
we'11 .have some informal discussion after those two
speakers or others that wish to volunteer are done*
The first speaker, speaking for the National
8 Academy of Sciences, is Myron Uman.
9 MR. UMAN: Thank you Joe. I speak for the
10 Academy only when I describe results of the Academy
11 studies. Anything else I say is for myself.
12 ' My remarks in fact, are based on the work
13 performed by the Academy's recent study of the implemen-
r
14 tation of the PSD program as published in the report
titled "On Prevention of Significant Deterioration of
16 Air Quality."
1" The report contains a detailed analysis of
18 the existing PSD provisions and, of special interest to
this Conference, a discussion of air quality modeling
for use in PSD decision making.
For the record, I submit excerpts from the
Academy report on the questions of accuracy and compar-
ability of models and of a statistical.approach for
managing air quality.
The report points out that there are likely
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
g
fc
\ *-
always to be uncertainties in the results of air quality
2
3
to expect.we shall be able to develop the definitive
deterministic model, one that is manageable and affordable
and also accounts for every detail of the dynamic
7
meteorology in any terrain.
Q
We shall, of course, try. But decision makers
9
will, in all likelihood, have .to use a predictive tool of
limited capability. It behooves policy makers to take
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
modeling for specific events, such as the highest or
second highest concentration, because it is impractical
account of the limitations when they define the objectives
and rules of the decision process.
In the technical community, the classical
r
methods for dealing uncertainties incorporate the prin-
ciples of statistics and treat outcomes probabilistically
Hence, the Academy's PSD report suggests that modeling
would be a more effective tool for PSD purposes, if the
short-term increments had more robust statistical criteri
for compliance than one permitted violation per year.
The Academy also believes the same conclusion
can be reached about the NAAQS. The draft recommendations
of the American Meteorological Society, presented earlier
by Mr. Fox, echo the Academy's conclusion. The work in
EPA on expected exceedances is in this spirit but does
not address the problem of uncertainties in modeling.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
6
7
8
9
10
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
While the Committee did not go farther than
this, because its charge was limited, I would like per-
sonally to suggest that EPA take another tack in approach-
ing the problem of statistical analysis applied to air
quality management, an approach that is more in keeping
with the Academy's conclusions, the recommendations of the
World Health Organization, and, I believe, those of the
AMS.
The analytical approach described by Mr.
Steigerwald does not deal with the uncertainties inherent
in atmospheric modeling, but only with time variations in
emissions. Thus, modeling for expected number of annual
violations of one, even over a period of ten years, is
r
still looking for rare events, way out on the tail of
frequency distribution.
The "exex" concept as incorporated in the ozone
standard also raises the specter of having to regulate the
frequency distribution of emissions, as Mr. Steigerwald
suggested.
I think EPA would be on firmer ground if it
concentrated, at least for now, on the problem of uncer-
tainty in atmospheric modeling. The data being presented
at this Conference, an elsewhere, strongly suggests that
it is much easier to validate models for frequency distri-
butions of concentrations than models for worst-case
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 734-4433 WASHINGTON, O.C. 70005
-------
1
2
3
4
5
6
- 7
8
9
10 .
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
events. The NAS report, the AMS draft, and a WHO position
paper all argue for a statistically robust measure of
compliance which in regulatory terms would be expressed as
an ambient concentration that could, on the average, be
exceeded a specified number of times per year, preferably
more than once.
So, based on effects data and understanding of
the stochastic nature of atmospheric processes, for ex-
ample, a 24-hour concentration'could be selected that
could be exceeded 18 days per year without significant
risks to air quality.
When I use the term "significant" in this
context, I mean statistically significant. The concen-
trations of pollutants representing the levels of ac-
ceptable rist would be chosen on the basis of effects
data. Other, lower concentrations that could be exceeded,
say, five per cent of the intervals per year, would be
determined based on the statistical characteristics of
atmospheric and meteorological phenomena.
This idea, of course, is not new. The value
of moving towards statistically robust measures of com-
pliance has been pointed out by many individuals and
groups. I hope there is some way this Conference can
convey this message to those whose responsibility it is to
develop the rules within which the modeling community
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
operates. Thank you.
MR. TIKVART: No questions from the panel. Any
questions from the floor? Comments? Okay, thank you.
The second speaker we have this morning is
Richard Kerch from the National Coal Association.
MR. KERCH: My name is Richard Kerch with
Consolidation Coal Company and I'm here today representing
the National Coal Association. Keep this very brief and
let us all get back out into the rain.
I'd like to take the opportunity to make some
brief, extemporaneous comments this morning, primarily"
to alert the Hearing Panel that NCA. will be providing
more extensive written comments to the record after we've
had a chance to reflect on some of the things we've heard
today, and get our thoughts together as an organization
in responding to the theme of this Conference.
