DETERMINATION OF ATMOSPHERIC DILUTION
FOR EMERGENCY PREPAREDNESS
A JOINT EPA-DOE TECHNICAL WORKSHOP
OCTOBER 15-17, 1986
Chaired by
Dr. Francis S. Binkowski, U.S. Environmental Protection Agency
Dr. Harry Moses, U.S. Department of Energy
Edited by
Ms. Sharron E. Rogers
Proceedings Summaries
Prepared by
Research and Evaluation Associates, Inc.
-------
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATMOSPHERIC SCIENCES RESEARCH LABORATORY
PEER REVIEW AND WORKSHOP MANAGEMENT SERVICES
Contract Number 68-02-4129
Project Officer
Ronald K. Patterson
Prepared by
Research and Evaluation Associates, Inc.
1030 15th Street, N.W., Suite 750
Washington, D.C. 20005
(202) 842-2200
100 Europa Drive, Suite 590
Chapel Hill, N.C. 27514
(919)968-4961
-------
DISCLAIMER
Although the workshop results described in this report were funded by
the U.S. Environmental Protection Agency through contract Number 68-02-4129
with Research and Evaluation Associates, Inc., this report has not been
subjected to Agency peer review. It does not necessarily reflect the views
of the Environmental Protection Agency or the Department of Energy, which
co-chaired the workshop, and no official endorsement should be inferred.
However, it should be noted that the individual presentations and the papers
included in the appendices have been peer reviewed by each contributor's
organization and are reproduced as received.
ii
-------
FOREWORD
This report summarizes the Joint EPA-DOE Technical Workshop
Determination' of Atmospheric Dilution for Emergency Preparedness, held in
Research Triangle Park, NC, at the Meredith Guest House on October 15 and
16, 1986. The individual contributions to this document were prepared from
verbatim transcripts of the oral presentations made at the workshop. The
transcribed presentations included in this report were reviewed by the
authors and their organizations. Several articles were prepared for
presentation at the workshop, and these have been included as appendices in
this document. In the opinion of the chairpersons, the detail contained in
these appendices was important to the overall objectives of the Workshop.
We would like to thank the chairpersons and all who worked to put these
proceedings together. Our special thanks go to the editor of this report,
Ms. Sharron Rogers at Research and Evaluation Associates, Inc.
Ronald K. Patterson
Project Officer
Atmospheric Sciences Research Laboratory
U.S. Environmental Protection Agency
iii
-------
ABSTRACT
The Joint EPA-DOE Technical Workshop on Determination of Atmospheric
Dilution for Emergency Preparedness was held in Research Triangle Park, NC,
in October of 1986. The objectives of this workshop were to review the
current methods of determining atmospheric dilution for use in hazard
identification, emergency preparedness planning, and emergency response, to
provide recommendations for choosing among these methods, and, finally, to
define the role of the meteorologist in hazard identification, emergency
planning, etc. Several invited papers were presented, and panel discussions
were held to meet the objectives. The results are presented and discussed.
IV
-------
CONTENTS
FOREWORD ................................ "H
ABSTRACT ............................... 1V
INTRODUCTION ............ ................. 1
TECHNICAL OVERVIEW
R. A. Cox ............................. b
THE BHOPAL GAS TRAGEDY
K. Shankar Rao
NATIONAL RESEARCH NEEDS FOR EMERGENCY RESPONSE
IN THE WAKE OF CHERNOBYL
Joseph Knox ............................ Zb
ACCIDENTAL RELEASE SCENARIOS FOR ANALYSIS
Jane Crum Bare ........................... 37
SHORT-TERM TOXIC RELEASES FROM CHEMICAL
MANUFACTURING SITES
Robert E. Rosensteel ....................... *7
FATE OF TOXIC RELEASES IN THE ATMOSPHERE-ATMOSPHERIC
RELEASE ADVISORY CAPABILITY (ARAC)
Marvin H. Dickerson ........................ 53
SOURCE STRENGTH MODELING
Jerry M. Schroy ......................... 59
FUTURE NEEDS FOR DISPERSION MODELS IN HAZARD EVALUATION,
EMERGENCY PREPAREDNESS, AND ACCIDENT PREVENTION
James L. Makris ......................... 65
ISSUES IN REGULATORY APPLICATIONS OF MODELS
David E. Lay! and ......................... 71
COMMUNITY NEEDS FOR HAZARD EVALUATION TOOLS
Fred Millar ............................ 77
MATHEMATICAL MODELS FOR ATMOSPHERIC DISPERSION
OF HAZARDOUS CHEMICAL GAS RELEASES: AN OVERVIEW
Jerry A. Havens .......................... °'
DENSE GAS DISPERSION MODELS
Donald L. Ermak .......................... 95
-------
CONTENTS (Cont.)
ON THE DEVELOPMENT OF REACTIVE, DENSE GAS MODELS
Bruce B. Hicks and Will R. Pendergrass 105
DISPERSION MODELS FOR NEUTRALLY BUOYANT
AND POSITIVELY BUOYANT GASES
Thomas E. Pierce 107
LAB-SCALE EXPERIMENTS
Robert N. Meroney 113
LARGE-SCALE EXPERIMENTS OF THE DOE LIQUEFIED
FUELS PROGRAM
Ronald P. Koopman 121
FLUID MODELING OF DENSE GAS DISPERSION OVER A RAMP
William H. Snyder . . . . 129
ATMOSPHERIC SCIENCE AND EMERGENCY RESPONSE AT THE
SAVANNAH RIVER LABORATORY
Allen H. Weber and R. W. Benjamin 137
EMERGENCY PREPAREDNESS AND RESPONSE IN THE
U.S. AIR FORCE
Captain Lawrence E. Key 141
POISONOUS GASES FROM LAKES: THE CAMEROON DISASTER
Daniel A. Livingstone and George Kling 151
SUMMARY OF PANEL DISCUSSIONS 161
APPENDIX A - The Bhopal Gas Tragedy A-l
APPENDIX B - On the Development of Reactive, Dense Gas Models B-l
APPENDIX C - Mathematical Models for Atmospheric Dispersion of
Hazardous Chemical Gas Releases: An Overview C-l
APPENDIX D - Workshop Agenda D-l
APPENDIX E - Workshop Participants E-l
vi
-------
-------
KEY TO PHOTOCOLLAGE OF SPEAKERS
Upper row: (left) Jane Crum Bare, (right) Jerry Schroy.
Middle row: (left) Tony Cox, (center) K. Shankar Rao, (right) Marvin
Dickerson.
Lower row: (left) James Makris, (right) Robert Rosensteel.
viii
-------
INTRODUCTION
This workshop was designed as a response to a request from U.S.
Environmental Protection Agency (EPA) Headquarters for a forum where
the current state of modeling of atmospheric dilution for use in
emergency preparedness could be discussed. It quickly became obvious
that a joint effort between the EPA and the U.S. Department of Energy
(DOE) was the most effective way to proceed. As the designated
representatives of the two agencies, we built an agenda for the
workshop that would meet the following objectives:
1. To review the current methods of determining the release
characteristics, source strength, and rate of dilution of
atmospheric contaminants for use in hazard identification and
evaluation, emergency preparedness, and emergency response,
and to assess the specific strengths and weaknesses for these
methods and make recommendations for their improvement.
2. To provide recommendations for choosing among the current
modeling methods for immediate use in the identification and
evaluation of potential hazards, in the preparation of
emergency preparedness scenarios, and in actual emergency
response situations.
3. To define the role of the meteorologist in hazard evaluation,
emergency preparedness planning, and emergency response.
The participants contributed to these broad objectives from the
perspective of their individual situations and experience. One
fascinating aspect of the workshop was the interaction between those
who had dealt with releases of radioactive material and those who dealt
with chemical releases. The radioactive releases have historically
involved concerns of large-scale contamination, while the chemical
releases have involved concerns of a more local nature. In view of
this disparity of scale, it was surprising to see how quickly a common
1
-------
view of the problems associated with potential and actual hazardous
releases developed. One strong point of difference was, however, the
attitude toward a source term. For radioactive releases, monitoring
data taken after the incident will often give a good estimate of the
source strength and character. This was shown for the Chernobyl
accident.
For hazardous chemical releases, similar monitoring data seldom
exist, and sufficient time is seldom available to make such
measurements. Only in cases where routine monitoring is done on site
is this approach possible for chemical releases. This difference
comes from the fact that the hazards of radioactivity were perceived
many years ago and appropriate instrumentation developed and installed.
It is only recently that the hazards of toxic chemicals have been
widely recognized. General agreement has not been reached on the type
of on-site measurements necessary, nor in some cases, the need for such
measurements.
Because this report consists of transcriptions of oral
presentations, no bibliographic citations are given unless provided by
the author. To obtain reference information not provided, contact the
author.
We wish at this time to express our appreciation to all the
participants for their contributions and their patience in the
preparation of this document. We wish especially to thank Sharron
Rogers, Charlotte Coley, Brenda White, Rebecca Peer, Laura Saeger,
Chanya Harris, and Linda Cooper of Research and Evaluation Associates,
Inc., who were responsible for the production of the workshop and for
-------
technical writing, editing, and production of this document. We also
wish to thank Don Cox of Research and Evaluation Associates, who
designed the cover; Mannie Currin, who made the verbatim transcript and
had to wrestle with an unfamiliar technical jargon; and the staff of
the Meredith Guest House for their help and hospitality.
Francis S. Binkowski
Environmental Protection Agency
Harry Moses
Department of Energy
-------
TECHNICAL OVERVIEW
R. A. Cox
Technica, Ltd.
London, England
In this technical overview for what I consider to be a very
important meeting, we need to cover a whole range of topics. What is a
source? What is the source of a release of hazardous material into the
environment, and how is it characterized? How do the characteristics
of sources affect the way that the hazardous material is diluted in the
atmosphere?. What are the key issues for research today regarding those
characteristics and the behavior of hazardous materials? To what use
do we put this information when we have that scientific knowledge?
What are we really aiming to do in the real world of decision making--
both in regulatory decision making and the sort of decision making that
industry must do in facility design and siting? Specifically,
regarding emergency response planning, we have to address how our
knowledge of atmospheric dispersion phenomena can be used in helping us
make decisions prior to an emergency and during one.
Examples of Real Sources
First, what is a source? I will be predominantly talking about
hazardous chemicals, although others will be addressing radioactive
chemicals in this conference as well. The major difference between the
two generically is that with hazardous chemicals one typically is
-------
concerned only with relatively short-term effects, while radioactive
chemicals typically require both short- and long-term concern.
A source that is very typical of hazardous releases into the
atmosphere, although an extraordinary one in terms of magnitude, is
illustrated by the rupture of a cross-country ammonia pipeline caused
by a large bulldozer striking, penetrating, and shearing off the
pipeline. An enormous cloud formed. In fact, a cloud of ammonia vapor,
ammonia liquid droplets, and condensed water vapor was formed in the
atmosphere. This source is not anything like what those of us who work
in the field of air pollution dispersion modeling are used to
considering. The differences include: 1) it is massive; 2) the cloud
emerged under high pressure in a jet-like stream; 3) rapid mixing was
induced near the source, particularly at the start of the event; and
4) finally, the cloud becomes denser than air. As the cloud sinks and
starts to spread laterally, the emission momentum is lost and the
effect of the wind starts to take hold on the behavior of the plume.
We must always consider how complicated the phenomena are near a
source. They can be very complicated.
A second example is that of a puncture in a chlorine cylinder.
This is the small end of the range of events we are considering here.
Consider that a very small hole occurs in a 1-ton cylinder, so that the
amount of inventory released is not very large. The release occurs as
a jet-like plume with a lot of initial mixing. What was the
orientation of the release? It could have been horizontal * vertically
downwards, or vertically upwards. The hole might have been in the
vapor space of the cylinder, or it might have been in the liquid space.
-------
In contrast,;.the hole in the cylinder could have been large and, thus,
the releaselof the contents of the cylinder could have been essentially
instantaneous.
When the brakes failed on a Mexican train comprised of chlorine
tank cars, the cars derailed at approximately 80 kilometers per hour
and five or six of the cars were ruptured in the resulting pileup. The
total amount of chlorine released was about 100 tons. Compared with
the previous two cases of holes in vessels, this is an extremely
complicated source. One of the tank cars was torn apart, and the
release of its contents was instantaneous. While most of the tank cars
were caught in the pileup, the head of one car failed and the car
rocketed away from the scene, spilling liquid chlorine under pressure
behind it. Trying to describe this source of dispersion, we must
consider both instantaneous and elongated releases.
When a pipe failed in 1974 in the Flixborough plant in England, a
flammable substance, cyclohexane, was released under pressure of about
10 bars. The release occurred as two opposing high-pressure jets with
flash vaporizing and aerosol formation; in other words, a very
complicated thing from the standpoint of estimating rates of dilution.
The result of this particular failure was total destruction of a rather
large chemical complex.
Characterization of Sources
Obviously, there are a far wider range of types of releases than
these examples cover. A source parameter that is really important to
-------
us is the amount of initial dilution that occurs during the quasi -
spherical expansion phase of the burst. In the chemical industry, we
are dealing typically with materials that are stored under pressure,
thus, having a vapor pressure considerably higher than atmospheric
pressure. These materials will flash vaporize, and there will be a
simultaneous process of flash vaporization and mixing with the
atmosphere. The phenomena we can expect to arise are: liquid droplet
formation, aerosol formation, liquid droplet coalescence, and rainout
of liquid droplets. The liquid droplets, being rather cold relative to
the ambient environment, may hit the ground and revaporize or stay
there as a liquid pool. These phenomena are extremely complicated and
have an important effect on the quantity of material that ends up in
the air and, therefore, on the range of hazard faced from such an
event.
Transportation container bursts are obviously of a similar
character. Another type of rapid, sudden release is the possibility of
run-away reactions that can lead to bursting of reactor vessels.
These are all relatively large events. When we look at smaller-
scale events, one can list such events as: breakage of pipework, pump
seal leakage, discharges from relief valves, and so on. These are also
more complex phenomena that result in material being released into the
atmosphere without any pressure behind it, such as boil-off from liquid
pools or seepage through the ground or through water.
From a real process plant, the releases can take on many forms
that seem almost endless. We must describe which of the parameters of
these releases are of particular importance to us. Most important is
8
-------
the quantity of the substance that ends up in the atmosphere. When you
have a typical chemical release, not all of the chemical ends up in the
atmosphere, and there may be many factors that limit the quantity that
actually does enter the atmosphere. "These factors include orientation
of the leak and the presence of any secondary containment.
To consider the effect of orientation, if a liquid were released
that is flash vaporizing, but the orientation is vertically downward,
it is possible that the flashing two-phase flow will hit the ground, be
deflected, and a portion of the liquid centrifuged out. That liquid
will cool the ground locally beneath the release point and a pool of
liquid will likely form. This situation is even.more likely in'those
cases where secondary containment exists. Conditions such as this are
typical of situations where the drain valve of a pressure vessel is
somehow stuck open. I submit to you that the processes of liquid
fallout from a jet of this sort are not well understood at the present
time, and that they are one of the most important subjects of research.
People are currently planning and deliberately designing secondary
containment in order to have precisely the effect described above.
The quantity of chemical released into the atmosphere also depends
on the amount of flashing,that is occurring, and how high the vapor
pressure of the material in storage is relative to the atmospheric
pressure. If it is very high, relatively more of the material will
vaporize and more will end up in the atmosphere. Flash vaporization
and the process, pressure generally both have an influence on what is
called initial dilution, either by inducing mixing directly through the
turbulence associated with the initial velocity or by the mixing that
-------
is associated with the expansion occurring with flash vaporization.
Another important modifying factor in actual chemical plants is the
presence of obstacles and their effects on initial dilution. An
example would be a leak of a fully refrigerated liquefied gas spilled
into a bounded area; the presence of an adjacent large tank structure
and the effects of the wake of that structure can induce a very large
degree of initial dilution of the vapor boiling off from the liquid
pool.
All of these factors influencing the amount of chemical released
into the atmosphere are extremely complicated. In many cases, the
effect of these source-related characteristics on the subsequent
dispersion of the cloud is very marked, particularly the effect of the
initial pressure. However, this is perhaps more true of flammable than
of toxic substances, particularly those toxic materials that are still
hazardous at concentrations of the order of only a few parts per
million.
Dispersion Modeling
In cases where releases occur with pressure, dispersion models
that fail to consider the pressure will seriously overestimate the
travel distances of the discharges. In cases where the release occurs
without or with almost no pressure, dispersion models, such as typical
dense gas dispersion models or even Gaussian dispersion models may be
quite relevant to these events.
10
-------
Please note that the vast majority of hazardous materials are, in
fact, in pressurized containment. Therefore, in order for the momentum
of such releases to be reduced, one must postulate some continuous
mechanism for abstracting the kinetic energy from the initial release.
Processes like seepage through the ground or bubbling up through water
are the sorts of processes that one has to postulate in order to ignore
the effects of the initial momentum when the original source is
pressurized.
In the area of subsequent dispersion, those of us who are
meteorologists know that the subsequent dispersion is highly dependent
on the weather conditions and on factors such as surface roughness and
topography. Topography is particularly important because a typical
cloud is initially extremely dense relative to the atmosphere and,
therefore, particularly strongly affected by topographic effects. In
those situations with an absence of topographic effects, such as flat
plains, the state of the art in dense vapor cloud dispersion modeling
is very well advanced and does not merit very much additional research.
Research areas that do require additional attention are the effects of
the following: topographic effects on dense vapor clouds, obstacles
other than topographic ones, flash vaporization, and rainout.
Hazard Analysis
We need to.look for a moment at the uses of hazard analysis and
the ways in which the decisions we must make depend on our knowledge of
the consequences of hazardous chemical releases. In practice, we find
11
-------
that there are only three uses for hazard analysis for which there is a
real need by real decision makers in the real world:
• Land use planning/site selection,
• Plant design, and
t Emergency response.
The first, and most basic, is in zoning and land use planning
activities such as selecting a site for a hazardous installation. In
such a case, it is obvious that we are dealing with a hazard that has
the potential, if the worst were to happen, to have a physical impact
over a distance of perhaps several hundred meters. Thus, siting of the
facility should be very carefully considered because of the potential
risks. The siting decision must involve a balance between the
magnitude of the consequences that might occur and the probability of
the event occurring. In practice, we find that it is very difficult to
site typical industries far enough away from vulnerable populations or
other vulnerable entities to eliminate the hazard to the people. In
any case, the employees of the facility will be exposed to the risk.
Thus, whatever we do about land use planning and site selection, we
must accept some residual risk.
There is an analogous argument with respect to plant design (the
second of the three uses of hazard analysis). Industries do have an
intense interest in designing their facilities to minimize major
hazards. They do not want to lose their facilities or injure their
employees. They do this not only by site selection, but also by plant
layout and by choice of process routes, level of instrumentation
detail, emergency shutdown valves, check valves, and other items of
12
-------
this nature, which are really a matter of design detail. Hazard
analysis techniques can be, and are being, used to address these
detailed matters related to major hazards, but implementing such
controls for major hazards may cost money. One can conceive of design
features that are impractical simply because they are ludicrously
expensive. Those things are not done if it can be shown that the
probability of the particular event occurring for which that design
provision was being considered was sufficiently low. Thus, there is
some acceptance of residual risk.
The third application of hazard analysis techniques, for emergency
responses, is the principal topic of our workshop. I would like to
make several points about the way I see emergency response fitting into
the hazard analysis framework. First, there is one crucial decision
that has to be made, how large a physical provision are we going to
make for emergency response? How much foam are we going to provide?
How much provision needs to be made for breathing apparatus and escape
masks? What will be provided to the local population in anticipation
of an emergency situation? This type of decision must be essentially
risk-based. It definitely is impractical to provide a sufficient
emergency response capability to prevent all casualties in the worst
possible accident one can postulate. Therefore, we make provisions
that fall short of the worst case, and the question is: how far do we
go? We make the decision on a risk basis, discounting the worst
scenarios provided that we can demonstrate they are sufficiently
improbable.
13
-------
For effective emergency response, we need good preplanning, by
evaluation of a representative range of scenarios using dispersion
models and tools of that nature. By representative, I mean that we
must look at everything from the small leaks to the most extreme cases.
Even though we may not be able to cater to the extreme cases, they
should be evaluated so that deployment of existing response capability
will be the most effective possible. In preplanning for emergency
response, the decision-making algorithm is particularly difficult. It
is not a simple thing to decide in advance whether the advice given to
local populations should be to evacuate or to stay indoors. Depending
on the duration of exposure and on the particular chemical released,
the advice might be different in different conditions. Such decisions
can be very difficult, but can be improved by the prior application of
very good models for consequence evaluation.
Whatever emergency plans we select must be extremely practical.
The decision-making algorithms have to be presented to the people who
will give the orders during the real emergency in such a way that they
do not hesitate about the decisions they must make.
Concluding Remarks - Future Priorities
In the recent past, there has been some misdirection of resources
by concentrating too much on the niceties of vapor cloud dispersion in
the atmosphere. There are in fact many more phenomena that are really
involved in chemical hazards than simply the atmospheric dispersion
14
-------
effects. In the future, the balance of effort and interest should be
moving towards these other phenomena.
15
-------
THE BHOPAL GAS TRAGEDY
K. Shankar Rao
National Oceanic and Atmospheric Administration
Atmospheric Turbulence & Diffusion Division, Oak Ridge, TN
and
M. P. Singh and S. Ghosh
Center for Atmospheric Sciences
Indian Institute of Technology, New Delhi, India
In the early hours of December 3, 1984, about 40 tons of highly
toxic, volatile, and reactive methyl isocyanate (MIC) gas leaked in
about 90 minutes through a 33-meter-high atmospheric vent into the cool
night air of Bhopal and quickly spread in a fog-like lethal cloud over
a large populated area. Thus began the world's worst industrial
disaster, which killed over 2000 people and injured more than 200,000.
Here we will sketch the accident scenario and outline the events of the
night leading to this catastrophe.
The meteorological and topographical features of Bhopal and the
physical, chemical, and toxicological properties of MIC will be
described in the context of the cloud dispersion and its effects on
human and biological life. A simple analytical dispersion model
emphasizing aqueous-phase conversion and deposition of MIC will be
presented. This model, based on solution of the atmospheric
advection-diffusion equation, gives estimates of ground-level and
vertically integrated concentrations. The model estimates
qualitatively correlate with recorded human fatalities and injuries and
the contours of observed damage effects on trees and vegetation in the
affected^areas.
17
-------
The rehabilitation efforts and some preliminary results of the
ongoing medical and toxicological studies will be summarized. We will
comment briefly on the legacy of Bhopal, including its implications for
siting and operation of plants handling toxic chemicals, worker
training, risk assessment and emergency preparedness, occupational
health and environmental regulations and enforcement, and other
relevant issues.
The city of Bhopal, with a population of about 900,000, is the
capital of the State of Madhya Pradesh, India's largest state, with an
agriculture-based economy. Bhopal's central location (about
600 kilometers south of New Delhi), communications, resources, and
hospitality attracted industry and people from all parts of the
country.
Union Carbide, India, Limited (UCIL), which is the corporation
involved in this accident, is one of the top 20 Indian companies. They
produce about 1500 tons per year of MIC-based pesticides, Sevin
(carbaryl) and Temik (aldicarb).
The UCIL pesticide plant is located in the northern part of
Bhopal. To the east of the plant is an industrial and warehouse
district. To the north are mostly wooded lands that are very sparsely
populated. However, the plant is surrounded by residential areas on
three sides. Unfortunately, on the night of the accident winds
directed the plume toward the populated part of the city.
Bhopal includes hills to a height of about 600 meters and two
lakes. It is likely that nocturnal drainage winds and land breezes
altered the local surface wind patterns on the night of the accident.
18
-------
Detailed meteorological data are not available, but various sources
indicate that the winds that night were initially from the northwest
and subsequently from the north.
There are three MIC storage tanks on the UCIL property. MIC is
the raw material used to produce the pesticides carbaryl and aldicarb.
At the time of the release, tank 610 contained about 41 metric tons of
MIC. A shift change occurred about 10:45 p.m. on the night of December
2, 1984, which was a Sunday. There were about 75 people on duty. The
incoming shift was unaware that a runaway chemical reaction was taking
place in tank 610.
About 11:00 p.m., the staff noticed the pressure in tank 610
increasing from 3 to 10 pounds per square inch. They mistakenly
attributed the increase to nitrogen pressurization of the tank by the
previous shift. About 11:30 p.m., the staff noted eye irritation from
MIC. Since minor leaks of MIC were not uncommon, they were
unconcerned. About 12:00 a.m., they noticed the pressure in tank 610
had increased to 30 pounds per square inch. This was reported to the
production supervisor, who checked it almost immediately and found that
the tank's rupture disk (designed to rupture at 40 pounds per square
inch) had burst, and the safety valve had popped.
At 1:00 a.m., MIC was observed to be escaping through a 33-meter
high atmospheric vent line into Bhopal's cool night air. Within a
90-minute period, about 40 tons of gaseous MIC leaked out. None of
several plant safety systems in place fully functioned that night. A
scrubber, charged with caustic soda, contained only enough to
neutralize about 7 tons of MIC released. A flare tower did not work
19
-------
because it was undergoing maintenance. A water curtain designed to
hydrolyze escaping MIC reached only a height of 15 meters, far below
the release point. The facility's refrigeration plant was down.
The public siren was turned on for a few minutes at 1:00 a.m., but
was not sounded again until 2:00 a.m., to alert the public to the
danger. By then, thousands of people were fleeing.
About 200 people died in their beds in the immediate vicinity of
the plant where a huge white cloud formed and moved southward towards
the densely populated areas. Near the plant, the gas was so thick that
visibility was very low. These people awoke to a nightmare of
suffocation, burning eyes, and panic. They rushed out into the streets
to join thousands of other people who were running for their lives.
Many people became overcome by fumes and collapsed along the way.
Rescuers converged on the area around 1:30 a.m. By dusk of that first
day, the death toll mounted to about 1000 and over 100,000 were
injured. As the magnitude of the disaster became apparent, medical
equipment and relief were rushed to the stricken city.
MIC is known to be a very reactive, toxic, and volatile chemical.
It undergoes exothermic and vigorous reactions with a variety of
compounds containing active hydrogen atoms. Analysis of sludge from
tank 610 after the accident showed that two major exothermic reactions
took place. One was MIC reacting with water; the products of MIC
hydrolysis with water made up 16 to 28 percent of the samples. The
second, and much faster reaction, was MIC reacting with itself,
catalyzed by iron. More than 50 percent of the sludge consisted of MIC
trimer resulting from this second reaction. A small amount of water
20
-------
with a trace of contaminants, such as rust or salt, could catalyze an
autoreaction. The heat liberated during an induction period lasting a
few hours could generate a reaction of explosive violence. It is not
inconceivable that a small trace contamination could have occurred in
the MIC tank in spite of the best prevention efforts.
Once gaseous MIC escaped into the air, it would undergo advection,
diffusion, and chemical transformation, with atmospheric moisture. The
major reaction product would be methyl amine, which is absorbed and
stays in the soil until slowly broken down by biological reactions and
weathering.
We used an analytical dispersion model based on the atmospheric
advection-diffusion equation. The source strength was established at
40 metric tons in about 90 minutes. MIC vapor is twice as heavy as
air. This suggests that density effects might be important for plume
dispersion. However, the vapor escaped at a very high velocity, which
leads to considerable entrainment. Also, the reaction of MIC with
moisture releases about 1.36 x 10^ joules per kilogram of heat that
reduces the plume density. With these considerations, the density
effects were not considered, but gravitational spreading was included.
Post-episodic reconstruction of the accident by scientists established
that the wind was initially northwesterly during the accident, and
subsequently northerly.
According to our model, most of the high concentrations occurred
in the first kilometer southeast of the plant. A region extending
about 7 kilometers downwind of the plant was affected. Birds, cattle,
cats, dogs, goats, and all other animals were killed over a 65-square-
21
-------
kilometer area in the southeast quadrant of the plant. Based on the
model estimates, we have divided the affected area into four zones.
Zone 1 consists of concentrations >50 parts per million, zone 2
consists of concentrations >15 parts per million, zone 3, >1.5 parts
per million, and zone 4, <1 part per million. In the first few hours
after the accident, zone 1 (severely affected) out of a population of
6173, sustained 360 fatalities. Initial fatalities in zone 2 (with a
much larger population) amounted to 508. In zone 3, there were about
11 initial fatalities, and in zone 4, about 5. The correlation with
the fatalities provides some indirect evidence that these concentration
estimates are reasonable, although the large population movement that
night makes these comparisons somewhat subjective. Most of the
fatalities, however, occurred close to their place of residence.
Perhaps the best concentration monitors for the effects of the
accident at Bhopal are the plants and vegetation in the area. The
Indian Agricultural Research Institute and the Central Board for
Control of Pollution conducted a vegetation damage survey that showed
most of the effects occurred in the southeast, southern, and eastern
directions from UCIL's boundaries. The same species were affected
differently depending on location relative to the plant and the amount
of moisture available in the ground at the time of exposure.
Vegetation that had been irrigated on the previous day was much less
affected. Some tall trees were completely defoliated. Much of the
vegetation in affected areas looked as though it had been burned.
Additional detailed studies are still under way to study the impacts on
22
-------
the vegetation. Studies are addressing the histological effects on the
lipids of the leaves and on the microorganisms attached to the foliage.
Very little is known about MIC's toxicological effects, especially
the long-term ones. What little data are available were from tests
done on rats. The permissible MIC skin exposure limit, a voluntary
standard, is 0.02 parts per million. This is the amount of MIC to
which a person can be safely exposed over an 8-hour workday. This
threshold limit value (TLV) has been set by the American Conference of
Governmental Industrial Hygienists (ACGIH). The lethal dose is
estimated to be between 15 and 30 parts per million, depending on age
and sex. Before the Bhopal episode, the effects of MIC had never been
observed on such a large and diverse population. Isocyanates were
known to attack the respiratory system, eyes, and skin.
Laboratory analyses have confirmed that most of the deaths in the
immediate vicinity of the plant occurred from cyanide poisoning.
Cyanide was probably released because of the high temperatures reached
in the MIC tank, causing the MIC to break up-into hydrogen cyanide,
methyl amine, carbon monoxide, and other organic cyanide products.
While it will be some time before the studies at Bhopal are
completed, clinical and toxicological studies are being coordinated by
the Indian Council of Medical Research. Medical treatment has
dramatically improved the condition of people with eye injuries. Lung
problems persist in many victims. A large number of victims with
pulmonary edema have recovered, but many were left with considerable
damage to their small airways and alveoli. Children were the worst
affected, with about 39 percent showing moderate to severe pulmonary
23
-------
disability. Miscarriages and stillbirth were frequent in the pregnant
women exposed to the gases. In the first 20 weeks, 436 spontaneous
abortions occurred out of 2600 pregnancies; the normal rate for Bhopal
was 6 to 10 percent.
On the night of the accident, the doctors of Bhopal, supported by
medical students, interns, and the nursing staffs, rallied to respond
to this unprecedented, sudden disaster. Because there was so little
known about the toxicology of MIC and its reaction products, the
medical personnel had to improvise in the beginning. There certainly
was and is a real need for information on the processes, potential
hazards, biological effects, and the medical treatments that should be
made available in any kind of industry involving toxic chemicals.
The Union Carbide plant was located close to the population in
Bhopal. There should be at least a 2- to 3-kilometer buffer zone
between hazardous industries and the population centers. Standards,
enforcement, safety, health, and environment are very important.
Bhopal offers many lessons that have to do with basic human
concerns for safety. If these lessons are lost, it is all too likely
that an accident like this can happen again.
Note: A prepared paper by the authors on this topic appears in
Appendix A.
24
-------
NATIONAL RESEARCH NEEDS FOR EMERGENCY RESPONSE
IN THE WAKE OF CHERNOBYL
Joseph Knox
Lawrence Livermore National Laboratory
Livermore, CA
The material presented today is a distillation of materials from
about 3 feet of shelf space. We will consider what actually happened
at Chernobyl, the nature of the warning that the free world received,
and how that warning came. The first tasks that then faced the United
States were assessment and making sense out of the inconsistent and
sometimes conflicting information.
First, let us briefly review the setting of the Chernobyl
accident. The reactor, in operation approximately 2 years, was a
1000-megawatts (electric) generating plant with about 3200 megawatts of
thermal energy. The reactor is graphite moderated, water cooled, and
has confinement, but not containment in the United States sense of this
term. The reactor core is in an interior building with what the
Russians call a protective shield around it, built of steel and
surrounded by an inert gas. Thus, there is or was a barrier between
the oxygen in the confinement building and the material inside the
reactor.
The accident sequence has been covered in detail in various
publications and will not be repeated here. During the turbine
experiments and the shutting down of the safety devices, there was an
excursion and a gradually increasing instability of the reactor with a
large steam pressure buildup inside the interior structure. In the
25
-------
interior of the reactor, inside of the protective shield, a pressure of
over 1000 pounds per square inch ripped off the tops of some
1650 pressure tubes exposing the graphite and fuel elements to the
atmosphere. Shortly thereafter, the graphite, which was at 750°C,
ignited. As a result of the preceding explosions, the plumbing and
instrumentation were probably disrupted. At that time, water and steam
could readily contact the zirconium, and water and steam could also
contact the hot graphite, generating large amounts of hydrogen and
carbon monoxide, both combustible. The source term at Chernobyl can be
described as a chaotic fire, multiple explosions, gas generation, and
various sources of water contacting the hot graphite. In terms of
initial distribution of the released inventory, we must think of
burning gases, burning graphite, burning carbon monoxide, and perhaps
multiple explosions, driving a plume up through the atmospheric
boundary layer. This is an "image" of Chernobyl at 1:23 a.m. on
[Saturday,] April 26, 1986.
Our first warning came from Scandinavia late Sunday and on Monday
morning. We learned by telephone that a suite of radionuclides were
being observed on particulate filters in Scandinavia; radioiodines and
radiocesiums were present in copious amounts. Other isotopes (e.g.,
Co) were coming from neutron activation of the cooling system, which
confirmed a disruption and at least partial destruction of this system.
The particles of ruthenium were spherical, suggesting'that at least
part of the fuel released had seen temperatures of 2500°C. (Later
interpretations indicate that the Ru particles came from shattered fuel
elements entrained into the plume or explosively injected into the
26
-------
along the route of cloud passage, and an estimate of the curies by
isotope that were released into the atmosphere was developed. Nuclear
engineers from both the civil and military communities made estimates
of the core inventories and the relevant nuclides of greatest interest.
At the time of reactor shutdown, 80 million curies of I were
estimated in the core inventory with 6 million curies of Cs. During
the first 24 hours, it is estimated that all of the noble gases were
released when the 1650 pressure tubes were destroyed. A high fraction
of the volatiles were released. Our assumption is that 40 percent of
the volatiles were released on the first day into the Zone A cloud and
10 percent into that zone on each of the subsequent 5 days. The amount
of material going into Zones A and B of the cloud over the 6-day period
was developed from experience based on fallout clouds. However, these
estimates were checked by cumulus cloud simulative models that normally
use a radius of 5 kilometers for the scale of the base of the cloud.
The Chernobyl fire plume was approximately 20 meters in radius—the
world's smallest convective column; it was driven by a flux of 62
megawatts across the surface of the 20-meter radius.
Over 100 maps of material dispersed and deposited over Europe and
Asia were generated in the simulations. One of these maps shows the
integrated exposures to an individual standing in the path of the plume
at 2 days (48 hours) after the accident. The highest calculated dose
was 133 rem to the adult thyroid in the vicinity of the plant. The
exposures at the distance of the coast of Sweden were equivalent to
0.1 rem at the end of 48 hours. From these simulations, it was learned
how Kiev was fortunate, indeed. The plume from the accident rose up
28
-------
over a relatively dry warm front, so there was not much of a problem
from rainout of radioiodine. There was, however, significant
convective activity in the area of Kiev at the time, although the plume
was not significantly impacted by that activity. Observations in
Sweden were significantly affected by the lifting of the plume over the
warm frontal area in the region of the Baltic Sea.
The results of the LLNL simulations were sent to the President's
Commission through the Department of Energy. These results are in
relatively good agreement with the calculations and observations of
Scandinavia and Eastern and Southern Europe, that is, by a factor of
two with measured values for deposition by isotope. Agreement was not
good in Western Europe, possibly due to the gliding up over the warm
front (later calculations show better agreement when meteorological
data from all of Western Europe were used). The calculations and
measurements showed that there were no acute health effects outside of
the Soviet Union.
To put Chernobyl in perspective, the accident was compared to a
20-kiloton nuclear test and the released inventory of the Three-Mile
Island (TMI) reactor accident in the following table. The estimated
COMPARISON* OF CHERNOBYL WITH OTHER RADIONUCLIDE RELEASES
Nuclide
13l!
137Cs
9°Sr
Noble gases
Chernobyl
10-50
1-6
.001-. 07
100-200
TMI
.00002
None detected
None detected
10
20-kiloton
Nuclear Test
2
.004
.004
5
*Megacuries of radioactivity 3 days after shutdown or test.
29
-------
release inventories for Chernobyl indicate that radioiodine releases
were approximately 20 times greater than at a 20-kiloton nuclear test.
No cesium was detected at TMI; it stayed inside the containment. No
90Sr was detected at TMI, but was at Chernobyl. While virtually all of
the noble gases were released at TMI and Chernobyl, the amount was far
smaller at TMI. The differences in long-term dose occur because the
dose at TMI came mainly from the noble gases, while the long-term dose
at Chernobyl comes mainly from the 137Cs. The maximum individual dose
at TMI was 11.5 millirems, while the maximum dose at Chernobyl was
above the lethal dose for half the exposed population (LD50), delivered
by fission products, not the noble gases. Rainout was not important at
TMI because noble gases are not scavenged. At Chernobyl, the iodines
were scavenged and are of continued importance to the Soviets. The hot
spots in Europe were created by rainout. These rainout hot spots run
roughly 10 times above areas subject to dry deposition alone. The
message from these findings is that containment, design, and safety
devices are immensely important in terms of explaining the differences
between Chernobyl and TMI. All of the diagnostic work shown today in
this presentation was completed in the first 14 days following the
event and reported to the President's Commission.
The New York Times reported that Chernobyl released much more
radioactivity than all of the atmospheric testing by both of the super
powers. It is important that we understand that this was an error.
The correct number is that Chernobyl's release of 137Cs is 6 percent of
the Cs released from super power testing.
30
-------
In preparing to model the upper cloud from Chernobyl, we quickly
learned that we did not need the whole radioiodine inventory to explain
the readings in Scandinavia. So the question was, where was the other
half of the radioiodine? It would have been easy to say that the
models were no good, but this was not the reason. When the United
States' sophisticated sampling aircraft measured radioiodine at 5500
meters above Japan, we quickly redirected our focus. Taking the long-
range transport models from nuclear test days and back calculating from
Japan, we found that half the radioiodine had to be at higher
elevations over Chernobyl on the first day. Calculating forward from
Japan, it was possible to estimate how much radioiodine would be
measured off the coast of California when the cloud began its passage
over the United States. The models estimated that 30 picocuries per
cubic meter would be found at the California coast. On cloud arrival,
the aircraft measured 12 picocuries per cubic meter. That is a
reasonable comparison in this business. We now believe that an upper
cloud was created by the early explosions, the intense fire, and
residual heat from the reactor. A lower cloud was produced by the
lingering fire that had a lower heat flux.
From the California coastal estimate, transport was projected over
the continental United States, assuming a reasonable vertical
distribution, calculating a vertical integral, and assuming immediate
scavenging of 100 percent of the vertically integrated iodine.
Assuming this radioiodine to be deposited on the ground and eaten by
the family cow, the maximum amount of radioiodine in cow's milk in the
United States was estimated to be approximately 900 picocuries per
31
-------
liter. As you will recall, guidelines are established at 15,000
picocuries per liter. Thus, we estimated that there would be no health
problems in the United States as a result of radioiodine deposition,
even in the deposition hot spots. In areas actually measuring
radioiodine concentrations in milk, the maximum measured was 600
picocuries per liter in one sample. Most of the measurements were from
100 to 200 picocuries per liter or less across the United States. From
these simulations and the many measurements, we believe that the upper
cloud from Chernobyl did exist.
From all of the northern hemispheric measurements and simulations,
we now estimate that there were 2.4 million curies of Cs released,
which represents just less than half of the core inventory.
Approximately half of the 131I inventory of 80 million curies, or 47
million curies, was released. Our final estimate is that half of the
core inventory of cesium and iodine was released with 100 percent of
the noble gases, 10 percent or less of the tellurium, and 1 percent or
less of the other materials.
The experiences from Chernobyl lead us to identify several
research needs. One of the most important things we can do to improve
emergency response is to improve the flow of information. This means
early warning by the country involved with the accident; openness
concerning what was released, the nature of the accident, severity and
early measurements; in fact, everything known about initial
distribution of the material released. In the United States, we need
to improve our own flow of information. There are nine Federal
32
-------
agencies involved in emergency response. Each one claims a role, does
quality control differently, and perceives a different jurisdiction.
Additional research is also needed in the area of meteorological
prediction. Models can be driven with input from numerical weather
prediction models that operate on the global scale down to mesoscale
models on the smaller scale, i.e., state-size, regional-size, or
continental in scale. Yet, each serves different purposes. On the
larger scale, for a massive event like Chernobyl, the question is where
is the released material going and what will the concentration be when
it gets there? For smaller chemical releases, the prediction serves a
different purpose. You may have a hazardous condition to engineer your
way out of and, through prediction, one can select optimal conditions
in which the material may not flash, or other risks can be minimized.
Thus, prediction serves different purposes depending on the manager's
and decision maker's needs.
We need to know the physics and chemistry of the release. At
Chernobyl, we needed to know the combustion products, the heat fluxes,
and the isotopes released to the environment, in order to know what the
exposure and doses to man would be. To say so many curies of activity
are released is meaningless. You must know the number of isotopes and
the amount of each isotope in order to know the mode of exposure and
those doses that will finally be delivered to the human community.
Our particle-in-cell techniques need to be expanded to handle
complicated and differing sets of isotopes. It is easy to do particle-
in-cell calculation for one to five different isotopes. To do
simulations for a suite of 15, 16, 40, or 50 and do these calculations
33
-------
quickly, requires improved techniques. We would call this improvement
a hybrid particle approach. In the last 5 years, two proposals have
been developed to address this specific need, but have not been funded
(by any Federal agency).
We need additional information about scavenging properties. The
radioiodine from Chernobyl was carried on a host aerosol. Iodine
exists both in gaseous and particulate form as it moves downwind. The
scavenging is mainly of the particulate. It can be scavenged very
efficiently if the iodine is coating the host aerosol particle, because
iodine is a cloud-seeding material and is most effective in the ice-
producing parts of clouds, which probably existed over Europe and Japan
where the "hot spots" were created by wet deposition.
We need standardization of action levels of contamination in
products to know what actions to take consistently. In Europe, part of
the confusion was due to the lack of uniformity of standards.
Vegetables were needlessly destroyed, and populations were needlessly
disturbed. The countries tended to overreact because of nonuniformity
of standards and extremely low standards set by people who wanted to
show what good they were doing for the environment. When, in fact, it
became "crisis time," they created confusion in the population. The
Soviets claimed that no damage occurred outside their country and that
the Westerners took unnecessary precautions. Yet, some Westerners,
such as the shepherds of Wales, are seeking compensation from the
Soviet Union. Such compensation will not be possible in international
law circles until there is standardization of action levels and
protective action guides. There needs to be standardization of
34
-------
measurements, analysis, and reporting procedures. A specific problem
with reporting occurs with iodine levels. When a given iodine level is
reported from a specific station, one must also know how it was
measured. It makes a difference whether the iodine is a gas or
attached to particulates.
We need to study the biomedical history of the 200,000 severely
dosed people in the Soviet Union. This is not the acute radiation
victims who saw more than a few hundred rem in the first day or so
after the Chernobyl accident; they will certainly be studied. The
200,000 who saw from 50 down to a few rem should be medically tracked
and will constitute a valuable database for understanding radiation
effects.
The Soviets proved to be ingenious in minimizing the impacts of
the accident. We should learn from how they did it. They minimized
resuspension of the radioactive material in the deposition pattern
around and up to some distance from the plant. They controlled runoff
to the streams and reservoirs by building levies so as not to
contaminate the water resources of the city of Kiev. They did rainfall
suppression in the region of Kiev in order to reduce the problem of
surface runoff in the neighborhood of the plant. They also took other
actions too numerous to list here.
There is one area where the Russians were very lucky. The initial
plume from Chernobyl went northwest to an area of relatively
uninhabited boreal forest. As a result, a relatively large fraction of
the release inventory is resting in and on the pine needles of the
boreal forest some meters above the ground. As the pine needles die
35
-------
from the beta radiation, they will fall as fallout during the next
several years northwest of Kiev. This is an ideal laboratory for the
study of long-term ecological effects of radiation, which the world has
not seen.
Thank you very much. This discussion outlines my ideas of
research needs in the wake of Chernobyl.
36
-------
ACCIDENTAL RELEASE
SCENARIOS FOR ANALYSIS
Jane Crum Bare
U.S. Environmental Protection Agency
Air and Energy Engineering Research Laboratory
Research Triangle Park, NC
Accidental releases are a new area of study with most activity
occurring in the United States only since the 1984 accident at Bhopal,
India, where approximately 2500 people were killed and thousands more
injured. Until this accident, very little recent effort was put into
modeling the dispersion of acutely toxic gas clouds that could cause
effects due to short-term emissions. The purpose of this workshop is
to gather expert dispersion modelers to analyze various scenarios. The
purpose of this paper is to provoke this discussion and present actual
scenarios for analysis. The specific subjects discussed were chosen by
those organizing the workshop who felt that the attending dispersion
modelers needed a background in this area to which they might not have
had previous exposure. The first section will discuss past releases
that have occurred in the United States. The second section will
provide more of a statistical side by presenting chemicals most
frequently involved. The third area will discuss some of the many
variables that affect source strength and dispersion modeling, and the
final section will present three representative scenarios for analysis
at the workshop. None of these sections will provide an exhaustive
account of the topics, but will simply touch on major points.
37
-------
Case Histories
Several questions were presented for this section. These
questions included: Have acutely toxic chemicals similar in toxicity to
methyl isocyanate (MIC) been released in quantities comparable to the
quantities released in Bhopal? Has anything equivalent to Bhopal
happened in the United States in the past? Are releases occurring
today? To find the answers to these questions case histories were
reviewed.
Have acutely toxic chemicals similar in toxicity to MIC been
released in quantities comparable to the quantities released in Bhopal?
EPA's Acute Hazardous Events Database contains a computerized record of
6928 incidents that have occurred in the United States in the last 6
years. The toxic load was calculated for Bhopal and 300 other events
in the database, where toxic load is defined to be the amount of
chemical released divided by the IDLH (immediately dangerous to life
and health) of the chemical. For Bhopal, approximately 90,000 pounds
of MIC was released, where MIC has an IDLH value of 43 milligrams per
cubic meter. This corresponds to a toxic load of approximately 2100
pounds of MIC per IDLH value. IDLH represents the maximum level at
which a healthy male worker could escape after 30 minutes exposure
without irreversible effects or loss of life. Some chemicals without
established IDLH values use calculated IDLH from other health effects
data. Here the IDLH is expressed in units of milligrams per cubic
meter for consistency in the toxic load calculations.
38
-------
Six events within the database were determined to have toxic loads
greater than Bhopal's. In other words, if all other factors had been
equal (e.g., population density, meteorology, physical state, and
emergency response), these events would have represented at least as
much potential for toxic effects as the event at Bhopal. In each of
the six cases, however, the factors were not equal and not a single
life was lost.O'2)
Has anything equivalent to Bhopal happened in the United States in
the past? In doing research, the following three events seemed not of
equal magnitude, but had serious consequences.
• Cleveland, OH, October 20, 1944. A liquid natural gas
storage tank ruptured at 5 pounds per square inch and -250°F
in a residential and industrial area. Liquid natural gas
flowed down the storm sewers, mixed with air, and exploded.
'A total of 128 people were killed, and 200 to 400 were
injured.
• Texas City, TX, 1947. Bagged ammonium nitrate fertilizer on
board a ship caught fire; 2280 tons of fertilizer burned and
a nearby ship exploded. After the explosion and fires, 330
dwellings, 130 business buildings, and 600 automobiles had
been destroyed. A total of 3000 people were injured, and 552
lost their lives.
• Brooklyn, NY, 1944. Chlorine was released from a gas
cylinder into the subway system. A total of 208 people were
injured; no one was killed.^3)
These events happened approximately 40 years ago. Are releases
occurring today? The following events from the Acute Hazardous Events
Database occurred within the last 6 years in the United States.
• Niagara Falls, NY. One ton of liquid chlorine spilled from a
process vessel in a chemical plant. The plume that followed
resulted in the evacuation of a nearby sports stadium where a
game was in progress. A total of 76 people were injured.
t Elizabeth, NJ. At the Newark Airport, a release of
isobutane, phosgene, hydrogen chloride, and hydrogen cyanide
39
-------
resulted in the evacuation of the entire airport and 15 miles
of the New Jersey Turnpike.
San Francisco, CA. A liquid and gaseous release of
polychlorinated biphenyls (PCBs), natural gas, and oil
resulted in the evacuation of 19 buildings containing 30,000
people at the
Event Statistics
What are the statistics? What are the chemicals most frequently
involved in releases?
The best source of information at this point in time for data on
accidental releases in the United States is the Acute Hazardous Events
Database. The database provides an analysis of the frequency with
which chemicals are involved in fatal and injurious events, the causes
of the events recorded in the database, and the locations of the
releases. Even this database has many caveats associated with it,
however. For example, the deaths and injuries may not be related
directly to exposure to toxic substances, but may have resulted from
collision, fire, explosion, or other related causes. Also, the method
of collecting the data may have been biased more toward significant
events, rather than toward less significant events. This database may
not have statistical significance, but this is the best source of
information available at this time.
What do the data indicate? For events causing death or injury,
chlorine was the chemical released in 9.6 percent of the cases,
anhydrous ammonia was released in 6.8 percent, followed in decreasing
40
-------
frequency by sulfuric acid, PCBs, hydrochloric acid, nitric acid,
toluene, methyl alcohol, and sodium hydroxide.0)
Range of Release Characteristics
What are the ranges of parameters that need to be considered when
modeling an accidental release? Some of the major factors in chemical
releases are: the chemicals involved, the physical state of the
chemicals, the conditions of release, and meteorological data. Most
source strength and dispersion modelers are familiar with the
differences due to different chemicals and meteorological data, but may
not be familiar with some of the physical states and conditions of
accidental releases.
The physical state includes solids, liquids, gases, or two-phase
releases. Low boiling point liquids are known for a slow, steady
evaporation rate. Superheated liquids and refrigerated gases are known
to have an initial flashing condition, and then a steady rate of
evaporation. After evaporation takes place, liquid spills may be
modeled as light, neutrally buoyant or dense gases, depending on the
characteristics of the gases formed. Usually, the heavier-than-air
clouds are considered the most serious because they tend to stay near
the ground and, thus, in the breathing zone of the surrounding
populations. Sometimes, however, due to meteorological conditions,
lighter-than-air gases can act as heavier-than-air gases and then
warm up to be neutrally buoyant and then lighter-than-air gases.
41
-------
Two-phase releases are mixtures of gases and liquids. These often
result from a safety valve release from a high-pressure vessel or an
explosion in a chemical plant. A boiling liquid expanding vapor
explosion (BLEVE) is a subset of this group. Solids can be divided
into four major groups: powders, molten solids, solids in solution,
or brick-like solids. Of these, powders often pose the largest problem
in terms of fires and explosions, while brick-like solids are very
rarely considered a problem.
Many other parameters affect short-term releases, including:
unexpected reactions (including products of combustion, which may
themselves be toxic) and characteristics of the release (e.g., jet vs.
passive releases). For more information on factors involved in
accidental releases or equations used to describe releases, see the
book by Frank Lees, Loss Prevention in Process Industries.(3)
Representative Credible Scenarios
The final purpose of this paper was to establish credible
scenarios for analysis at the workshop. The scenarios chosen are all
historical accounts from a review of individual articles, the Acute
Hazardous Events Database and the Loss Prevention in Process Industries
by Frank Lees. These cases represent some of the worst situations
involving the chemicals chosen.
The first scenario is a chlorine release. Chlorine is a
heavier-than-air chemical involved in 9.6 percent of the accidents in
the Acute Hazardous Events Database and has been credited with causing
42
-------
2 deaths and 382 injuries.(*) Chlorine is stored as a liquefied gas in
amounts up to 450 tons and in larger quantities at low temperatures and
atmospheric pressures.(4) On December 10, 1976, chlorine was released
in Baton Rouge, LA, at a plant undergoing start-up after shutdown for
maintenance. An explosion overturned the tank, puncturing its sidewall
on a sharp object. Over the next 6 hours, 100 tons of chlorine were
released.(3) What model would be recommended for the emergency
response personnel making evacuation/shelter-in-place recommendations?
The second case study involves an ammonia release. Ammonia was
the third highest chemical in production in 1985.(5) It is credited
with 271 injuries and 4 deaths in the Acute Hazardous Events
Database.U) Ammonia is lighter than air but can act as a'heavy gas in
very cold conditions.(3) The Mid-America Pipeline system crosses seven
states covering 720 miles with a 6-inch pipe and an 8-inch pipe. It
has a pumping capacity of 1180 tons per day.(^) Three releases have
been reported on this system in .the last 10 years.U»3) The latest
release occurred on July 31, 1981, in the countryside near Willowbrook,
KS. An 8-inch pipeline of anhydrous ammonia ruptured at high pressures
releasing 700,000 pounds before being stopped by the shut-off valves on
both ends. The release occurred in an unpopulated portion of the
country and resulted in no deaths or injuries.(l>2,6) The book, High
Risk Safety Technology, states that in the event of a catastrophic
release of liquefied gas under pressure the chemical may be thrown into
the air as a mixture of vapor and very fine liquid droplets. Air may
be entrained in the cloud and evaporate the liquid, thus cooling the
cloud. This will cause the cloud to act as a dense cloud.(7) Would
43
-------
this have occurred In this case? If so, when would the cloud rise?
Would a different model be used for summer and winter conditions? What
is the meteorological cut-off point for modeling of this type of
release? Do the models consider ammonia leaks that initially act as a
dense cloud release, then as a neutrally buoyant cloud, and then as a
lighter-than-air cloud?
A third scenario is the release on January 4, 1986, at the
Kerr-McGee Corp. uranium processing plant at Gore, OK. A 27,000 pound
cylinder of uranium hexafluoride was overfilled by 2000 pounds. In
response, employees heated the cylinder to liquify the contents and to
remove the material. The cylinder burst, instantaneously releasing 14
tons of uranium hexafluoride, which reacted to form hydrogen fluoride
and radioactive uranyl fluoride particles upon release into the
atmosphere. The hydrogen fluoride then reacted with the moisture in
the air to form toxic hydrofluoric acid. Although the media emphasized
the radioactive aspects of the release, the one person killed and the
100-plus injured were attributed to the release of hydrofluoric
acid.(8) In this case, hydrogen fluoride has a molecular weight less
than air but may act as a dense gas. It may become associated and
consist of a mixture of hexamers and monomers with an effective
molecular weight of 70. Dissociation takes place as the gas is diluted
with air, but the heat required to dissociate cools the mixture and can
keep it denser than the surrounding air.(7) Given this scenario, would
the hydrogen fluoride form a dense cloud? What models should be used
to predict the concentration of hydrofluoric acid under these
conditions? Would the models have predicted an 18-mile vapor cloud?(8)
44
-------
Summary
The purpose of this paper was to answer questions posed by
identifying actual case histories and statistics of releases in the
United States and to present three scenarios for analysis during the
workshop in which it was presented. The scenarios chosen included two
of the chemicals most frequently involved in accidents resulting in
death or injury (i.e., chlorine and ammonia), and a complex release of
hydrogen fluoride.
Table of Conversions
1 mile
1 Ib
1 psi
1 inch
Tf
1.6093 km
0.454 kg
6.895 kPa
2.54 cm
1.8 x Tc + 32
REFERENCES
1. U.S. Environmental Protection Agency. Acute Hazardous Events
Database Executive Summary, EPA-560-5-29(a) U.S. EPA Office of
Toxic Substances, 1985.
2. Conversation with Fred Talcott, EPA, Office of Toxic Substances,
Aug. 13, 1986.
3. Lees, Frank P., Loss Prevention in Process Industries. Butterworth
and Company (Publishers) Ltd., 1980.
4. Kirk-Othmer "Encyclopedia of Chemical Technology," Third Edition,
1980.
5. "Top 50 Chemicals Output Declined 3% Last Year," Chemical and
Engineering News, April 21, 1986.
6. Conversation with Max Young, Mid-American Pipeline,
Sept. 11, 1986.
45
-------
7. Green, A. E., High Risk Safety Technology, John Wiley and Sons,
Ltd., 1982.
8. U.S. Nuclear Regulatory Commission Asse!^F5t ^^tt
Health Impact from the Accidental Release of UF6 at the
Fuels Corporation Facility at Gore, OK, Docket No. 40-8027, License
No. SUB - 1010, U.S. NRC, 1986.
46
-------
SHORT-TERM TOXIC RELEASES FROM CHEMICAL MANUFACTURING SITES
Robert E. Rosensteel
Office of Air Quality Planning and Standards
U. S. Environmental Protection Agency
Research Triangle Park, NC
The U.S. Environmental Protection Agency's Office of Air Quality
Planning and Standards in 1985 published in the Federal Register
notices of intent to list eight organic pollutants under Section 112 of
the Clean Air Act. Inclusion on this list was based primarily on
carcinogenic risk potential to the public and, to a lesser extent, on
other health effects. Exposure estimates for noncarcinogenic health
effects for the intent to list decisions consisted of utilizing
information on the continuous emission rates in conjunction with
various combinations of terrain and meteorological conditions. An
acute risk focus may have resulted in a different group of organic
compounds. The eight compounds undergoing evaluation for short-term
health effects are:
1. Butadiene
2. Carbon tetrachloride
3. Chloroform
4. Ethylene dichloride
5. Ethylene oxide
6. Methylene chloride
7. Perch!oroethylene
8. Trichloroethylene
47
-------
A program was initiated in 1986 to gather information from
processes producing or using one or more of these chemicals to evaluate
the potential for noncarcinogenic health effects associated with
releases of one or more of the eight organics. The production and
related processes considered up to the current time are:
0 Butadiene production
• Chlorinated hydrocarbon production
• Chlorinated solvent-using processes
• Chlorofluorocarbon production
• Epichlorohydrin production and user processes
• Ethylene dichloride production
• Ethylene oxide production
• Neoprene production.
This paper summarizes the screening procedure we designed to
evaluate nonroutine releases from chemical manufacturing and using
source categories. Data are being collected for several types of
releases:
• Process vent emissions resulting from start-ups, shutdowns,
and control device bypasses
• Pressure relief events
• Equipment opening losses
• Handling and storage emissions
• Accidental releases.
Detailed information to characterize the releases was obtained with a
questionnaire. The questionnaires were fairly extensive, asking for
information to allow us to pull together the necessary information for
submission to our Source Receptor Analysis Branch for modeling.
48
-------
For process emissions, the questionnaire addresses information
associated with start-up, shutdown, or controlled-bypass conditions.
Meeting with the Chemical Manufacturer's Association, we received input
that helped us to identify those actual process events that result in
the higher emissions. There is, of course, a possibility that some
events could occur beyond the limits of the profiles that were
established, but we are confident that we captured the vast majority of
events. Another area where we obtained information concerns pressure
relief events. There may be safety relief valves that are designed to
automatically relieve overpressure process conditions that may result
in relatively high emission rates for short periods.
In the area of equipment openings, the questionnaire inquired
about experience and procedures, e.g., in preparing equipment for entry
for either maintenance or inspection purposes and where residual
material left in the equipment could be released during the procedure.
We have found some equipment maintenance activities that represent
significant emission sources. Handling and storage of materials also
represent a potential source of release. For example, working losses
are associated with such activities as pumping into a storage tank. At
some facilities, such transfers can involve fairly large vessels.
Typical activities include transferring of products to rail cars or
tank trucks and the evaporation from wetted material that is not
properly sealed.
The final category of nonroutine releases included in the
questionnaire involved accidental releases. We have been discussing
49
-------
this category throughout this conference. Examples include failure of
a pump seal, liquid spills, and gas or vapor releases.
The data collected via the questionnaire have been used as inputs
to a screening procedure to predict ambient concentrations for worst-
case conditions. We are working with the Source Receptor Analysis
Branch and the Pollutant Assessment Branch of the Office of Air Quality
Planning and Standards to combine the information we obtain from the
questionnaires along source category lines, as we typically do in our
standards development work. For example, for each of the plants that
utilize a carbon tetrachloride scrubber on their chlorine tail gas, we
are taking the events that have been reported and combining them into
what we characterize as a worst-case scenario. Our modelers are also
using conservative meteorological and atmospheric stability conditions
for evaluating this worst case. If this worst case analysis does not
result in a significant health concern, then no further refinements in
that analysis are undertaken.
Development of modeling cases incorporating several simultaneous
emission events to evaluate potential ambient exposure for a chemical
production category is under way, and some preliminary modeling results
have been completed. We obtain a plot plan from the source to locate
the emission points from the process. Generally, we are looking at
averaging times of 15 minutes, 1 hour, and 8 hours, which are typically
used for the health effects evaluations being considered. It is also
necessary to compare the relative contribution of the various modes of
release, i.e., maintenance versus background equipment leaks. If this
type of evaluation suggests that health effects may be experienced by
50
-------
the public, then a refined analysis will have to be developed to
evaluate short-term releases from the source category.
51
-------
FATE OF TOXIC RELEASES IN THE ATMOSPHERE--
ATMOSPHERIC RELEASE ADVISORY CAPABILITY (ARAC)
Marvin H. Dickerson
Lawrence Livermore National Laboratory
Livermore, CA
The Atmospheric Release Advisory Capability (ARAC) is an emergency
planning, response, and assessment service, developed by the U. S.
Departments of Energy and Defense, and focused, thus far, on
atmospheric releases of nuclear material. For the past 14 years, ARAC
has responded to over 150 accidents, potential accidents, and major
exercises. The most notable accident responses have been the COSMOS
954 reentry, the Three Mile Island (TMI-2) accident, the uranium
hexafluoride accident at the Kerr-McGee Plant at Gore, OK, and the
Chernobyl nuclear reactor accident in the Soviet Union. At this
workshop, Joe Knox presented a paper based on the ARAC analysis of the
Chernobyl accident. I will discuss the ARAC project in general.
I would like to give you a brief history of the ARAC program. The
concept for the project was developed about 16 years ago essentially by
three people: Rudy Engleman, now with the National Oceanic and
Atmospheric Administration (NOAA) and present here today; Joe Knox,
still with ARAC and here today; and Todd Crawford, now with the
Savannah River Laboratory. I have had the good fortune to be involved
with the project for more than 14 years.
ARAC directly supports the Department of Energy (DOE) and the
Department of Defense (DOD) by providing real-time assessments of the
consequences that may result from atmospheric release of radioactive
53
-------
material. As with any similar program, there are three key components:
the staff, the models, and the computer systems. We have a number of
meteorologists on the team. Several times during this workshop I have
heard comments that meteorologically trained staff are the persons
everyone turns to for advice during an accident. The core of the ARAC
response team is composed of meteorologists.
The types of incidents that we address are quite varied, and they
have numbered approximately 150. They include accidents/incidents in
transportation, at DOE facilities, at DOD facilities, at power
reactors, extortion threats, satellite reentries, and atmospheric
weapon tests. We continue to be involved in a number of tests and
exercises for our DOE and DOD facilities and conduct exercises with
states and nuclear power plants. As a continuing link to the research
community, we participate in planned atmospheric tracer releases.
We serve many masters depending on the type of accident that
happens. We serve seven DOE sites (with the potential to serve up to
about 15): Livermore National Laboratory and Site 300, Sandia
Livermore, Savannah River Laboratory, Pantex, Mound, and Rocky Flats.
We directly serve these facilities by maintaining databases for terrain
and geography and by having meteorology data coming into our emergency
response center in real time. We serve approximately 40 nuclear-
capable DOD sites and their commands. During the Three Mile Island
stabilization, we worked for both DOE Headquarters and the EPA. We
also worked closely with these agencies for the Chernobyl accident. We
have worked with several states regarding nuclear material:
California, Pennsylvania, New York, Virginia, Florida, and Texas. In
54
-------
fact, we have a fairly broad customer base and during any one incident
may work with multiple customers.
While the source term is very important to modeling efforts, one
of the first things we learned in the emergency response business is
that we will not know the source term at the beginning of an accident.
What we do is normalize the source term, which becomes a useful piece
of information in our early calculations. Then the people in the field
begin to receive data to bracket the effects using the normalized
values as multipliers.
In recent years, we have been involved in several toxic chemical
releases: a 1976 train accident involving uranium hexafluoride (UFs),
a Titan II accident (even though there was a weapon on board it was not
damaged so this was a toxic event because of the missile fuel), a
hydrogen sulfide leak during a material transfer at the Savannah River
Plant in 1981, the Gore, OK, UFg accident, and the white phosphorous
rail accident in Miamisburg, OH.
The Miamisburg incident lasted for 5 days, and our colleagues at
Mound Laboratory, near the accident, provided input to the evacuation
planning process during those days.
The models that we maintain to support our customers must cover
widely different scales. For nuclear weapons tests, we have to cover a
global scale. As it turned out, we also used global-scale modeling for
the Chernobyl event. ARAC's operational models are highlighted as
fol1ows.
55
-------
ARAC OPERATIONAL MODELS FOR NUCLEAR MATERIAL
Model
Type
Simple
Intermediate
Complex
Global
Scale
=20,000 km
--
2BPUFF
PATRIC
Synoptic
Scale
=2,000 km
--
2BPUFF
PATRIC
ADPIC
Regional
Scale
=200 km
--
MATHEW/
ADPIC
KDFOC2
MATHEW/
ADPIC
Meso
Scale
=20 km
Gaussian
MATHEW/
ADPIC
KDFOC2
MATHEW/
ADPIC
MATHEW/ADPIC is a three-dimensional, mass-consistent, wind field
model coupled to a particle-in-cell transport, diffusion, and
deposition model. It is generally applied to regions 200 by
200 kilometers or less. However, for the Chernobyl accident it was
expanded to 2000 by 2000 kilometers. The 2BPUFF is a two-dimensional
long-range transport and diffusion model, most often applied to nuclear
weapons tests. PATRIC is a three-dimensional, particle-in-cell
diffusion model that was specifically designed to treat continental and
hemispheric scales; it is a simplified version of the ADPIC model.
KDFOC2 is a fallout model used for surface or partially buried nuclear
detonations.
These are the models that we have operational. For our on-line
sites, where we have databases, we can produce calculations in about 40
to 45 minutes from the time we are notified (new computer capabilities
will reduce this time to 15-20 minutes). This is a full three-
dimensional calculation. If a site does not have its own capabilities,
56
-------
then we provide them with a system in which they can get simple
calculations almost instantly from tower data.
Our regional meteorological data come from the established
national meteorological network. For the DOE and DOD sites, the system
that remotely collects the data is a DEC PC 350/380. Some sites are
connected directly to a local tower with the data coming directly into
the system1. Others have smaller towers and the data are hand entered.
Still others depend on nearby airport data. The minicomputers collect
the data, display it locally, and transmit it to the center in
Livermore. We also have direct links with Air Force Global Weather
Central and a private meteorology company. We receive data from the
National Weather Service as well. Thus, we operate a large data
collection, database, data analysis, and model-execution facility at
Livermore.
We are currently expanding our capabilities in several areas. We
are acquiring geographical information from the U.S. Geological Survey
for the entire United States. We are currently experimenting with our
own local terrain data, including the overlay of model particles on the
terrain data. In complex terrain, we are able to overlay the
geography, the terrain data, and the particle behavior. Eventually, we
hope to be able to accomplish this in real time as part of our normal
operational response.
We think some of the high payoff research areas are:
1) development of a local- to regional-scale meteorological forecast
model or dynamic model, not for an extremely complex terrain to begin
with, but perhaps for rolling terrain; 2) development of a realistic
57
-------
rainout module for both the regional- and global-scale models; 3)
continued model evaluation studies; and 4) expansion into toxic
chemical response, including addressing the issues of source term,
dispersion, and reactivity. The models that we are using today to deal
with nuclear releases will be applicable to the toxic chemical events
that are nonreactive and at ambient density. We will need to include
new capabilities, however, to deal with the chemicals that are reactive
and those that are heavier than air. We believe these capabilities
should be integrated into the ARAC system in the future as part of a
total emergency response capability.
58
-------
SOURCE STRENGTH MODELING
Jerry M. Schroy
Monsanto
St. Louis, MO
Most chemical companies maintain up-to-date material balances
around processes in newly designed facilities. They understand what
TM
the emissions are and maintain access to models, such as Flowtran ,
(which Monsanto uses) or other similar simulation models to keep the
data current.
For the most part, when we talk about process emissions, people
normally think of routine emissions. However, I will discuss
nonroutine events, such as runaway reactions and fire situations. They
will represent elevated sources for which the questions related to
dispersion are a lot clearer. Most of the work I will discuss today
was done in response to the acrylonitrile standard that the
Occupational Safety and Health Administration (OSHA) developed in 1979.
Industry attempted to get a handle on whether or not it could meet the
objective and criteria OSHA was attempting to address. In response,
Monsanto developed a series of models to assist in that understanding.
The models were developed in general terms that allow their use with
any chemical for which physical and chemical properties are known.
In 1979, Monsanto put its model on-line for any company engineer
to use, worldwide, on a computer system over which the developers of
the model have no control. The engineers can use it for whatever
purpose they wish. While the model has been used for many purposes,
most are related to accidents.
59
-------
Most of the accidents that I will discuss today, relating to
releases of organic and inorganic materials, are based on situations
such as overfilling tanks or cutting of small pipes. Others have
discussed bursting tanks and instantaneous releases today.
Our model best addresses issues dealing with releases where there
is a release point, flashing, liquid, and a pool formation. The
potential presence of aerosols in the flashing material must also be
considered. An engineer, able to use our company model, can usually
make a judgment of whether or not an aerosol is involved. We do not
know how to model the aerosol yet. The American Institute of Chemical
Engineers (AIChE) has initiated some work in an attempt to qualify
aerosol behavior. Monsanto's models can, however, handle questions
related to flashing, pool evaporation, and release rate.
All of the equations and criteria used in our model are presented
in the paper, "Emissions from Spills" by John Wu and me (Air Pollution
Control Association [APCA] Specialty Conference, 1979). I will not
cover those details today, but will cover other issues. You should
understand that the Monsanto emission model can handle different tank
configurations, whether they are horizontal or vertical arrangements
and whether the structure is a fixed tank or a tank wagon, on rail or
on truck. It makes no difference.
In dealing with questions related to diking, the model was
designed to deal with the question of how fast the surface of the pool
grows. If the release occurs in a diked arrangement, the surface will
grow out to a maximum area, and then pool calculations are relatively
straightforward. If the material is released into an undiked area, how
60
-------
big will the pool grow? Part of the model and part of the equations
presented in our paper deal with that issue. The model will calculate
a rate of growth. When dealing with chemicals that volatilize rapidly,
such as hydrogen cyanide (HCN), the model will malfunction, because it
does not know how to deal with pool shrinkage. The model will reach a
point where there is more volatilization than there is incoming
material into the pool, and it will fail. However, you can avoid this
problem in most cases. By cutting study time back to the time interval
that only allows pool growth to the diameter where the amount entering
the pool equals the amount evaporating, you can define the critical
area. Actually correcting the model is a minor issue that simply
requires time.
The question of model validation comes up frequently. One area,
of which many of the meteorologists in the audience are aware, is the
abundance of pan and lake evaporation data in the world. When we have
attempted to match the data for big lakes and ponds, we have found a
fairly close comparison.
Our model is dynamic. It solves heat and mass balances
simultaneously and provides temperature responses of the pool. Some
questions related to cooling arise when dealing with releases of a
chemical like ammonia.
A second area, where issues related to source terms come into
play, occurs in cases such as spilling material into a sewer or
attempting to control a spill by washing it into a waste treatment
pond. Randy Freeman and I and a few others have done some work related
to surface-aerated ponds and systems using diffused air systems. Short
61
-------
of a long explanation of the models, this requires a steady state
model. In this case, we are trying to solve simultaneous differential
equations for competitive processes, and we do not know how to do that
dynamically yet, at least not with our model and not on a computer
system we can afford.
We deal with the question of competitive removal in the model in
terms of biological, chemical oxidation, or chemical reaction in
competition with air stripping. The user must input a first guess
answer to initiate the model. The better the guess, the lower the cost
to run the program. The worse the guess, the more expensive it is to
run the program. So far, we have done fairly well on closure of the
balance. When biological oxidation is going on, you have to have
actual information regarding pure material kinetics versus the amount
stripped. In the absence of biological oxidation, the model air
stripping predictions are within about 3 percent of actual air
stripping rates.
In those cases where we must use a "lattice of models," as Dr. Cox
discussed earlier, we have the question of what is occurring with
respect to seepage into the ground. We have a temperature model that
has been built into our "dioxin" model. We refer to the base model in
terms of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD) only because that
is the first chemical we modeled with it. Now, quite a few chemicals
have been studied.
The temperature model allows variation of temperature with time at
the surface and at different depths within the soil. Behavior of
chemicals in the ground has always been messy to profile. The chemical
62
-------
migrates both into and out of the ground. Our focus in using a model
for source strength is straightforward, however. We are trying to
understand the impact on people by accurately estimating the
concentration of the material in the airspace.
Using dioxin, the model was tested with 1984 data from a 1972 Air
Force experiment. The dioxin started at about 10 centimeters below the
surface and moved up and down. This is unsaturated zone behavior.
Dioxin has a vapor pressure of about 10"9 millimeters. Additionally,
Dr. Nash from the U.S. Department of Agriculture at Beltsville, MD,
provided us with data from experiments on dieldrin, heptachlor, and
lindane with which we tested the model. Lindane, for example, has a
vapor pressure about five orders of magnitude higher than dioxin. If
we were looking at benzine or HCN, the same migratory behaviors would
be occurring. The only difference is that the migration would not be
spread out over 12 to 30 years, it would be over a matter of minutes or
hours.
From a safety point of view, the spills model will also model down
to something equivalent to TCDD, and we can use it for something that
extreme. It does provide conservative values, but does not consider
downward migration. Therefore, from the people point of view, if you
model only the evaporation from the pool, you have a conservative
approach, and that is the central question regarding source terms in
active models.
Another source term question area, which we all face, is the
runaway reaction question. There you must go to other models, whether
it be Sapphire or one of the other models available from AIChE or their
63
-------
contractors. The models available from AIChE and the new ones under
development appear to be relatively realistic in providing results. At
Monsanto, we use data from an accelerated rate calorimeter to drive a
reaction kinetics model for both conditions of runaway reactions and
fire impingement on a vessel to determine the rate of release from a
rupture disc or a relief valve.
We have heard some about routine emissions in terms of
understanding background. We must add fugitive emissions to the list
because background makes a difference. If background levels are
considered, pool evaporation may be suppressed. Fugitive emissions in
the work place have been documented. However, we feel that EPA's work
is conservative by a factor of at least ten in a lot of cases for small
plants, but realistic for refineries. Fugitive emission estimates
provide an idea of the kind of losses that will occur from sources and
the impacts that they will have on workers.
Regarding the conservative nature of EPA's estimates of emissions
from small plants, we define small plants as those that have small
piping and use chemicals of lower volatility than ethylene and propane,
which are some of the lighter chemicals found in a gasoline plant.
Comparison of actual measured mass emissions from an overall process
within a structure versus our estimate versus EPA's estimate will give
you an idea of the kinds of ranges available.
There are copies of our paper available, and I would be available
to discuss our work with those interested.
64
-------
FUTURE NEEDS FOR DISPERSION MODELS IN HAZARD EVALUATION,
EMERGENCY PREPAREDNESS, AND ACCIDENT PREVENTION
James L. Makrls
U.S. Environmental Protection Agency
Washington, DC
Someone In the audience observed that there are about 55
dispersion models available at this time, none of which are great and
some of which are not any good at all. So, the subject we are dealing
with here is certainly very timely. Since this time last year, many
events have occurred that have impact on our discussions today. The
Chemical Emergency Preparedness Program Interim Guidance has been
released in draft form and commented on extensively by the public. It
is now being combined with an older Federal Emergency Management Agency
(FEMA) document and will be released as the Hazardous Material
Emergency Planning Guide. Previously referred to as FEMA-10, the
earlier document dealt with contingency planning for hazardous material
accidents on a state and local level. The new unified document will be
published by the National Response Team and made available through
notification in the Federal Register by March 17, 1987.