^sT-"
- -.- -.c^^waj
We find that this whole process a very laudable
effort by EPA and we want to let them know that we support
the notion of formally quantifying the various uncertain-
ties associated with modeling and finding how these un-
certainties can be incorporated into policy and regulatory
decision making.
We would note, however, that the modeling
community which can probably be relied on to quantify
uncertainties, perhaps is not the proper group to rely on
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17"
18
19
20
21
22
23
24
25
- ~2S-_ -
wholly in integrating uncertainties into decision making.
Perhaps EPA needs to broaden their audience in this area
in soliciting advice.
The second point I'd like to make today is
more specific to the coal industry. I apologize for not
hearing the last two speakers yesterday afternoon, and I
may be covering some ground that was covered yesterday
afternoon. Most of the emphasis on, if not all the empha-
sis, in any modeling conference that I've ever been to has
been on the bouyant, elevated release from point sources.
In the coal mining industry, we have the unique
situation of having non-bouyant plumes released at, or
near ground level from area sources. We think that the
fact that very little effort has, by comparison, gone into
developing appropriate models to treat the emissions from
surface coal mining is, in fact, a source of uncertainty.
I'd like to point out three areas that, in
particular, that we have some concern: One, as an indus-
try, we're not satisfied with the treatment of deposition
at the ISC model gives; we think that there would be more
realistic ways to handle that.
Secondly, we think that there probably are
better methods of handling the non-buoyant, ground-level.
releases from area sources. I heard some information
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
19
20
21
22
23
24
25
"..:= " - - - 8
given today, I forget by which, or during this conference,
2 I forget by which speaker, that related to the fact that
3 if you took an area source and calculated the concentra-
4 tions and then you divided the area source in two pieces,
5 and modeled it, you got a different result. That doesn't
6 sound right to me.
7 The third area that we would try and create
8 some attention to is the problem of emissions which are
9 generated below ground level, typically in a Western
10 surface coal mine, we have pits which are anywhere be-
ll tween 100 feet to 400+ feet beneath the surface. A good
12 fraction of the total emissions coming from that surface
13 mine are generated in that pit and we know that they all
r
14 don't get out. But today's models that are available to
15 us pretend like the earth is flat and all of these fu-
16 gitive emissions are accounted for in the model.
17- We would urge EPA to place some of their
IS resources in addressing these areas.
The third broad comment I would like to make.
A group from NCA environmental committee met informally
together last night. We'd like to inform the panel that
we plan to take up the challenge of doing some creative
thinking that was thrown out at the audience by Dennis
Trout yesterday and, really, which is the underlying theme
of this Conference , to attempt to address some of the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
"how-tos" rather than the "shovjlds". Or, if we are not
very successful in coming up with some ideas on the "how
-tos", we plan to take a couple of steps backward and look
at .the whole issue from a little more broad focus and see
if there might be some alternatives that might allow us
to finesse the whole question.
ri
The last comment I'd like to make today is
c
primarily a personal opinion and not so much representing
q
the National Coal Association. We heard some information
.11
12
13
14
15
16.
17
18
19
20
21
22
24
25
presented at this Conference. I know Mr. Tikvart made a
presentation to one of the Senate or House Subcommittees
on"modeling-recently, Mr. Rhoads sent out to the partici-
f
pants a statement on the accuracy of models, and as a
scientist, I take little comfort from the fact that models
such as the CRSTER model which has been identified to be
accurate +_ 10-40% when it is comparing distributions of
high concentrations to distributions of observed concen-
trations cannot predict the spatial and temporal location
of those high concentrations and does a significantly
worse job at doing that.
I think while the highs estimated and the
highs predicted some place around the source match, that
gaives me no real comfort that we know what's going on in
the model. You're being essentially right for the wrong
reasons, and I think that in the PSD process, it's even
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
-------
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
10'
more scary that we don't know where the high is located
or when the high is located.
For a remote source, that may not be a problem.
But, when you start to bring in the second source or
multiple sources, I think it's important if we're going to
have any chance of keeping track of increment, and not
placing undue control requirements on the subsequent
sources coming into an area, that models do better than
just predict what the high, or second high, or distributio
is in the immediate area.
We've got to know what the location-of that
high is; we need to know when the time is.
" That concludes my remarks today.
MR. TIKVART: First an observation and then
a question. Regarding your last point, I don't think we
know for sure that it's the models that cause the problems
with time and location. It may well be our knowledge of
what the atmosphere is doing. . ± suspect it's a contri-
bution of both. But I don't think the results that have
.
been presented indicate that the models cannot accurately
estimate the concentrations in time and space, but that
certainly the data bases are insufficient.to describe what
the atmosphere is doing.
Now, unless you can describe what the atmos-
phere is doing, the models can't estimate concentrations
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(2021 234-4433 WASHINGTON. D.C. 20005
-------
11
correctly. But I think that there's contributions from
both .in that and, perhaps, another contribution is our
knowledge of what the emissions are.
Okay. The question. You suggested that we
should bring in a broader group of people to look at the
inclusion of uncertainty in decision making rather than
just modelers. Could you expand on that? What other
sort of people should become involved in this and, perhaps
in what form?
1° MR. KERCH: Okay. Let me respond to your ob-
11 servation first while I'm trying to think of an answer-.to
your question.
13 I think I used the term modeling semantically
different in context than, perhaps, you do. I look in the
way I used it then, I used the word modeling to mean the
. . 4
16 whole process of starting with some data of some sort and
I? ! coming out with some computer sheets at the end. And, in
that sense, modeling doesn't give me the comfort.
Whether you have the right, actual algorithm
in there, and I think that's what you meant when you used
the word modeling, I was not prepared to argue that with
you.
MR. TIKVART: Okay, I think semantics problem is
that many technical people don't appreciate. In other
18
19
.20
21
22
23
24
25
words, when they say modeling, they think of the models
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(207) 234-4433 WASHINGTON. D.C. 20005
-------
.'.1
f"
f-
1"
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
_j 12
right away and forget about the data base problems.
MR. KERCH: With respect to your question,
I think you got me. I guess my comment is a reaction
against placing all of this burden on the modeling
community. I think that it can provide some service in
terms of identifying the technical questions.
But when you begin to ask a technical body to
deal with non-technical issues, then I'm not as comfor-
table with the results that are likely to come out of that
exercise.
Whom I would suggest to be contacted, I think
I'll further think on that, and include that in our
written comments to you.
r
MR. TIKVART: Okay, because as a technical
person, I share your concern. We have to get the right
people to ask the right questions, and we can't, as tech-
nical people, tell the decision makers what questions
they should ask us.
We would like to provide them with the infor-
mation they think they need, so, if you give us some
further input on that, I think that would be good.
MR. KERCH: Thank you very much.
MR. TIKVART: Any questions or observations ;
from the audience?
Okay, thank you, Dick. At this time, is there
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
j '
1 anyone else who would'like- to make a presentation, ask a
2 questions, or ask a observation? Name an affiliation,
3 please.
4 MR. HELLUMS: Lloyd Heliums, Phillips Petrol-
^ eum Company. Yesterday, EPA and the other Governmental
6 agencies issued a request requiring outside organizations
1 or the public in general to make contributions in order to
8 help out with the question of uncertianty in modeling.
9 May I address that this morning?
10
18
19
20
21
22
23
24
25
MR. TIKVART: Yes, sir, you may. Would you
11 care to use the lectern?
12 MR. HELLUMS: It doesn't make any difference
13 to me.
14
MR. TIKVART: Why don't you come up?
lo MR. HELLUMS: May I break this problem of
1R
uncertainty into three areas. The first area I'd like to
17
mention is source inputs. These are emission-type of
variations and characteristics of emissions and the
uncertainty involved in the emissions whether it be
a petroleum plant like we in the oil companies understand
or whether they be a coal mining operation, there are
always uncertainties in the emissions and the type of
emissions being involved. This is the first point I
would like to mention.
The second area of uncertainty is in the
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS "
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
i
14
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
meteorological variables themselves, because there is a
certain amount of stochastic processes involved in the
atmosphere yet, however, I think that without question,
patterns and certain types of trends can be shown to hold
in meteorlogical conditions, as some of our meterological
friends yesterday have pointed out.
So that the thing is not completely without
patterns and discernible type of phenomena going on that
can be characterized to some degree.
The third area of uncertainty is in the model
itself. Now, there are many types of models. You can
have the numerical partial differential equation models
in,, 3-D, three-dimensional space, or you may have a Gaus-
sian Plume model which is a derivation of the partial
differential equation models under certain assumptions,
and is an algebraic type of model.
You may have many different variations of the
Gaussian Plume model itself.
So let me back up. Let's, first start, if you
want to start pinning down the uncertainties, you've got
to break it down and use some control and separate out
these variables. And, the best way I know of doing this,
is to start with the Gaussian Plume model since it is the
recognized equation being used presently by EPA, and which
includes, of course, the Briggs' Plume Rise formulation
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
* . . J.D
4 " " "i
1 f _ in withv-that and ±o/take a -look at that .under, very con-. "
2 trolled type of situations using the best data that we
3 have available.
4 The best data we have available, at this point
5 in time, is that which has been collected over the past
6 30, 40, 50 I don't know how far this goes back because I
7 first started taking a look at these questions in '77,
8 so I'm not a meteorologist, I'm not, my background is not
in this area, my background is in complex modeling and
10 non- linear optimization.