Recently, a couple of major accidents reminded us that we are
dealing with a serious business. The Miamisburg incident caused the
evacuation of 40,000 people following a rail accident that released
white phosphorous. The Kerr-McGee release of uranium hexafluoride
brought an entire new group of facilities into our area of interest,
i.e., materials-handling facilities from the nuclear industry. The
Kerr-McGee incident also alerted us to areas where we had thought that
65
-------
all government agencies were functioning in a unified way. We learned
that there were still some activities that the Department of Energy and
the Nuclear Regulatory Commission had not quite aggregated together.
This has been corrected, and the Federal agencies now meet on a monthly
basis to address issues of common concern regarding accidents
specifically.
This year, state and local governments, including the States of
Hawaii, Illinois, New Jersey, and California, passed legislation aimed
at preparing to provide the public with information on the chemical
risks that they face, as well as requiring the state to move forward in
preparing for an accident before it occurs. We may even have new
Federal legislation in the form of the pending Superfund
reauthorization bill. When or if the President signs the bill, it
contains an important right-to-know component in Title III.
Basically, it instructs the governors to appoint commissions, who
in turn will appoint local planning districts and committees. Those
committees would undertake immediate contingency planning around the
list of 402 extremely,hazardous chemicals (originally published in the
Chemical Emergency Preparedness Program Interim Guidance). Facilities
that exceed the baseline quantity limits for storage established by EPA
must submit information and participate in a broad-based community
contingency planning effort. EPA is currently rushing to establish
those levels.
If EPA fails to establish another level, the new Superfund bill
requires that a base level, i.e., quantity, of 2 pounds be established
as the amount of a listed chemical in a facility requiring the
66
-------
submission of contingency planning information. Therefore, two pounds
of any one of 402 chemicals will potentially require that a contingency
plan be developed for the local community. Certainly, the base level
for some chemicals will be increased. However, the important thing is
that this will require a contingency plan that will deal with such
issues as evacuation, emergency organization, notification, and public
health protection. The local committee must include the specific
elements of evacuation, identification of hazardous chemicals,
identification of the organizations and individuals in charge, and must
in fact develop a plan. One of the key issues is going to be what
happens when a chemical release takes place? Who makes judgments on
how far to evacuate and when? All of these issues are going to depend
upon some unified Federal guidance being provided to that local
community, in order to assure there is no division, or diversion, or
confusion at the state or local levels, as they proceed to comply.
Other important aspects of this legislation are that it comes with no
Federal funding and that it requires no Federal approval. It says to
the governor that it is the state's responsibility and lists several
activities the state must undertake to meet that responsibility.
This brings us directly to the issue we are discussing today. The
states are going to be clamoring for specific information on modeling.
It seems appropriate that the Federal government, in conjunction with
public interest groups, industrial associations, technical institutes,
scientific groups, and other agencies, can meet together to deal with
this issue and to provide state and local governments with information
67
-------
around which they can develop specific contingency plans for those
chemicals that are in their community.
There was some skepticism earlier about whether industry would
really come forward with the information that would be required to do
this contingency planning effectively. The law is going to require
that industry not only disclose the existence of the chemical, but also
how much they have, and for specific chemicals, where they keep them,
and what protective systems exist.
Despite all of this planning, however, a key issue will always
remain. That is, what is going to happen when a release occurs in
spite of all the precautions that precede, and all the mitigation
planning that had occurred? When an event occurs, what is the
evacuation route? What is the dispersion probability? What is the
plume of the release and what are the effects going to be? It is
around these very issues that this and other groups must apply their
greatest efforts. It is clear that a lot of dispersion models
currently do not take into account issues of topography and weather.
When we first sent people to Cameroon, carbon dioxide had not been
established as the cause of the deaths, although it soon was. It was a
shocking thing to learn that people and all animal life were felled in
their tracks 25 kilometers away from a point source by this heavier-
than-air chemical. It is hard to grasp that a chemical as common as
carbon dioxide could disperse for a distance of some 15 miles and still
remain in such intense concentration to kill. Obviously, we have a lot
to learn about the dispersion of chemicals in the atmosphere.
68
-------
We are involved in policy-in-the-making as we work together trying
to deal with this problem. Our solutions will provide state and local
governments with the ability to respond quickly and accurately for the
protection of public health, which is clearly a state and local
responsibility. Title III of the Superfund bill is somewhat vague. We
will need information on how to set thresholds, add chemicals, or
eliminate chemicals; that is certainly within our prerogative. There
is, however, no mandatory prevention program included in Title III.
While it is not legislatively mandated at this time, EPA is gathering
information and reviewing technologies that we think will be able to
lead to an improved capability of preventing chemical accidents. While
we have a fair amount of activity going on in the prevention area, it
is catalytic, not definitive. Most of our efforts deal with prodding
others to do their jobs and being sure that there is an adequate
transfer of technology.
In this same vein, we believe that dispersion modeling well suits
the kind of forum we are having here, where ideas can be shared. In
February, 1987, we are having an international symposium in Washington,
DC, where the technology that exists throughout the world dealing with
the prevention of chemical accidents and the dispersion issue will be
brought together. Further, the American Institute of Chemical
Engineers (AIChE) is having a major dispersion modeling conference in
this country in November, 1987. These kinds of activities are to be
commended for supporting the coalescence of information and guidance
that can be provided to state and local governments, as they try to
69
-------
make the determinations of how to protect their citizens from future
chemical accidents.
We all need to work together, and that is what we are doing.
Everybody has a job. The public sector has a job of being a catalyst,
a cautious neutral, an informed neutral. The private sector has a job.
It is their chemical; their liability. State and local government has
a role. They have been responsible for protection of public health and
safety on the local level for a long time, and they continue in that
role. Federal agencies like FEMA, the Department of Transportation,
and the Coast Guard all have a role. Agencies like the Centers for
Disease Control and the Poison Control Centers have responsibilities as
they review these chemicals and chemical profiles and reach conclusions
about how to deal with them. The scientists have a job to continue to
move forward in these very complicated technology areas. The citizen
has a job, not only a right to know, but a right to become informed.
The job requires active participation and learning, not only in
assuring that there is a high level of safety, but also in passing
appropriate local ordinances where they best prevent chemical accidents
from having an impact on people.
Most of all, we all have a job together to try to increase
chemical safety in this country. In the event of a chemical accident,
we must be sure that the proper information is available for state and
local officials who have to make key decisions to protect the public.
70
-------
ISSUES IN REGULATORY APPLICATIONS OF MODELS
David E. Layland
U.S. Environmental Protection Agency
Research Triangle Park, NC
The Source Receptor Analysis Branch of U.S. EPA's Office of Air
Quality Planning and Standards (OAQPS) is responsible for supporting
the development of air regulations under the Clean Air Act and, more
recently, the Resource Conservation and Recovery Act, as well as
supporting activities of other EPA groups in the area of dispersion
modeling. OAQPS views modeling needs and issues primarily from the
perspective of planning for sudden, unplanned, routine release events,
rather than catastrophic releases per se. These issues and needs will
be discussed from the standpoint of the following:
t Regulatory and technical background giving rise to the need
for dispersion modeling,
• Various technical issues remaining to be resolved, and
• Considerations in devising a framework to approach meeting of
those needs.
Although OAQPS is mostly concerned with more routine releases, it
is worth mentioning two aspects of planning for the management of
catastrophic releases. One is prevention, and the other is emergency
response planning. Looking at prevention, specifically as related to
hazard analysis (the question of who gets hurt), this is really the
place where dispersion models find their best use. Included here are
other modeling techniques: source modeling, health assessment, and
identification of the release mechanism itself.
71
-------
Section 112 of the Clean Air Act (CAA) deals with establishing
National Emission Standards for Hazardous Air Pollutants (NESHAPS). It
requires a listing of hazardous air pollutants first and a setting of
standards second. A hazardous air pollutant is defined as one that
could cause an increase in mortality or an increase in serious
irreversible or incapacitating reversible illness. Thus, from a
regulatory perspective, not just situations that cause immediate danger
to life, but also those that cause sublethal acute and chronic health
effects are to be considered. Section 112 requires that emission
standards be set at a level that would provide an ample margin of
safety to protect the public health.
OAQPS is currently looking at several types of sudden releases
from the standpoint of dispersion modeling. These types of releases
may result from:
• Process overpressurization; e.g., pressure relief valve
events;
• Transfer and handling operations, e.g., leaks and spills of
different magnitudes and sizes and vapor overflow from the
filling of large tanks or vessels;
• Malfunction of control devices, e.g., scrubbers and
incinerators; and
• Maintenance operations, i.e., typically smaller releases,
tank flushing, and equipment opening.
These types of releases have some general characteristics that
must be considered. Duration of releases can vary from quasi -
instantaneous releases lasting only a few seconds to quasi-steady
releases lasting several minutes or longer. There are three
depressurization characteristics: 1) rapid jetting action from relief
valves and from other pressurized releases, 2) adiabatic expansion and
72
-------
cooling, and 3) two-phase flow. Other characteristics of concern
include formation and spreading of liquid pools and evaporation from
those pools, along with heating effects associated with the substrate
or the atmosphere. Density-related effects include the sinking motion
of negatively buoyant elevated plumes, the formation of secondary
source clouds with associated dampening of turbulent motion, and the
gravity spread of dense gas clouds with terrain contour interaction.
Technical issues of a regulatory nature can be grouped into three
basic areas: 1) model selection, 2) model application, and 3) model
verification. Model verification should probably be first on this
list, but experience shows it tends to be dealt with last.
One of the most important criteria in model selection is the
availability of the model in the public domain. For regulatory
purposes, proprietary models must generally be excluded. Other aspects
related to model acceptability are code standardization and
maintenance, and the availability of users manuals and related
documentation. The theoretical soundness of the model is also
extremely important; i.e., is the model consistent with current
knowledge?
Consideration must also be given to whether a model is intended
for screening or whether it is a more refined model. Typically, a
screening model is both simple and conservative, while a refined model
is characterized by being less conservative and more data and computer
intensive. One must consider how conservative the screening model
should be and how much better an answer is obtained by the more refined
models for a given application.
73
-------
Another important aspect of model selection is the release
characteristics that were discussed above, including buoyancy effects,
momentum effects, aerosol effects, and reaction effects. A final
selection characteristic is terrain, in terms of a generalized surface
roughness that affects the rate of dispersion and in terms of specific
terrain features that affect wind flow patterns, e.g., upslope flows,
drainage winds, and ridge effects. Downslope flows of dense gas clouds
may also be an important modeling consideration.
The first issue with respect to model application is the
meteorological conditions to be used for the analyses. OAQPS tends to
perform two types of applications. First is a screening application
where worst case meteorological conditions are used to establish a
worst case. Identification of the worst case meteorological conditions
generally requires some sensitivity analysis. Second, for more refined
applications, a time series of meteorological data is used that
reflects autocorrelation inherent in the meteorological data.
Questions here are, "What is a reasonable worst case?" and "What is the
joint probability that the release event and the meteorological event
will occur simultaneously?"
Another issue of model application is averaging time. It would be
logical to look at the duration of the release and call that the
averaging time, but in modeling, we must be concerned if the emission
is varying very rapidly with time. We may need to focus on the period
of peak emission. Generally, however, it is the occupational or
community exposure standards or criteria that truly dictate the
averaging time.
74
-------
Integration of the source model with the dispersion model is the
final issue of model application. As the models and the events to be
modeled become more complex, there will be greater need to integrate
these two aspects. Of course, there are models available now that do
that. Two complex releases, previously mentioned, that may require
source modeling in order to specify the initial conditions for the
dispersion phase are relief valve discharges and liquid spills.
There are, of course, a number of ways of going about model
verification. One can form a theoretical evaluation of a model by
simply looking at the model formulation, evaluating it, and evaluating
the assumptions of the model on a theoretical basis. Validation of the
model code is also important; that is, checking how various numerical
techniques in the code are implemented and checking the model code over
a range of conditions of the parameters. It is important to adequately
validate the model code itself before the model is used. Validation
against real data may include both laboratory experiments and field
trials. Laboratory experiments tend to be most useful for model
development. For verification purposes, their applicability to the
real atmosphere may be suspect. In the design of field trials, scale
effects (small releases versus large releases), source-related effects,
individual chemical effects, and meteorological conditions must all be
considered. A very large matrix of field trial experiments is
therefore possible, but only a very few are feasible to be actually
carried out. •
In deciding where to go from here, we must ask ourselves, first,
what models do we have that are ready now for regulatory applications,
75
-------
to what situations do these models apply, and with what degree of
confidence? Then, we can ask what models have been formulated but are
incomplete at this point, to what extent can they be adapted to actual
situations, and how can this be accomplished? We must also identify
those situations for which no appropriate model exists, whether the
absence results from a lack of scientific knowledge or from simply a
failure to formalize a mathematical model from existing knowledge, and
what can be done to develop the needed models.
In summary, this has been an overview of dispersion modeling needs
and issues from the OAQPS perspective in the area of hazard analysis.
Some of the issues of model selection, model application, and model
verification were discussed, and some suggestions on where we might go
from here have been made.
76
-------
COMMUNITY NEEDS FOR HAZARD EVALUATION TOOLS
Fred Millar
Environmental Policy Institute
Washington, DC
The Environmental Policy Institute, which has been working on the
issues of major hazardous accidents for about 6 years, believes that
there are now some new opportunities and some new resources that need
to come together. Specifically, what can the public sector do within
the next 4 or 5 years to improve chemical safety?
A major public problem is that people have not been told what
risks they are being exposed to now. There really has not been a
communication of potential hazard to the people. Yet, people are going
to want to know to what risks and hazards they are exposed. It is fair
to say that people in Bhopal were not told of their risks, nor were
people at Chernobyl. Certainly, no one in Europe was told that a
release event at the Chernobyl facilities could dump its radionuclide
inventory on them.
Let me relate a story of how a few months ago I took my
4 and 1/2-year-old daughter to a skating rink in Fairfax County, VA.
As we walked up to the rink, I noticed a gigantic petroleum tank farm
looming over the rink. Also adjacent to the tank farm are a shopping
center and a residential development. I said to myself, "That doesn't
look like very smart zoning." Fairfax County is one of the 10
wealthiest counties in the United States and one of the most heavily
developing at this time. Obviously, there had been encroachment by the
community on an existing tank farm. I decided to approach the local
77
-------
fire chief and inquire about this situation. He reported that this was
only one of what he referred to as his "target hazards," and offered to
give me what turned out to be a 4-hour tour.
First, we went to an underground propane storage facility. Going
into the facility, I knew that this one had leaked. Propane has been
found in adjacent creeks. It is good to go to these facilities with
the fire chief, because you simply walk in and talk with the operators.
A residential community is about 100 yards away from the storage area,
separated by only a small fringe of trees. The operator related that
when the facility is to be filled, they bring in from three to nine
jumbo tank cars simultaneously and fill the cavern. I asked the
operator what the community thinks of the facility. He related that
several of the people in the community had been told when th^ey were
buying their houses that the line of trees was a park where their
children could play. In fact, the trees are part of the storage
facility. It appeared to me that this was a "Mexico City" accident
waiting to happen. Asked why nine tank cars were on site and unloaded
simultaneously, the operator related that it took the same time to
unload nine as to unload one. Since the people were paid the same
overtime, the operator simply chooses to take as many rail tank cars as
he receives at one time. It appears that this is an opportunity for
some operational improvements to reduce risk.
The fire chief took me to another propane facility, where there
were a lot of storage tanks, and where there had been a couple of
leaks. He reported that, to get water at the facility, he had to line
up fire trucks with over 6000 feet of hose. When he asked the
78
-------
operators to get water to the facility, they refused. It is a pre-
existing facility and is protected by a grandfather clause from the
current safety regulations. The fire chief related several other
similar problems. One is a water treatment plant built over 10 years
ago with a chlorine facility, but no scrubber. A similar new plant has
a scrubber. It appears that a lot of retrofitting is required, and
even some fairly simple adjustments in operating procedures would be
beneficial.
Part of the problem is that people have not been told what risks
they are exposed to. The people of Bhopal and Chernobyl were not told
of their risks. There really has not been the communication of hazard
to people. In this country, one can illustrate this by looking at the
U.S. Department of Transportation (DOT) Emergency Response Guidebook--
the "orange book"--that is in every fire truck. There are some "Stone
Age," well actually, more "Model T" tools in that book. For example,
the phosgene entry says that in case of a spill of phosgene in
transportation the fire chief is supposed to evacuate the city 5.2
miles downwind and 3.3 miles wide. There is no fire chief in the
country that believes he or she could do that kind of evacuation.
Further, there are not 10 city council people in the United States that
have ever been told that this material is coming through their city, or
that it could require a 17-square-mile evacuation. The guidebook is
currently under revision, but that is not the entire message here.
There is a basic lack of communication of hazard to people.
People are going to want to know what risks they are exposed to,
and, from a community point of view, they are going to need some decent
79
-------
models and some decent techniques to evaluate the risk. These may be
either paper and pencil techniques or computerized ones.
At a recent conference at New York University, executives from
Johnson and Johnson were talking about their drug case, and Warren
Anderson of Union Carbide was talking about Bhopal, when someone in the
audience said, "Look, you're describing these major industrial crises
from the perspective of a fisherman, rather than the fish." We need to
try to look at things the other way around. There is a need for a
reversal of the way people look at hazard assessment. It is fair to
say that most of the hazard assessment done up to now has been from the
perspective of the facility. The facilities have done a lot of hazard
assessment already, but little has been done from the perspective of
the community. This will be a different perspective, and some
simplification will be required. This may be a bit painful for those
who have had the resources and the perspective of the facility, and
where you want to know exactly what is going on.
The U.S. Department of Housing and Urban Development (HUD)
discovered, when they started subsidizing low income housing around the
country, that a lot of communities, including Columbia, SC, thought the
ideal place to locate subsidized housing for the poor was on what they
called "marginal land," meaning right next to the propane facility or
right next to the rail yard. The development in Columbia was (to be)
right next to (within 150 feet of) 44 propane tanks. HUD did a quick
analysis and found that the explosion of just one of those tanks would
destroy 60 percent of the proposed 1250 housing units. HUD scrambled
and established some "acceptable separation distances" between things
80
-------
that explode and some radius of fire risk, but they still do not even
consider toxic clouds. This is the kind of response we can expect.
Public safety officials are going to establish various kinds of
regulations based on the immediate problems they are faced with.
The National Transportation Safety Board (NTSB) did a rail yard
safety study in 1985. They discovered that of eight major rail yards
reviewed in the United States, not one of them had an emergency
response plan that was coordinated with the local community.
What kind of tools are available to the local community? You can
take a 7 and 1/2-minute U.S. Geological Survey (US6S) quadrangle map of
any area of the United States, a map of the corresponding city, locate
the hazardous facility of interest, and lay down the plume from the
NTSB on it. If we use the Miamisburg, OH, chlorine accident as an
example, the chlorine would be expected to go 2 miles downwind in 10
minutes at a toxic concentration of 100 parts per million. This is
something you can work with, and these are the kind of tools that
people are starting to work with now. The city of Denver is doing a
rail yard risk assessment right now. After Miamisburg, five cities in
Ohio became interested in what was coming through their area.
To many of you who understand the new advances in dense vapor
cloud technology, this is going to seem like "Model T" technology. But
the fact is that we are going to need something tomorrow, not 2 years
or 10 years from now. Local officials are going to need a workable
system, tomorrow to aid them in hazard assessment, not necessarily
quantitative probability analysis. Tools will be needed where normal
81
-------
logic allows you to identify the chemical and the kind of event; then
you can identify the consequences and try to engineer solutions.
Some interesting developments are occurring in the United States
following the Bhopal incident. Union Carbide came forward with a
statement that, "Hey, we ship methyl isocyanate around the United
States to five different places much safer than Federal regulations
require." They reported that they were using better tank cars, better
placards, and special routing to avoid densely populated areas. As a
result, the NTSB jumped on DOT, and now they have revised their
regulations for methyl isocyanate (MIC) and other similar materials.
It is interesting to note that Union Carbide considers the Federal
regulations as minimal. Thus, the better companies have been using far
better safety technology than those "minimal Federal codes." We can
also lump National Fire Protection Association (NFPA) codes, consensus
codes, and other industry codes together. Many of the industries go
beyond these, right? What has happened is a lot of "grandfathering".
Things have been kept at a minimum level. What we need to do is
determine what is the state of the art; what is the best that can be
done with a particular technology? Then, when opportunities arise, the
regulations can be ratcheted up, so that performance is improved.
In 1985, EPA conducted an inspection of -an Allied plant in Baton
Rouge, LA, that is using hydrogen fluoride. .Allied told the inspectors
that they are using state-of-the-art handling of hydrogen fluoride.
The inspectors then go down the river to the next facility and learn
that that is not state of the art. As a result, Allied must agree to
fix things up, and they get a $10,000 fine from the State Department of
82
-------
Environmental Quality (DEQ). With some sort of public input, there
will be some sort of ratcheting up of some of the standards that are
not as adequate as others.
The NTSB took the standard chlorine plume and the USGS map of the
Potomac rail yard in Washington, DC, and overlaid them. They learned
that if the chlorine plume moved east, it would take out National
Airport. If the plume moved west, it would take out six schools and a
hospital in downtown Alexandria. In the case of the plume moving
north, it would take out the Crystal City business complex and the
Pentagon. It would seem to be a fair question to ask if the Pentagon
or the Crystal City office complex have a system of shutting off
incoming air in case of a serious release at the rail complex?
In Middleport, NY, there was a mini-precursor to Bhopal when an
FMC facility released some MIC, and it drifted 100 yards to a school.
The students started gasping and choking. Since the principal did not
have a way to shut off the air ventilation system, he evacuated the
students outside into buses that had been waiting to take children on a
field trip. They went out into the cloud and were evacuated to a fire
station where 200 children had their eyes flushed because of methyl
isocyanate poisoning. Now, the principal has a system where pushing a
button shuts off the ventilation system, and there is some possibility
of sheltering in place. This is the kind of simple solution we must be
putting into place.
What are our priorities? What if the public sector is going to
have a role? What do we need to do first? This is where we need a lot
of help from the experts--those people who have been working in this
83
-------
area for a long time. We need help to identify the kinds of situations
that have the most potential. The National Oceanic and Atmospheric
Administration (NOAA) developed a real-time emergency response system
on an Apple Macintosh computer and placed the system at the Seattle
Fire Department for a year's trial. I am very skeptical about this
working for emergency response in real time. We may be learning how
not to do it. Union Carbide missed the mark with their system during
the aldicarb oxime release at Institute, WV. People are going to be
developing these real time emergency response systems, whether they are
adequate or not.
After Bhopal, we are having a rush to emergency response planning
on the part of all the major actors. For example, the Chemical
Manufacturers Association (CMA) is touting its Community Awareness and
Emergency Response (CEAR) program. CEAR is being implemented in many
parts of the country. In the Lake Jackson/Brazosport, TX, area, the
CEAR program placed full-page newspaper advertisements about emergency
planning. In that area a siren program was developed that covers the
entire community with sirens that have a range of one-half mile. A map
was provided to the community showing the location of the sirens whose
coverage overlapped. When they were asked how the siren locations were
selected, they indicated that a map of the industrial facilities was
first developed and the location of the major chemicals of concern
noted (Map A). Then, the kinds of plumes that might be expected from
accidents were mapped, and potentially affected community areas
identified (Map B). From this information, locations for sirens
(Map C) was established. When asked how the community responded to the
84
-------
map of the potential plumes and effects, the developers of the system
responded that they never showed that map to the community. State and
local regulators are going to want the whole thing (Map A, Map 6, and
Map C). They think they have a right to know this type of information.
. 1 We have to come up with hazard assessment technology that is
workable at the local level and not so refined that it can only be used
by a company with $2 million to put into it. We have to have some kind
of a hierarchy of hazardous assessment techniques and they have to be
acceptable to the local people.
Finally, there is a bit of a double-edged sword in the changes
that hazard assessment and emergency response planning are going to
precipitate in industry. Union Carbide announced they were using a
special transportation planning system (similar to one developed by ALK
Associates of Princeton, NJ) for MIC shipments. Union Carbide
indicated that special low risk routes are selected and population
centers avoided. However, they will not reveal which areas they do go
through or what routing technology they use. They know that revealing
this information will result in demands being placed on them for other
chemicals to be routed with similarly stringent rules. One day before
the Bhopal incident, Carbide shipped MIC to five different locations in
the United States. One year after Bhopal, there are no MIC shipments
in the United States, because people have found safer things to use.
This raises all kinds of interesting questions about the potential for
safety improvements.
At the AIChE meeting in New Orleans, LA, last year the final
session was titled, "Accident Case Histories and Miscellaneous Topics."
85
-------
At the conference, the session chairman, Dr. Ian Swift, got up and
shook his finger at a large audience of engineers and told them that
they were going to have to listen to a whole lot of "miscellaneous"
papers because "not one accident case history" had been submitted. He
said, "I guess your liability lawyers have told you to shut up. This
is a very dangerous trend in our industry." It is understandable, but
it is dangerous. It is unlikely that the public will think that such
an industry has its act together on safety. Bhopal gave them a shock.
It is highly probable that there is going to be even more public sector
involvement. In Cincinnati, when they passed a right-to-know law,
companies submitted 8000 chemicals and declared that all of them were
proprietary. The city appointed an advisory board, and now the claims
for proprietary information have plummeted from 8000 to 6.
We are going to have a lot of good dialogue between the companies
and the public sector in the next few years. We need to work out some
workable hazard assessment systems. I am encouraged by the number of
people from industry who agree. We need to get to work on the
dialogue; it is inevitable.
86
-------
MATHEMATICAL MODELS FOR ATMOSPHERIC DISPERSION
OF HAZARDOUS CHEMICAL GAS RELEASES: AN OVERVIEW
Jerry A. Havens
Department of Chemical Engineering
University of Arkansas
Fayetteville, AR
We need to consider our ability to make reasonable dispersion
predictions for use in emergency response. There are two main issues
in hazardous chemical (gas) consequence analysis that require the use
of atmospheric dispersion models. One is associated with fire and
explosions, the other with toxicity. The flammability hazard
associated with accidental release of a gas usually disappears when gas
has been diluted (with air) to about 1 percent. The toxicity hazard
may extend to concentrations four or five orders of magnitude lower.
We have spent much more time working on the development of dispersion
models for flammable gas concentration problems than we have for
toxicity problems.
Analysis of a large number of accidents leads us to consider the
important phenomenology in the dispersion process. In releases of
large quantities of denser-than-air gases, there will be a part of the
process in which the flows are gravity-dominated, i.e., flows driven by
the density of the gas rather than the atmospheric wind field. This
process is particularly important with flammable gases because
considerable dilution sometimes occurs. If we are interested in
predicting gas dilution down to 1 percent (the flammable gas issue),
these effects are important. However, these processes are much less
important if we are trying to predict distances to toxic gas
87
-------
concentration levels. A gas that is heavier than air is stably
stratified, and vertical mixing is inhibited. The cloud hugs the
ground, spreads laterally, and moves some distance downwind. We are
trying to predict the distance traveled before dilution occurs to a
given concentration. In the late stages of the event, the dispersion
is driven entirely by atmospheric turbulence. We need a model that
considers these different phases of dispersion. The model should be
able to make a gradual transition starting from the gravitational phase
through subsequent phases without the operator having to make the
switches.
We have tried to identify the main questions requiring study in
dense gas dispersion applications. Our approach has been to isolate
different parts of the problem for study. One controversial problem we
have addressed occurs with a nearly instantaneous release of a large
volume of heavy gas. We know that the gas slumps to the ground and
flows out laterally, but no one had (previously) agreed on whether the
gas is rapidly diluted during the process. We ran many experiments in
the lab involving instantaneous releases of heavy gas volumes,
measuring the lateral gravity spreading and dilution. We looked at the
scale of the release, the effect of the initial density, and the effect
of the initial height-to-diameter ratio of the release. Under these
conditions, a heavy gas cloud behaves in an unusual way. Upon release,
it rolls up in what has been described as a doughnut or torus
/
formation. Based on the laboratory experiments, we are able to
estimate the gravity spreading and the dilution that occurs. Thus, we
-------
can now locate the cloud front and estimate peak concentrations that
will occur as a function of distance from release.
In the last 10 years, a number of large field-scale experiments on
heavy gas dispersion have been completed. At China Lake, CA, the U.S.
Department of Energy (DOE) released around 30 cubic meters of liquefied
natural gas (LNG) onto water, forming a cloud (visible due to condensed
moisture) that moved off downwind. Shell Research, Ltd., conducted
similar experiments at Maplin Sands in England. At Maplin Sands, the
material was spilled onto water and the dispersion occurred over water;
whereas at China Lake, the material was spilled on water and the
dispersion occurred over land. The Thorney Island experiments
conducted by the British Health and Safety Executive involved the
instantaneous release of nominal 2000 cubic-meter gas volumes of
heavier-than-air Freon™-air mixtures. These experiments resemble our
laboratory experiments; a container is filled with gas and the
container (nearly instantaneously) removed. At Thorney Island, a large
tent-like structure was filled with about 2000 cubic meters of test
gas, and the side walls of the tent were dropped. The field
experiments showed responses similar to the lab-scale ones; the dense
gas fell down and rolled up in a doughnut shape. With results of these
field experiments corresponding so closely to the laboratory ones, we
gained confidence that we can scale our laboratory results to describe
field-scale events. We can also take mathematical models that are
applicable to a laboratory scale and apply them to a field scale,
further demonstrating the scaling relationships derived in the
laboratory.
89
-------
In the last few years, several accidents involving catastrophic
releases of hazardous gases have occurred. Considering the liquefied
petroleum gas (LPG) terminal disaster in Mexico City in 1984, do we
have mathematical models that would give us a reasonable picture of
what occurred? I think the answer is yes. Considering the methyl
isocyanate release at Bhopal, would application of available models
draw a picture consistent with the experience? I think the answer is
yes. However, accidents that are difficult to model (even after the
v
fact) have occurred, and consideration of these difficulties leads to
identification of areas where additional research is needed.
An accidental release of refrigerated ammonia, occurred during a
river barge-to-storage tank transfer in Blair, NE, in 1970.
Approximately 150 tons of ammonia overflowed onto the tank roof, ran
down the tank wall, and evaporated to form a large cloud. Reports of
the incident indicated the visible cloud extended 9000 feet from the
tank. Because of the prevailing winds, the cloud missed the town of
Blair. This is an example of an accident that we do not know how to
characterize very well. The release of refrigerated ammonia exhibited
all of the characteristics of a heavy gas cloud (the cloud was very
wide and shallow and hugged the ground), and it occurred in near-calm
conditions.
A high-pressure ammonia pipeline was ruptured by a heavy equipment
operator near Enid, OK, several years ago, and a wide plume path,
characteristic of a heavy gas cloud was observed. These accidents
raise basic questions for which we do not have adequate answers.
Ammonia is a material that, because of its molecular weight, would not
90
-------
be expected to form a heavier-than-air gas cloud. Even at the boiling
point, saturated ammonia is less dense than air. There was some
physical mechanism at work in these two accidents that caused the
ammonia clouds formed to be heavier than air. It was probably
associated with aerosol formation and evaporation, which cooled the
mixture and increased its density. In any case, we have a situation
where a heavier-than-air cloud is formed under some release conditions
and not under others.
There are two main types of dispersion models: mathematical
models and physical (wind tunnel) models. Mathematical models can be
subdivided into those using ordinary or partial differential equations.
Similarity models (involving ordinary differential equations) are the
mathematical models that, in my view, are most appropriate for use in
emergency response applications. We assume a regular shape for the
cloud, talk about some integrated values that assign the dimensions of
that cloud, and calculate how those things develop during the
dispersion process. That is, in fact, what Gaussian models do. Three-
dimensional mathematical models (involving partial differential
equations) can, in principle, simulate the spatial and temporal
dispersion process without artificial separation of the flow into
separate regimes and may be able to provide for effects of terrain and
wake turbulence. Evaluation of three-dimensional models is under way,
but most of the work is biased strongly toward evaluation against data
for dispersion in the absence of terrain and wake turbulence effects,
even though description of these effects is a primary motivation for
their use.
91
-------
DEGADIS (a similarity model we developed for the Coast Guard)
distills what we consider the best features of several heavy gas
dispersion models. It describes the dispersion of denser-than-air
gases on flat terrain and accounts for steady or transient releases.
The DEGADIS model, which was based on laboratory-scale experiments and
theoretical considerations, has been applied to data from all of the
major field-scale tests: the DOE series, the LNG series, the Shell
Maplin Sands series, and the Thorney Island series. With these
results, a statistical analysis has shown a confidence level with which
we can predict the results of these field tests. For all of these
field tests, the ratios of observed-to-predicted maximum distance to
the 1.0 percent concentration level ranged from 0.88 to 1.33 for a 99
percent confidence interval. In my judgment, this degree of accuracy
is good enough for most of the applications we will find in emergency
response.
\
DEGADIS and similar simple models do have shortcomings however.
They cannot adequately provide for situations involving obstacles in
the flow field, chemical reaction effects, or phase change effects
(such as formation of aerosols, evaporation, and rainout). There have
been some field tests conducted to address some of these questions. In
the Thorney Island tests, some experiments were conducted with (gas)
barriers erected around the release site. This particular test was
aimed at addressing the situation where a line of trees forms a barrier
at the boundary of a hazardous material installation. Simple models
also are inadequate in cases where a reactive gas is released. In such
cases, one may need to consider the importance of reaction effects. An
92
-------
example is the series of field test releases of nitrogen tetroxide
conducted by the Air Force.
For modeling large gas releases on flat terrain, when no important
chemical reaction or phase-change effects are involved, we have a
number of techniques that are adequate for risk analysis and emergency
response application. Future research should be directed to the
verification of models that provide for the effects of terrain and wake
turbulence and for description of jetting releases, aerosol formation,
chemical reaction, and deposition.
Note: This is a summary of the verbatim transcript of Dr. Havens'
remarks at the workshop. A complete copy of a paper of the same name,
which was presented at the AIChE CCPS International Symposium on
Preventing Major Chemical Accidents, February 3-5, 1987, in Washington,
DC, appears as Appendix D.
93
-------
DENSE GAS DISPERSION MODELS
Donald L. Ermak
Lawrence Livermore National Laboratory
Livermore, CA
There have been numerous efforts to model the dispersion of heavy
gas releases and most of these have been reviewed in at least one model
comparison study or another. Models have been proposed that vary
considerably in physical completeness and numerical complexity, and
generally, these two model characteristics increase (or decrease)
together.
The models that provide the most physically complete description
of dense gas dispersion are those that are based on the three-
dimensional, time-dependent conservation equations. Examples of this
type of model include FEM3, SIGNET, MARIAH, and ZEPHYR. At the
intermediate level of completeness and complexity are the similarity-
type models. These models use simplified forms of the conservation
equations that are obtained by averaging the cloud properties over the
crosswind plane. Quasi-three-dimensional solutions are obtained by
using similarity profiles, that is, by assuming a crosswind profile for
the concentration and other cloud properties. Examples of this type of
model include SLAB, HEGADAS, and DEGADIS. At the simplest level are
the modified Gaussian plume models. These models are usually used to
simulate continuous releases and employ a variety of modifications to
include the effects of dense gas dispersion within the Gaussian plume
model for trace gas releases. Top-hat models, which are used to
simulate instantaneous releases, fall into either the intermediate or
95
-------
simple category depending upon the complexity of the model regarding
the number of conservation equations to be solved.
In this presentation, a brief description of the representative
model types is given with emphasis on the FEM3 and SLAB models being
developed at the Lawrence Livermore National Laboratory (LLNL). The
details of the various heavy gas dispersion models will not be given
here except to point out the main differences between them.
The LLNL has been conducting research for a variety of sponsors in
the field of dense gas dispersion in the atmosphere. This research has
centered about a number of field-scale test series that were performed
in order to gather data to develop and validate dense gas dispersion
models. LLNL performed liquefied natural gas (LNG) field tests for the
U.S. Department of Energy in 1978 and 1980, and again in 1981 with
additional sponsorship by the Gas Research Institute. In 1983, LLNL
performed ammonia dispersion tests for the U.S. Coast Guard and The
Fertilizer Institute and nitrogen tetroxide spill experiments for the
U.S. Air Force. During this same time period, LLNL developed two
state-of-the-art atmospheric dispersion models called FEM3 (Chan, 1983;
Gresho et al., 1984) and SLAB (Morgan et al., 1983; Ermak and Chan,
1985).
Both of these computer models incorporate mathematical
descriptions of the physics of heavy gas dispersion. These include:
1) gravity spread, 2) effect of density stratification on turbulent
mixing, and 3) ground heating into the cloud and its effects on density
stratification and turbulence. Of the two codes, FEM3 provides the
96
-------
more detailed and complete description of the physics Involved In dense
gas flows.
FEM3 simulates the dispersion of a released gas by solving the
time-dependent, three-dimensional, conservation equations of mass,
momentum, energy, and species along with the Ideal gas law for the
equation of state. In addition, it can treat flow over variable
terrain and around obstructions such as cylinders and cubes.
Turbulence is treated by using a K-theory submodel. Since it is fully
three-dimensional, FEM3 can simulate complicated cloud structures such
as: 1) the vortices that are typical of dense gas flows, 2) cloud
bifurcation that has been observed during heavy gas releases under
stable, low wind speed conditions, and 3) cloud deflection caused by
sloping terrain. \
The SLAB model solves the crosswind-averaged equations for the
conservation of mass, species, momentum, and energy along with
additional equations for cloud width and the ideal gas law equation of
state. The current version of SLAB also includes the steady-state
assumption for continuous releases. Thus, the code is one-dimensional
with downwind distance being the independent variable. However, since
cloud width and cloud height are also calculated, the model is, in this
sense, quasi-three-dimensional. The crosswind concentration
distribution is determined by using similarity profiles based on the
calculated crosswind height and width. Mixing of the cloud with the
ambient atmosphere is treated by using the entrapment concept. The
main advantage of the SLAB code is its low computing cost. Typical
97
-------
simulations require only a few seconds on a CDC 7600 computer or a few
minutes on an IBM microcomputer.
The various types of models differ considerably in their approach
to simulating the atmospheric dispersion of a dense gas release.
Perhaps the most obvious differences are related to the degree to which
each type of model incorporates the basic conservation laws and three-
dimensional effects. The modified Gaussian plume model is based on the
single conservation of species equation and either neglects momentum
and energy transfer or attempts to include them in some ad hoc manner.
On the other hand, the SLAB model includes the conservation equations
of mass, momentum, and energy* in addition to the species equation, but
only in an average way. Variations in the crosswind plane are
neglected, and all properties of the vapor cloud are expressed as
crosswind averages that vary in the downwind direction only. The
conservation equation model FEM3 includes the most complete description
of the conservation laws by treating them explicitly in three
dimensions.