However, it is a mathematical problem .and
whether you're modeling the atmosphere, atmospheric
problems, or whether you're modeling problems in physical
sciences as .in container reactor vessels or other type
13
14
15
16
17
IS
problem.
19
20
21
22
23
24
25
thing, the mathematical formulation and then resolving of
the solution to determine an optimal best set of a that
model to the data is still the same basic mathematical _
.So-, the problem I want to tackle with you this
morning is a fundamental problem using a control data base
and today, that data base the best we have is Prairie
Grass, Hanford, Ringo, Cross-through, Ocean Breeze.
API, American Petroleum Institute, has collec-
ted -about 1.7, in fact, I think 17 exact is the number,
data bases in which the best information that is available
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
18
19
20
21
22
23
24
25
K. . -. 16
y t '
k'^T ' - ' - - '-'... "' ' "
1 today, in the area of modeling, has been assimilated into
2 a computer-library type format and we are now proceeding.
3 to analyze this data base.
4 In this particular data base, you do have the .
5 meteorological variables defined as probably as best as
any base to date. There are still some areas of unreli-
ability even in that data base in the fact that wind
o
direction does vary, and the actual wind direction or
g
effective wind direction, if you're looking at tracer
10 releases where you have arcs then, out, proceeding away
from .the source of release, your effective wind direct-ion
12
at a given arc will vary according to what has happened in
10
the process from the time of release to the point where
particular pollutant rates that particular arc.
So, this is an unknown within the system and
i fi
it must by itself be determined in the optimal analysis.
17
Now, I want to define now what I mean by optimal analysis.
I want to determine, since the Gaussian Plume equation
has coefficients, dispersion coefficients for one, I
want to eliminate the Briggs1 Plume-rise formulation that
has some more coefficients which must be determined, so
if we go to the data base I'm talking about, there all
released at some particular elevation, either ground, or.
above ground, so we can eliminate the Briggs' Plume rise
equations from our formulation, and we can look strictly
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C 20005
-------
6
7
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
1.7
at the basic, fundamental equation of the Gaussian Plume
formulation.
And the major coefficients within that are
Sigma y's and Sigma z's. They are a function of distance
and, also, stability.
Let's make a mathematical formulation, a
least squares mathematical formulation in which we can
optimate, determine the coefficients that will best fit
a Gaussian Plume formula to the data base, to a given test
Now, I want to bring up a point here. In the
past, when people have calculated from the basic data
Prairie Grass, Hanford, you can go down the list, to the
best of my knowledge, and I must say that because my
j-
knowledge is limited as far as the literature in this
field, to the best of my knowledge, what they have done is
they've calculated Sigma y's in a directform way and then
they've gone and calculated the crosswind integrate con-
centrations or CWIC's, if you'll allow me to use the
abbreviation, and when they do this, they separate them.
They break apart the Sigma y and the Sigma z
and say it's independent because that's what they assume
in the way that they make the analysis, the mathematical
analysis.. If you will lo;qk at the Gaussian Plume formu- .
lation, you can mathematically prove that it is impossible
to break apart the Sigma y and the Sigma z determination
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NVV
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
18
j 7 f _ , , _, ;
independently of each other like this. It's a very simple
little proof, you take your least squares formulation and
then simply by taking the derivatives and setting them
equal to zero, you can come up with what is in the op-
timal solution for fitting a Gaussian Plume formulation
to a given data set and you can very quickly ascertain
that the problem is inseparable for a Sigma y and Sigma z
Therefore, Sigma y and Sigma z in order to
determine coefficients for the Gaussian Plume formulation
which will optimally match the data must be done simul-
taneously.
The problem is further complicated in that
Sigma y's and Sigma z's must be formulated in such a
manner that they're a function of downwind distance.
Okay, this is our basic problem. About a
little over a year ago, in studying what had been done and
looking at the problem, I realized that the only way
we were ever going to come up with a really accurate
determination of what the uncertainties in Gaussian Plume
model is, is this particular exercise would have to be
done. It's a rather complex, constrained, least square
optimization problem. So, I proceeded to, I'm on a API
task force, AQ7 non-reactive pollutant task force, so I
looked .at this and the questions came up to me about
separability and there was a lot of question about it, so
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17-
18
19
20
21
22
23
24
25
I said give me a little time and the next meeting I came
back and I had produced a mathematical proof form showing
that the Sigma y's and Sigma z's are not separable and
that, in fact, if you wanted to determine Sigma y's and
Sigma z's which will optimally, in a best-fit manner,
represent the data, it must be done, they can not be
calculated as has been done in the past, but must be
calculated using different type of procedures more in-
volved mathematical techniques-.
And, off the top of our head, it looks like a
fairly simple little problem, three or four weeks, I can
set up a mathematical analysis package to do that, Well,
one year later, the problem was much more difficult than
r
I had anticipated. It was a very, very, nastywhen
actually I got into it I found it was a very, very,
nasty, curving ridge, back surface in space.