A unique feature of the SLAB model is that it calculates only
crosswind-averaged properties and characterizes the cloud shape by the
height, h, and half-width, B. The parameters B and h do not correspond
to any particular concentration level; rather, they can be considered
to describe a surface that encloses the bulk of the cloud, for example
90 percent. Consequently, the crosswind concentration distribution is
not uniquely defined, making it difficult to compare the predicted
cloud shape from this model with the contour plots obtained from the
experiments. To overcome this difficulty, one generally assumes a
98
-------
distribution (such as Gaussian, exponential, or quadratic) for the
crosswind horizontal and vertical vapor cloud concentration.
There are other important differences related to the manner in
which each model treats the effects of gravity and turbulence. As
indicated above, the modified Gaussian plume models use ad hoc formulas
with "empirical" coefficients to describe the gravity spread and
turbulent dispersion of the cloud. In contrast to this, the SLAB and
FEM3 models use conservation principles to treat the effects of
gravity. This is done in the FEM3 model by solving the three momentum
conservation equations, including the buoyancy term and variable
density, while the SLAB model solves two layer-averaged momentum
equations and uses the hydrostatic approximation.
These two models differ considerably in their approach to
turbulence. The SLAB model uses the somewhat artificial concept of
entrainment across the cloud-air interface and essentially neglects any
explicit treatment of turbulence within the vapor cloud. Air is
entrained into the cloud at the surface and then is assumed to mix
rapidly in the cloud, creating a nearly uniform layer in the crosswind
plane. Thus, there are two separate regions: the cloud and the
ambient atmosphere. Mixing between the two is assumed to occur at the
interface and is governed by an entrainment velocity that depends on
the local properties of both the cloud and the surrounding atmosphere.
The FEM3 model assumes that turbulence can be described as a diffusion
process 'and uses a continuous diffusion coefficient that depends on the
local properties of the dense gas cloud. While the entrainment and
diffusion concepts are peculiar to the SLAB and FEM3 models
99
-------
respectively, the choice of a particular entrainment or diffusion
submodel is not an essential aspect of the models. Several submodels
have been proposed in the literature and could be used without changing
the whole model.
Finally, a three-dimensional, time-dependent, conservation
equation model is required in order to describe the distribution of gas
concentration in space and time from a heavy gas release into an
atmospheric boundary layer with speed and directional wind shear in the
presence of complex terrain and man-made structures. The versatility
of the three-dimensional conservation equation model in treating more
realistic situations and providing a more detailed description of the
flow is somewhat balanced by the increased computer costs in running
these models. Conversely, similarity models, while giving up some
degree of realism and detail, are much faster to run on computers.
Consequently, at LLNL we tend to view the three-dimensional
conservation equation model more as a research tool that helps us
discover new things about the flow, and the intermediate, or similarity
model, more as an operational model for situations where computing
costs and time are of the essence.
Over the past few years, the predictions from both the FEM3 and
SLAB models have been compared with the data obtained from a variety of
field-scale experiments. These include the Burro and Coyote series of
LNG dispersion experiments and the nitrogen tetroxide spill tests
conducted by LLNL and the refrigerated liquid propane tests conducted
by Shell Research, Ltd., at Maplin Sands, England. The models have
been observed to perform quite satisfactorily within the limits of each
100
-------
model and have been very helpful in Identifying the important phenomena
associated with dense gas dispersion in the atmosphere.
Basically, three major effects can be observed in the dispersion
of dense gas clouds that are not observed in the dispersion of trace
emissions. The first effect is a reduction of turbulent mixing within
the vapor cloud due to stable stratification of the dense layer. The
second is the generation of gravity-spreading and self-induced vortices
due to density gradients in the horizontal direction. The third
effect, cloud lingering, occurs when the dense gas cloud travels
downwind at a slower rate than the ambient wind speed. It is due to a
lack of mixing between the dense gas layer and the ambient atmosphere.
All of these effects are most pronounced when the ambient wind speed is
low and the atmospheric stability conditions are stable.
Perhaps the most dramatic example of these effects was observed in
the Burro 8 LNG experiment. In this test, the dispersion cloud
developed a distinct bifurcated shape with two lobes on either side of
the cloud center!ine. This bifurcated structure was due to the
combined effects of gravity spreading of the dense gas cloud and a
reduction in turbulence allowing this structure to develop without
damping out the bifurcated shape. In addition, the vapor cloud
lingered over the source region for more than 2 minutes after the spill
ended. In all of the other experiments conducted under higher wind
speeds and less stable conditions, the cloud drifted downwind from the
source within 10 to 15 seconds after the spill had ended. The FEM3
simulation of the Burro 8 experiment simulated both of these effects,
namely cloud bifurcation and cloud lingering, quite accurately.
101
-------
Similarity models have been able to reasonably simulate the average and
peak concentration versus downwind distance for the Burro 8 experiment;
however, prediction of the cloud structure is beyond the scope of these
models because cloud structure is a basic assumption of the models.
One final comment is important regarding the use of dense gas
models in operational settings. When an accident occurs, the actual
amount of material released and the release rate are often unknown. In
order to make dispersion predictions, a unit source simulation is
generally performed, and the resultant concentrations are then modified
when actual source strength values are obtained. For dense gas flows,
the unit source simulation technique is generally inappropriate because
of nonlinear effects that are occurring. One cannot simply double the
predicted concentration because the measured source strength was found
to be double that used in the original calculation. Consequently, this
complication of dense gas flows will require a different way of
planning for potential accidents.
REFERENCES
Chan, S. T. FEM3 -- A Finite Element Model for the Simulation of Heavy
Gas Dispersion and Incompressible Flow: User's Manual. UCRL-
53397, Lawrence Livermore National Laboratory, Livermore,
California, 1983. 83 pp.
Ermak, D. L., and S. T. Chan. A Study of Heavy Gas Effects on the
Atmospheric Dispersion of Dense Gases. UCRL-92494, Lawrence
Livermore National Laboratory, Livermore, California. Paper
presented at the 15th NATO/CCMS International Technical Meeting on
Air Pollution Modeling and Its Applications, St. Louis, Missouri.
April 15-19, 1985. 23 pp.
Gresho, P. M., S. T. Chan, C. Upson, and R. Lee. A Modified Finite
Element Method for Solving the Time-Dependent, Incompressible
102
-------
Navier-Stokes Equations: Part 1 - Theory; Part 2 - Applications.
Int. J. Num. Meth. Fluids, 4:557-598 and 619-640, 1984.
Morgan, D. L., Jr., L. K. Morris, and D. L. Ermak. SLAB: A Time-
Dependent Computer Model for the Dispersion of Heavy Gases
Released in the Atmosphere. UCRL-53383, Lawrence Livermore
National Laboratory, Livermore, California, 1983. 15 pp.
103
-------
ON THE DEVELOPMENT OF REACTIVE, DENSE GAS MODELS
Bruce B. Hicks and Will R. Pendergrass
National Oceanic and Atmospheric Administration
Atmospheric Turbulence and Diffusion Division
Oak Ridge, TN
A summary of this presentation was not prepared at the authors'
request. The abstract of a formal paper is given below. The text of
this paper can be found in Appendix C.
Chemical reactions between released pollutants and
atmospheric constituents can be affected or mechanical
turbulence damped. If the reaction is sufficiently
exothermic, plume rise can be increased; if sufficiently
endothermic, then plume rise can be suppressed. It is
necessary to consider the potential influence of such
reactions on the mixing rates characteristic of the ambient
air. Reaction rates are also a critical issue. If reactions
are completed rapidly, then in concept it is possible to
assimilate the consequences of the reactions in the source
term initialization or in the first time step of a numerical
model. On the other hand, if reaction rates are slow in
comparison with the time step of a simple model, then a more
complicated and detailed modeling effort may be required.
105
-------
DISPERSION MODELS FOR NEUTRALLY BUOYANT
AND POSITIVELY BUOYANT GASES
Thomas E. Pierce
U.S. National Oceanic and Atmospheric Administration
and U.S. Environmental Protection Agency
Research Triangle Park, NC
It is my good fortune to be involved with buoyant plume modeling
which seems to be much more manageable than reactive and dense gas
modeling. I am assigned as a National Oceanic and Atmospheric
Administration (NOAA) meteorologist to the Environmental Operations
Branch of EPA. Our branch helps to provide a technology transfer link
between the research side and the regulatory side of EPA. One of our
jobs is to develop, adapt, and evaluate atmospheric dispersion models.
These models tend to be operational and not research-grade models like
those presented by others at this workshop. We make these models
available through the User's Network for Applied Modeling of Air
Pollution, which is commonly known by its acronym, UNAMAP.
Since its beginning in 1973, hundreds of people have used UNAMAP
either through EPA's UNIVAC computer or through the National Technical
Information Service (NTIS), where computer tapes can be purchased. The
most recent version of UNAMAP, Version 6, was released in August, 1986,
with more than 23 models and programs.
Today, I would like to focus on three programs on UNAMAP that were
developed by the Environmental Operations Branch:
• TUPOS - Dispersion estimates using on-site turbulence data
• MPDA - Meteorological processor for dispersion analysis
• INPUFF - Integrated puff model.
107
-------
TUPOS represents one of the first operational models that
incorporated on-site turbulence data. It operates like a routine
Gaussian plume model in its steady-state assumptions. TUPOS,
therefore, deals only with continuously emitting sources.
In addition to handling on-site turbulence, TUPOS has several
other special features. It incorporates layer-by-layer plume rise and
partial penetration of a plume into an elevated temperature inversion.
It also incorporates recent ideas by Briggs on the modeling of .a
buoyant plume in a convective boundary layer.
The modeling of partial penetration is particularly important for
some sources. Current routinely used Gaussian plume models assume an
all-or-nothing approach to plume penetration. If the centerline of the
plume is predicted to remain below the elevated inversion, then 100
percent of the plume is estimated to remain in the mixed layer.
However, if the centerline is predicted to be above the mixed layer
(even by 1 meter), then the entire plume is assumed to loft above the
mixed layer. Clearly, partial penetration is an improvement that we
feel should be included in Gaussian plume models.
Another important development is the modeling of hesitant or
bumping plumes. Willis and Deardorf noted that when a buoyant plume,
which is in a convective boundary layer, encounters an elevated
inversion, that it can bump against the inversion and spread out
laterally. In the Journal of Climate and Applied Meteorology. Briggs
discussed a method for predicting diffusion for this situation. The
most recent version of TUPOS, Version 2, includes this methodology.
108
-------
It is hoped that with models like TUPOS we can do a better job of
modeling atmospheric pollutants, and certainly there is plenty of room
for improvement in this area. If we are to use more complicated
models, however, we should have a means for incorporating on-site
meteorological data and estimating similarity parameters.
Based on ideas of Irwin, we have developed a meteorological
processor for dispersion analysis (MPDA). MPDA allows input of three
types of meteorological data: upper-air data from National Weather
Service stations, surface observations from National Weather Service
stations, and on-site data. MPDA processes all of these data and
produces a merged or clean data set. The processor also determines
several parameters, such as the solar angle, friction velocities,
Monin-Obukhov length, turbulence profiles, and vertical variation of
wind speed and wind direction. A great deal of flexibility has been
programmed into this model. For many of the parameters, several
options are available. For example, there are three options for
specifying the sensible heat flux. They are the Irwin-Binkowski
scheme, the Holtzlag-van Ulden scheme, and direct determination from
on-site turbulence measurements.
INPUFF, developed by Peterson, resulted from EPA's need to
evaluate dispersion from incinerator ships. The problem with modeling
incinerator ships was that a steady-state Gaussian model could not be
used. Therefore, the integrated puff model was developed. In recent
years, demand for INPUFF has grown as people have sought programs to
model toxic gas releases.
109
-------
INPUFF simulates individual puffs from one or more sources.
Although many sources can be included, the model's speed decreases as
the number of time steps increase.
Briefly, INPUFF works in the following manner. At the first time
step, a puff is released and disperses according to the specified
meteorology for that time step. At the next step, another puff is
released so that it and the previous puff disperse according to the new
meteorological conditions. INPUFF proceeds in this step-wise manner
and is able to calculate contributions at specified receptors from each
puff for each time step. Hence its name, the integrated puff model.
Inputs to INPUFF include miscellaneous options, modeling region
specification, and source characteristics. Source characteristics
include emission rate, time of emission rate, and how the emissions
vary with time. Receptor points and meteorological data are included
as inputs. Dispersion algorithms included in the model are provisions
for initial dispersion, short travel-time dispersion, and long travel -
time dispersion.
Currently, the model is designed so that meteorological data are
entered for every time step for a user-designed meteorological grid.
Eventually we hope that a processor such as MATHEW can be front-loaded
into INPUFF to allow a three-dimensional wind field projection.
In reviewing these three programs that are available on UNAMAP,
hopefully, I have presented some ideas that may be useful to those
involved in the modeling of toxic gas releases.
110
-------
REFERENCE
Briggs, Gary A. Analytical Parameterization of Diffusion: The
Convective Boundary Layer. J. Climate Appl. Meteorol. 24:1167-
1186, 1985.
Ill
-------
LAB-SCALE EXPERIMENTS
Robert N. Meroney
Fluid Mechanics and Wind Engineering
Civil Engineering Department
Colorado State University
Fort Collins, CO
As a physical modeler in a room full of numerical modelers, I must
confess I feel like a chicken in a fox house. So, I will do the best I
can, and you must decide whether or not you are convinced. I want to
talk about the capabilities of physical fluid modeling with respect to
meeting the needs of the hazardous materials community. There are many
experiments that can be classified as basic fluid mechanics experiments
associated with the mixing process. There are tests that have been
performed for the meteorological community associated with pre-field
test planning experiments. There are concept testing experiments
where, for example, a certain mitigation device can be tested for
feasibility. There are validation experiments that can be performed
specifically to determine whether physical modeling can provide a
viable approach to solving some particular problem. Finally, there are
experiments that have been directly applied to hazard analysis. The
goals of all of these experiments is to test, calibrate, and validate a
numerical code.
One might note that if a code is not capable of predicting the
behavior of an idealized laboratory-controlled experiment, one should
not feel the right or the ability exists to predict a far more chaotic
field experiment. As a proponent, I will say a couple of words about
the advantages of fluid modeling. Wind/water facilities are in effect
113
-------
analog computers, and they have special attributes. They have near-
infinitesimal resolution. Grid sizes are not a concern. The transport
processes, one could argue, go down to the molecular level. The
facilities have near infinite memory. They also have the ability to
look at very large three-dimensional grid regions. Fluid modeling
incorporates real fluids, not models of fluids. We, therefore, start
out using the right stuff in the right place, not someone's concept of
how the atmosphere or fluid behaves. Implicitly, this analog computer
is nonhydrostatic, non-Bosinnesque, capable of compressible effects,
thermal effects, and includes variable property. It includes a non-
slip boundary condition, effects of dissipation, and many nonlinear
processes. Also, it inherently includes full conservation equations
without truncation.
On the down side, there are some limitations. At smaller scales,
one must recognize that some similarity is lost in the mixing
processes. As speeds drop to handle stratified fluids, the Reynolds
numbers decrease, and it is possible that the Reynolds number can drop
below some critical value in different phases of the mixing process.
Depending on the phenomena, this may result in a minor or a major
error. When running experiments at very low speeds, one reaches a
point, perhaps associated with the fluid number dominance, where the
ratio described as the Peclet number over the Richardson number first
proposed by Colenbrander and Puttock from Shell Research, Ltd., is less
than some critical value. This means, basically, that you are
operating at such a low rate of mixing that the molecular mixing
exceeds the scaled turbulent mixing. This is the only phenomenon that
114
-------
tends to give a nonconservative result in physical modeling. Almost
all other errors tend to give conservative results.
As smaller scales are used in the laboratory, the separation we
are familiar with in the atmosphere between the integral scales, Taylor
scales, and Kolmogorof scales of turbulence, will bunch up. Depending
on the kind of mixing process under study, this may or may not be
important.
There are a number of basic fluid mechanics experiments one might
wish to consider. Anyone who has produced a numerical model that uses
an entrainment rate at a box-model or a slab-model level or who has
worked with various K-theory-type models has probably drawn from
physical modeling for basic turbulence coefficients. Some of the
earliest work by Lofquist, associated with overflow of fresh water over
saltwater, generated the information we use today on the entrainment
variation of Richardson numbers. Basic experiments by Kantha, where a
surface plate was dragged around and around on top of a circular
channel filled with stratified salt water, the merry-go-round
experiments, have provided us with additional information. Some errors
were later found in these experiments, and meteorologists did
experiments to improve this work (Willis and Deardorff and, later,
Lindberg). In England, McQuaid has done some basic experiments on
transport through dense shear layers associated with carbon dioxide
(C02) releases. Jerry Havens and I also have done some basic
experiments.
We should consider some simple idealized cases. Yang, a student
of mine in 1972, and I did some laboratory puff model experiments,
115
-------
where we released thousands of puffs of gas in a boundary layer shear
flow and were able to develop models for a transport similarity that
included probability density distribution and various coefficients
associated with lateral, vertical, and longitudinal variation. These
models and the coefficients very closely agree with models we are using
today. In 1980-81, Lohmeyer and I did some experiments with
instantaneous volumes of dense gases, releasing them at the wall of the
boundary layer. At that time, there were no field experiments to guide
us, but we identified the basic characteristics of the dense gas cloud.
The behavior of the arrival time, the departure time, and the
statistical deviations within a multiple sum average condition of a
torus-shaped cloud, I believe, are very important. These are critical
points for issues of flammability and toxicity. We cannot make
decisions based on average conditions, but must know about the
statistical range of conditions that exist. A gas cloud is not set on
fire by the average conditions that exist. It is the instantaneous
concentration that sets the fire.
There have been several somewhat complex studies done in the
laboratory, including both dense gas cloud effects and heat transfer
effects. The best known of these is the tank experiments on the
convective boundary layer concept by Willis and Deardorff. These were
heated water experiments that have had a revolutionary effect on both
numerical modeling and our field understanding of the convective
boundary layer. One of the key points from these experiments is that
we now know that clouds released through plumes at the ground tend, in
a short distance in a convective boundary layer, to rise. Also, clouds
116
-------
released at elevated points in the convective boundary layer can fall.
Maybe on average this point is not so important, but it can make a very
big difference in how you should be calculating the results of a toxic
or flammable cloud passage.
What kind of pre-field planning experiments are possible? We have
performed pre-field tests studying the effects of field terrain,
windward variation, and stability. We have looked at instrument
placement and cloud extent.
Post-field test experiments also can be useful. Recently in some
DOE and Gas Research Institute-funded work preceding the 1987 LGF vapor
barrier tests, we found some unexpected things occurred. In the vapor
barrier-contained region, we found that the gases sloshed up against
the end, hit the barriers like a wave, were caught in the air flow
passing over the barriers, and were transported downwind at fair
heights. Thus, we found bursts of higher concentrations at higher
elevations than were observed in the numerical experiments.
Physical fluid modeling also can play a very useful role in
testing of concepts. When the gas industry was looking at alternative
secondary containment schemes for large tank storage, there were
questions about berm heights for large volumes of cryogenic chemicals.
There were questions also about whether soil surfaces or insulated
concrete surfaces were better inside the berms and about the
comparative effects of these choices on the eventual dispersion of
resulting gases. With physical fluid modeling, we were able to test
many options and learn not only the answer to those questions, but at
117
-------
that time due to the early nature of those experiments (mid 1970s), we
were able to learn new fluid flow fundamentals as well.
In some work with Factory Mutual, funded by the Gas Research
Institute (GRI), we looked at water spray curtains and how they
mitigate cloud dispersion. In this case, simultaneous field tests were
run that were very confusing due to the unexpected results that
occurred. Our work on nozzle sizes, varying water pressure, and wind
speeds helped to explain the nature and extent of mitigation that did
occur in these experiments.
As mentioned earlier, idealized experiments have been used
extensively to calibrate modules of various models. Some numerical
models have been calibrated against both laboratory and field results.
Physical fluid modeling can serve as a useful method for evaluating
Federal regulatory-specified accident scenarios at existing or planned
facilities. For example, we were able to model such a scenario, a
guillotine of a pipeline with 10 minutes of spill at the maximum flow
rate, for the Brooklyn Union gas storage facility on Long Island. The
experiment showed a positive result, that no effects would be seen
beyond the facility boundary in the event of that accident. Since
there were resources remaining in the contract, they had us look at
larger, even less likely potential accidents, such as spilling the
entire tank. This time the results were potentially far more
catastrophic, even though unlikely.
Wind tunnel measurements can be very helpful during risk
assessment, licensing, or the regulatory process. DOT regulations
currently require the use of an extremely over-conservative numerical
118
-------
algorithm. The model does not account for roughness, obstacles,
terrain, or mitigation devices. To prepare for licensing hearings for
a liquefied natural gas (LNG) peak-shaving facility on Staten Island,
we examined spills at about a 1:250 scale model. We introduced
conservatism into the experiment by simulating larger, heavier spills
in an environment with reduced mixing. The facility appeared to meet
DOT requirements even when significantly more extreme conditions were
considered than required by DOT regulations.
Recently, field/laboratory validation experiments have been
completed for both instantaneous and continuous releases of dense and
cryogenic gases. In work for the GRI (Meroney, 1986), I examined some
26 field/laboratory data sets and found that the laboratory-predicted
distances to lower flammability limit (LFL) on the average to within
0.4 percent of actual values with a standard deviation of ±22 percent.
Pattern comparison plots of concentration isopleths could always be
matched by appropriate fluid modeling techniques with less than a 15°
shift in surface patterns. The British Maritime Technology group in
the United Kingdom scaled the recent Thorney Island field spills of
TM
Freon -air mixtures. They found no apparent lower limit for Reynolds
number or Peclet/Richardson number-scaling criteria for collapsing
dense clouds.
It is now apparent that fluid modeling can faithfully reproduce
the physics of transient dense gas cloud entrainment and motions within
the inherent variability of the process for many interesting
situations. Fluid modeling can contribute valuable input information
119
-------
for future siting and risk analysis models for the chemical and
petroleum industries.
REFERENCE*
Meroney, Robert N. Guideline for Fluid Modeling of Liquefied Natural
Gas Clouds, Volume II: Technical Support Doc^ent. Colorado
State University, Civil Engineering Report CER84-85RNM-50b,GRI
86/0102.2. Gas Research Institute, Chicago, Illinois, 1986.
266 pp.
*Author's/Editor's Note: All of the work by other researchers
mentioned in this presentation are referenced in the above citation.
120
-------
LARGE-SCALE EXPERIMENTS OF THE DOE LIQUEFIED FUELS PROGRAM
Ronald P. Koopman
Lawrence Livermore National Laboratory
Livermore, CA
The accidental release of hazardous gases into the environment
often results in what has come to be known as dense gas behavior, even
when the gases themselves may be nominally less dense than the
atmosphere into which they are released. This dense gas behavior can
dominate the consequences of the accidental release making it either
worse or better than might otherwise be expected and making prediction
of the consequences of the release difficult. There are a number of
contributing causes of dense gas behavior, including large-scale
release, low temperature, flashing, chemical reactions, evaporative
cooling, and molecular weight. Cloud behavior differences are due to
density or gravity-induced effects (turbulence damping); thermodynamic
effects (aerosol formation and flash vaporization upon release,
evaporative cooling, or heat transfer from the ground modifying cloud
buoyancy and turbulence); and chemical reaction effects (hydrolysis
with atmospheric water vapor, polymerization, or decomposition, which
also affects cloud density). These effects are very important close to
the release point, but also can have dramatic effects on the cloud as
it disperses downwind.
The study of atmospheric dispersion of dense gases began as a new
scientific field in the 1970s. Interest was prompted by a series of
accidents: Flixborough in 1974, Seveso in 1976, and the Houston
ammonia accident also in 1976. The interest in this field has always
121
-------
been more intense in Europe than in the United States. As a result,
much of the research has been done in Europe. Early work in the United
States centered around liquefied natural gas (LNG) safety, and this was
when Lawrence Livermore National Laboratory (LLNL) got involved,
starting in 1976. Many people at this conference got involved at about
this same time with the LNG safety program.
Work on dense gas dispersion continues to involve three
complementary areas:
• Mathematical modeling based on physical laws
• Near full-scale field experiments to:
1) Discover unknown and important effects in scaling laws
2) Simulate accidents or evaluate mitigation equipment
3) Validate models
• Scaled-down simulations in the laboratory, wind tunnel, or
water flume.
We have discussed all of these at this workshop except large-scale
experiments on which I will focus my remarks.
There have been a number of recent large-scale field tests with
dense gases. I will focus on the most recent and the best
instrumented. These are the tests that are currently being used in
developing databases for model comparisons and model validation. Most
of the good quality data have been obtained since 1980 when,
essentially simultaneously, Shell/NMI conducted a series of LNG and
liquefied petroleum gas (LPG) trials at Maplin Sands in England, and
DOE/LLNL conducted the Burro series of LNG tests at China Lake, CA.
These tests resulted in the discovery of previously unknown and
important effects. The Maplin Sands tests involved spills of about 5
to 25 cubic meters of LNG with spill rates between instantaneous and 1
122
-------
to 5 cubic meters per minute. DOE's Burro series, conducted at about
the same time, focused on dense gas dispersion and involved spills of
24 to 39 cubic meters at rates of 12 to 18 cubic meters per minute. A
follow-up series of DOE tests conducted in 1981 was called Coyote.
These tests were designed to get information on three different areas:
combustion, dispersion, and rapid-phase transition explosions.
There are a number of reasons for doing field-scale experiments.
Many of the effects that occur upon accidental release of dense gases
are poorly understood. In some cases, the effects have been observed
in the laboratory, but the extent of what will occur or what will be
important or dominant in accidental releases is unknown. The important
physical effects in dense gas dispersion include:
• Gravity spreading Low cloud
Displaced ambient wind field
Self-induced vortices
Increased surface area
Increased wind shear
Increased heat transfer
a Turbulence damping Stable stratification
Impenetrable interface
Negative turbulent mass flux
Evaporating droplets
t Heat transfer Increased ground contact
Decreased air entrainment
Enhanced evaporation and condensation
for cryogenics spilled on water
Chemical reactions with water vapor or
other species
• Turbulent entrainment Turbulence generated by the release
(internal)
Atmospheric turbulence excluded
Heat transfer generates convective cells
Jet releases entrain air
Instabilities
123
-------
Additionally, much of the physics dominant in dense gas dispersion
is highly nonlinear and, therefore, not easily scaled to larger size.
Very stable atmospheric situations and nonisothermal (i.e., heat
transfer, condensation, etc.) effects can often be important and are
also not easily scaled. In many cases, dominant behavior at large
scale was not even observed in small-scale tests. For instance, dense
gas behavior is not observed for small releases in the atmosphere;
turbulent dispersion dominates. Thus, two of the major reasons for
performing these tests are the discovery of unknown and important
effects and determination of their scaling laws. These well-
instrumented tests allow quantitative measurement of these effects so
that correct descriptions can be incorporated into the mathematical
dispersion models. This is necessary if accurate predictions are to be
made for circumstances different than those under which tests were
performed. As you can readily see, dense gas dispersion is very
complicated.
Other reasons for conducting field tests include accident
simulation or evaluation of mitigation equipment such as water or steam
curtains. These can be situations that are simply too complicated,
with too many unknown contributions to yield to mathematical or
physical model simulation. Other complicating effects that are
difficult to model or scale include chemical reactions and certain non-
reversible thermodynamic effects, such as flashing, two-phase flow.
Perhaps the most common reason to conduct large field experiments
has been to obtain basic data for dispersion model validation. This
requires extensive, carefully verified, quantitative data from well-
124
-------
documented and well-instrumented experiments. The need for such data
was demonstrated by McQuaid of the British Health and Safety Executive,
in 1983 when he invited the mathematical modeling community to predict
in advance the results of an instantaneous release of 200 cubic meters
of gas with an initial density two times that of air. Variations
between models of two orders of magnitude were present even for this
relatively small, isothermal release onto flat terrain without chemical
reactions, thermodynamic effects, or other complications. Models have
improved since that time, but the tasks we are giving them are much
more difficult and include many more complicated effects. Thus, model
validation continues to be one of the major reasons to conduct field
experiments. In spite of the number of tests that have been conducted,
data for model validation are still in short supply.
There are many difficulties associated with the conduct of large-
scale field tests. They involve huge instrument arrays with hundreds
of individual instruments and their attendant maintenance and
calibration. These extensive arrays are necessary because gas cloud
behavior is highly variable in space and time and complicated by the
presence of the many simultaneous and interactive effects mentioned
earlier.
The large-scale, well-instrumented spill tests that have been
conducted since the 1970s have contributed much to our understanding of
dense gas dispersion. Dense gas characteristics are particularly
important for large releases, low wind speed releases, stable
atmospheric conditions, low-level inversions, and proximity to the
release point. Proximity to the release point is very important for
125
-------
dense gas releases. In our ammonia release tests, we were still
measuring dense gas effects at a distance of 3 kilometers. Dense gas
behavior is caused by large size, low temperature, flashing two-phase
flow, chemical reactions, evaporative cooling, and molecular weight.
Many of these effects are nonlinear, and, therefore, it is important to
base calculations and predictions on size.
Not all of the things that people do can be scaled, depending on
which dominant effects are of interest. Dense gas releases exclude
ambient atmospheric processes. Gravity spreading and turbulence
damping decrease dispersion, while heat transfer and turbulent
entrainment can increase it. Dense gas effects do have profound
influence on our ability to predict the consequences of accidental
releases.
In summary, there are a number of reasons for doing field-scale
experiments. Many of the effects that occur upon accidental release
are poorly understood. In some cases, they have been observed in the
laboratory, but the extent of what will occur or what will be important
or dominant is unknown,. In addition, much of the physics dominant in
dense gas dispersion is highly nonlinear and, therefore, not easily
scaled to larger size. Thus, two of the major reasons for performing
these tests are the discovery of unknown and important effects and
determination of their scaling laws. Well-instrumented tests allow
quantitative measurement of these effects so that correct descriptions
can be incorporated into the mathematical dispersion models. Models
are necessary if accurate predictions are to be made for circumstances
different than those under which tests were performed. Other reasons
126
-------
for conducting field tests include accident simulation or evaluation of
mitigation equipment such as water or steam curtains. These can be
situations that are simply too complicated, with too may unknown
contributions, to yield to mathematical or physical model simulation.
Other complicating effects that are difficult to model or scale include
chemical reactions and certain nonreversible thermodynamic effects such
as flashing two-phased flow. Perhaps the most common reason to conduct
field experiments has been to obtain basic data for dispersion model
validation. Today's sophisticated models require extensive, carefully
verified, quantitative data from well-documented and well-instrumented
experiments.
127
-------
FLUID MODELING OF DENSE GAS DISPERSION OVER A RAMP
William H. Snyder
U.S. Environmental Protection Agency
and National Oceanic and Atmospheric Administration
Research Triangle Park, NC
Recently, we completed our first foray into dense gas dispersion
at the U.S. EPA Fluid Modeling Facility at Research Triangle Park, NC.
Rex Britter from the University of Cambridge drew up the plans for the
first research study. Unfortunately, he could not be here to present
the work today.
We have drawn upon our previous experiences in modeling and
simulating dispersion in the neutral atmospheric boundary layer. This
work has focused mostly on ground-level and elevated point sources. We
have also done numerous studies in the wind tunnel on the flow
structure and dispersion around isolated hills of simple geometry and
studies in the towing tank of flow structure over and dispersion of
plumes within stratified flow over isolated hills. The new addition to
these programs is our subject today, dense gas.
One of the biggest problems we have encountered was alluded to
today by Bob Meroney. That is the fact that when you simulate a plume
that has a density different from that of its environment, you need to
do Froude number scaling, which means you need to reduce the w.ind
speed. By reducing both the scale and the wind speed, you have reduced
the Reynolds number, which means that what would have been a turbulent
flow in the wind tunnel may now become a laminar flow, irrespective of
the density difference. In addition, if you are modeling a positively
129
-------
buoyant plume, i.e., a destabilizing influence, you may be able to
model that at a smaller-than-normal Reynolds number. However, in the
case of a dense gas plume, the dense gas has a stabilizing influence.
This stabilizing influence of the density gradient plus the reduced
wind speed due to the Reynolds number makes the simulation of dense gas
dispersion in the wind tunnel doubly difficult.
Our first exercise in the wind tunnel was thus a flow-
visualization study. We used carbon dioxide (C02) as the effluent in a
wind tunnel boundary layer, which had previously been used for point
source studies. Me fixed the mass flux of C02 at the rate planned for
later quantitative studies. Smoke was injected in order to be able to
see the plume. We visualized the plume in the tunnel at various wind
speeds, watched the plume to see that it remained turbulent, and
progressively reduced wind speeds until the flow became laminar. We
found that the plume remained fully turbulent to wind speeds as low as
1.0 meters per second. Thus, we selected this wind speed for the
remaining quantitative tests. We have prepared a videotape of these
flow visualization tests.
Our source configuration consisted of a metal can filled with
gravel up to the level of the tunnel floor to maintain a homogeneous
flat surface across the source. The CO* was emitted through the
gravel, and was thus emitted in a circular area through the flat
surface. We placed a honeycomb at the entrance of the test section of
the wind tunnel to stabilize the flow. A straight fence served as a
trip to generate a simulated neutral atmospheric boundary layer. From
previous experiments, we knew that this was a reasonably well-
130
-------
developed boundary layer* The source was placed about 8 meters
downstream of the fence.
For an initial series of tests, we emitted C02 and/or air from the
source and then measured the detailed structure of the plume downstream
in longitudinal, vertical, and crosswind or lateral concentration
profiles. Later a ramp was included in one of two positions either
1800 millimeters downstream from the source or 600 millimeters
downstream from the source.
Looking at the boundary layer structure of these experiments, we
found that for mean velocity profiles between 1 and 4 meters per
second, the flow is virtually independent of Reynolds number.
Longitudinal turbulence intensity profiles also show that structure to
be virtually Reynolds-number independent.
For the quantitative portion of the study, about 3 percent ethane
was added to the source, so that concentrations could be measured
downstream. These measurements, again, show that at 1 and 4 meters per
second the concentration and dispersion field in the wind tunnel were
independent of Reynolds number. Much additional data on the flow
structure are available due to this and previous work. There are hot-
wire anemometer measurements of mean velocity profiles, flow angle,
longitudinal fluctuating velocities, and other parameters.
Streamline patterns derived from mean velocity profiles show
divergence of the flow from upstream to the base of the ramp, followed
by convergence of the flow from the base to the top, and a slight
divergence downstream from there.
131
-------
Using C02 (52 percent heavier than air) with smoke added for
visibility, we have demonstrated that for flat terrain with free stream
wind speeds of 2 and 4 meters per second, plume buoyancy is
unimportant. However, at lower wind speeds, strong lateral spreading
of the dense plumes was observed. Thus, we conclude that under these
conditions, a significant buoyancy effect exists.
Repeating the above experiments with a ramp placed 1800
millimeters downstream of the source, a significant widening of the
plume is seen both in the neutral plume case and in the dense plume
case. When the wind speed was reduced to 0.75 meters per second,
laminarization of the plume was observed. At 0.5 meters per second
wind speed, obvious streaks, indicating laminar behavior, were
observed. We do not believe that a valid model of a full-scale problem
is achieved when laminarization occurs.
To compare the dense gas plume with the neutral plume, we plotted
the concentration normalized by the source concentration as a function
of downstream distance. The decay of the centerline concentration with
downstream distance in the dense plume is essentially identical to that
of the neutral plume. The model scale ratio used was approximately
1000 to 1. Far downstream, the slope is about -1.5, typical of the
downstream behavior of a point source in a neutral atmospheric boundary
layer. Closer to the source, the slope is -1.0, which has to do with
the finite width or area source. Thus, in spite of the differences in
the plumes observed visually, the centerline ground-level concentration
patterns changed very little. Therefore, this is not a good indicator
132
-------
in field measurements of whether a dense gas plume or a neutral gas
plume has been released.
In our studies, lateral concentration profiles of the plumes were
measured at 300, 600, 1800, 5000, and 8500 millimeters downstream.
Gaussian distributions were extremely well fitted to the lateral
concentration profiles of the neutrally buoyant plumes. Dense gas
crosswind lateral profiles are not Gaussian at all; they are
essentially flat topped.
The centerline concentrations for the dense gas and neutral gas
plumes are essentially identical. However, vertical concentration
profiles differ dramatically. At the 50-millimeter elevation, the
concentration for the neutral gas plume at 300 millimeters downstream
is about 50 percent of the surface concentration. In the dense gas
plume, the concentration at the 50-millimeter elevation is more than
two orders of magnitude lower than the surface concentration value.
Farther downstream, the dense plume tends more toward neutral or well
mixed at the surface.
In our experiments, the dense gas plume starts out substantially
wider than the neutral plume due to lateral spreading right at the
source. The dense gas issuing from the source actually creeps upstream
approximately one half of a source diameter. A very short distance
downstream, the lateral growth rates of the neutral and dense gas
plumes are essentially identical, i.e., equal slopes of the curves.
Comparing the centroids of the two distributions, a measure of the
vertical widths of the plumes, the negative buoyancy appears to be
significant as far downstream as you choose to compare.
133
-------
Looking at ground-level concentrations of the neutral gas plume as
a function of downstream distance on the center!ine with the ramp
installed, the ramp induces a ground-level reduction of approximately
20 percent. This is not a major difference. With the dense gas plume,
the ramp induced perhaps a 20- or 30-percent decrease in ground-level
concentration.
Anticipating that the buoyancy would be more important closer to
the source, we moved the ramp closer to the source. In that case, the
ramp induces perhaps a 30- to 40-percent reduction in ground-level
concentration for the neutral plume. With the dense plume, the ramp
induces about a 50 percent reduction in concentration at ground level.
An extensive experimental data set on the detailed structure of a
dense gas plume resulted from these studies. The model depicts a dense
gas plume that has significant buoyancy effects, but which,
nevertheless, will be insignificantly affected by molecular effects.
Data also were collected for the neutral plume in the same experimental
setting to support comparative analysis.
In summary, the longitudinal centerline concentration profiles of
the dense gas plume were strikingly similar to those of the neutral
plume. However, plume shapes were dramatically different; the dense
plume being much wider in the lateral direction and much narrower
vertically. Lateral distributions of the dense gas plume were top-hat
shaped, essentially uniform all the way to the edges. In contrast,
lateral distributions for the neutral plume were essentially Gaussian
in shape. Introduction of a ramp into the experiment induced a slight
reduction (20 to 30 percent) in the ground-level concentration when the
134
-------
ramp was far downstream. When the ramp was located close to the
source, the reductions were somewhat greater. Reductions were also
somewhat greater for the dense plumes than for the neutral plumes, but
not significantly so.