A very difficult problem to solve in the
optimization field which I have been working in ever since
1959. And, therefore, I found out that, also it had to
be a constrained optimization. It took me a year before
I finally resolved all the technical difficulties and the
numerical to resolve and solve this problem.
But, they have been solved. At present, within
Phillips, we have a air quality state of technology budget
The management at Phillips has been very receptive to the
NEAL R. GROSS
COURT REPORTERS AND . TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
20
if: !?>( -'-
proposition of trying;to do some fundamental work to help'
out and to looking into the technical problems in the air
quality field. And, with my making some presentations to
management, they have agreed to devote some money, some
computer, which it requires a lot of computer capability
to make these types of analysis. We do them on a 3033.
The programs are very large. To just get one little case
solved which may be just Hanford, for instance. In Han-
ford, we have 73 cases within that data set.
To just analyze one of those cases takes about
five minutes, however, that is five different analyses
within that. The program is now being set up and we're
starting to run. I just got through Hanford, the first
r
35 cases which is a zinc sulfide pollutant emissions, it
took 262 minutes of CPU time.
But we are going to process that entire data
base and what is coming out of it is a rigorous mathe-
matical analysis in order to determine what the exact
optimal coefficients Sigma y's Sigma z should be to give
an optimal fit to that data.
This will then allow us to tell you exactly
how good a plume formulation which is only one of many
models will actually fit the data, because these are
optimal solutions.
In other words, this is the best solution that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS'
1330 VERMONT AVtNUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
OS-
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
21
-a rv ' *' ' * "
' ,- i " ;
the Gaussian plume formulation can get through that
particular data base, and as you can see, we got hundreds
and hundreds and hundreds of data bases that we're looking
at individually in this analysis.
And it is an optimal fit. You're not going to
to be able to do better in this fit. Now you can go to
a different model and you can make a better fit. But,
with a Gaussian Plume, the standard that we're using, the
RAM, the basic RAM model"is what we're presently using in
the analysis, which is, by the way, the same as is, we
just got finished doing a comparison of ISC short of RAM
and of MPTER which we have in-house, on a data case which
involved a petroleum refinery, and if you use the same
technology that's common to all of them, for instance,
RAM does not have the Huber Downwash algorithm and it
does not have elevation possibilities, if you eliminate
that and compare them on equal comparison, all three give
the same result, which is what you would expect if you
have the same basic formulation and the same plume rise
equations. . .
If they're all the "same in three algorithms and
they're run and in a case in which the capabilities are
all the same, you would expect the same result, and that's
exactly what turns out. They are equivalent to within a
A
per cent or less, in some cases.
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
-. I . .__ ,- , ..':,_, '. . 22
1 Okay, let me go back then. So what we're
2 coming up with is an optimal capability on what the
3 Gaussian Plume model is capable of doing. This will
4 then tell us, and I can tell you right now that there
5 are cases we run into in which they are not Gaussian
6 the data is not Gaussian in nature. It is bineural.
7 I mean by this the concentrations will go up,
8 they'll dip and then, they'll go up again and you'll have
9 two maxima across the arc. What the optimal procedure
10 does, it puts a least squres optimal fit for all the
11 data points through that entire distribution.
12 What it does, it does an optimal fitting of a
13 Gaussian Plume mathematical characteristics to that type
of displacement. So it is a best fit and a best capabil-
15 ity of the Gaussian Plume to represent that type of
16 variation.
What I would like to do is, since Phillips
Petroleum Co. and API is going to extensive effort in
order to rigorously, mathematically, attack a problem
which has never been solved as far as I know, to the
best of my knowledge there has never been made concerning
this large data base collected over the past 50, I don't
know how many years it goes back, this data base, there
has never been a rigorous, mathematical analysis made to
19
20
21
22
23
24
25
determine what are the parameters of the model you are
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
. 20
21
22
23
24
25
23
. ~V»r- '' -"
-- - - i.t» v.
, __'__- -.-" *"" ,""'.., -- s
applying essentially"to determine an optimal (inaudible)
to represent that data base.
That is being done. Right now, my plans are
to finish that. I told Phillips management that I will
have that through by the end of this year. The output is
in microfiche because the output is massive, it is de-
tailed analysis and also what the program does is-, after
it determines the optimal fit, these dispersions Sigma y
and Sigma z coefficients as a function of downwind
distance, it then goes back and for each of the arc, it
breaks out equiva-lent stabilities.
That is very enlightening. Let me give you
some brief results of these types of analysis. First,
r
you will find out is that Sigma y's have a different
stability than Sigma z's. There are different stabilities
You will find that sometimes they will switch,
going from arc to arc, stabilities. Not more than
usually one clash, but there will be a change even in one
clash in stabilities. These are very basic, fundamental,
type of analyses which is throwing some tremendous in-
sight back into what is going on for definitions meteoro-
logically.