135
-------
ATMOSPHERIC SCIENCE AND EMERGENCY RESPONSE
AT THE SAVANNAH RIVER LABORATORY
Allen H. Weber and R. W. Benjamin
E. I. du Pont de Nemours and Company
Savannah River Laboratory
Aiken, SC
The Savannah River Plant (SRP) is a nuclear production facility
owned by the U.S. government and operated by du Pont. SRP is located
in South Carolina on the Georgia border near Aiken, SC. It is the
third largest nuclear materials production site of the Department of
Energy (DOE) and includes nuclear fuel production, fabrication,
irradiation, and reprocessing. A program for responding to unplanned
releases into the atmospheric or aqueous environment of radioactive
toxic chemicals is essential, and one has been in place since the
facility began operation in 1953. There has been an atmospheric
dispersion research program at SRP for almost 15 years. The
consolidation of emergency response and atmospheric research has
strengthened both programs.
The SRP encompasses 315 square miles of mostly wooded, gently
rolling terrain. Four operational reactors, two chemical reprocessing
plants with high-level waste tank farms, a tritium processing and
handling facility, and a fuel fabrication facility are operating. A
waste vitrification facility and a naval fuels fabrication facility are
under construction. Hazardous effluents that accidentally could be
released into the atmosphere from these facilities include tritium
(elemental and oxide forms), gamma-emitting aerosols, radioiodine,
137
-------
transuranic aerosols, radioactive noble gases, and chlorine gas. The
Savannah River runs across the southern border of the plant.
The Environmental Transport Group (ETG) is charged with
characterizing and calculating the transport and dispersion of
atmospheric and aqueous releases from SRP facilities. The group is
composed of 14 professionals and 8 technicians, including 5
meteorologists, an oceanographer, 2 health physicists, a nuclear
chemist, and a computer systems manager. At least two meteorologists
and a supervisor are on call for emergencies. Pertinent information is
reported to DOE and du Pont managers for final action. Environmental
impact and regulatory document matters are assigned to a separate group
although ETG offers expert assistance as needed.
The primary aim of emergency response activities at SRP is to
quickly identify and characterize any accidental release of materials
and to provide SRP decision makers with the information necessary to
respond wisely. The emergency response system has two major
components, the WIND system and the TRAC mobile laboratory. The
Weather Information and Display (WIND) system is used for data
acquisition, archiving, and model calculations. Data enter the WIND
system from the Remote Environmental Monitoring System that provides:
• Real-time wind and turbulence data collected from 8 single-
instrument, 60-meter on-site towers and a nearby 300-meter
television tower instrumented at 7 levels;
• Regional meteorological data from the Automated Field
Operations Services (AFOS) system;
• Data from temperature monitors in on-piant streams and'the
Savannah River;
§ Real-time source term information from stack monitors; and
r
138
-------
• Real-time plant boundary radioactivity measurements from
perimeter monitors.
The WIND system utilizes VAX 780 and 750 computers to process these
data, run dispersion models, and support remote terminals in each
operating area. The TRAC mobile laboratory is used for plume tracking
and real-time radioactive monitoring. These systems are operational 24
hours per day, 365 days per year.
The research programs at SRP emphasize mesoscale atmospheric
dispersion. The present field program, Mesoscale Atmospheric Transport
Studies (MATS), is designed to determine the accuracy of dispersion and
transport models and to .measure horizontal plume width. The downwind
distance range of greatest current interest is approximately 30
kilometers, which coincides with the major population centers
surrounding SRP, Augusta, GA, and Aiken and Barnwell, SC. Since 1983,
this program has been concerned primarily with measurements in unstable
atmospheric conditions, but this year the emphasis has shifted to
measurements in stable conditions. The experiments simulate short (15
minutes), intense puff releases. A sulphur-hexafluoride (SF-6) tracer
is measured downwind from a 60-meter stack with fixed, sequential
samplers and a mobile, continuous sampler in the TRAC vehicle. The
fixed samplers are deployed by many of the same team workers who are
called during actual emergencies. Sampling in the TRAC vehicle is done
to measure the instantaneous horizontal puff width during passage over
the 30-kilometer arc. To date, 26 daytime experiments and 3 nighttime
experiments have been completed in the MATS program. An additional 25
nighttime and several more daytime experiments will be completed in the
coming year.
139
-------
A spin-off from the MATS program is the SJable Atmospheric
Boundary layer Experiment (STABLE). The objective of STABLE is to
develop an understanding of turbulence and dispersion at mesoscale
range in the stable (nocturnal) boundary layer and to improve the
models for dispersion under these conditions. It is expected that the
largest offsite doses would occur from a release in stable conditions.
STABLE will include study of a 5-year turbulence database,
inter!aboratory field experiments, and extensive model evaluation and
development. The fact that both the Bhopal and Chernobyl accidents
occurred during stable nighttime conditions has given additional
impetus to this program.
The consolidation of operational emergency response and research
programs in one medium-sized group with a single group manager has
greatly benefitted both programs. Researchers can utilize the
laboratory and technical abilities of carefully trained emergency
response specialists. On the other hand, emergency response personnel
learn the latest observational procedures while gathering data for
research applications. Moreover, dispersion and transport codes are
tested and improved using results from research. Finally, operational
feasibility of research results can be tested during emergency response
exercises.
The information contained in this summary was developed during the
course of work under Contract No. DE-AC09-76SR00001 with the U. S.
Department of Energy.
140
-------
EMERGENCY PREPAREDNESS AND RESPONSE IN THE U.S. AIR FORCE
Captain Lawrence E. Key
U.S. Air Force Engineering and Services Center
Tyndall Air Force Base, FL
The U.S. Air Force Engineering and Services Center, where I am
assigned, is designated as the focal point within the Air Force for
environmental quality research and development (R and D). While we do
not do all of the R and D work, we do try to keep up with what is
happening in this area. When we recognize a need for research, we are
responsible for initiating a project to fill that need. We also are
responsible for coordination of research activities with other agencies
within the Air Force and the Department of Defense, such as
Headquarters, Air Weather Service, and Air Force Geophysics Laboratory.
We identify needs in the field and try to develop tools that people in
the operational areas can use to prevent an adverse impact on the
environment.
With regard to emergency response, the Air Force is interested,
basically, in two kinds of events—hot spills and cold spills. In more
technical terms, a hot spill refers to any event involving a fire or
an explosion, while a cold spill does not. That terminology comes from
the people actually working in the field in missile range safety and at
various missile sites. A fire or explosion generally involves some
type of missile explosion, either the vehicle itself or the stored
propellents used to fuel that vehicle. Cold spills may involve a
release of liquid missile propel1 ant or a release of aircraft fuel, but
do not involve fire or explosion.
141
-------
Focusing on hot spills, we are primarily concerned with two types
of propellants. Both the liquid and solid propellants are toxic. The
primary liquid propellants are nitrogen tetroxide (N204) and Aerozine-
50 (ABO), a 50:50 mixture, by weight, of hydrazine and unsymmetrical
dimethylhydrazine (UDMH). Both of these are very toxic and, because
they are stored in liquid form, are mobile in the environment when
spilled. The typical solid propellant in the Air Force is composed of
ammonium perchlorate and aluminum. The primary combustion material
resulting from the ammonium perchlorate-aluminum mixture is hydrogen
chloride, which, when compared with N204 and A50, is not very toxic.
However, we must be concerned with local and climatological effects of
a large hot spill of the solid propellants. We must also consider the
situation of a hot spill with both types of propellants. For example,
when the Titan 34-D blew up shortly after liftoff at Vandenberg earlier
this year, both liquid and solid propellants were involved.
A typical explosion involves a fireball that lifts, stabilizes at
a certain height, and then disperses. At the two largest launch sites,
Vandenberg Air Force Base and Kennedy Space Center, meteorological data
available at the time of the accident are extensive. There are a large
number of towers. Vandenberg currently has 26 towers on the
reservation. These towers are instrumented at various heights, have
instruments for mean wind speed and direction and temperature, and can
measure temperature differences at various heights to get an idea of
atmospheric stability. Vie also directly measure the standard deviation
of wind direction at various levels. Prior to launch at these
locations, a rawinsonde run is made to gather further information about
142
-------
the stability characteristics of the atmosphere, particularly at higher
altitudes.
At other sites, primarily Intercontinental Ballistic Missile
(ICBM) sites, instruments are much less numerous. At these ICBM sites,
we have the capability to get mean wind speed and direction at
approximately 10 meters above the ground. Also, we have the capability
to estimate the standard deviation in wind direction and determine a
value for the vertical change in temperature. That value for the
vertical change in temperature is the primary means that the Air Force
currently uses to characterize atmospheric stability.
The tools we have for modeling a hot spill are currently limited.
The Hypergolic Accidental Release Model (HARM) was developed
specifically for the Titan II missile system that is fueled with
hypergolic liquid propel1 ants. The Rocket Exhaust Effluent Dispersion
Model (REEDM) was designed to estimate the ground-level concentrations
of gaseous hydrogen chloride resulting ,from a Space Shuttle launch.
REEDM is used primarily for planning prior to a Shuttle launch;
however, it does have a module to handle a launch abort or explosion,
which would release fairly large quantities of hydrogen chloride.
The primary meteorological input to HARM is rawinsonde data. HARM
has a preprocessor that calculates the source term taking into account
the heat released by varying amounts of hypergolic liquids. It
calculates an initial fireball size and cloud rise, and the resulting
dispersion is modeled using a Gaussian puff model. There is an
algorithm that estimates the amount of deposition due to rainout. The
143
-------
primary outputs of the model are ground-level concentrations of the
combustion products of interest.
The role of the meteorologist in these hot spills depends on where
the accident occurs. At Kennedy and Vandenberg, a meteorologist is
heavily involved in planning and in responding to accidents. The REEDM
model is run by the meteorologists and is used extensively for risk
assessment. Risk assessment is generally done well in advance of a
launch and involves a large number of computer runs using climatology
to determine a worst case scenario. REEDM is used for planning just
prior to a launch using forecast winds and temperatures. It is used
for accident response as required; but there really is limited time and
many more things to worry about than running a model immediately after
an accident.
At other locations, such as the ICBM sites, the meteorologist's
role is less well-defined and tends to be up to the local commander.
For risk assessment and for planning and response, the heaviest
involvement is with the safety and operations personnel. This is
principally a condition of geography. The meteorologist is not at the
missile site that has a problem. He or she is at a central location,
and it is the on-site people that must do the planning and response.
However, the meteorologist is involved in post-accident analyses.
There are two broad areas of interest with respect to the Air
Force's response to cold spills; 1) a spill involving Air Force
materials and occurring on Air Force property or 2) a spill involving
industrial chemicals occurring off of Air Force property. The latter
category is included because the Air Force is obligated to assist if an
144
-------
accident occurs near an Air Force installation and the local
authorities asked for assistance. As a result, the Air Force
meteorologist occasionally gets involved with calculating toxic
corridors for non-Air Force materials and non-Air Force locations.
On Air Force property, our primary concerns are the missile
propel!ants, N^O. and A50, and other types of hydrazines used in
various aircraft. There are also relatively small quantities of
typical industrial chemicals on Air Force installations, such as
chlorine, which is used in water treatment plants.
Cold spills do sometimes occur in missile operations. For
example, in 1978, a Titan II site in Kansas experienced a N204 spill.
In such cases, the meteorological information available depends upon
the site. Kennedy and Vandenberg are, again, well instrumented. At
the missile sites, such as Titan II and Miniteman sites in the Midwest,
meteorological information is relatively limited, but they will have,
at a minimum, wind speed, direction, and the vertical change in
temperature. At a typical Air Force base, not supporting missiles, but
experiencing a hydrazine spill from an F-16 aircraft, the only
meteorological information available is usually mean wind speed and
direction, temperature at a height of about 6 feet above the ground,
and a very rough means of estimating the vertical temperature profile
and the standard deviation of wind direction. If a spill occurs near
Air Force property, but not on it, the amount and quality of
information available is highly variable and, typically, very limited.
Off base, the information available typically depends on the proximity
to the base or to a National Weather Service site.
145
-------
For cold spills, there are a few models that we have available.
One is the well known Ocean Breeze/Dry Gulch (OB/DG) equation, which
has been available since the mid-sixties. This is an empirical
equation derived from a large number of tracer releases. It was
developed by the Air Force Geophysics Laboratory. There are several
other similar equations that are not as well known, such as the
Mountain Iron and Sudden Ranch equations, which are used at Vandenberg
primarily. These are based on a series of releases of material at a
specific location and are used only at that location. Vandenberg is an
interesting and complex place meteorologically. Vandenberg is in a
coastal situation approximately 150 miles northwest of Los Angeles;
therefore, it has coastal meteorology over very rough terrain. The
southern half of the base happens to be the most complex, and that is
where the launch facilities are located. The northern half of the base
is less complex. That is the reason why there are two equations used
at Vandenberg.
For cold spills, we also use an Air Force-modified version of the
TM
Complex Hazardous Air Release Model (CHARM) developed by Radian
TM
Corporation. CHARM is used specifically for the current program of
TM
deactivation of the Titan II missile system. The modified CHARM
model was added to our pool of models following a thorough
investigation of a 1980 hot spill accident of a Titan II near Damascus,
AR. As a part of this investigation, a thorough review of the safety
procedures for the entire Titan II program was made. One of the
recommendations from that review was the need to replace the OB/DG
TM
equation. OB/DG does not handle dense gases, such as N«0.. CHARM
146
-------
does include a dense gas algorithm and has been shown to be useful in
the field for estimating a toxic corridor that would result from a
release of NgO^.
The OB/DG equation is fairly simple. It is based on a linear
regression of the data gathered during a large series of test releases.
As a meteorologist, it is a little disturbing that the equation does
not consider wind speed. This basic equation is even further
simplified for the typical Air Force base by adjusting the coefficients
TM
and exponents accordingly. We adapted CHARM for Air Force use
specifically for the Titan II deactivation. It is a Gaussian puff
model that includes a dense gas algorithm based on Eidsnik's work. It
has a source strength module, while OB/DG does not. Thus, CHARM
gives us an improved capability in estimating the source strength
resulting from a spill.
At the time of a cold spill, by regulation, the Air Weather
Service meteorologist on site calculates a toxic corridor. In
practice, the response varies. The on-site commander who responds to
the spill decides whether the corridor is calculated. This does not
mean that no action is taken to prevent loss of life or injury if the
modeling is not conducted. Typically, what happens, in my experience,
is that the commander responding to the spill directs that the entire
downwind area be evacuated. This is a very conservative, but effective
approach.
Just as a point of interest, there are several weather
restrictions on handling materials, and, frankly, one is a point of
contention with some of the maintenance personnel. This limitation was
147
-------
particularly troublesome during the Titan II deactivation that involved
quite a bit of transfer of propellants. Essentially, there could be no
transfer of propellants during calm winds and/or stable conditions, the
stable conditions being determined by the vertical temperature profile.
This limitation created a lot of arguments between meteorological and
maintenance staffs.
The Air Force has identified several research needs. We need,
like most others in this field, improved source strength models. We
need a better means of estimating the surface temperature in our
models. We are concerned about mixtures of chemicals released to the
environment mixing with other chemicals: water, other pollutants, and
naturally occurring materials. We do not really have a good model to
handle fires, particularly with solid propellants.
We are looking for an improved dispersion model for general use
and are still looking for a better replacement for OB/D6, which will
consider dense gas effects and incorporate a terrain wind flow model,
particularly at Vandenberg. Finally, we need better model validation
techniques. The commander on site needs to have some reliable means of
quantifying the uncertainty associated with the toxic corridor that is
given to him. Frankly, right now we do not have a good means of doing
that.
We have several activities under way to meet these research needs.
The Air Force Geophysics Lab is working on a new dispersion model. It
will be a relatively simple, Gaussian puff model based on the Shell
spills model, although we are making several modifications. Most of
this effort involves looking at the source strengths involved with
148
-------
spills of the various chemicals about which the Air Force is routinely
concerned. We are planning to include a dense gas algorithm. We are
also looking at various other models, including the DEGADIS model that
was discussed earlier in this conference. A wind flow model, based on
work done by the Army, is also being evaluated.
Also, we have initiated a study at Lawrence Livermore National
Laboratory to recommend a methodology that could be used by the Air
Force when we are faced with the need to do an evaluation of whatever
model we finally select. This validation technique will involve
compiling the dense gas data that are available in a form suitable for
model evaluation and, then, recommending certain techniques to use in
comparing our model calculations to the actual field data. We also
have a small effort under way to look at quantifying the uncertainty
associated with toxic corridor calculations. Additional hot spill
models are being considered to fill the current gaps in our capability
in that area.
This discussion presents a brief overview of the current
capabilities of the Air Force and of the research we have under way to
improve our emergency response modeling support capabilities.
149
-------
POISONOUS GASES FROM LAKES: THE CAMEROON DISASTER
Daniel A. Livingstone and George Kling
Duke University
Durham, NC
In 1960, I became interested in poisonous gases because of an
experience while on vacation with my family in the crater lake district
of Ruwenzori Park in western Uganda. We smelled rotten eggs and
tracked the smell about five miles overland with a Landrover, ending up
at a small lake that was purple with sulfur bacteria. There was a
whitish deposit of salts around the shore. Upwind, the stench was not
too bad. We stood there and saw no dead elephants along the edge of
the lake, but then, elephants are not usually attracted to drinking
water that is supersaturated with halite or trona. It seemed to us the
better part of valor not to venture down into the crater to look for
smaller dead animals. Elephants might steer clear; limnologists might
not.
This incident stayed in the back of my mind for 10 years. When
one of my students, John Melack, went to the same lake in 1971, it no
longer stank of rotten eggs. The climate of Uganda had changed, with a
number of wet years occurring in succession. When Melack got to the
lake, it was saturated with respect to sodium chloride in the deep
water, but there was a meter and a half of rainwater lying on the
surface. There was no evident evolution of gas. On our first visit,
we had not observed any bubbles either, but we were probably too far
away to see any but catastrophically large ones.
151
-------
I have since spent several months on Lake Bosumtwl In Ghana, a
meteorltic explosion crater, where there are many traditional reports
of large bubbles of gas coming up out of the water. There are even a
few reports in the scientific literature of such degassing. Although
few people have been killed by these bubbles of gas, accidents have
happened with sufficient frequency to generate an interesting folklore
regarding the lake and the proper way to propitiate its dangerous
resident deity.
The bubbles of gas tend to rise in August. I was there in August,
1966, when the lake was circulating freely. There was no smell of
rotten eggs, but the lake had an odor to suggest that some form of
reduced sulfur, perhaps mercaptan, was percolating slowly from it. We
saw no bubbles of gas burst from the lake.
Last year, Kling and I went to Cameroon with Curt Stager, Jean
Maley, and support from National Science Foundation (NSF), Centre
National de la Recherche Scientifique (CNRS), and Office de la
Recherche Scientifique et Technique d'Outre-Mer (ORSTOM) to study the
history of Cameroonian lakes and make a limnological survey of them.
Ndoni Paul was our Cameroonian assistant. We arrived shortly after a
very big bubble of gas had burst from Monoun, killing 37 people. That
happened on August 15, 1984, at approximately 11:30 p.m. On August 21,
1986, at 9:30 p.m., a bubble of gas came out of Lake Nyos in the
western part of Cameroon killing almost 1700 people and a great number
of cattle. This was a major disaster, comparable to Bhopal and
Chernobyl.
152
-------
When the first reports of a gas discharge came from Lakes Monoim
and Nyos, it was regarded as a mystery. Various explanations, natural,
artificial, and supernatural, were advanced to explain the mystery.
Most of these explanations were less satisfactory to us than they were
to others. Now, after the post-event observations of Kling and the
other members of the Agency for International Development (AID)-
supported Foreign Disaster Team, there is little mystery left. We
think we understand the phenomenon quite well. We can probably predict
it in a broad sense, and we can almost certainly control it.
These natural events differ from most catastrophes we have been
discussing at this workshop. Meteorological processes dissipate this
problem, but meteorological processes probably also trigger the
catastrophe. They must also be considered in preventing its
recurrence.
In a tropical lake, there can be many layers. This layering can
be very stable. The first layer is the air-water interface. Next,
there is a layer of surface water, usually fairly low in dissolved
salts, that circulates freely during the season of minimal thermal
stratification in the lake. Underneath that layer, there are usually
one or more layers that are stabilized by dissolved substances. These
deep layers are not stabilized thermally, and may even display reverse
thermal stratification, but are chemically so stabilized that they may
be perennial. During the season when the surface water is at its
coolest, when the thermal stratification is weakest, a strong wind,
low-pressure system, landslide, subterranean slump, or anything else
153
-------
that will disturb the free water surface produces standing waves at all
the deep discontinuities.
In a small lake, such as Singletary Lake in North Carolina, the
surface rises in the downwind direction on even a fairly still day by
something like 5 centimeters. The lakes in Cameroon are also small,
with a similar surface setup; but deep in a stratified crater lake, the
next discontinuity is set up much more than 5 centimeters in the
opposite direction. These two slopes, that of the air-water interface
and that of the first water-water boundary, are not the same. The
surface slope involves air with a density close to zero and water with
density close to one, while the internal slope involves two layers of
water with density differences about 1 percent of the density
difference at the surface. So, when a wind sets up a lake, the
internal slope is some hundred times that of the air-water interface.
The vertical displacement will be 5 meters or more.
If a deep layer of the lake, for whatever reason, is charged with
a very soluble gas, Carbon dioxide (COp), hydrogen sulfide (HpS),
methane, or ammonia to the point of saturation at the ambient pressure,
and the layer is displaced by a large set up, then the deep water can
effervesce violently. When this happens, we can have a tragedy like
that in Nyos or Monoun.
Kling has detailed stratification data from Barombi Mbo, a lake
near Kumba in southwestern Cameroon, which fortunately does not have
much gas dissolved in its deep waters. Because of the annual pattern
of cloudiness, southwest Cameroonian air temperature reaches its annual
minimum in August. The surface water of Barombi Mbo also reaches its
154
-------
annual minimum around August. This kind of thermal regime is typical
of a tropical lake, which typically has a season of maximum thermal
stability with warm surface waters. Stratification involves a
temperature range where water density changes quickly with temperature,
so small temperature differences are sufficient to produce stable
stratification. The upper part of the lake is thermally stable during
most of the year, and the deep waters are perennially stabilized by
dissolved salts even to the point of allowing a thermal inversion. In
June, July, and August, the shallow stability is largely removed by
cooling of the surface water. The surface layer and the next layer
develop very similar densities. When something sets up the air-water
interface, the discontinuity between the surface and deep waters will
be set up enough to generate large-scale effervescence of dissolved gas
if the deep waters are close to saturation at the ambient pressure.
That was apparently the situation at Lake Nyos when it killed 1700
people.
The discharge of gas at Lake Nyos occurred near an 80-meter-high
promontory. The initial wave rose over the top of the promontory; and
stripped the shallow layer of soil and vegetation from its granitic
rocks. That initial displacement generated a surface seiche. The
amplitude of the surface seiche was about 6 meters at the far end of
the lake. This is obviously a very serious kind of event. It takes a
great deal of force to move such an amount of even fizzing water to so
great a height.
Maps of the area are poor, and the distribution of dead animals
and people have not been plotted with a high level of accuracy. The
155
-------
village of Old Nyos, which is 100 meters or more above the lake was not
affected. Observers in Old Nyos, who were awake at 9:30 p.m., claim to
have seen the burst of gas and water from the lake, but no one died
there. On the other side of the lake at Sawe, which is not so high
above the lake, no deaths occurred either. However, downstream from
the lake the village of New Nyos was wiped out. The fairly heavy
settlements along the road following the river, as far as Su Bum, were
also wiped out. At that point, the river takes a turn and passes
through an area that is lightly inhabited, until it reaches the village
of Mashi, where again there were very heavy deaths. The distance along
the course of the river from Lake Nyos to Mashi is about 34 kilometers.
The width of the valley is not well known, but from the map contours we
estimated it to be about 2 kilometers wide. If the valley were
2 kilometers wide and if it were filled with gas to a depth of 10
o
meters, then we are talking about something like 7 x 10 cubic meters
of gas.
Autopsy reports indicated that the people died of C0« poisoning.
There was no indication of carbon monoxide (CO) or hydrogen cyanide
(HCN), nor was there any indication of HpS in the victims. They died
of COp poisoning.
If the lake, which is very deep for its size (1 kilometer by 2
kilometers, with a maximum depth of 220 meters), were fully charged
with C02, it would hold about 10*° cubic meters of C02- This is
sufficient to fill the valley more than 10 times. An amount of gas
that would fill the valley, if expanded from 22 atmospheres to
atmospheric pressure, could do a lot of work. It could easily lift a
156
-------
destructive wave of water to a height of 80 meters. Remember, the
entire volume of the lake is not being lifted, but only a small portion
of one side. It struck the promontory with enough force to splash over
the top, even though the water surface generally was not raised that
high.
What was the source of the gas? The limnologist thinks first of
detritus, the corpses and feces of lake-dwelling plants and animals
that decompose as they settle to the bottom. This decomposing material
generates C02 by aerobic respiration and by fermentation together with
methane when the oxygen is depleted. Ammonia and ^S are also
generated under anoxic conditions. It turns out that the COg in the
lake cannot come solely from biological processes. There is
proportionately too much CO* and not enough of the other products of
.decomposition in the gases of Nyos and Monoun. There is not enough
methane, FLS, or ammonia. If the gases in the deep waters of the lake
contained appropriate amounts of biologically derived ammonia, water
samples from the lake would bring tears to your eyes. While there is
ammonia in Lake Nyos waters, there is not enough to support a theory of
biological origin for the COg.
The isotopic ratios of gases in the lake waters also provide
clues. The first U.S. team to sample Lake Monoun last year found that
the C02 in the water had a radiocarbon age of 18,000 years. This
suggests that perhaps one-eighth of the C02 came from biological
processes. The remaining seven-eighths came from some other, non-
biological source. These ratios are consistent with various mixtures
of primeval- CO-, i.e., COg coming from magma. Overall, the evidence
157
-------
indicates that approximately one-tenth of the C02 was coming from a
biological source, and the remaining nine-tenths, probably, from deep
springs. Looking at Cameroon's springs, we find that most,
particularly those of West Cameroon, are effervescent and that most are
a good source of C02«
In each of the papers we have heard at this conference, we learned
about opportunities to select sites, to design plants, or to develop
emergency response plans as an effective means of dealing with
environmental emergencies. Such plans are not helpful in Cameroon. If
you are ever caught in a Cameroonian CO^-release emergency, the only
thing you can do is hold your breath and run uphill. You are very much
better off suffocating from lack of air than breathing dense
concentrations of CO*. Breathing the C02 causes severe acidosis, your
brain shuts down in seconds, your glottis goes into spasm, and you die
quickly.
What can you do in a situation where you cannot select your site
by moving away and raising your crops elsewhere? You may be able to
modify the site. It is possible to take a pipe, of perhaps 6 inches in
diameter, put it down into the deep water of a lake that is fully
charged with CO-, and pump the water up with a pump until it
effervesces. Then, you can shut off the pump and sit back and watch
the lake degas. It would take about 10 years by this method to
discharge a lake the size of Nyos to saturation at surface pressure.
But do not forget meteorology while degassing the lake. First, you
must know where your discharge plume will travel. Second, you must
take care how fast the lake will degas: too fast, and you might
158
-------
produce your own small catastrophe; too slowly, and the springs can
keep up with you, and you will never get rid of the C02. There also
must be a limnological portion to your plan. When the pipe is running,
you must dispose of the waste water carefully. If it is returned to
the lake in a turbulent fashion and mixes freely with the upper layer
of the lake, you may generate a catastrophe worse than any happening
naturally.
159
-------
SUMMARY OF PANEL DISCUSSIONS
Panel discussions were organized to directly address the goals of
the workshop and to provide a forum for free discussion of the issues
raised by the goals. The panelists and participating workshop
attenders sometimes followed an independent agenda, but the resulting
discussions were very fruitful.
Panel 1: Summary of Current Status of Source Strength and Dispersion
Models, Their Strengths and Weaknesses, and Recommendations
for Improvement
Panel 2: Recommendations for Selection from Among Current Models for
Immediate Use in Hazard Identification and Evaluation,
Preparation of Emergency Preparedness Scenarios, and in
Response to Emergencies
Panel 3: The Role of the Meteorologist in Hazard Evaluation and
Emergency Response
One major accomplishment of these panel discussions was the
fostering of strong interactions between two technical communities
present at the workshop. The people involved with chemical releases
seldom have interacted directly with those involved in radioactive
releases. The difference in viewpoint was profound. For example, Dr.
Knox related his experience with Chernobyl and showed how the source
term could be inferred from measurement of the radioactive debris.
This procedure took several days, but as his major concern was the
impact upon the United States, this time interval was not important.
Even if the concern was for European exposure, there was time to take
measurements and estimate the source term. On the other hand, Dr. Gait
showed that when one is dealing with an overturned vehicle, immediate
161
-------
estimates of the source term and the relevant species are required
immediately, especially if evacuation of people is contemplated.
Between these two extremes was a continuous spectrum of concerns. The
panels allowed for a free expression of these concerns and thus proved
useful.
Panel 1: Summary of Current Status of Source Strength and Dispersion
Models, Their Strengths and Weaknesses, and Recommendations
for Improvement
Chairperson: Frank Schiermeier
Reporter: Steve Perry
Panel Members: Gary Briggs, Tony Cox, Don Ermak, Dan McNaughton,
Bob Meroney, Jerry Schroy
This panel summarized the current status of source strength and
dispersion models, especially their strengths and weaknesses, and
recommended necessary improvements. Four major areas were addressed:
source term models, dispersion models, evaluation and validation of
models, and practical application of models.
Knowing how much material is emitted in a given emergency
situation is of primary importance. Most often the emission rate is
estimated by means of a model calculation. The panel reported on five
source term model categories: process emissions, contained emergency
releases, spills, burning sources, and density impacts. Process
emissions are usually point sources from established stacks and are
thus well documented. Contained emergency releases result from
components designed to release material under certain conditions.
Examples of these are discharges from relief valves and rupture disks.
162
-------
The material released could be from a runaway reaction, which requires
careful engineering judgment. If the material is accidentally or
deliberately ignited, there would be a burning source, which is
discussed further below. Finally, if the material is stored under high
pressure, the release would be a jet that might produce a multiphased
flow. This last case is not well understood at all and requires a
great deal of careful research. Spills of pure components and binary
mixtures that undergo, simple flashing and pool formation are reasonably
well understood. Spills that involve multiphase flows are, again, not
well understood. Steady-state spill models are useful for situations
that are most nearly steady (obviously). As the conditions depart from
steady-state conditions, such as occurs with pressure drops, the
steady-state models decrease in usefulness. There is a very strong'
coupling between source models and dispersion models for spills that
involve aerosols? and evaporation from these aerosols. In fact, it is
virtually impossible to separate the two for some conditions.
Burning sources are another category where there are major
unknowns and much research to be done. A relatively simple burning
source would be a flare that resulted from igniting a contained
emergency release. The input products are known; so presumably would
be the combustion products. This is not always true because combustion
may be strongly influenced by the ambient meteorology and any
mitigation techniques that might alter the combustion temperature. The
reaction products often are controlled by the combustion temperature.
Thus, even a flare is not well enough understood. With an open burning
source^ the?unknowns multiply very quickly. The combustion temperature
163
-------
is unknown, the product species are unknown, and the effect on material
that is not burning but adjacent to the fire is unknown. Furthermore,
attempts to extinguish the fire may make the situation worse by making
the species emitted more toxic than the original combustion products.
Major research is needed here.
The density impact source category is a general category for which
parameters like aerosol formation from condensation, partial
vaporization, chemical reaction, initial dilution, and source
temperature are important. The single most important point about
density impact sources is the very strong nonlinear effect between the
size of the source release and the subsequent behavior of the
dispersing plume. Here again, there is a very close coupling between
the source term and appropriate dispersion model. The panel reported
three major dispersion modeling categories related to plume buoyancy:
nonbuoyant, positively buoyant, and negatively buoyant plumes. The
nonbuoyant or passive plume category applies to all plumes after enough
dilution. The character of the planetary boundary layer (PBL) is the
most important variable here. If the PBL is unstable, i.e., large
convective eddies driven by the surface heat flux dominate the
turbulent exchange, the dispersion of passive plumes is quite well
understood. The models developed for this case do much better than the
Gaussian models. This is rarely a worst case (where very high
concentrations occur) since dilution is quite rapid. For the neutral
case where the turbulent exchange is dominated by eddies generated by
the vertical wind shear, the Pasquill-Gifford type Gaussian model does
an adequate job. But, this is only true for the near-ideal cases in
164
-------
smooth flat or gently rolling terrain. This, also, is seldom a worst
case, because there is often adequate mixing and dilution.
The passive plume in stable conditions is often the worst case
scenario, as well as the most difficult case to model. The
Pasquill-Gifford method only works for smooth terrain and unrestricted
mixing height (an unlikely occurrence). The passive plume in stable
conditions is the hardest case to generalize because of surface
roughness effects, limited vertical mixing, flow meander, and drainage
flows in sloping terrain with the possibility of channeling. This case
requires an extensive research effort.
Buoyant plumes from stacks are reasonably well modeled with
current methods. Surface releases are very difficult to model because
of surface deposition, high turbulence levels, and entrainment effects.
If the plume is buoyant enough, it will lift from the ground. Buoyant
plumes interact with the PBL and will reach greater heights during
convective conditions and will be restricted in stable conditions,
especially if the buoyancy is insufficient to completely penetrate the
nocturnal stable layer.
A special type of buoyant plume that has not been well studied is
the flare. Often, a mitigation technique for a release is to burn the
material in a flare. An example is a release of hydrogen sulfide,
which would be converted to sulfur * dioxide after combustion. The
detailed consequences of such an action under all possible PBL
conditions are unknown at present.
Dense gas plumes are currently the subject of intense interest
because of their potential for immediate health consequences. A dense
165
-------
plume is one that is negatively buoyant under all atmospheric
conditions. Such plumes often modify the atmospheric conditions
locally that can inhibit dilution. The current view is that in sloping
terrain, with slopes as small as 0.1 percent, these plumes will travel
downslope. With an adverse wind, this might not be the case, however.
In complex terrain, such as canyons with steep walls, a dense gas can
travel great distances as the event in Cameroon showed. Much work
needs to be done to characterize dense plume behavior in complex
terrain, especially with different wind fields. In terms of accidents,
however, the most important questions concern the direction a dense gas
cloud will travel in real terrain with real wind speeds and directions,
and what the concentration field will look like.
One technique that should be explored for all possible source and
dispersion situations is the use of virtual sources that can be
nondimensionalized. One mathematical model then could be used for a
variety of releases. Some problems, of course, cannot be treated in
this manner because of their inherent nonlinearity, e.g., a burning
source.
Dispersion models cannot deal satisfactorily with complex terrain
and winds of very low speed to calm, nor can they deal with the
transition from near source density-dominated dispersion to far-field
passive dispersion. There has been little effort to include spatial
and temporal variability into current dispersion models.
Introduction of new or improved models into the inventory of
operational models requires a demonstration that the new models are
indeed improvements. For cases where there is no current model, a
166
-------
rigorous comparison of model results with data is most important. For
cases where there are current models, the new models must show a clear
improvement. Further, all new models should undergo extensive
sensitivity testing. In addition to the sensitivity analyses, a set of
cases where previous models have failed should be developed and the new
model tested on these cases to ensure correct model performance.
The panel concluded that in selecting a suitable model for
application, several issues must be considered. In hazard
identification, the emphasis is on source characteristics, i.e., the
possible ways in which the source could emit hazardous material. The
models required for this phase of emergency response concern the
different types of accidental spills that can occur along with a good
engineering estimate of the probability of such an accident. Emergency
action planning requires the results of the hazard identification study
followed by consideration of all possible ways in which the release can
be influenced by weather conditions. The models required must account
for the transport and dilution of the released effluent. Chemical
transformation of some species must also be considered.
A family of concentration isopleths for a variety of release
scenarios and environmental conditions are developed. This information
can be viewed together with appropriate demographic data to estimate
human exposures to the hypothetical releases. This last step should
result in a plan for responding to accidental releases. Emergency
response requires immediate answers for alleviating the risk to the
public. The models used must be able to be run on very short notice
with unambiguous results. If hazard identification and preparedness
167
-------
planning processes are done correctly, then there should be a menu of
recommended actions from which to choose. If enough cases were
examined In the planning stage, then there should be a close analogy to
the actual accident. Appropriate action can then be taken.
Panel 2: Recommendations for Selection from Among Current Models for
Immediate Use in Hazard Identification and Evaluation,
Preparation of Emergency Preparedness Scenarios, and in
Response to Emergencies
Chairperson: Ray Hosker
Reporter: Tom Pierce
Panel Members: Harry Allen, Jerry Gait, Jerry Havens, Reed Hodgin,
Gene Runkle, Bruce Turner
The second panel considered recommendations for selection of
models for immediate use in hazard identification and evaluation,
preparation of emergency preparedness scenarios, and for response to
actual emergencies. The panel did not recommend specific models, but
rather agreed to a list of features and technical criteria for model
choice.
The first general feature was scientific defensibility. The
scientific basis of a model for a particular application is of great
importance. The model should be operated only for those situations
where scientific credibility has been established. The model should be
widely available with full documentation on both the scientific basis
and on test results showing how the model performs over the full range
of applicability. The documentation should also include all technical
criticism and responses to this criticism. Further, the model
structure should be organized into modules that allow updates to be
168
-------
easily installed. Each model should have been subjected to rigorous
peer review and all results of the review available to the user
community.
The panel recommended a number of technical points that models
should deal with:
1. Near calm to very light winds
2. Wind shear in the vertical (speed and direction)
3. All stability conditions
4. Site-specific meteorological data
5. Local terrain and local flow conditions
6. Local obstructions
7. Dense gas effects
8. Effects of hydrometeors (rain, snow, etc.)
9. Momentum and buoyancy effects at the source
10. Time-varying release rate
11. Chemical reaction of effluent
12. Real time on-site concentration measurements.
As is obvious from this list, the panel determined that the problem of
application of models was very much dependent upon the individual case
and that no global recommendation of a specific model for all cases was
possible. Rather, a suite of models, each carefully matched to the
others in the suite, was the best approach.
The panel next recommended that potential model users examine the
suitability of a given model with the following guidelines. First, for
application to toxic releases, the effects of the toxicity and a
probabilistic statement of associated risk should be incorporated into
169
-------
the model output so that the user is able to relate this output to the
problem at hand without going through another step. Second, the model
should be capable of producing results commensurate with the level of
accuracy actually required in the application.