Nbwy -thereis'been a lot of concern about does .
terrastability or your wind fluctuations, your sigma
| thetas, these other type. What are the best characteri-
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
24
25
I' .""".-' , 24
of stability? This type of analysis is throwing some
direct insight into this.
So, this output is on microfiche. To give
you an example of the volume of microfiche, on the 35
cases I ran of the Hanford (inaudible), it produced 24
microfiche frames, I don't know how many pages in a frame
now, but that's a tremendous amount of output.
I got the output by paperback also, and it was
two boxes full and that's less than half of the Hanford
system. And I have 10 of those data bases. Of the 17 we
have at API, there's only 14 amenable to this analysis"
because some of them are puff release and some of them
ar,e line releases and we're only interested in point
source releases.
So we're going through in a detailed analysis
- a
of this data base and the output's coming back. Now, you
can do the analysis, it's being put out in output form
and a very compact microfiche form, but still the job of
analyzing because people still have to go through.
They have to know the trends, they have to knovv
the patterns and all of these things. It takes people to
do this. It is a time-consuming job to go back and
analyze and get insight into the fundamental studies what
is saying. And you cannot do this statistically.
You cannot do it because still, the best
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW .
(202) 234-4433 WASHINGTON. D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
' 20
21
22
23
24
25
25
pattern recognition is not statistical procedures, but
.the human mind.
Now, the thing about it is, the mind can do
things like integrating as you go through all type of
variables, so we need help, and we'd like to see EPA join
in this effort. If EPA is really interested in finding
out what is the uncertainty of models, we're making an
attempt to do that.
- 3
But there is a lot of work involved people
involved. Just to go through and to analyze this data
and ascertain the patterns, the trends, and come up then
with the final recommendation on what is the ability of
the Gaussian Plume formulation to represent meteorological
data.
Now, this is coming out. It's going to come,
-. a
I would love to see it come out. Phillips Petroleum Co.
-«£
individual, I cannot speak for API, they have to let me
speak for Phillips and I have not discussed this with my
mates, but from what they've already indicated to me, they
are perfectly happy to cooperate, with EPA in any way in
this basic data analysis in a cooperative effort, not in
an antagonistic sense at all.
Because I have heard, as I listen to people
talk, I sensed there was some antagonism in this meeting.
That's not our feeling. My feeling is one of mutual
NEAL R. GROSS *
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
. ... j. '- 26
cooperation. If we have a problem, let's do it in such
a manner that we jointly resolve it and use the best
3 technical capabilities we have available.
4 EPA has experts. Industry has people who have
5 technical capabilities that can be utilized. We can
6 jointly approach the problem in a cooperative deal to
7 come up as the Environmental Act has stated with a better
8 and the most accurate model that we can produce in order
9 to evaluate the environmental issues and come up with
10 solutions acceptable to both the public at large and the
11 industrial concerns, and if at all possible, we hope, in
i
12 a cost effective manner.
13 I asked my management this question. I said
14 which would you rather have: would you rather have not
15 knowing the truth and have over-conservative models, or
16 would you rather have not knowing the truth and having
t
I7 models -that-actually would allow us off easy, or permit-
ting, because we're constantly permitting and getting
applications for processes.
Here's the answer they came back with. We
would like the most representative, accurate information
18
19
20
21
22
23
24
25
we can get to make our decisions on for a very simple
reason. If we put in a plant in which the model has not'
correctly predicted, and let's say for instance, and these
are their remarks, that the model has actually said that
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
27 Zv
the concentrations were a lot more than they actually in
practice turned out to be after monitoring that work to
be put in.
It's very expensive going in, then, and cor-
recting after the design is finished. The most economical
place to correct design information is in the design
stage. Once you've built that plan and you have to go
back in and find out through monitoring networks around
your plant that it is out of regs, then it is very costly
to make modifications at that point.
The cheapest place to make modifications is in
the design stage. So, management at Phillips is inter-
ested in having as accurate models and meeting a real-
r
world condition. If we have to put the equipment in to
meet what the public wants, we'll do it, but we do not
want to put in equipment based oh models which are in-
accurate which cost us money and in turn that cost is
being passed on to the public and the products and ser-
vices that they use and unnecessarily charging them.
So, this is an issue I want to bring up before
EPA and the other Governmental concerns, Departments that
are concerned here, to throw this open to a cooperative
venture. The output's coming out. Is EPA really inter-
ested in seeing what is the actual limitations of the
Gaussian Plume model and what are the best capabilities
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
5
6
7
8
9
10
11
12
13
14
15
16
that we can put as far as determining paramemters in that
model that we can really best optimally match.
Now, we still have not, after we have done
this, we still have the whole .problem of meteorological
variation. And it's there. All we've done in this
particular step, is taken a very rigorous analysis of the
efficiencies of a Gaussian Plume model formulation.
Now, another thing this analysis is doing,
by the way, is looking not only downind distance, but
crosswind distance as Sigma y and Sigma z, I beg your
pardon. Sigma z is only a function of downwind distance,
but on the Sigma y, we're looking at it as a function not
on-ly of downwind distance, but also crosswind distances.
We found out that within the data that there
are cases many times, the sooner you can get a much,
- A
.much better fit to the data when you include crosswind
17
18
19
20
21
22
23
24
25
distance. In general, I say that's true.
I'm open to questions.
MR. TIKVART: Okay, thank you very much. I can
assure you we are very interested in tracking what you're
doing. I've seen the interim report on the 17 sets of
tracer studies and we looked at that in preparing the
review that Mr. Cox presented here on Monday. So we are
very interested in tracking the work that you're doing.
Thank you. Norm?
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(702) 234-4433 WASHINGTON, D.C. 20005
-------
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18.
19
.20
21
22
23
24
25
**...;. 29
.J^ ') ~- _ ,, ". ._
MR. BOWNE: _Norman Bowne, TRC, I have a '
question for Mr. Heliums and their fit. Are you using
the source term for the Hanford?
MR. HELLUMS: We are using the data that Dick
has TRC has put together IPO data base for us and
they are maintain it and the information that I have was
gathered and sent by Dick and it does have the source
information in there as far as emissions. Is that what
you're referring to, Norm? "°
MR. BOWNE: Is the emissions part of your
solution? --Do you have to have the emis-sions to get your
>
Sigma z?
MR. HELLUMS: Oh, yes. You must have the
r
emissions and you must have the meteorological data that
was associated with that data set as measured in the field
These wind directions are measured. " F*or in-
stance , for Hanford, they have releases at 26 meters,
56 meters, 111 meters. They have wind speeds measured at
that height.
MR. BOWNE: I would suggest, then, that there1
a third variable that needs to be accounted for here,
especially in the particulate tracer tests because it
appears to us, and in our report to you, that there might
be significant deposition affecting these and this would
make considerable difference in the sigma z that you
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
i:
7
8
9
10
11 -
12
13
14
15
16
17
IS
19
20
21
22
23
2-1
25
3.0 -
calculate and the-^S-i gin zv-zthat" you might infer from the i
data if you have not accounted for that deposition.
MR. HELLUMS: We've looked at that, Norm. Dick
did initially mention that. I looked at it in detail.
In fact, at the last meeting we had, I gave a result of
our month analysis where I had 'looked in my spare time.
This is a secondary project. This is not my
main effort. The state of the technology is not my main
effort. This is a side effort for me when I do my other
functions within Phillips.
No, we looked at this in detail. As a
gravitational settling is, I think we can mathematically
prove according to the ISC type of information, it is not
r
a factor. And, in fact, the analysis then that we're
making sure that is also it is not :.a._f actor. There is the
settling velocity and the friction . at .ground., surface .
And, of course the friction goes (inaudible).
The ultimatizor is also determining of these reflectors.
There is some difficulty and in doing this and it takes
certain types of information, .certain types of data sets
in order to do it because of the variations withing the
data itself and because also of other types of things.
This is in about six pages of analysis that I
gave, handed out, at the last API task force meeting and
if you check with Dick, Norm, you can see some very
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
3:r
V
1 *
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
1
18
19
20
21
22
23
24
detailed information' 'concerning -thisv "Right now, I- think-i
I have proved fairly conclusively to my own satisfaction,
that gravitational settling is not a problem within these
data sets because settling velocities are so small, that
they literally make no difference and that the main
factor moving the particular particles down to the ground
is not gravitational settling, but it is the eddy cur-
rents within the air.
MR. BOWNE: I'll look forward to looking at
that, thank you.
MR. HELLUMS: Okay, thank you.
MR. TIKVART: Thank you very much. I'd like
to encourage you to think about two things. Dick Kerch
r
mentioned that they're going to, National Coal Associatior
is going to go back and look some more at the should
versus the how to.
As I mentioned yesterday, we've heard an
awful lot of shoulds. We should do this, we should do .
that. Decision makers should consider these things, etc.
But, we're talking about things that haven't been done
systematically, and quantitatively in the past, so we
would really encourage you to think about this and to
submit written comments on this if no one has any further
observations to make here, today, on it.
How do you go about expressing quantitative
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
32
ffi
quantitatively the accuracy of models, and given that
quantitative expression of their accuracy, how is that
information used by decision makers to come to reason-
able determinations relative to the emission limits and
locations of new and existing plants?