The panel considered the issue of possible categories for model
use. The first use is for hazard identification. Does a given
situation pose a hazard to the public? The best approach here is to do
an analysis of possible release scenarios, then use the model most
applicable for that case. Since rapid turnaround of results are not
necessary for this situation and realistic estimates are desired, the
model should be as complete a representation of the situation as is
scientifically possible. The second possible use identified was
preplanning for the possible occurrence of the releases that posed
hazards. As before, the model should be as complete and realistic as
possible. Response plans should be developed using these results. The
third model use category occurs when and if an actual release occurs; a
simple model with quick turnaround is most appropriate. Of course,
this quick turnaround model should have been run in the planning stage
so that a record of its performance relative to the most complete model
is available. Thus, a check on the uncertainty of the simple model's
results are available to the user, and safety factors are incorporated
into response plans that are to be activated. The final model use
category is for post-event analysis. The most complete model should be
used along with simple model results to assess the performance of the
entire emergency response system.
170
-------
' The panel recommends that all emergency response plans be
thoroughly tested. These plans can be tested by holding simulation
exercises to ensure familiarity with all components.
Panel 3: The Role of the Meteorologist in Hazard Evaluation Emergency
Response
Chairperson: Lew Nagler
Reporter: Jack Shannon
Panel Members: Marvin Dickerson, Allen Weber
The role of the meteorologist in hazard identification, emergency
preparedness planning, and in emergency response is multifaceted but
must remain closely connected with his or her expertise. This
expertise should include some knowledge of allied fields such as
industrial hygiene and health physics because the advice sought most
often is related to these fields.
In hazard identification, the role of the meteorologist is to
advise on how a given potential hazard might be affected by weather
conditions. For example, in analyzing the risk for a release from some
storage or processing facility, the wind field and atmospheric
stability determine the rate of dilution for the released material.
The details of the source configuration as well as any proposed
mitigation techniques are very likely affected by meteorological
conditions. Thus, a meteorologist should be consulted at the very
start of a hazard identification exercise.
In terms of emergency preparedness, the meteorologist will select
the appropriate models for the problem and will be able to advise on
171
-------
the most probable exposures arising from a given release. He or she
will also select the models to be used in the event of a real emergency
and demonstrate how the operational model handles the same situations
simulated by the more sophisticated planning models. Finally, the
meteorologist will prepare weather scenarios for initial training in
emergency response and will also provide real data for further training
exercises. The meteorologist will also advise on the most appropriate
placement of on-site instrumentation (both weather sensors and chemical
or radiological).
For actual emergency response situations, the meteorologist will
have to provide the weather data and model predictions for use by the
appropriate authorities. If the meteorologist has been included in
this process from the beginning, then this last phase, which is the
most difficult, will be that much more reliably done.
172
-------
APPENDIX A
The Bhopal Gas Tragedy
A-l
-------
THE BHOPAL GAS TRAGEDY
K. SHANKAR RAO
Atmospheric Turbulence and Diffusion Division
NOAA/ARL, P. 0. Box-2456, Oak Ridge, TN 37831
and
M. P. SINGH and S. GHOSH
Center for Atmospheric Sciences
Indian Institute of Technology, New Delhi-110016
Abstract
In the early hours of December 3, 1984, about 40 tons of highly
toxic, volatile, and reactive methyl isocyanate (NIC) and other
poisonous gases leaked from a storage tank at a pesticide plant in
Bhopal, India. The gases escaped in about 90 rain through a 33 m high
atmospheric vent into the cool night air, and quickly spread in a
foglike lethal cloud over a large populated area. Thus began the
world's worst industrial disaster involving toxic chemicals which
killed over 2,200 people and injured more than 200,000.
In this paper, we sketch the accident scenario and outline the
events in the aftermath of this catastrophe. The topographical
features and meteorology of Bhopal, and the physical, chemical, and
toxicological properties of MIC and other gases in the leak are
described in the context of the dispersion of the cloud and its
effects on human and biological life. A simple atmospheric dispersion
model, emphasizing aqueous phase conversion and deposition of NIC, is
presented. This model, based on an analytical solution of the 3-D
advection-diffusion equation, gives estimates of ground-level
concentrations and deposition fluxes of MIC. The model estimates are
qualitatively correlated with recorded human fatalities and injuries,
and the observed damage effects on trees and vegetation in the
affected areas.
In the epilogue, the medical care, relief and rehabilitation
efforts, and some preliminary results of the ongoing medical and
toxicological studies are summarized. The legacy of the Bhopal
Tragedy including its implications for information, siting and
operation of plants that handle toxic chemicals, safety, health, and
environmental standards and enforcement, emergency response programs,
and other relevant issues, are briefly discussed.
Presented at the Special Joint Session of the Ocean and Atmospheric
Sciences on "Paths and Fates of Toxic Pollutants in the Atmosphere and
Oceans," American Geophysical Union Fall Meeting, Dec. 7-12, 1986, San
Francisco, CA.
A-3
-------
INTRODUCTION
On the night of December 2-3, 1984, at around midnight, methyl
isocyanate (MIC) and other poisonous gases, escaping from the Union
Carbide India Ltd. (UCIL) pesticide plant in Bhopal, silently spread
out into a huge cloud that enveloped a large populated area of the
city. Hundreds, mostly children and old people, died in their beds.
Thousands more awoke to a Kafkaesque nightmare of suffocation,
blindness, and chaos, and stumbled into the streets, joining human
stampedes fleeing the gas cloud. In the confusion, some ran in the
direction of the wind and died along the way. Many others died later
in hospitals. Thousands more were injured due to exposure to the
toxic gases. In all, over 2,200 people died, and more than 200,000
were injured. Thus, "Bhopal" came to be associated with the world's
worst industrial disaster involving toxic chemicals.
TOPOGRAPHY AND METEOROLOGY
Bhopal was an unlikely setting for this tragedy. A city of about 1
million people, 600 km south of New Delhi, Bhopal (lat. 23°17'N) is the
capital of Madhya Pradesh, India's largest state, with an agriculture-
based economy. Because of its central location, resources, and hospi-
tality, the state traditionally attracted people from all parts of the
country to work in its 8000 industrial plants of various sizes. Bhopal
was once a center of India's mogul past, and the city still reflects
this history despite a skyline of modern state office buildings,
colleges and hospitals built around two ("Upper" and "Lower") lakes.
The UCIL pesticide plant is located in the northern parts of the
city (Figure 1). North of this facility, there are hillocks, woods,
and stretches of residential areas extending out to 24 km from the
plant. The Indian Agricultural Research Insitute is situated about
4 km north of the UCIL factory. The latter is surrounded by residen-
tial areas on the other three sides. The approximate distances and •
locations, with respect to (w.r.t.) UCIL, of some Bhopal landmarks
shown in Fig. 1 are as follows: Electrical and Mechanical Engineers
(EME) Center (1.5 km H), airport (6 km N), Hamidia Hospital (3 km SW),
Sultania Hospital (3 km S), the State Secretariat (Vidhan Sabha) and
Police Headquarters (PHQ) (4 km S), and the Railway Station (1.5 km
SB). An industrial and warehouse area extends out to 15-20 km on the
plains ESE of the plant along the railroad. {
Table 1 shows the elevations and locations (w.r.t. UCIL) of some
of the important topographical features of Bhopal. Some of these
features could have influenced the local wind and dispersion patterns
on the night of the accident.
Detailed meteorological data are not available regarding the wind
speed and direction, and the vertical temperature structure at the
site on the night of the accident. According to Ekalavya (1985), a
private non-profit organization based in Bhopal, the winds that night
were initially from the northwest and subsequently from the north.
Post-episodic wind data indicate that the most probable wind speeds
were of the order of 3 m/s. The temperature was reported to be 14°C.
A-4
-------
Table 1. Elevations and Locations of Bhopal's Topographical Features
Elevation (m) Distance (km) & Direction
above MSL (w.r.t. UCIL)
Singarcholi Hillock
Idgah Hill
Ahnadabad Hill
Dharampuri Hill
Birla Temple Hillock
Shahpur Hill
Upper Lake
Lower Lake
625
575
513
601
562
525
-
-
4.5
2.0
2.5
5,5
4.5
8.0
4.0
4.0
W
NSW
WSW
SW
SSE
SE
SW
S
This is consistent with the minimum winter temperature of about 10°C
in Delhi. Since both cities are located in the Indo-Gangetic plains,
a mixing height (H) of 200 m, close to its climatological-mean value
in Delhi (Kumari, 1985) for a December night, was assumed for Bhopal.
The gas-affected areas in Bhopal are characterized by a gently
rolling terrain surrounded by hillocks and lakes. It was likely that
nocturnal drainage winds and land breezes altered the local surface
wind patterns on the night of the gas leak. In the absence of data,
the best way to decide on the wind direction is by charting the course
of the catastrophic imprint of the poisonous gases over the city.
Interviews with survivors and rescue personnel confirmed that the wind
direction varied during the night. Based on this information, Singh
and Ghosh (1985) adopted the following sequence of wind directions,
apportioned appropriately for the duration of the episode: (i) 285°
for 45 min, (ii) 340° for 30 min, and (iii) 45° for 15 min. This
scheme gave realistic model simulation results, as described later.
THE ACCIDENT AND ITS AFTERMATH
The UCIL factory in Bhopal employed 950 people to manufacture and
market Sevin (carbaryl) and Temik (aldicarb), two MIC-based pesticides.
The accident began on the night of Sunday, December 2, 1984, when about
75 workers were on duty. As reported in the biweekly India Today.
which investigated the accident and reconstructed it with approximate
timings, the sequence of events (in local time) is as follows:
10:45 pm - A shift change occurred at the factory. Unknown to the
incoming shift, a runaway chemical reaction was already under way in
storage tank 610, which contained 41 metric tons of liquid NIC.
11:00 pm - Pressure in tank 610 was observed to increase from 3 to 10
psi. The new staff mistakenly thought this was due to nitrogen pres-
surization of the tank by the previous shift.
11:30 pm - Operating staff noted eye irritation but ignored it, because
tiny leaks of NIC in the utility area were not unusual.
00:15 am - MIC control room staff reported the high pressure (30 psi)
in tank 610 to the Production Assistant; he checked and found that the
tank's rupture disk (designed to rupture at 40 psi) had burst, and the
safety valve had popped.
01:00 am - Untreated MIC vapor was seen escaping through the 33 m
high atmospheric vent line. Within a period of 90 min, about 40 tons
of MIC leaked out in gaseous form.
A-5
-------
The escaping gases overwhelmed the scrubber designed to detoxify
process vent gases. Other elaborate safety systems that failed
included the flare tower, the refrigeration system, and the water
curtain (which was designed to shoot water to a height of only 15 m),
and some of the other systems which had been out of operation for
repairs. The public warning siren was turned on for a few minutes
around 1:00 am, but was not sounded again until 2:00 am. The city
police control room was alerted. Just after 1:00 am, by a town
inspector on patrol near the factory. The police, in turn, alerted
the area hospitals, the army, and the state medical director at 1:30
am.
The poisonous gases emerging from the vent stack formed a huge
white cloud that moved south towards the densely populated areas. In
the vicinity of the plant, according to survivors' reports, the gas
was so thick that visibility was very low. More than 200 people died
in their beds in the area immediately south of the plant. Others
awoke to a feeling of suffocation and burning eyes, and rushed out of
their homes In panic. In the narrow streets, they joined tens of
thousands of half-blinded people running for their lives. It must
have been a terrifying spectacle. Hundreds were overcome as they
fled. Many collapsed on the roads, only to be removed to the
hospitals later. The cloud engulfed Bhopal's railway station as it
spread southeast from the plant. Passengers and station personnel
collapsed. As humans died, so did animals. The carcasses of dogs,
cats, chickens, birds, goats, cows, and other animals were later found
strewn over an area of 65 km2 in the quadrant SE of the plant.
Starting about 2:00 am, the first victims began arriving at
Bhopal's hospitals. Buses, army trucks, ambulances, and private
vehicles were pressed into service. All seven of the city's hospitals
were beseiged by thousands of gas victims. Doctors, joined by
interns, nurses, and medical students, began a seemingly unending
battle to save lives. As the hospitals became jammed, doctors, relief
agencies, and private groups set up emergency clinics on sidewalks, in
stores, and wherever they could find space. Bhopal's medical
community rose to the occasion to handle the emergency resulting from
the unprecedented disaster and performed heroically in saving many
lives.
A house-to-house search of the affected areas turned up hundreds
of victims. By late afternoon, the city morgue was overflowing with
bodies, many of them unidentified. There were mass burials and
cremations for the dead. As dusk fell the death toll mounted to over
1000 and the number of injured rose to more than 100,000. As the
magnitude of the tragedy became apparent, additional medical
personnel, equipment, and relief supplies were rushed to the city.
A-6
-------
CHEMISTRY AND TOXICOLOGY
Methyl isocyanate (CH3N=C=0) is a reactive, toxic, volatile and
flammable compound which is commercially used to make carbamate pesti-
cides. MIC, an ester of isocyanic acid, is industrially produced by
the reaction of phosgene (COC12) with the base methylamine (CH3NH2):
COC12 + CH3NH2 > CH3N=C=0 + 2HC1 (1)
It should be noted that MIC is stored with a purity of 99.5%, and
contains up to 0.1% phosgene to inhibit MIC polymerizations and
reaction with water. Thus, 40 tons of MIC (in tank 610) could also have
contained up to 40 kg of phosgene.
MIC undergoes exothermic and vigorous reactions with a
variety of compounds containing active hydrogen atoms. The cumulative
effect of the adjacent double bonds adds to the instability of the
compound. From a practical point of view, only a limited number of
those reactions could conceivably have taken place in the plant at
Bhopal (depending on the presence of specific reactants during the
episode). These possible reactions are shown in Table 2. Post-
episodic analysis of sludge samples from tank 610 showed that two
major reactions took place in the tank — that of MIC reacting with
water, and that of MIC reacting with Itself, catalyzed by iron, which
is a much faster reaction. More than 50% of the sludge was found to
consist of MIC trimer resulting from the latter reaction, which
liberates about 1.26 x 106 joules of heat per kg of MIC. About 16 to
28% of the sludge was made up of compounds produced by MIC reacting
with water, which starts off slowly at room temperature, producing
about 1.36 x 106 joules of heat per kg of MIC. Just a small amount of
water with a trace of contaminants such as rust and salt, for example,
could catalyze an autoreaction. The heat liberated during an induction
period lasting a few hours could generate a reaction of explosive
violence. It is not hard to imagine situations in which such trace
contamination could occur, despite best efforts at prevention.
Though the initial mix of poisonous gases that escaped from the
factory into Bhopal's air primarily consisted of MIC. it apparently
also included smaller amounts of phosgene, hydrogen cyanide, methyl-
amine, and other gases such as carbon monoxide. Though officials
initially denied any possibility of cyanide presence, scientists
detected cyanide near the MIC storage tank, and 50 m downwind, three
days after the accident. Further, autopsies confirmed symptoms of
cyanide poisoning among some of the victims. During the runaway
reaction in tank 610, it is possible that temperatures exceeding 200°C
were attained. At these high temperatures, MIC can break down into
hydrogen cyanide, carbon monoxide, »ono-methylamine, and some organic
cyanides. The physico-chemical, and toxicological properties of
various poisonous gases in the Bhopal tragedy are given in Table 3.
The behavior and fate of a MIC plume in the environment is schema-
tically shown in Figure 2. In the atmosphere, MIC reacts with
moisture to yield methylamine (MA) and C02:
CH3N-C»0 + H20 > CH3NH2 + C02 (2)
A-7
-------
Table 2. Some Important Chemical Reactions of MIC.
>
oo
1 . Reaction with alcohols and phenols
o
ROH
ROCNHCHg
An N-methylcarbamate
2. Reaction with water
(a)CH3N = C = O + H2O (excess)
(b) CH3N = C a O (excess) + H2O
3. Reaction with itself
catalyst
3CH3 N = C = O
CH3NHCNHCH3 + C02
1,3-Dimethylurea
O CH3 O
II I II
CH3NHC-N —CNHCH3 H
1,3,5-Trimethylbiuret
CO
I
CH3
Trimethyl isocyanurale
-------
Table 3. Physico-chemical and Toxicological Properties of NIC and
Other Poisonous Gases In the Bhopal Tragedy.
Chemical
formula
Description
Nol. Height
Boiling
Point (°C)
MIC
CH3N=C=0
colorless
liquid with
powerful
odor
57.05
39.1
Phosgene
coci2
colorless
gas or
volatile
liquid
98.92
7.6
Hydrogen
Cyanide
HCN
colorless
liquid with
faint odor
of bitter
almonds
27.04
25.7
Methylamine
CH3NH2
colorless
liquid or
gas with
strong odor
of ammonia
31.06
-6.3
Liquid Density 0.96
(w.r.t. water)
(i 200/4eC)
1.38
0.69
0.66
Vapor Pressure
(mm of Hg)
Vapor Density
(w.r.t. air)
LC50 (ppm)
(rats)
MSO
(humans)
TLV-TWA
(ppm)
References r Sax
348
(§ 20°C)
2.2
5
(4 hr)
>20*
ppm
0.02
(1979), Weast
1180
(§ 20°C)
3.4
1482
(1 min)
3200
mg/m3
0.1
(1982).
400
(@ 9.8°C)
0.93
544
(5 min)
200**
mg/m3
10
1.07
-
-
10
LC50: The concentration lethal to 50k of a specified population.
TLV-TWA: Threshold Limit Value-Time Weighted Average.
(The average concentration for an 8-hour workday to which
workers may be repeatedly exposed without adverse effect).
* Estimated as 1000 times TLV-TWA value.
** Lowest published lethal concentration for a 10 min inhalation.
A-9
-------
Similarly, phosgene reacts with moisture to yield hydrogen chloride
and C02. In the presence of sunlight, MIC degrades into polymers and
gases such as CO and N2. In soil and water, methylamine is the major
product. It is held tenaciously by soil particles until decomposition
is accomplished by weathering and biological reactions. In plants,
MIC could have competed with C02 in the process of photosynthesis. It
could have been transformed into methylamine and other carbamylated
derivatives of the plant constituents.
One of the disturbing aspects of the chemical disaster at Bhopal
is that scientists know relatively little about the toxic effects,
especially the long term ones, of MIC. Good toxicological data on MIC
were non-existent in the open literature. The permissible MIC skin
exposure limit is 0.02 ppm averaged over 8 hrs. This threshold limit
value (TLV) has been set by the American Conference of Governmental
Industrial Hygienists (ACGIH). It represents the time weighted
average concentration (for a normal 8 hour day or 40 hour work week)
to which nearly all workers may be repeatedly exposed, day after day,
without adverse effect. Isocyanates long have been known to attack
the respiratory system, eyes, and skin. They can injure the lungs and
bronchial airways, and cause permanent eye damage. But until the
Bhopal episode, the effects of MIC had never been observed on such a
large and diverse population. Most of the deaths have been attributed
to various forms of respiratory distress.
Indian scientists have initiated a variety of studies to inves-
tigate the long-term consequences of exposure to MIC. Many of these
research efforts are focused on damage to specific organs, such as the
eye or lung. Others address possible genetic and carcinogenic
effects. The value of all these clinical and laboratory studies,
which may take up to 3 to 5 years, will largely depend on building a
data base of high-quality epidemiological information. For this
purpose, researchers need to answer at least three key inter-connected
questions:
(i) Where were the victims when they were exposed to the gas?
(ii) What dose did the victims encounter?
(iii) What were the health effects from exposure to the gas?
The most difficult part of this investigation will be to estimate the
concentrations of MIC that individuals encountered. One indirect way
to estimate the MIC concentrations is by charting the degree of damage
to sensitive trees and plants, in different parts of the city; this
will enable one to determine the level of MIC to which people in those
areas were exposed. Scientists from the Indian Agricultural Research
Institute and the Central Board for Prevention and Control of
Pollution studied the trees and crops in Bhopal immediately after the
disaster.
Another indirect way to estimate the concentrations is through the
use of a suitably formulated atmospheric dispersion model. Singh and
Ghosh (1985, 1987) have developed a simple analytical dispersion model
including the aqueous phase transformation and deposition of MIC, as
described below, for use in preliminary assessments.
A-10
-------
DISPERSION MODEL, RESULTS, AND CORRELATION WITH OBSERVED EFFECTS
The density of MIC vapor is about twice that of air. This
suggests that density effects might be important for plume dispersion.
However, the velocity of ejection of the gas was very high and, hence,
there was probably considerable entrainment and mixing. Exothermicity
of the MIC-moisture reaction suggests that a great deal of heat was
generated in the process, and this could further reduce the density of
the plume. As a first approximation, therefore, the density effects
were not explicitly considered in the model. The plume dispersion
was assumed to consist of spreading due to passive diffusion as well
as spreading due to gravitational effects; this would imply larger
a values than in the case of a purely passive diffusion. Singh and
Ghosh (1985) found that Briggs1 (1973) urban dispersion parameters for
stable conditions (E-F) give concentration estimates that were
consistent with the observations. Other assumptions used in their
model formulation were as follows: (i) the time-dependent emissions
are specified by a uniform emission rate over the duration of the
episode, (ii) the terrain is flat and spatially homogeneous, and
(ill) removal by dry deposition and chemical transformation is a
first-order irreversible process.
The time-dependent three-dimensional atmospheric advection-
diffusion equation is given by
3c + uac = D a!c + D afc_kc
at 3x y 3y2 z 8z2 *
where C(x,y,z,t) is the pollutant (MIC) concentration, U is the
constant average wind speed along the x-direction, k^ is the total
removal rate, and Dy and Dz are the constant eddy diffusivities in the
crosswind and vertical directions, respectively. The source condition
is described by
C(0,y,z,t) = p 5(z-hs) 6(y) (4)
i
Q , 0 < t < t
where Q(t) ' ° = e
0
In Eq. (4), Qo is the MIC emission rate, hs is the effective release
height, and te = 5400 s is the duration of the release. The initial
and boundary conditions are given by
C(x, y, z, 0) =0
' z=0,H
C(x ,± oo, z, t) = 0
A-ll
-------
The total removal rate (kt) of MIC is given by
kt = kc * kd
where kc is the chemical reaction rate and k
-------
mated by the power-law profile from the wind speed (3 m/s) at 10 m
height, using an exponent value of p = 0.35 for stable conditions.
The wind direction and mixing height H were specified as discussed in
Section 2. The spreads, ay and ffz, were given by Briggs1(1973) urban
curves for stability classes E-F. The stack inside diameter was 0.2 m,
and the estimated gas exit-velocity was 88 m/s. The effective release
height hs was estimated to be about 40 m from simple plume rise
calculations (e.g., Holland, 1953) under stable conditions.
Figure 3 shows the variation of the ground-level concentrations
(GLC) as a function of downwind distance along the plume centerline
after 45 min of time integration. The model simulation results are
given in Table 4, which lists the total time-integrated GLC, surface
deposition fluxes, and the approximate times of initial impact (i.e.,
plume front arrival) at the various affected sites. These sites are
marked in Figure 4 which shows the concentration contours demarcating
various zones of the affected areas based on the estimated total GLC.
In general, the estimated GLC distribution and the plume arrival times
seem to correlate well with the extent of affliction at the various
sites and information obtained from interviews with survivors.
It is clear (from Fig. 4) that within zone I, encompassing sites
1-3. extremely high concentrations of the gas prevailed. The calcu-
lated maximum GLC were greater than 50 ppm. According to local
residents, these colonies were engulfed by copious dense white fumes
with a bitter sweet smell; visibility was very low. These were also
the worst affected areas, with total mortalities adding up to 360 (out
of a population of 6,173) by the time the smoky pall lifted at 5:30
am. The major medical symptoms in the survivors were severe eye irri-
tation, coughing, vomiting, and excruciating chest pain. Zone II,
which includes sites 4-6. was also a very badly affected zone. Here
the estimated GLC were above 15 ppm. About 508 casualties were
reported within this area (the toll was higher in this zone since it
includes many more colonies with a much larger population than zone
I). About 87% of the initially recorded casualties were within these
two zones. Model results (Table 4) suggest that the plume front
arrived at these sites within the first 6 min of the episode.
Zone III, the "zone of moderate effects", encompassed sites 7-11,
14, 22, and 23. The immediate fatalities were 11, much lower than
in zones I and II. People experienced severe eye irritation and
breathing difficulties. The estimated gas concentrations in this
zone were between 15 and 1.5 ppm. Zone IV, encompassing sites 12, 13,
15-21, and 24-26, is the "zone of marginal effects" with only five
fatalities. The estimated concentrations here were less than I ppm.
Some residents experienced slight eye and throat irritation, though no
traces of the fumes were visible. As described earlier, many people
fled their homes during the episode and, hence, there is an inherent
subjectivity involved in this correlation study. However, many of the
casualties occured close to their normal place of residence. The
effects of the gas faded significantly with increasing downwind
distance. Though humans and animals inhabiting the region extending
out to 6 km downwind of the plant were affected by the gas, the
heaviest toll occured within the first 200 m to 1 km distance.
A-13
-------
Table 4. Model Estimates of GLC and Deposition Fluxes of NIC.
Site GLC Surface Deposition Time of Initial
(ppm) Flux (|*g/m2/s) Impact (min)
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
26
93.5
74.2
57.4
25.9
18.5
15.5
8.0
3.4
2.5
8.2
8.6
0.1
0.2
1.7
0.9
1.0
0.4
1.0
0.3
0.6
0.2
3.4
1.8
1.0
0.8
0.7
4133
3280
2537
1145
818
685
354
150
111
362
380
5
9
76
39
43
18
44
13
26
9
150
80
44
35
30
1.2
1.6
1.6
3.9
5.4
5.8
8.6
10.9
13.7
49.7
50.4
51.6
57.4
57.4
58.6
60.1
61.3
61.7
61.7
63.6
63.6
78.9
80.8
82.8
82.8
85.2
This agrees with the calculated downwind variation of the GLC (see
Fig. 3). The total of early fatalities listed above is rather low by
present estimates, since many of the deaths were not recorded.
The best concentration monitors available in the Bhopal gas
disaster were the affected trees, plants, crops, and soil, which were
studied by Indian scientific organizations. The imprint of the toxic
gas could be discerned on all types of vegetation including tall trees,
shrubs, and ground-level crops. Vegetable and spice crops were most
affected in the eastern, southeastern, and southern directions from
the factory. Some vegetation looked as if it had been burned. It was
observed that the same species of plants, which were badly damaged by
the gas in other areas, were mostly unaffected near the water. Also,
the same species of crops showed relatively less damage in fields
which were irrigated on the evening preceding the accident. This
could be due to the great affinity of NIC to water and its rapid
hydrolysis resulting from the increased moisture content in the soil,
vegetation, and surface air. Some of the large trees were totally
defoliated, while others showed acute to mild effects, depending on
the species and location relative to the factory. Table 5 shows the
extent of the vegetation damage in terms of area. These vegetation
A-14
-------
Table 5. Classification of Vegetation Damage Due to Exposure to the
Toxic Gases.
Zone
A
B
C
D
Classification
Severely affected
Badly affected
Moderately affected
Mildly affected
Area
3.5
10
6
5
(km2)
damage zones, shown in Figure 5, indicate that trees and vegetation
within zones A, B and C could indeed be affected by exposure to the
spreading gas, with estimated concentrations ranging from very high to
moderate values. Zones A and B correspond reasonably well to zones I,
II, and III shown in Fig. 4. The vegetation-damage contours are also
tilted towards the SE, similar to the concentration contours.
MEDICAL CARE,*RELIEF, AND REHABILITATION
Inhalation of MIC, mixed initially with unknown amounts of other
poisonous gases (see Table 3) from the leak, is considered to have
resulted in the large number of deaths in humans and animals. The
initial symptoms displayed by the survivors were burning sensation in
eyes, nose and throat, breathing difficulty, coughing, vomiting,
headache, lethargy, and disorientation. The main target organs were
the eyes and the lungs. Lung damage.was the cause of death in most of
the casualties.
While vision and condition of the injured eyes have dramatically
improved following medical treatment, lung problems persist in many
victims. Pulmonary edema resulting from the exposure to MIC gas was
treated with steroids, bronchodilators, oxygen, and antibiotics; a
large number of patients have recovered. However, many of the sur-
vivors were left with considerable damage to their small airways and
alveoli. The worst affected among this group were children. About
39* of the patients showed moderate to severe pulmonary disability.
Those who display normal functional status may still have significant
lung pathology. Open lung biopsy performed in some of the severely
affected patients has led to the detection of bronchiotitis obliterans
and interstitial fibrosis. Attempts are being made to Identify
lesions in victims in the early stages, and treat them using drugs and
physiotherapy.
Miscarriages and stillbirths were frequent to the approximately
3000 pregnant women who were exposed to the gases. In the first 20
weeks, 436 spontaneous abortions occurred out of 2600 pregnancies; the
normal rate for Bhopal was 6 to 10%. Gynecological disorders in
various forms were reported in a large percentage of the women exposed
to the gases.
A-15
-------
Cyanide poisoning in the survivors resulted from direct exposure
to HCN gas, during the first hours in the areas closest to the plant,
as well as indirect exposure owing to unusual generation of toxic
amounts of cyanide in the body after exposure to NIC. Some experts
believe that the "cyanogen pool" resulting from the latter can persist
in the human body for a very long time. The Indian Council of Medical
Research (ICMR) studied this controversial issue, and recommended
continued treatment of symptomatic patients with sodium thlosulfate,
the antidote for cyanide poisoning, under careful medical supervision.
The state government has opened 16 dispensaries, a new 30-bed
hospital with modern equipment, and 25 mobile clinics in the affected
areas, in order to provide effective medical care to the Bhopal gas
victims. A 60-bed ward and MIC clinic, equipped with a computerized
blood gas analyzer, was established in the Medical College Hospital to
undertake various sophisticated investigations and develop the best
methods of treatment. In addition, 100 more beds were reserved
exclusively for the gas victims in other hospitals in Bhopal.
Disaster relief measures undertaken by the government include
financial assistance to affected families, cash grants to low income
group victims, free rations of essential commodities, and supplies of
nutritious food to pregnant women, lactating mothers, and infants. A
comprehensive rehabilitation plan consisting of housing and urban
development, vocational training and employment schemes, and
establishment of work centers and community facilities, is being
implemented in the affected areas. Many voluntary organizations are
supplementing the government efforts in medical care and
rehabilitation.
Numerous studies on the Bhopal gas disaster have been carried out
in India, and many others are still in progress at various government
laboratories, universities, medical colleges, and hospitals. The
biomedical and toxicological studies on the long-term effects of MIC
gas on the survivors are being coordinated by the ICMR. The Council
of Scientific and Industrial Research is coordinating the scientific
and technical projects. Various issues raised by the Bhopal tragedy
(such as development policy, Industrial hazards and public safety,
occupational health, environmental protection, legislation, and other
topics) are being studied and addressed by a number of government
agencies, and by scientific and public interest groups. The Central
Insecticides Board has banned the use of 25 highly toxic pesticides,
including the two MIC-based products, Sevin and Temik.
CONCLUSIONS
The most important legacy of the Bhopal tragedy is its enormous
worldwide impact in raising the level of public awareness and concern
over the hazards of chemical plants to the general public, especially
in those plants involving toxic chemicals. This will inevitably
generate change in safety procedures, equipment to prevent and contain
toxic releases, and the kind of Information given to the public about
chemical hazards. Most chemical companies have perceived the need to
reassure the public of the safety of their operations after Bhopal.
A-16
-------
The Bhopal disaster bears many important lessons of immediate
global relevance, which can be briefly summarized as follows:
(1) Information: Industries with potentially dangerous processes
should work together with state and local officials to maintain a
current understanding of what they are doing, giving details of the
toxic substances, potential hazards, toxicological effects, medical
treatment, and other pertinent information.
(2) Industrial siting and urban planning: Hazardous industries should
be located far away from population centers with due regard to poten-
tial impacts on the total environment. Existing chemical plants in
urban areas should be relocated, if possible, or isolated from dense
residential areas by a 2-3 km radius green buffer zone planted with
sensitive tree and vegetation species to act as chemical monitors.
(3) Standards and enforcement: Industrial safety, worker and community
health, and environmental regulatory standards appropriate for each
industry should be established and strictly enforced. Regulatory
agencies should carry out frequent inspections of hazardous plants to
examine the safety of processes and equipment, and to verify compliance
with regulations.
(4) Emergency response: This consists of several components.
The scientific program should address atmospheric behavior of toxic
chemicals including transport, dispersion, transformation, and
deposition, ultilizing appropriate models and monitoring. The medical
response should require that hospitals and physicians be knowledgeable
about the toxicology and treatment for exposure to locally used toxic
chemicals, and should be prepared to handle any emergency. The state
and local authorities, in cooperation with the concerned plant, should
develop effective communications, evacuation plans, and community
education programs.
(5) Management: The management of chemical companies at all levels,
in view of their responsibilities for industrial and public safety,
should play a leading and constructive role in most of the areas
listed above, especially in providing information and developing
appropriate emergency response programs. Procedures involving the
manufacture, storage, transportation, and disposal of toxic chemicals
should be periodically reviewed to minimize potential hazards to
public health and safety. Worker training and plant safety audit
programs should be intensified.
Bhopal holds out lessons that have to do with social
responsibilities of science and technology, and basic human concerns
for safety, for standards, and for good common sense. If these
lessons are lost, then it is all too possible that accidents such as
this may be repeated in another factory in another city in another
way. A fitting memorial to the victims of the Bhopal Gas Tragedy
would be to make sure that this shall not happen again.
A-17
-------
ACKNOWLEDGEMENTS
This work was performed under an agreement between the National
Oceanic and Atmospheric Administration and the U.S. Department of
Energy. The literature search was accomplished while the first author
was visiting the Center for Atmospheric Sciences, New Delhi, during
the summer of 1985 under the ATDD/IIT Collaborative Research Program
in Air Pollution, which is part of a multi-institutional project
supported by the U. S. National Science Foundation. Many Individuals
and publications including technical books and reports, scientific
Journals, news papers and magazines were consulted in the course of
this work. It is impossible to list all of these articles and their
authors; only the key references and general sources are given here.
DISCLAIMER
This paper has been reviewed by the Air Resources Laboratory
(ARL), National Oceanic and Atmospheric Administration (NOAA), and
approved for publication. Approval does not signify that the contents
necessarily reflect the views and policies of NOAA or any other U.S.
Government agency, nor does mention of trade names or commercial
products constitute endorsement or recommendation for use by NOAA/ARL.
Use of information from this publication for publicity and advertising
purposes is not authorized.
A-18
-------
REFERENCES
Briggs, G. A., 1973: Diffusion estimation for small emissions. ATDL
Annual Report, NOAA, Oak Ridge, TN, 83-146.
Ekalavya, 1985: Bhopal - A reconstruction of the gas tragedy, its
background and aftermath, from press reports and local informa-
tion. Ekalavya Report-El/208, Arera Colony, Bhopal.
Hicks, B. B., 1985: Personal communication.
Holland, J. Z., 1953: USAEC Report ORO-99, Washington, D.C., 584 pp.
Kumari, M., 1985: Diurnal variation of mean mixing depths in different
months at Delhi. Mausam 36, 71-74.
Reid, R. C,, and T. K. Sherwood, 1958: The Properties of Gases and
Liquids. McGraw Hill Book Company, Inc., New York, NY.
Sax, N. I., 1979: Dangerous Properties of Industrial Materials. Van
Nostrand Reinhold Co., New York, NY, 1118 pp.
Singh, M. P., and S. Ghosh, 1985: Perspectives in air pollution model-
ing with a special reference to the Bhopal Gas Tragedy. CAS Report,
Indian Institute of Technology, New Delhi.
Singh, H. P., and S. Ghosh, 1987: Bhopal Gas Tragedy - Model
simulation of the diffusion scenario. Accepted for publication in
J. Hazardous Materials.
Schwartz, S. E., 1984: Acid Precipitation Series. Vol. 3, Chapter 4.
J. I. Teasley, Series Ed., Butterworth Publishers, Stoneham, MA.
Weast, R. C., 1982: CRC Handbook of Chemistry and Physics. 63rd Edi-
tion, CRC Press, Inc., Boca Raton, PL.
MAGAZINES AND NEWS PAPERS (Unreferenced)
Chemical and Engineering News, Washington, D.C., February 11, April 8,
and December 2, 1985.
India Today, New Delhi, December 31, 1984.
The New York Times, New York, December 10, 1984.
The Illustrated Weekly of India, Bombay, September 1, and September 15,
1985.
The Statesman, New Delhi, March 28 and May 4, 1985.
The Sunday, New Delhi, April 13, 1985.
The Times of India, Bombay, December 16, 1984, March 25 and April 17,
1985.
A-19
-------
Bhopal
HAMIDlAv'.. .'i
: '-"SULTANIA
^-="== LOWER
ZFLAKE:
::::--..::::::....:::::—:. HAM ID IA V
""—^COLLEGE
MINISTERS'
BUNGALOWS
SCALE
Kilometers
Figure 1. A partial Map of Bhopal showing the location of the DCIL
pesticide plant and its environs. The dotted area shows the
localities affected by the toxic gases.
A-20
-------
ATDL-M86/624
'ro
MOISTURE
ADVECTION BY - .-:,^-w:i»rJl^ •• CHEMICAL •...•:.; - -. .
^"* ^^ ^^ * ***** *"••*• * "• • * ^^^^^^jBH^^ i™ £3 AM O r***/^ Q ft M A ~r" i /*x K i "
WIND; BENT-0/ER >LUMEy;.;'^J//.'.^^^nMATION;:.;.. -
DEPOSITION
WATER,
Soil and water:
MIC
Adsorption
in soil surface
Humans and Animals +• MA (major metabolite)
Figure 2. A »che»atic diagraa ahcmine the atmospheric behavior and fate of
MIC plu»e in the environment. Methylanine (MA) is the major
product of MIC reactions in toil and water.
-------
ATDL-M 86/622
ro
ro
E
ex
ex
Z
60
40
cc
£ 20
UJ
o
Z
o
o
0
200
(a]
400 600 800
x (m)
1000
30 i r
Figure 3.
-
45 .in of ti«e integration: (a) near field, (b) far field.
-------
Bhopal
FIRDAUS
NAGAR
E.M.E. CENTER
CHOLA
KENCHI
SHAHAJAHA^
NABAD _ __
/ 22
J.P. NAGAR
2
KAZI CAMP
ILWAY KHAJANCHI
COLONY BAGHV 1
6 I
II II /
CHANDBAD/
RAILWAY-
STATION' *
SINDHI
COLONY
STRAW
PRODUCT
OLD
SECRETARIAT J
BUS
• STAND
HAMtOIA
HOSPITAL
AISH BAG
STADIUM
SULTANIA
HOSPITAL
.CENTRAL
IL1BRARY
JAMA 11
MASJID
~~~~~~ ~~LAKE I'.'."."""
JAHANGIRAi?
BAD
..- _~ LOWER r
"LAKE
::~~HAMIDIA
--—COLLEGE
CENTRAL
SCHOOLX
LAL PARADE
GROUNDS
CHAR
BUNGALOW
BIRLA X
MANDIR
LEGEND
ZONE I > 50ppm
ZONE II > 15ppm
ZONE III > I.Sppm
ZONE IV < 1 ppm
MINISTERS'
BUNGALOWS
SCALE
Kilometers
Figure 4. Modeled concentration contours demarcating the various zones of
the gas affected areas based on the estimated 6LC. The numbers
1-26 denote the localities in Bhopal affected by the gas leak,
where the 6LC are specifically calculated and given in Table 4.
A-23
-------
Bhopal
™^ LOWER_=
""=_" LAKE:
MINISTERS-
BUNGALOWS
LEGEND
SEVERELY AFFECTED (3.5 km2) °
BADLY AFFECTED (10 km2)
MODERATELY AFFECTED (
MILDLY AFFECTED (5 km2)
SCALE
Kilometers
Figure 5. Vegetation danage contours demarcating the various zones of
injury to plants by the toxic gases. These contours are baaed
on a detailed survey of vegetation and crops in Bhopal conducted
by the Indian Agricultural Research Institute and the Central
Board for Prevention and Control of Pollution, following the gas
leak.
A-24
-------
APPENDIX B
On the Development of
Reactive, Dense Gas Models
B-l
-------
On the Development of
Reactive, Dense Gas Models*
B. B. Hicks
and
W. R. Pendergrass
NOAA, Atmospheric Turbulence
and Diffusion Division
P. 0. Box 2456, Oak Ridge, TN 37831
ATDD Contribution No. 86/26
March 31, 1987
ODR/16
* For presentation at the Joint EPA/DOE Technical Workshop on
Determination of Atmospheric Dilution for Emergency Preparedness, Research
Triangle Park, NC, 27711, and for inclusion in the proceedings.
B-3
-------
ABSTRACT
Chemical reactions between released pollutants and atmospheric
constituents can influence atmospheric dispersion if the heats of reaction
are such that buoyancy can be affected or mechanical turbulence damped.
If the reaction is sufficiently exothermic, plume rise can be increased;
if sufficiently endothermic, then plume rise can be suppressed. It is
necessary to consider the potential influence of such reactions on the
mixing rates characteristic of the ambient air. Reaction rates are also a
critical issue.. If reactions are completed rapidly, then in concept it is
possible to assimilate the consequences of the reactions in the source
term initialization or in the first time step of a numerical model. On
the other hand, if reaction rates are slow in comparison with the time
step of a simple model, then a more complicated and detailed modeling
effort may be required.
B-4
-------
1. INTRODUCTION
The wide variety of needs for numerical models of atmospheric
dispersion imposes requirements for models of a wide range of
complexities. It is clear that simple models giving conservative
estimates are most suitable for purposes of real-time emergency response,
whereas more sophisticated models can be used for assessment of risk, for
planning response strategies, or for evaluation of the consequences of
previous accidents. The range of modeling capabilities must also be
adequate to.encompass the spectrum of circumstances to be addressed,
involving terrain complexity, different meteorological conditions, and
source configurations.
In general, the complexity of the model best suited in some special
application is increased by increasing complexity of the circumstance and
limited by the time interval available' in which to run the model.
Considerable additional complexity can arise if the pollutant involved is
dense, or if it reacts chemically after emission. The case of dense gas
dispersion has received extensive attention, largely as a result of
problems associated with the transport and storage of liquified gaseous
fuels. In essence, the density interface is a stabilizing influence,
tending to isolate the underlying "bubble" of dense gas from the
turbulence of the ambient atmosphere. The volume of dense gas itself can
then be free (if dense enough) to respond to its own dynamical forces,
largely independent of the atmosphere passing over it. Spreading may be
rapid, but dilution may be correspondingly slow.
Endothermic chemical reactions impose a further set of complications,
that can serve to make a pool of emitted trace gas act as if it were a
B-5
-------
genuine dense gas, even though quite dilute. On the other hand, some
chemical reactions are exothermic, and will tend to increase vertical
mixing between the trace gas and the ambient air.
The present purpose is to explore the options available to modelers
faced with the need to consider the consequences of exothermic and
endothermic reactions. We start with the assumption that a suitable
modeling framework is available, including dense-gas formulations
appropriate to the case under consideration, with which chemically-induced
dynamic factors must then be combined.
2. SOME THEORETICAL CONSIDERATIONS
Consider a trace gas released into the air and reacting with some
atmospheric constituent, such as atmospheric water vapor. The chemical
reaction is then of the general type
kc
K! + H20 > X2 + X3 -i- J (1)
where species X^ reacts with H20 (in this example) to form species X2 and
X3 with rate constant kc (s~l) and releasing thermal energy J (Joules
mole"1). If J is positive, then the reaction enhances the mixing due to
buoyancy. In this case, it is informative to consider the role of the"
additional thermal energy in the context of related features of the
ambient air.
In a convective atmosphere with sensible heat flux H, the rate of
generation of turbulent energy associated with the heat flux is
HB = (H/cp).(g/e)
where cp is the specific heat of air at constant pressure, g is the
B-6
-------
acceleration due to gravity, and 0 is potential temperature (absolute).
The equivalent mechanical term is
HM = -T«(3u/3z), (3)
where T is the surface momentum flux and 3u/3z is the local wind gradient
at the height in question. The negative sign is a consequence of the sign
convention, with positive T being directed away from the surface and 8u/3z
positive when wind increases with height.
In classical micrometeorology, the importance of buoyancy relative to
shear-produced mechanical diffusion is quantified by the ratio of (2) to
(3):
C = -(gH/cp0)/(T-3u/3z)
<|>m-l
(4)
Here, standard micrometeorological relations have been invoked to relate
the local wind gradient and shear stress to the friction velocity, u,f
involving air density p and height z, and introducing the von Karman
constant k and the stability-dependent dimensionless wind shear m. The
quantity L is the Monin-Obukhov length scale of turbulence, initially
derived from dimensional arguments. The quantity C is thus an index
of dynamic instability associated with buoyancy, much like the familiar
index z/L.
In this context, it is useful to consider a second buoyancy term,
like !IB( DUt representing the consequences of an exothermic chemical
reaction.
B-7
-------
nc = kcJpv/Mv
where pv is the partial density of the species Xj in air (kg m~3) and Mv
(kg mole"1) is the molecular weight of the species Xj.
It is then apparent that if the ratio
CB - nc/nB
- kcJpvcpe/(HgMv) (6)
is large (with respect to unity, then the reaction is sufficiently
exothermic to modify buoyancy significantly, in unstable conditions.
In unstable stratification, an endothermic reaction will tend to
reduce the rate of buoyant mixing. In general, therefore, a modified
index of the net effective instability can be postulated:
-?' = (HB + nc)/HM (7)
where nc is positive for an exothermic reaction (positive J) and negative
for endothermic.
Although the modified stability index C1 has been developed here for
unstable conditions, its generality is not so constrained. Just as z/L
and Ri are stability indices that extend across the range of stable and
unstable stratification, so does C' provide an equivalent mechanism for
modifying such standard quantities in the cases of exothermic and
endothermic chemical reactions.
Equations (6) and (7) provide a basis for "screening" chemical
reactions for potential concerns related to modifying atmospheric mixing.
Consideration of dense-gas effects can be included in the same general
framework. In this case, consideration of relevant time scales provides
useful physical insight. At a density interface, characterized by a
B-8
-------
difference Ap, the restoring force (per unit volume) associated with the
displacement of a unit volume of denser fluid into the (upper) less-dense
medium is
Fr = g.Ap (8)
A time scale (T
-------
density variations must be used to correct ambient stratification
calculations when considering local stability= Such matters have been
dealt with extensively elsewhere.
3. DISCUSSION
Inspection of the relations involving J reveals a few hidden
difficulties. In particular, the chemical reaction rate kc is not usually
a constant, but depends on variables which might include temperature,
pressure, and solar radiation. Furthermore, the reaction may be
equivalent to a gas-phase titration, in which the reaction is controlled
by the rate of delivery of one gaseous reactant or the other. Thus, from
the present viewpoint the specification of kc is far from trivial.
Inspection of equations (6) and (7) reveals several intriguing
conclusions.
(a) Plume rise enhancement/suppression
• The practical effect of the release of heat of reaction is
likely to be greatest in near-neutral conditions (when H is
small). The diurnal cycle of H is large, such that H varies
—2 -2
from -10 to -20 W m at night to more than 200 W m at midday,
typically. The influence of J is therefore likely to be
greatest near dawn and dusk.
(b) Mechanical mixing with the ambient air
• The suppression of dilution by an endothermic reaction is a
strong function of the friction velocity, suggesting an
inverse cubic dependence on wind speed.
B-10
-------
• The roughness of the surrounding surface is important, insofar
as it controls u,,, in given wind speed conditions. Minimum
boundary dilution rates will occur in light winds, over smooth
terrain.
(c) The role of the reaction rate
• The reaction rate enters as a first-order factor in both
CB and CM- The appropriate reaction rate is the effective
value in the conditions of interest, which will generally be
less than the rate based on chemical consideration alone. The
problem that arises is associated with turbulent mixing;
chemical reaction consumes material available in any specific
volume, and hence the rate of resupply of reactions must be
considered. Thus, if the rates kc in equations (6) and (7)
are based on chemistry alone, then the properties Cn and
o
C« should be considered as indices of the potential
M
importance of the heat of reaction, rather than an indication
of actual importance in any specific instance.
4. A PRACTICAL APPROACH
The need ,to consider the special characteristics of reactive gases in
dispersion models leads immediately to two fundamental questions:
(a) Does the reaction cause the dynamic behavior of the atmosphere to
be modified?
(b) Is the reaction completed fast enough that the consequences can
be accommodated in the source term of a relatively standard
dispersion code, or in its first time step?
B-ll
-------
The answers to those questions will determine the complexity of the
model which must then be used. In some instances, such as when a reaction
is strongly exothermic yet quite slow, a complicated model is likely to be
required in all situations, and the need for rapid computation as for
real-time emergency response are unlikely to be realized.
An objective approach is to use a detailed simulation of the chemistry
to derive the necessary answers concerning reaction rates and
exothermicity in natural conditions. If the results of these explorations
are satisfactory, then perhaps simpler models can be designed without loss
of either generality or applicability.
This general philosophy is presently being tested in the development of
a model of dispersion of uranium hexafluoride. This material is used in
uranium processing and isotopic enrichment. Upon release into the
atmosphere, it reacts exothermically with water vapor, generating
particulate uranyl fluoride and gaseous hydrogen fluoride. The amount of
heat released and the rapidity of the reaction are clearly dependent on
ambient humidity. In this instance, consideration of scaling properties
such as were discussed above is not completely satisfying, since factors
not included in the development of the scaling properties might be critical.
The chemical reaction is as follows:
UFg + 2H20 > U02F2 + 4HF + J. (14)
The quantity J is a function of temperature and pressure, but in typical
conditions involving gaseous UFQ and water vapor, J * 5.86 x 104 kJ/kg
based on chemical considerations alone (see summary by Just et al.. 1985).
B-12
-------
A sophisticated numerical model of UF6 reactions in air was developed
earlier (Just, et al., 1985). This includes a detailed description of the
turbulent resupply of reactants as material is consumed (Varma. 1982), and
provides a convenient tool for investigating the rapidity of UF6 reactions
in air. Figures 1 and 2 illustrate the results. In Figure 1, the reaction
rate is seen to be short in comparison to the time step of typical
puff-dispersion models (usually about five minutes), except for very dry
conditions. Figure 2 shows that there is little need to consider
background atmospheric stability as a controlling factor, since in even
the most stable case considered the reaction appears to be completed in 30
seconds. Hence, in this particular case it appears adequate to assimilate
the chemistry and its consequences in the first time step of a standard
puff model, and to use the more detailed chemical model to initialize the
chemically-simpler scheme.
5. CONCLUSIONS
Two major questions arise when adding chemical reactions into an
existing dispersion scheme (either of possible trace gases or dense gas).
First, the rapidity of the reactions must be considered. If the
reactions are completed quickly in comparison with the time step of the
model under consideration, then the relevant chemical reactions can be
accommodated within the initial time step, or perhaps assimilated in the
specification of the source term.
Second, consideration must be given to the consequences of the heat of
reaction. If exothermic, the reaction is essentially a destabilizing
factor, working to increase plume rise and to enhance buoyant mixing (and
subsequent dilution rates).
13-13
-------
If endothermic, the reaction is essentially a stabilizing influence,
increasing local stratification, and hence decreasing mixing (and
dilution) rates. If all other factors are equal, a bubble of endothermic
material must maintain its integrity for a longer time than a bubble of
exothermic material.
An examination of these factors can be based either on first
principles, by consideration of the relative magnitudes of the dynamic
effects of the heat of reaction, or on the use of a detailed model. In
practice, if the reaction involves a trace gas constituent of the ambient
atmosphere (e.g., H20 or C02), then the use of a detailed chemical model
appears especially beneficial. In such cases, the reaction can proceed as
a gas-phase titration in which the rates involved and the thermodynamic
consequences are controlled by the rates at which reactant gases are fed to
the reacting volume.
B-14
-------
6. REFERENCES
Just, R. A. and W. R. Williams. 1985: A computer program for
simulating the atmospheric dispersion of UF6 and other
reactive gases having positive, neutral, or negative
buoyancy. Martin Marietta Energy Systems. Inc., Engineering
document No. K/D-5694. Oak Ridge, TN.
Varma, A. K. , 1982: Development of models for the analysis of
atmospheric Releases of Pressurized Uranium Hexafluoride,"
ARAP Report No. 482. Aeronautical Research Associates of
Princeton, Inc., Princeton, NJ.
B-15
-------
ATDL-M86/647
100
V)
10
i i i i LL
10 100
WATER MIXING RATIO (g/kg)
Figure 1. The humidity dependence of the exponential time scale of
reaction of UFg released into air, as computed by a sophisticated
chemistry/turbulence mixing model (Just and Williams, 1985). Time
scales correspond to (1-1/e) completion of the reaction.
B-16
-------
CO
CO
o
H
_J
CD
CO
0
•—•
1
ATDL-M86/648
T
•—•
10 20 30
TRAVEL TIME (s)
40
Figure 2 The dependence on stability class of the exponential
characteristic time scale, as plotted in Figure 1.
B-17
-------
APPENDIX C
Mathematical Models for
Atmospheric Dispersion of
Hazardous Chemical Gas Releases:
An Overview*
*Presented at the AIChE Center for Chemical Process Safety,
International Symposium on Preventing Major Chemical Accidents,
February 3-5, 1987, Washington, DC.
C-l
-------
MATHEMATICAL MODELS FOR
ATMOSPHERIC DISPERSION OF
HAZARDOUS CHEMICAL GAS RELEASES:
AN OVERVIEW
Jerry Havens
Department of Chemical Engineering
University of Arkansas
Fayetteville, Arkansas
Mathematical models are required for prediction of
atmospheric dispersion of heavier-than-air gases to
assess hazards of unconfined vapor cloud combustion
and toxicity. The dispersion process involves three
more-or-less distinct regimes of fluid flow:
buoyancy-dominated, stably stratified, and passive
dispersion. For releases on uniform terrain,
similarity models are available for the description
of all three regimes with sufficient accuracy "for
most risk assessment and emergency response
requirements. 3-D mathematical models can, in
principle, simulate the spatial and temporal
dispersion process without artificial separation of
the flow into separate regimes and may be able to
provide for effects of terrain and wake turbulence.
Evaluation of 3-D models is underway, but most of the.
work is biased strongly toward evaluation against
data for dispersion in the absence of terrain and
wake turbulence effects, even though description of
these effects is a primary motivation for their use.
Future research should be directed to the
verification of models which provide for the effects
of terrain and wake turbulence, and for description
of jetting releases, aerosol formation, chemical
reaction, and deposition.
C-3
-------
INTRODUCTION
The recent accidents at Bhopal and Mexico City illustrate
the potential dangers of acutely toxic or combustible gas
clouds. These catastrophes have intensified pressures on
Government to "do something" about major industrial accidents.
The European Community's Seveso Directive calls for tighter
regulatory controls to reduce the likelihood and consequences
of "major hazards", and the response has been to implement
tighter regulations of activities using hazardous chemicals.
The regulations require provision of evidence that hazards to
man and the environment have been identified and that suitable
precautions have been taken to minimize occurrence and
potential consequences. Many chemical manufacturing, storage,
and transportation activities with large inventories of
hazardous materials have had specific requirements imposed,
including
• submission of safety assessments
• preparation of on-site emergency response plans
• preparation, in cooperation with the local authorities,
of off-site emergency plans
• provision of information to people in the vicinity who
might be affected by a major accident
Pressures for similar regulatory controls are increasing in the
United States.
The concept of separation distances, by which the
consequences of major accidents are ameliorated, is implicit in
regulations being considered world-wide. Methods are required
for prediction of atmospheric dispersion of hazardous gas
releases, of which the main hazards are vapor cloud combustion
and acute toxicity, to aid in the determination of appropriate
separation distances. Separation distances are required for
siting fixed hazardous chemicals operations, and for the
C-4
-------
preparation of emergency response plans. Methods are also
required for predicting atmospheric gas dispersion in support
of emergency response to transportation accidents.
Prediction of separation distances for use in risk
assessment and emergency response is controversial. The
inclusion of this review of mathematical models for atmospheric
dispersion of hazardous chemical gas releases, and lesser
consideration of modeling methods applicable to predicting
other accident phenomena such as fire/explosion damage or
toxicity effects, indicates the importance attached to this
question by the chemical industry and regulatory parties. The
Environmental Protection Agency and Department of Energy
jointly sponsored a "Workshop on Determination of Atmospheric
Dilution for Emergency Preparedness" in October 1986 with the
following principal objectives.
• to review current methods for determining release
characteristics, source strength, and dispersion
of hazardous gases released into the atmosphere for use
in hazard evaluation and emergency response
• to provide recommendations for choosing among the current
methods for immediate use
• to assess the specific strengths and weaknesses of
available methods and make recommendations for their
improvement.
The controversy can be traced to predictions in the mid-70's of
the extent of the hazard zone around an accidental release of
liquefied natural gas (1). A postulated accident of concern
was the rupture of one or more 25,000 cubic meter LNG ship
cargo tanks. Since the vapor cloud which would be formed
following the spillage of such a large quantity of LNG onto
water would be cold, it would be expected to be significantly
heavier than air. Air pollutant dispersion models (primarily
Gaussian similarity models with ad hoc provision for gravity
spreading of the heavy gas) were applied, with widely disparate
results, to predict the extent of flammable gas-air mixtures
C-5
-------
from these "worst credible" releases (2). The disparity of
such dispersion predictions made more uncertain the risks
associated with LNG importation, and fostered an extensive
research program aimed at better understanding the atmospheric
dispersion of large quantities of heavy gases which are
accidentally released.
Since van Ulden's proposal (3) of a simple mathematical
model to describe the gravity spread and subsequent dispersion
of a large volume of heavy gas instantaneously released into
the atmosphere, there have been numerous other models proposed
and extensive laboratory and field test experimental programs
to provide data for model validation have been completed.
Several recent reviews which describe the mathematical models
are available (4,5,6). The field test programs completed at
China Lake, California (7), Maplin Sands, England (8), Thorney
Island, England (9), and Frenchman Flat, Nevada (10) to provide
data for dispersion model validation have also been reported.
For dispersion scenarios where the gas cloud can be
represented as having a regular shape, similarity mathematical
models (such as the familiar Gaussian models in widespread use
for passive atmospheric pollutant dispersion) are applicable.
Similarity approximations are also justified for modeling
accidental releases of large quantities of heavy chemical gases
under certain conditions.
Complex (three-dimensional) mathematical models can, in
principle, be used to simulate the spatial and temporal
dispersion processes without the "artificial" separation of the
flow into separate regimes required by the similarity models.
Predictions can me made of dispersion effects due to nonuniform
terrain and wake turbulence. It may also be possible to
provide useful estimates of concentration fluctuations around
the predicted mean (time average) values.
This paper reviews the mathematical modeling methods which
have been developed for predicting atmospheric dispersion of
hazardous chemical gas releases. The emphasis is on
C-6
-------
mathematical models for dispersion of heavier-than-air gases.
Such emphasis is justified because the majority of hazardous
chemical gases form heavy gas clouds following accidental
release, and because such clouds remain near the ground for
longer times and to greater distances.
PHENOMENOLOGY OF HEAVY GAS DISPERSION
AND.MODEL APPLICABILITY
The typical heavy gas dispersion process involves three
more-or-less distinct regimes of fluid flow. Immediately
following rapid release of a large quantity of heavy gas, a
cloud having similar vertical and horizontal dimensions may
form. Initial slumping and lateral spreading motion continues
until the kinetic energy of the buoyancy driven flow is
dissipated. As the dispersion proceeds, stable stratification
due to the negative vertical density gradient decreases until
the process can be represented as a neutrally buoyant plume (or
cloud) embedded in the wind flow.
For heavy gas releases at ground level on uniform terrain
with unobstructed wind each of the three flow regimes
(buoyancy-dominated, stably stratified, and passive dispersion)
can be approximated as a cloud (or plume) having a regular
shape, and can be described with similarity models.
Similarity Models
When the cloud formation time tf is very small compared to
the time of cloud travel to the maximum distance exposed to the
concentration of interest tt, an "instantaneous puff"
representation is applicable; if tf » tt, a stationary "plume"
representation of the cloud is indicated.
Figure 1 depicts the cloud shape which has been most
frequently used in similarity heavy gas dispersion models for
instantaneous releases. An initial volume of gas, usually
C-7
-------
Wind Direction
Source
FIGURE 1. Cloud shape for instantaneous release.
represented as a vertically oriented cylinder, is placed in the
flow field. The cloud is moved downwind with a velocity
determined from the wind vertical profile. Figure 2 depicts
the cloud shapes which have been most frequently used in models
for continuous releases. The principal dimensions of the puff
or plume change as a result of gravity spreading (assumed to
occur crosswind only in the plume representation) and
entrainment of air across the top and/or side boundaries.
Buovancv-Dominated Flow Regime. For heavy gas releases
with initially similar vertical and horizontal dimensions there
is conclusive evidence that the rapid gravity-driven flow
results in large-scale turbulent structures which effect rapid
dilution of the cloud (11,12,13,14,15). Since this initial
turbulent motion can result in a ten-fold to one hundred-fold
dilution, it must be accounted for in heavy gas dispersion
predictions. The lateral spreading that follows such releases
is most frequently modeled as a gravity current with a frontal
velocity calculated from the relation
uf - CE (g H)1/2 (1)
Equation (1), which reflects the assumption of quasi-steady
exchange of cloud potential and kinetic energy, indicates a
C-8
-------
C(x,y,j
.y >«>
C(x,y,i) » CA(X) exp - (~rT
• |y|*«>
\ '(
• ISO CONCENTRATION \ S
CONTOURS \ /
FOR C«CU V
\l
FIGURE 2. Cloud shapes for continuous releases.
C-9
-------
step change of the front velocity to its maximum value at the
instant of release. The heavy gas must accelerate from rest,
and van Ulden (16,17) and Meroney and Lohmeyer (18) have
proposed methods for modeling the acceleration phase. Air
entrainment at the spreading cloud front during the buoyancy-
dominated flow is often modeled by specifying an entrainment
velocity proportional to the front velocity:
ue - GI uf (2)
Havens and Spicer (19) reported laboratory instantaneous
releases, in calm air, of right-circular cylinders of Freon-12
with initial volumes 0.035 and 0.51 m3. Figure 3 summarizes
the measured cloud front position (radius/V1/3) vs. time (t/T).
The cloud front position is well represented, for t > « 20 T,
by the solution of Equation (1) with Cg - 1.16. Figure 3 also
indicates the predicted cloud frontal position vs. time
obtained using van Ulden's model for the acceleration phase.
Figure 4 shows ground level, peak-measured concentration as a
function of distance from the release center for the
instantaneous Freon-12 releases. The volume-averaged
concentrations of the clouds, determined by spatial integration
of vertical and horizontal cloud concentration profiles,
indicate a value for C± in Equation (2) of about 0.6.
Stablv Stratified Flow Regime. An intermediate phase of
the typical heavy gas dispersion process (between the buoyancy-
dominated flow regime and passive dispersion) is similar to a
stably stratified plume embedded in a mean flow. This regime
is characterized by the persistence of a lateral (crosswind)
gravity-driven flow and vertical density stratification which
damps turbulent mixing. The lateral gravity spread can be
modeled using Equation (1). The vertical mixing is usually
modeled with a vertical entrainment velocity which is a
function of the friction velocity of the flow and the
C-10
-------
100-
10-
A- 0.035 m
O- 0.054
D- 0.135
A- 0.530
model fit
(Havens and
Spicer, 19E5)
t/ g&
~^/6"
l I I
t*
FIGURE 3. Cloud front position vs. time, Freon-12,
H/D = 1.0, instantaneous release.
100-
X
-------
stabilizing effect of the density gradient. The stabilizing
effect of the density gradient is determined from a bulk
Richardson number for the flow:
ve -
O)
The function ^ in Equation (3) is chosen to agree with
laboratory experimental measurements of mixing in density-
stratified flows. Figure 5 shows vertical entrainment velocity
data vs. the bulk Richardson number of the flow from the
experiments reported by McQuaid (20), Kantha, Phillips, and
Azad (21), and Lofquist (22). The plotted line represents a
curve fit of the three data sets, which cover a Richardson
number range from near zero to about 105. This range
encompasses heavy gas dispersion scenarios of interest. It is
noted that questions have been repeatedly raised about the
interpretation of both KFA's data and McQuaid' s experiments,
and there exist data reported by Ellison and Turner (23) and
more recently by Deardorf f (24) , Kranenberg (25) , and Stretch
(26) which may justify some modification of the entrainment
velocity specification shown in Figure 5. However, it is
anticipated that such modifications would favor an increase in
the entrainment velocity, and Figure 5 should reflect
conservatism. Furthermore, modifications which may be
necessary appear to be for higher Richardson numbers than are
likely to be experienced in heavy gas dispersion scenarios of
interest (27).
C-12
-------
.s
10-'.
10-*.
io-J
»er«
10*
• McQuaid (20)
I K»nth« et «1. (21)
* Lefquist (22)
IB4
Rl,
FIGURE 5. Correlation of vertical entrainment velocity
with bulk Richardson number.
Passive Dispersion Regime. Vertical passive dispersion
from ground level sources is conventionally modeled as a
gradient transfer process by application of similarity
principles developed by Monin (28) and Batchelor (29) . The
velocity profile in a stratified flow against a rough wall
boundary is determined from
du u.
(4)
where the function ^ has been determined from experimental
measurements of vertical momentum transfer by Businger (30) .
For the limiting case of neutral stratification, \4 - 1, and
Equation (4) indicates a logarithmic velocity profile with
roughness height zr. The corresponding vertical diffusivity,
defined as the ratio of momentum flux to the mean velocity
gradient, is given by
K -
k u* z
(5)
C-13
-------
and (invoking the Reynolds analogy), the equivalent vertical
entrainment velocity is
v -* (6)
e z
For neutral stratification, $ - 1 and ve/u* - k, the von Karman
constant, which is about 0.4. This result is consistent with
the extrapolation of data summarized in Figure 5 to zero
Richardson number.
A General Purpose Similarity Model. Havens and Spicer (31)
have developed a general purpose heavy gas dispersion
similarity model for incorporation in the U.S. Coast Guard
Hazard Assessment Computer System. The model is designed for
simulating dispersion from ground level sources on water or
level, unobstructed terrain. The DEGADIS (DEnse GAs
Dispersion) model is an adaptation of the Shell HEGADAS model
described by Colenbrander (32,33). The buoyancy-dominated flow
regime is simulated using a box model to predict a "secondary"
heavy gas source which is input to the downwind dispersion
model. The box model of the buoyancy-dominated flow regime
incorporates air entrainment at-the gravity-spreading front
based on the data correlation shown in Figures 3 and 4. The
downwind dispersion phase of the calculation assumes a power
law concentration distribution in the vertical and a modified
Gaussian profile in the horizontal direction, with a power law
specification for the wind profile. Vertical mixing
(entrainment) is modeled using the data correlation shown in
Figure 5. Horizontal dispersion1in the stably stratified flow
regime and the ensuing passive dispersion regime (a smooth
transition, based on the vertical mixing data of Figure 5, is
effected by the model) is forced to reflect experimental data
on horizontal dispersion of passive plumes from point sources,
such as the power law correlation of av developed by Pasquill
(34). DEGADIS provides for treatment of transient (including
instantaneous) releases as a series of pseudo-steady state
releases. The model also provides for heat transfer from the
C-14
-------
underlying surface to the cloud, as well as enhancement of
vertical mixing by the unstable temperature gradient which
results from heat transfer to the cloud. The convective
turbulence is modeled using an approach adapted from Zeman and
Tennekes (35).
DEGADIS has been used to simulate a wide range of field
heavy gas releases, including small to intermediate LPG (0.1 to
1 kg/s) and LNG releases (1 to 100 kg/s) on land, large-scale
releases (10 to 150 kg/s) of LPG and LNG on water, and
instantaneous releases of approximately 5000 kg Freon/air
mixtures on land. A comparison between the predicted and
"observed" distance to the 5%, 2-1/2%, and 1% concentration
levels for the Burro/Coyote, Maplin Sands, and Thorney Island
field tests has been made. The observed values were determined
from reported maximum concentrations for each experiment by
drawing a visual best-fit straight line through the reported
points in the concentration range of interest; all of the
measurements used were made at heights at or below 1 m. The
predicted distance to a given concentration level was based on
the ground level centerline concentration calculated by
DEGADIS; for the concentrations and conditions of interest, the
predicted concentration level is essentially constant for
heights below 1 m. From these values, a ratio of the observed
to the predicted distance for each experiment was calculated.
Table I summarizes ratios of the observed to predicted
distances for the concentration levels of interest. As well, a
90% confidence interval of these ratios is included for each
test series and for all of the experiments together. (For
instance, the Maplin Sands comparisons indicate the ratio of
the observed to the predicted distance to the 2-1/2%
concentration level would be between 0.91 and 1.20 in nine out
of ten realizations.) For all of the comparisons in Table I,
the ratio of observed distance (DBS) to predicted distance
(PRE) for a given ensemble-averaged concentration level ranged
from 0.73 to 0.96 for the 5% level nine out of ten times (i.e.
C-15
-------
TABLE I. Comparison between "Observed" and DEGAD1S-
Predicted Maximum Distance to Gas Concentrations
in the Flammable Concentration Range.
Test Series
Thorney Island 7
8
9
11
13
15
Thorney Island
90X Confidence
Interval 0.71 S
Haplin 22
27
29
34
35
39
43
46
47
49
50
54
56
Haplin 90X
Confidence
Interval 0.80 s
Burro 3
7
8
9
Coyote 5
6
Burro 90X
Confidence
Interval 0.46 s
Ratio of
Distance
v « 5X
1.17
1.09
1.00
0.55
0.70
1.05
(06S/PRE)5 s 1.11
0.47
0.95
0.89
1.2S
1.28
0.46
0.73
1.09
0.77
1.27
0.92
1.12
0.93
5 S 1.03 0.95 S (OBS/PRE), S 1.24
Sunroary 99X
Confidence
Interval 0.64 s (OBS/PRE)j £ 1.03 0.78 s
-------
90% confidence interval); for the 2.5% level, (OBS/PRE) ranged
from 0.82 to 1.03 for a 90% confidence interval; for the 1%
level, (OBS/PRE) ranged from 0.95 to 1.24 for a 90% confidence
interval.
3-D Mathematical, Models
The Gas Research Institute in the United States is
sponsoring a research project (36) to evaluate the SIGMET-N
(37), ZEPHYR (38), MARIAH-II (39), and FEM3 (40) 3-D
mathematical models for LNG vapor dispersion. 3-D mathematical
models have been also reported by Schreurs (41), Rioux (42),
Betts et al. (43), and Deaves (44). These models have had
limited use, partly because of hesitation by regulatory
authorities to allow their use without a thorough technical
evaluation.
Comparison simulations were made in the GRI evaluation
project with SIGMET-N, ZEPHYR, MARIAH-II, and FEM3 of the Burro
9 LNG spill test (7). Figure 6 shows the centerline vertical
and one meter elevation horizontal 5% gas concentration
contours predicted at 80 seconds from spill start by the four
models. (Also shown are the grid cell sizes used, the total
number of cells computed, and the VAX 11-730 cpu time required
for the models.) Important differences were observed in the
predicted maximum downwind extent as well as the maximum
lateral extent of the LFL contour. These differences were
attributed primarily to two factors:
• numerical diffusion (truncation) errors
• turbulent mixing (closure) models
The model evaluation program has focused on these two model
characteristics.
C-17
-------
SICXET-N
MAKIAB-II
TOO
*
»
S 10
S
•
4J
X
--•N ['
d* V^ ,j ' \'N^'rV-'->' i
• 0 100 200 300 400 500
K
e
"100
*J
•
«4
o
t • 80 •
600
t - 80 •
* 0 100 200 300 400 500
Dovnvind Distance, n
LTL Diet. • ix. « iy, » t«- t
ZEPHYR >440 10 .67 6
S1CMEI-H 215 10 .67 6
MUUAH-IX 310 10 .67 6
TOO 375 * * *
600
Cells CPU, h
11,250 173
11,250 79
11,250 6.*
3,680 -60
•variable, expanding grid
FIGURE 6. Comparison predictions of vertical centerline
and one meter height horizontal 5% gas
contours—Simulation of Burro 9 experiment
with four models.
Numerical Diffusion Properties. The numerical solution of
advection-diffusion equations (the momentum, mass, and energy
balance equations) using finite differences or elements results
in truncation errors which cause a "numerical diffusion"
component in the solution (45). The numerical diffusion errors
must be maintained small compared to the contributions from the
actual diffusion terms in the balance equations. The severity
of numerical diffusion depends on the finite difference scheme
used and the discretization (spatial step size).
The SIGHET-N model uses first-order, forward-time, centered
space finite differences for the advection terms of all balance
equations. The method used is equivalent to using "upwind"
differencing, and test calculations for representative heavy
gas dispersion scenarios demonstrate the severity of numerical
diffusion (Havens and Schreurs, 46).
C-18
-------
ZEPHYR uses second-order, centered-space finite differences
for the advection terms in the momentum balance equations. The
contaminant gas balance equation is simulated with a Lagrangian
model in which particles representing the contaminant gas are
transported in the finite difference grid by advection at the
local (interpolated) flow velocity and diffused via a random
walk prescription. The particle treatment of the gas
contaminant balance equation effectively avoids the problem of
numerical diffusion. However, the random walk method used to
simulate the turbulent mixing (diffusion) of the gas is not
suitable for application to the strongly varying diffusivity
field which is expected in an LNG vapor cloud (Schreurs
et al. (47)).
MARIAH-II uses a second-order Crowley difference
approximation for the advection term in the gas contaminant
balance equation. This method substantially reduces numerical
diffusion errors but can result in oscillatory behavior in
regions of high concentration gradients. In MARIAH-II these
oscillations are identified and locally damped using Chapman's
FRAM (filtering remedy and methodology) method (48).
FEM3 uses finite elements to approximate the spatial
variation of the problem variables. With the first-order
function approximations which have been used in applications to
LNG vapor dispersion, FEM3 can be compared to the finite
difference methods.
An extensive computational exercise to evaluate the
numerical diffusion properties of the four models being
evaluated in the GRI project has been reported by Havens and
Schreurs (49). Based on the results, it was determined that
satisfactory control of numerical diffusion errors in LNG vapor
dispersion predictions was not practical with the SIGMET-N
model, and it has not been evaluated further. The numerical
diffusion properties of the ZEPHYR model and the development
efforts which would be required to adapt (and verify) the
Langrangian particle treatment of turbulent gas mixing for the
C-19
-------
typical heavy gas dispersion scenario also led to its
elimination from the evaluation project. The evaluation of
MARIAH-II and FEM3 has continued, with emphasis on testing by
comparison with wind tunnel test data which reflect strong
density stratification effects and for which the turbulent
mixing can be quantified. Particular use has been made of two
wind tunnel heavy gas release data sets:
• two-dimensional, continuous, heavy gas releases in a
wind tunnel
• three-dimensional, continuous, heavy gas releases in a
wind tunnel. In the experiments selected for analysis,
the dispersion appears to have been confined to a zone
adjacent to the wind tunnel floor where turbulence was
damped and the mixing was essentially molecular (50,51).
McQuaid (20) reported measurements of the vertical
dispersion of carbon dioxide (p/pa - 1.52) introduced as a
floor level, near-line source in a fully developed wind tunnel
flow of width 0.3 m, height 0.9 m, and working length 5.5 m.
The mean velocity, ua, in the tunnel (=» 0.8 u^^) ranged from
0.82 to 3.52 m/s. The friction velocity (in the absence of any
dense gas) was reported as u* — 0.0412 ua. The Reynolds number
(based on the mean velocity and the tunnel hydraulic diameter)
ranged correspondingly from 2.5 x 10^ to 1.1 x 10 . Three of
McQuaid's experiments have been selected for simulation. The
conditions for the three experiments, which were chosen to
represent near-passive dispersion to stably stratified flow
dispersion, are given in Table II.
C-20
-------
TABLE II. Summary of Experimental Conditions--McQuaid (20)
Average velocity over
channel (m/s)
Friction velocity (m/s)
Source strength (kg/m s)
Source exit (vertical)
velocity (m/s)
Richardson number1 range
Gas density (kg/m3)
Source dimensions width,
m - length, m
Roughness length (m)
(calculated from
velocity profile)
1
3.6
0.144
0.0113
0.124
0.5-0.9
1.83
0.3-0.05
1 x lO'5
Experiment
2
1.83
0.073
0.0142
0.155
4.0-8.0
1.83
0.3-0.05
1 x 10'5
3
1.28
0.05
0.0226
0.248
28.0-44.0
1.83
0.3-0.05
1 x 10'5
. H --
L - v ££ eff
* & p 2
a u,
*
McQuaid's experiments have been used to evaluate numerical
diffusion in the MARIAH-II and FEM3 models in an indirect way.