That's one point. The second point is we had
some discussion yesterday of what the process should be of
submitting a permit for a new source or a siprevision for
... 3
an existing source when one has a better technique as the
gentleman just now indicated he was working on.
What is the process by which this better
technique should be submitted to the control agencies
whether it's EPA or a State or local government? And,
j how should .the agency and the source, the industry invol-
ved come to a, again, a reasoned approach to determining
. . ^
whether that new technique is acceptable for this particu-
lar application.
What is that process, and we've heard that it's
a good idea to conduct such a process but not much infor-
mation on the mechanics of that process.
At this time, are there any other comments on
those issues, or any other issues? Does anybody else wish
to make -a statement? /
MR. MOE: Rod Moe with the State Department of
Highways and Public Transportation, Austin, Texas. I just
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
1
i
- J,
^33
want to reiterate something that was said yesterday after-
noon.. I feel that there are too many models and not
enough data to support them.
And I think there has been too much piece-
mealing of research on model validation studies. I think
that some types of sources haven't been validated at all,
as far as trying to validate models, like the mines and'
things like that that have been discussed here.
We've done some work in line source modeling
to try to establish data bases and once you've established
the data bases then to try to validate'them. I'd like
to suggest that EPA have a national research effort to
fund some-of'their research in a program to study how they
r
can assemble these data bases, existing data bases, and
how they can come up with new data bases to amplify on
these things .
I think that quite often there is no adequ'ater^
data base for validating models. None, for quite* a few
types of sources. I think that the measurements have
been incomplete and I think that the monitoring has been
poorly dome.
I think quite a bit of the original data has
been lost and I don't think the data has been acquired
that should have been acquired. And I think that instead
of piecemealing the effort, there should be a coordinated
NEAL R. GROSS -
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON, D.C. 20005
-------
1
2
3
6
7
8
9
10
11
12
13
14
15
16
17
IS
19
, 20
21
22
23
24
25
it
effort, a joint effortV and there could be quite a lot' ,
of contribution from the private sector in this regard.
Then, I think there should be national sites
set up for continuing monitored studies. And I think that
methodologies should be up-dated and revised and new
instruments brought in, things like this.
Then, with a good data base, if somebody comes
up with a new modeling, idea, then you can test it against
this data base, not against a data base that that fellow
comes up with some sort of inadequate monitoring network
to try to -do this, but with a really " complete net-^
work to test that data, that particular model, and to
establish uncertainty and to apply a standard set of
f
statistical tests across the board.
That's my suggestion, any comments on this?
MR. TIKVART: Thank you, anybody else?'
MR. FEIN: My name is Richard Fein, I'm a ..... ~"~^~
Senior Research Associate with Texaco at Beacon, Sew York.
And I would like to add to your list of things that we all
ought to think about.
I'd like to add the point that was made by
myself and a number of others yesterday. That is, that
we look very carefully at the air quality management pur-
j poses of each of the applications of modeling, and then
see if there isn't some better way to achieve these
NEAL R. GROSS . .
COURT REPORTERS AND TRANSCRIBERS
j 1330 VERMONT AVENUE, NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17'
18
19
20
21
22
23
24
25
35
[ »'« .. t .'<!
ambient air quality purposes than using modeling the way
we've presently used it.
And, I think this would be an important thing
to do for everybody in trying to put together their final
comments for this Conference. Thank you.
MR. TIKVART: If there are no other questions
or comments, I would like to remind you that if you wish
an individual copy of the transcript of the proceedings
of this Conference, you should' contact the court reporter
directly, Mr. Miles Anderson.
The transcript of this meeting will be put in
the Docket A-80-46. I would guess that it would appear
ill. there some time early in September. The record will
remain open until September 14 for written comments.
Those comments should be submitted directly to the Docket,
so that by September 14, there should be a complete
record in the Docket of all written and oral comments
made pertinent to this Conference on Air Quality
Modeling.
If there are no other comments or statements,
I would like to thank you for your attention and parti-
cipation in this Conference and draw the 2nd Conference
on Air Quality Modeling to a close. Thank you.
(Whereupon the meeting adjourned at 9:50 a.m.)
NEAL R. GROSS
COURT REPORTERS AND TRANSCRIBERS
1330 VERMONT AVENUE. NW
(202) 234-4433 WASHINGTON. D.C. 20005
-------
36;,
CERTIFICATE OF REPORTER
I hereby certify that the foregoing transcript
represents the full and complete proceedings of the 8_i2_8l
aforementioned matter, as renorted and reduced to type-
writing under my direct supervision.
NEAL R. GROSS
NEAL R. GROSS
COURT REPORTERS AND TRANSCRICERS
1330 VERMONT /VENUE. NV/
(202) 234-4433 WASHINGTON. D.C. 2000S (301) 261-4445
------- |