McQuaid has shown that the vertical gas concentration
distributions in the three experiments chosen for evaluation
are well represented as Gaussian. The measured gas
concentration distribution has been used to estimate the
vertical air entrainment velocity and a diffusivity profile
consistent with the measurements specified. This approximate
knowledge of the "correct" turbulent eddy diffusivity for these
experiments allows comparison calculations with either zero
(MARIAH-II) or very small (FEM3) diffusivities to determine the
relative importance of numerical vs. "physical" diffusion.
These computations have also been reported (49). For the three
experiments chosen for analysis, which represent near-passive
to strongly stratified flow dispersion, the slope of the
C-21
-------
maximum (ground level) concentration vs. distance which is
attributable to numerical diffusion is satisfactorily small
compared to the experimentally observed concentration decay.
Kothari and Meroney (KM) (50) and Meroney and Neff (KN)
(51) reported measurements of the lateral gravity spread and
dispersion of isothermal heavy gases released continuously from
floor level area sources in an unobstructed wind tunnel flow.
Two experiments have been carefully studied. Table III lists
the experimental conditions of KM 75 and MN 42. It appears
that in these two experiments the heavy gas dispersion
processes were confined to the zone adjacent to the wind tunnel
floor where mixing was molecular. Accordingly, both
experiments were simulated using the molecular diffusivities
for the gases involved (C02 and Argon, respectively) for the
diffusion calculation.
Figure 7 compares the measured and predicted ground level
centerline gas concentration and the ground level horizontal
concentration distributions at four downwind distances for KM
75 using MARIAH-II and for MN 42 using MARIAH-II and FEM3.
Since these three-dimensional experiments involve gas flows
TABLE III. Summary of Experimental Conditions--
Kothari-Meroney (50) and Meroney-Neff (51)
KM 75
KM 42
Tunnel velocity (m/s)
Friction velocity (m/s)
Source strength (kg/s)
Source exit velocity (m/s)
Gas density (kg/m3)
Source diameter (m)
Roughness length (m)
Gas molecular diffusivity
(m2/s)
0.24 @ 0.4 ra
0.017
2.68 x 10
2.11 x 10
-4
-3
.82 (C02)
0.3
0.00015
1.5 x 10
-5
0.20 @ 2.1 cm
0.015
2.77 x 10'4
9.6 x ID'3
1.63 (Argon)
0.15
0.0001
1.5 x 10
-5
C-22
-------
Kothari-Meroney 75
Diffusivity Sensitivity
Meroney-Keff 42
20-
I 0
Predicted
MARIAH-H
data
3v
.5 1.0 1.5 2.0 2.5 a
S
0 Exp. ,
Pre. MARIA?-II
1.0 2.0
3.0 4.0
x=.3 m
20
101
0
\
0
x=.8 a
x-1.2 m
20
.3 10!
g °
0)
o
io)
X-.6 m
x-1.2 m
FEM3 same as MARIAH-II
20
10
x-2 n
.2 .4 .6 .8
Crosswind Distance, o
20i
10
x=2.4 m
FEM3 not at steady state
here, but appears same as
MARIAH-II
°ooonnnn 0-0-0-^^-0—
0 .2 .4 .6 .8
Distance from Centerline, m
FIGURE 7. Comparison of measured and predicted gas
concentrations for Kothari-Meroney 75 and
Meroney-Neff 42.
C-23
-------
with a strong lateral gravity- driven component, and since the
molecular diffusion coefficient is well known, the agreement of
these simulations with the experimental data provides important
assurance of the general applicability of the model equations
used in MARIAH-II and FEM3 as well as the adequacy of the
numerical methods used.
Turbulent Mixing (Closured Models. The four models being
evaluated in the GRI project assume the gradient transfer
hypothesis with ad hoc specification of eddy diffusivities.
The primary specification is of eddy diffusivity for transfer
in the vertical direction; horizontal diffusivities are
obtained from prescribed ratios with the vertical diffusivity.
SIGMET-N, ZEPHYR, and MARIAH-II use the diffusivity
specification method incorporated in the original SIGMET model
(52). The vertical diffusivity is specified as a function of
height, "local stability" (based on temperature or density
gradient), and local velocity as shown in Figure 8. The
selection of a "local" Pasquill stability category is based on
the correlation of stability with vertical temperature gradient
specified in the USAEC Safety Guide 23 (53). The Pasquill
stability categories can be correlated with a bulk Richardson
number for the atmospheric flow (54), and the vertical
diffusivity coefficient can be expressed in terms of a vertical
entrainment velocity, v' , as
k u*
Figure 9 shows representative values of nondimensionalized
vertical entrainment velocity with these models, based on
Equation (7), for bulk Richardson numbers greater than 0.1.
Figure 9 also shows nondimensionalized vertical entrainment
velocity as a function of bulk Richardson number (data points)
obtained from analysis of the laboratory stratified fluid
mixing experiments of McQuaid (20); Kantha, Phillips, and Azad
(21); and Lofquist (22). The solid line curve plotted on
C-24
-------
10
x.l
.01
1 10 100
Height, m
FIGURE 8. Vertical diffusivity specification in MARIAH-II,
SIGMET-N, and ZEPHYR.
0.5
10
-1
... . McQuaid (1976)
SIGMET Turbulence
Closure
KPA (1977)
B,y».01-10a
Ravens and
Spicer
DEGADIS (1985)
FIGURE 9. Vertical entrainment as a function of bulk
Richardson number.
C-25
-------
Figure 9 is a curve fit (of the laboratory experimental data)
which has been incorporated in the DEGADIS model. The DEGADIS
correlation of nondimensional entrainment velocity with bulk
Richardson number is also similar to that used in FEM3.
The vertical diffusivity calculated in a simulation of LNG
vapor dispersion with the MARIAH-II, SIGMET-N, and ZEPHYR
models will have a lower limit corresponding to "G" stability
when the bulk Richardson number is greater than about 0.1, and
there is no provision for further damping of turbulence which
would be indicated for the stronger temperature (and density)
gradients expected in an LNG cloud. Bulk Richardson numbers,
based on release conditions, can also be estimated to
characterize laboratory and field experiments. Figure 9
indicates the Richardson numbers estimated for three wind
tunnel heavy gas releases reported by McQuaid (20), the Burro 9
field experiment, and continuous LNG releases of 160, 980, and
5250 kg/s (representing maximum source rates from instantaneous
LNG spills onto water of 10, 100, and 1000 m3 respectively) in
a 5 m/s wind. There is a clear indication that the turbulent
mixing model used in MARIAH-II, SIGMET-N, and ZEPHYR will tend
to give too low vertical diffusivities for Richardson numbers
between about 0.1 and about 20, and too high vertical
diffusivities for Richardson numbers greater than about 20. It
is also significant that the turbulence model may fortuitously
represent vertical mixing reasonably well for Rig around 30,
which is probably appropriate for characterizing the Burro 9
experiment. Based on these observations, the turbulence
closure model used in MARIAH-II, SIGMET-N, and ZEPHYR is not
expected to scale adequately for LNG releases which have widely
different ratios of vertical shear (high friction velocity) and
density stratification. It follows that vertical turbulent
mixing will be underestimated for releases with low bulk
Richardson numbers and overestimated for releases with large
bulk Richardson numbers. The former cases include many small-
scale wind tunnel experiments, and the latter include large LNG
C-26
-------
releases in low-wind conditions, such as Burro 8. The
underestimation of vertical diffusivities for the small-scale
McQuaid 2-D wind tunnel heavy gas experiments has been
demonstrated by Havens and Schreurs (46).
A simplified second-order turbulence model has been
developed by Freeman (47) for simulation of LNG vapor cloud
dispersion. The principal simplifying assumptions invoked in
the model are:
• The theory is local: all turbulent quantities are assumed
point functions depending only on the state of the mean
field.
• The mean field employed in the turbulence terms is
simplified to retain the dominant terms describing a
stratified boundary layer; turbulent kinetic energy is
dependent on the magnitude of the shear of the flow
field
S - [(3u/5y)2 + (aw/ay)2]1/2
and the gradient Richardson number
MARIAH-II with the turbulence closure method developed by
Freeman and modified to provide for turbulence kinetic energy
diffusion has been used to simulate HcQuaid's experiments.
Figure 10 shows the experimental vs. MARIAH-II-predicted ground
level gas concentration decay with distance. The agreement is
now reasonably good for all three experiments.
MARIAH-II, with Freeman's turbulence closure model, has
also been used to simulate Havens' and Spicer's laboratory
instantaneous releases of heavy gas. Figure 11 shows a MARIAH-
II simulation of a 55 -liter Freon release with an initial
height- to -diameter ratio of 1.0. The cloud profile (half of
the radially symmetric section of the cloud is shown) , defined
by the 1% volume fraction, is shown at zero time and at 0.5, 1,
2, 3, 4, 5, and 6 seconds after release. Figure 12 shows the
C-27
-------
100
g
2
O
B
10
McQuaid,
- measured
No. 3
lo. 2
No. 1
-MARIAH-II, prediction
0.1 1 10
Distance, m
FIGURE 10. McQuaid 2-D wind tunnel experiments—MARIAH-II
predictions with improved air entrainment
model.
0
40
0
40
0
40
. 0
, 40
0
40
0
40
0
40
0
1
t « 0
t • 0.5 8
t " 1 8
t " 2 8
t - 3 8
t " 4 8
t - 5 8
t - 6 8
40 BO 120 160 200 240 280 320
Radial Distance, cm
FIGURE 11. Laboratory, calm-air, 55-liter Freon release,
(H/D)i = 1.0—MARIAH-II prediction of cloud
boundary vs. time.
C-28
-------
20
10
r-l -rl
«
B!
Measured
' *MARIAH-II
10
100 200
FIGURE 12. Laboratory, calm-air, 55-liter Freon release,
(H/D). - 1.0—MARIAH-II prediction of cloud
radial extent.
100.
10
0
-------
predicted vs. measured radial cloud extent for the same 55-
liter Freon release. Figure 13 shows predicted and measured
maximum values of the cloud gas concentration and the cloud
spatial average gas concentration as a function of
dimensionless time for the 55-liter Freon release. The
predicted cloud average gas concentrations were obtained by
spatial averaging the MARIAH-II prediction using cloud boundary
concentration limits of 1% and 2.5%. The DEGADIS model cloud
average concentration prediction is shown for comparison (31).
MARIAH-II and FEM3 are now being tested by comparison with
selected data from the Burro/Coyote, Maplin Sands, and Thorney
Island experiments.
CLOSURE
Similarity models are available for heavy gas releases at
ground level on uniform terrain with unobstructed atmospheric
flow. Some models have been verified for the prediction of
dispersion to flammable concentration levels (order 1%) and are
sufficiently accurate for most risk assessment and emergency
response applications.
Most field heavy gas dispersion tests have provided data on
cloud dispersion to fuel flammable-limit concentration levels;
such data are not sufficient to verify models for application
to toxic gas dispersion to parts per million concentration
levels. It is expected that models which have been verified
for flammable fuel dispersion prediction and also incorporate
an appropriate description of the passive dispersion regime
should be useful for prediction of dispersion of dense gases to
toxicity levels.
It appears likely that 3-D mathematical models can be
verified for flammable gas dispersion prediction, but current
work is biased strongly toward evaluation using data without
terrain or wake turbulence effects. Since the primary
C-30
-------
motivation for using these models is to describe such complex
effects, research should be directed to their validation for
such use.
3-D mathematical models are not likely to prove useful for
predicting heavy gas dispersion to very low concentration
levels because of difficulties in managing numerical diffusion
in calculations with several orders of magnitude change in
concentration.
Application of dense gas dispersion models to toxic gas
dispersion is expected to require additional research to
provide for description of jetting releases, aerosol formation
and evaporation, chemical reaction, and deposition.
The participants in the recent joint EPA-DOE workshop
agreed that source specification in an accident may be a
greater concern than the accuracy of a dispersion model. A
technology transfer problem was identified; many of the
improvements in dispersion models have not been transferred to
the potential user. There was a consensus that atmospheric
dispersion predictions should be done to plan emergency
response whenever possible. No selection or recommendation of
models was made.
LIST OF SYMBOLS
Cg constant in density intrusion (spreading) relation
C-^ constant
f\
g acceleration of gravity (m/s'')
H height or depth of density intrusion or cloud (m)
H£pp effective cloud depth (m)
K vertical turbulent diffusivity (m2/s)
k von Karman's constant, 0.35
C-31
-------
Ri* Richardson number associated with density differences
T characteristic time scale, V1/6/7g(P - /»a>//>a
tf characteristics time of cloud formation (s)
tt cloud travel time to the maximum distance exposed to
a given concentration (s)
u velocity along x-direction, (m/s)
ua ambient average velocity, (m/s)
ue horizontal or frontal entrainment velocity, (m/s)
uj cloud front velocity, (m/s)
Ujj wind velocity, along x-direction, (m/s)
umax maximum velocity
u* friction velocity, (m/s)
V cloud volume, (nr)
ve vertical entrainment velocity associated with HT,
(m/s)
w velocity along z-direction, (m/s)
x,y,z Cartesian coordinates, (m)
a constant in power law wind profile
A Monin-Obukhov length (m)
P density (kg/m3)
fla ambient density (kg/m3)
logarithmic velocity profile correction function
C-32
-------
REFERENCES
1. General Accounting Office (GAO), "Liquefied Energy Gases
Safety," EMD-78-29, July 31, 1978.
2. Havens, J. A., "Predictability of LNG Vapor Dispersion
from Catastrophic Spills onto Water: An Assessment,"
Department of Transportation--Coast Guard Report CG-M-09-
77, April, 1977.
3. van Ulden, A. P., "On the Spreading of a Heavy Gas
Released Near the Ground," 1st International Loss
Symposium, The Hague, Netherlands, 1974.
4. Havens, J. A., "A Review of Mathematical Models for
Prediction of Heavy Gas Atmospheric Dispersion," Institute
of Chemical Engineers Symposium Series No. 71, 1982.
5. Havens, J. A., "Evaluation of 3-D Hydrodynamic Computer
Models for Prediction of LNG Vapor Dispersion in the
Atmosphere," Gas Research Institute Contract No. 5083-252-
0788 with the University of Arkansas, March, 1983 -
September, 1986.
6. Wheatley, C. J. and D. M. Webber, "Aspects of the
Dispersion of Denser-than-air Vapors Relevant to Gas Cloud
Explosions," Safety and Reliability Directorate, United
Kingdom Atomic Energy Authority, Warrington, England,
1984.
7. Koopman, R. P. et al., "Burro Series Data Reports,
LLNL/NWC 1980 LNG Spill Tests," Lawrence Livermore
National Laboratories Report UCID-19075, December, 1982.
8. Colenbrander, G. W., A. E. Evans, andJ. S. Puttock,
"Spill Tests of LNG and Refrigerated Liquid Propane on the
Sea, Maplin Sands, 1980: Dispersion Data Digests," Shell
Thornton Research Center, May, 1984 (confidential).
9. HSE--British Health and Safety Executive, Research and
Laboratory Services Division, Red Hill, Sheffield, UK--
Heavy Gas Dispersion Trials, Thorney Island 1982-83, Data
Digests.
10. McRae, T. G., R. T. Cederwell, H. C. Goldwire, Jr., D. L.
Hippie, G. W. Johnson, R. P. Koopman, J. W. McClure, and
L. K. Morris, "Eagle Series Data Report: 1983 Nitrogen
Tetroxide Spills," Lawrence Livermore National
Laboratories Report UCID-20063, June, 1984.
11. Hall, D. J., "Experiments on a Model of an Escape of Heavy
Gas," Warren Springs Laboratory, UK, Reports LR 217 (AP),
1976, and LR 312 (AP), 1979.
12. Hall, D. J. et al., "A Wind Tunnel Model of the Porton
Dense Gas Spill Field Trials," LR 394 (AP), Warren Spring
Laboratory, Department of Industry, Stevenage, UK, 1982.
C-33
-------
13. Havens, J. A. and T. 0. Spicer, "Gravity Spreading and Air
Entrainment by Heavy Gases Instantaneously Released in a
Calm Atmosphere," Proceedings I.U.T.A.M. Symposium on
Atmospheric Dispersion of Heavy Gases and Small Particles,
Delft University of Technology, The Netherlands, August
29-September 2, 1983.
14. Picknett, R. F., "Dispersion of Dense Gas Puffs Released
in the Atmosphere at Ground Level," Atmospheric
Environment r 15. 1981.
15. Spicer, T. 0. and J. A. Havens, "Modeling the Phase I
Thorney Island Experiments," Symposium on the Thorney
Island Heavy Gas Trials, sponsored by the British Health
and Safety Executive, Sheffield, UK, April, 1984.
16. van Ulden, A. P., "The Unsteady Gravity Spread of a Dense
Cloud in a Calm Environment," 10th International Technical
Meeting on Air Pollution Modeling and its Applications,"
NATO-CCMS, Rome, Italy, October, 1979.
17. van Ulden, A. P., "A New Bulk Model for Dense Gas
Dispersion: Two-Dimensional Spread in Still Air,"
I.U.T.A.M. Symposium on Atmospheric Dispersion of Heavy
Gases and Small Particles, Delft University of Technology,
The Netherlands, August 29-September 2, 1983.
18. Meroney, R. N. and A. Lohmeyer, "Gravity Spreading and
Dispersion of Dense Gas Clouds Released Suddenly into a
Turbulent Boundary Layer," Draft Report CEF82-8RNM-AL-7 to
Gas Research Institute, Chicago, Illinois, August, 1982.
19. Havens, J. A. and T. 0. Spicer, "Development of an
Atmospheric Dispersion Model for Heavier-than-Air Gas
Mixtures," Volume II, "Laboratory Calm Air Heavy Gas
Dispersion Experiments," Report No. CG-D-23-85, U.S. Coast
Guard, Final Report, May, 1985.
20. McQuaid, James, "Some Experiments on the Structure of
Stably Stratified Shear Flows," Technical Paper P21,
Safety in Mines Research Establishment, Sheffield, UK,
1976.
21. Kantha, L. H., 0. M. Phillips, and R. S. Azad, "On
Turbulent Entrainment at a Stable Density Interface,"
Journal of Fluid Mechanics. 7_9_, 1977. pp. 753-768.
22. Lofquist, Karl, "Flow and Stress Near an Interface Between
Stratified Liquids, Phvsics of Fluids. 3, No. 2, March-
April, 1960.
23. Ellison, T. and J. S. Turner, "Turbulent Entrainment in
Stratified Flows," Journal of Fluid Mechanics. 6, 1959.
24. Deardorff, J. W., 1983, "A Multi-Limit Mixed-Layer
Entrainment Formulation," Journal of Physical
Oceanography. 13. pp. 988-1002.
25. Kranenberg, C., "Wind-Induced Entrainment in a Stably
Stratified Fluid," Journal of Fluid Mechanics. 145. 1984,
253-273.
C-34
-------
26. Stretch, D. D., R. E. Britter, and J. C. R. Hunt, "The
Dispersion of Slightly Dense Contaminants," I.U.T.A.M.
Symposium on Atmospheric Dispersion of Heavy Gases and
Small Particles, Delft, University of Technology, The
Netherlands, September, 1983.
27. Britter, R. E. , "A Review of Mixing Experiments Relevant -
to Dense Gas Dispersion," The Institute of Mathematics and
Its Applications Conference on Stably Stratified Flow and
Dense Gas Dispersion, Chester, England, April, 1986.
28. Monin, A. S., "Smoke Propagation in the Surface Layer of
the Atmosphere," in Atmospheric Diffusion and Air
Pollution, ed., F. N. Frenkiel, Academic Press, 1959.
29. Batchelor, G. K., An Introduction to Fluid Mechanics.
Cambridge University Press, Cambridge, UK, 1967.
30. Businger, J. A., J. C. Wyngaard, Y. Izumi, and E. F.
Bradley, Flux-Profile Relationships in the Atmospheric
Surface Layer," Journal of the Atmospheric Sciences. 28.
March, 1971.
31. Havens, J. A. and T. 0. Spicer, "Development of an
Atmospheric Dispersion Model for Heavier-than-Air Gas
Mixtures," Volume I, Report No. CG-D-23-85, U.S. Coast
Guard, Final Report, May, 1985.
32. Colenbrander, G. W., "A Mathematical Model for the
Transient Behavior of Dense Vapor Clouds," 3rd
International Symposium on Loss Prevention and Safety
Promotion in the Process Industries, Basel, Switzerland,
1980.
33. Colenbrander, G. W. and J. S. Puttock, "Dense Gas
Dispersion Behavior: Experimental Observations and Model
Developments," International Symposium on Loss Prevention
and Safety Promotion in the Process Industries, Harrogate,
England, September, 1983.
34. Pasquill, F., Atmospheric Diffusion. 3rd ed., Halstead
Press, New York, 1983.
35. Zeman, 0. and H. Tennekes, "Parameterization of the
Turbulent Energy Budget at the Top of the Daytime
Atmospheric Boundary Layer," Journal of the Atmospheric
Sciences. January, 1977.
36. Havens, J. A., "Evaluation of 3-D Hydrodynamic Computer
Models for Prediction of LNG Vapor Dispersion in the
Atmosphere," GRI Contract No. 5083-252-0788, Annual Report
(March 1984-February 1985), University of Arkansas.
37. Su, F. Y. and P. C. Patniak, "SIGMET-N: Near-Field
Solution of Contaminant Dispersion Phenomena; Model
Description and User's Guide," Science Applications, Inc.,
July, 1981 (proprietary).
38. Hertel, J. and L. Teuscher, "Advances in Heavier-than-Air
Vapor Cloud Dispersion Modeling," Proceedings of the AGA
Transmission Conference, Chicago, Illinois, May, 1982.
C-35
-------
39. Taft, J. R., M. S. Ryne, and D. A. Weston, "MARIAH: A
Dispersion Model for Evaluating Realistic Heavy Gas Spill
Scenarios," Proceedings of the AGA Transmission
Conference, Seattle, Washington, May, 1983.
40. Chan, S. T., "FEM3--A Finite Element Model for the
Simulation of Heavy Gas Dispersion and Incompressible
Flow--User's Manual," Lawrence Livermore Laboratory, UCRL-
53397, February 1983.
41. Schreurs, P. J., "Mathematical Modeling of the Dispersion
of Accidental Releases of Heavy Gases at Ground Level in
an Industrial Environment," Catholic University, Leuven,
Belgium, 1983.
42. Riou, Y., "The Use of a Three-Dimensional Model in
Simulating Thorney Island Field Trials," The Institute of
Mathematics and Its Applications Conference on Stably
Stratified Flow and Dense Gas Dispersion, April, 1986,
Chester College, Chester, England.
43. Betts, P. L. et al., "Finite Element Calculations of
Transient Dense Gas Dispersion," The Institute of
Mathematics and Its Applications Conference on Stably
Stratified Flow and Dense Gas Dispersion, April, 1986,
Chester College, Chester, England.
44. Deaves, D. M., "Application of Advanced Turbulence Models
in Determining the Structure and Dispersion of Heavy Gas
Clouds," I.U.T.A.M. Symposium on Atmospheric Dispersion of
Heavy Gases and Small Particles, Delft University of
Technology, The Netherlands, August 29-September 2, 1983.
45. Roache, P. J., Computational Fluid Dynamics. Hermosa
Publishers, Albuquerque, NM, 1976.
46. Havens, J. A. and P. J. Schreurs, "Evaluation of 3-D
Hydrodynamnic Computer Models for Prediction of LNG Vapor
Dispersion in the Atmosphere," Proceedings, Eighth
International Conference on Liquefied Natural Gas, Los
Angeles, California, June, 1986.
47. Schreurs, J. A. et al., "A Lagrangian Particle Model for
Atmospheric Dispersion of Heavy Gases," submitted to
Atmospheric Environment. Spring, 1986.
48. Chapman, M., "FRAM--Nonlinear Damping Algorithms for the
Continuity Equation," Journal of Computational Physics.
44, 1981.
49. Havens, J. A. and P. J. Schreurs, "Evaluation of 3-D
Hydrodynamic Computer Models for Prediction of LNG Vapor
Dispersion in the Atmosphere," Annual Report, Contract No.
5083-252-0788, (March 1984-February 1985), with the
University of Arkansas.
50. Kothari, K. M. and R. N. Meroney, "Accelerated Dilution of
Liquefied Natural Gas Plumes at the Source," Final Report
to Gas Research Institute, Contract No. 5014-352-0203,
1982.
C-36
-------
51. Meroney, R. M. and D. E. Neff, "The Behavior of LNG Vapor
Clouds: Wind-Tunnel Tests on the Modeling of Heavy Plume
Dispersion," Final Report to Gas Research Institute,
Contract No. 5014-352-0203, 1982.
52. Havens, J., "A Description and Assessment of the SIGMET
LNG Vapor Dispersion Model," U.S. Coast Guard Report No.
CG-M-3-79 (available through NTIS), 1979.
53. U.S. Atomic Energy Commission Safety Guide 23, 1972.
54. Randerson, D., Editor, "Atmospheric Science and Power
Production," U.S. Department of Energy, Report DOE-TIC-
27601, 1984.
C-37
-------
APPENDIX D
Workshop Agenda
D-l
-------
AGENDA
Determination of Atmospheric Dilution for Emergency Preparedness
A Joint EPA-DOE Technical Workshop
October 15-17, 1986
TIME
TOPIC
SPEAKER
Wednesday. October 15. 1986
SESSION 1
8:00 - 8:10 a. m.
8:10 - 8:40 a. m.
Two Recent Catastrophic Events
8:40 - 9:00 a. m.
9:00 - 9:20 a. m.
Meredith Guest House
Research Triangle Park, NC
Wei come
Technical Overview
Bhopal
National Research Needs for
Emergency Response in the
Wake of Chernobyl
SESSION 2
Characteristics of Sources of Releases of Toxic Material
Chairman: Joe Knox, Lawrence Livermore Laboratory
9:20 - 9:40 a. m.
9:40 - 10:00 a. m.
10:00 - 10:20 a. m.
10:20 - 10:40 a. m.
Historical Accounts and
Credible Scenarios
Short Term Toxic Releases from
Industrial Sites
Fate of Toxic Releases in the
Atmosphere - ARAC
Source Strength Modeling
SESSION 3
Reaulatorv and Operational Needs for Models
Chairman: Bob Hangebrauck, USEPA
Frank Binkowski, USEPA &
NOAA and Harry Moses,
USDOE
R. A. Cox, Technica
Shankar Rao, NOAA/ATDD
Joe Knox, Lawrence
Livermore Laboratory
Jane Crum, Bob
Hangebrauck, and
Bill Rhodes, USEPA
Robert Rosensteel, USEPA
Marvin Dickerson, LLL
Jerry Schroy, Monsanto
10:40 - 11:00 a. m.
Future Needs for Dispersion
Models in Hazard Evaluation,
Jim Makris, USEPA
D-3
-------
TIME
TOPIC
SPEAKER
SESSION 3 (Cont'dl
11:00 - 11:20 a. m.
11:20 - 11:40 a. ra.
11:40 - 1:00 p. m.
Emergency Preparedness, and
Accident Prevention
Issues in Regulatory
Applications of Models
Community Needs for Hazard
Evaluation Tools
Lunch
Dave Lay!and, USEPA
Fred Millar,
Environmental Policy
Institute
SESSION 4
Dispersion Models
Chairman: Dick Benjamin, Savannah River Laboratory
1:00 - 1:30 p. m.
1:30 - 1:50 p. m.
1:50 - 2:10 p. m.
2:10 - 2:30 p. m.
2:30 - 2:45 p. m.
Keynote Talk
Dense Gas Dispersion Model
Reactive Dense Gas Models
Jerry Havens, University
of Arkansas
Don Ermak, LLL
Bruce Hicks and Will
Pendergrass, NOAA/ATTD
Dispersion Models for Neutrally Tom Pierce, USEPA & NOAA
Buoyant and Positively Buoyant
Gases
Break
SESSION 5
Results from Dispersion Experiments
Chairman: Ray Hosker, NOAA/Atmospheric Turbulence & Diffusion Division
2:45 - 3:05 p. m.
3:05 - 3:25 p. m.
3:25 - 3:45 p. m.
Laboratory Scale Experiments
Experimental Work at the EPA
Fluid Modeling Facility
Large Scale Experiments of
the DOE Liquefied Gaseous
Fuels Program
Bob Meroney, Colorado
State University
Bill Snyder,
USEPA & NOAA
Ron Koopman, LLL
D-4
-------
TIME
3:45 - 4:05 p. m.
4:05 - 4:42 p. m.
IQPIC
Atmospheric Science and
Emergency Response at the
Savannah River Plant
Emergency Preparedness and
Response In the U. S. Air
Force
SPEAKER
Alan Weber, Savannah
River Laboratory
Larry Key, USAF
Thursday. October 16. 1986
SESSION 6
Panel Discussions
8:00 - 11:30 a. m.
(Break 9:30 - 9:45 a. m.)
1:00 - 3:30 p. m.
3:30 - 3:45 p. m.
3:45 - 5:00 p. m.
Summary of Current Status
of Source Strength and
Dispersion Models, Their
Strengths and Weaknesses,
and Recommendations for
Improvement
Recommendations for Selection
from Among Current Models for
Immediate Use in Hazard Identi-
fication and Evaluation, Prepara-
tion of Emergency Preparedness
Scenarios, and in Response to
Emergencies
Break
The Role of the Meteorologist
in Hazard Evaluation, Emergency
Response
Panel 1
Chairman: Frank
Schiermeier
Reporter: Steve Perry
Panel 2
Chairman: Ray Hosker
Reporter: Tom Pierce
Chairman:
Reporter:
Lew Nagler
Jack Shannon
Friday. October 17. 1986
8:00 - 11:30 a. m.
(Break 9:30 - 9:45 a. m.)
Presentation of Results of Panel
Discussions
D-5
-------
APPENDIX E
Workshop Participants
E-l
-------
DETERMINATION OF ATMOSPHERIC DILUTION FOR EMERGENCY PREPAREDNESS
A JOINT EPA-DOE TECHNICAL WORKSHOP
OCTOBER 15 - 17, 1986
Allen, Dr. Harry L.
USEPA-ERT
Edison, NJ 08837
(201) 321-6747
Bach, Dr. Walter D., Jr.
Meteorologist
Army Research Office
P. 0. Box 12211
Research Triangle Park, NC 27709
(919) 549-0641
FTS 629-3890
Barr, Dr. Sumner
Atmos, Sci. G-8, Mail Stop D 466
Los Alamos National Lab
P. 0. Box 1663
Los Alamos, NM 87545
(505) 667-2636
FTS 843-2636
Benjamin, Dr. Richard W.
Room 1015A, Bldg. 773A
E. I. Du Pont
Savannah River Laboratory
Aiken, SC 29808
(803) 725-3325
Binkowski, Dr. Francis S.
USEPA (MD-80)
Research Triangle Park, NC 27711
(919) 541-2460
Blaunstein, Dr. Robert P.
EPG, RCD, EH-33, USDOE
Washington, DC 20545
(301) 353-5849
Bresnick, Dr. Gerald I.
Amoco Corp., MC 4903
200 E. Randolph Drive
Chicago, IL 60601
(312) 856-7198
Briggs, Dr. Gary A.
USEPA (MD-80)
Research Triangle Park, NC
(919) 541-2606
27711
Catalano, Joseph A.
Aerocomp, Inc., Technical Director
3303 Harbor Blvd.
Cost Mesa, CA 92626
(714) 957^-6596
Catalano, Jude J.
Connecticut Dept. of Environmental
Protection
165 Capitol Avenue, Room 146
Hartford, CT 06106
(203) 566-2690
Chaplin, Anton
Manager, Environmental Meteorology
Unocal Corporation
1201 W. 5th Street
Los Angeles, CA 90017
(213) 977-5202
Compton, Harry
USEPA-ERT
Edison, NJ 08837
(201) 321-6751
Cox, Dr. R. A.
Technica, Ltd.
Lynton House
7/12 Tavistock Square
London WC1H9LT
England
011-44-1-388-2684
Crum, Jane
USEPA, MD-62B
Research Triangle Park, NC 27711
(919) 541-1528
E-3
-------
Cupitt, Dr. Larry
USEPA (MD-84)
Research Triangle Park, NC 27711
(919) 541-2878
Cushmac, Dr. George E.
RSPA, USDOT (DHM-21)
Washington, DC 20590
(202) 366-4493
De Wolf, Glenn B.
Radian Corp.
P. 0. Box 9948
Austin, TX 78766
(512) 454-4797
Dicker-son, Dr. Marvin H.
Associate Division Leader
Atmospheric & Geophysical Science
P. 0. Box 808
Livermore, CA 94550
(415) 422-1806
FTS 532-1806
Doelp, Louis C.
Air Products and Chemicals, Inc.
P. 0. Box 538
Allentown, PA 18105
(215) 481-6136
Eheman, Christie
Health Physicist
1600 Clifton Rd., NE, CMB-28N
Atlanta, GA 30333
(404) 452-4161
Elderkin, Dr. Charles
Atmospheric Science Dept., PNL
P. 0. Box 999
Richland, WA 99352
(509) 376-8639
Eltgroth, Mark
Radian Corp.
P. 0. Box 9948
Austin, TX 78766
Englemann, Dr. Rudolph J.
NOAA
11701 Karen Drive
Potomac, MD 20854
(301) 443-8721
Ermak, Dr. Donald L.
Modeling Group Leader
Liquefied Gaseous Fuels Program
Lawrence Livermore Laboratory
P. 0. Box 808, L-451
Livermore, CA 94550
(415) 423-0146
FTS 469-0146
Fairobent, James
Office of Environmental
Department of Energy
Washington, DC 20585
(202) 252-4852
Guidance, EH-23
Foster, Jerry R.
Union Carbide Corporation
P. 0. Box 8361
South Charleston, WV 15303
(304) 747-3615
Gait, Jerry A.
Hazardous Materials Response
Branch-NOAA
7600 Sand Point Way
Seattle, WA 98115
(206) 526-6317
FTS 392-6317
Gelinas, Gary
Safer Emergency Systems, Inc.
5700 Corsa Avenue
Westlake Village, CA 91362
(818) 707-2777
Hall, Loren
USEPA, Office of Toxic
Substances (TS-798)
Washington, DC 20460
(202) 382-3931
Hangebrauck, Robert
USEPA (MD-62B)
Research Triangle Park, NC 27711
(919) 541-4134
Havens, Dr. Jerry A.
Chemical Engineering Dept.
University of Arkansas
Fayettville, AR 72701
(501) 575-4951
E-4
-------
Hicks, Bruce
ATDD, P. 0. Box 2456
Oak Ridge, TN 37831
(615) 576-1232
Hodgin, C. Reed
Rockwell International
Rocky Flats Plant
P. 0. Box 464
Golden, CO 80402-0464
(303) 966-7084
Hosker, R. P., Jr.
NOAA-Atmospheric Turbulence &
Diffusion Division
2456 S. Illinois Avenue
Oak Ridge, TN 37831
(615) 576-1248
FTS 626-1248
Jann, Paul R.
E. I. Du Pont
Engineering Dept. (L-13W40)
Wilmington, DE 19898
(302) 366-3219
Jensen, Andreas V.
Motech Services Co.
220 South River, Suite A
Eaton Rapids, MI 48827
(517) 663-7502
Jersey, Gilbert
Mobil Research & Development Corp.
Billingsport Road
Paulsboro, NJ 08066
(609) 423-1040
Keith, William
USEPA (RD-680)
Washington, DC 20460
(202) 382-5716
Key, Capt. Lawrence E.
Research Meteorologist
Headquarters, Air Force Engineering
& Services Center (HQ FECS/RDV)
Tyndall AFB, FL 32403
(904) 283-4234
Knox, Dr. Joseph B.
Atmospheric & Geophysical Science
Lawrence Livermore Laboratory
University of California
P. 0. Box 808
Livermore, CA 94550
(415) 422-1818
Koopman, Dr. Ronald P.
Lawrence Livermore Laboratory
P. 0. Box 808, L-450
Livermore, CA 94550
(415) 422-7381
Koretzky, Herman
IBM Corporation, D/728, 4A-07
44 S. Broadway
White Plains, NY 10601
(914) 686-1822
Kornasiewicz, Dr. Robert
Office of Nuclear Regulatory Research
Mail Stop NL - 007
U. S. Nuclear Regulatory Commission
Washington, DC 20555
Lake, Robin
Standard Oil Research & Development
4440 Warrensville Center Road
Cleveland, OH 44128
(216) 581-5976
Lantzy, Dr. Ronald
Rohm & Haas Company
P. 0. Box 584
Bristol, PA 19007
(215) 785-7456
Layland, David E.
Monitoring & Data Analysis Division
USEPA (MD-14)
Research Triangle Park, NC 27711
(919) 541-5690
Lee, Dr. Gene K.
Corporate Engineering Department
Air Products & Chemical
Box 538
Allentown, PA 18105
(215) 481-6424
E-5
-------
Livingstone, Dr. Daniel
Dept. of Zoology
Duke University
Durham, NC 27706
(919) 684-3264
Machta, Dr. Lester
NOAA, ARL, 8060 13th St.
Silver Spring, MD 20910
(301) 427-7684
Makris, James L.
Director of Preparedness
Office of Solid Waste &
Emergency Response
USEPA (WH-548A)
Washington, DC 20460
(202) 475-8600
Matthiessen, Craig R.
USEPA (TS-779)
401 M Street, SW
Washington, DC 20460
(202) 382-3694
Mayer, Alice R.
Chemical Manufacturers Assoc.
2501 M Street, NW
Washington, DC 20037
(202) 887-1176
McNaughton, D. J.
TRC Environmental Consultants
800 Connecticut Blvd.
East Hartford,'CT 06108
(203) 299-8631
Meier, Gerald E.
Manager, Environmental Services
Bureau of Explosives
Association of American Railroads
50 F Street, NW
Washington, DC 20001
(202) 639-2136
Meroney, Dr. Robert
Fluid Mechanics &
Wind Engineering Program
Engineering Research Center
Colorado State University
Ft. Collins, CO 80523
(303) 491-8574
Millar, Dr. Fred
Environmental Policy Institute
218 D Street, SE
Washington, DC 20003
(202) 544-2600
Moser, Dr. James H.
Environmental Engineering Dept.
Shell Development Company
Westhollow Research Center
P. 0. Box 1380
Houston, TX 77001
(713) 493-7941
Moses, Dr. Harry
PTRD, ER-74, GTN, USDOE
Washington, DC 20545
(301) 353-5572
Nagler, Lewis
Air Programs Branch, EPA
AP&TMD, 345 Courtland Street, NE
Atlanta, GA 30365
(404) 347-4253
Papal ski, Ray
New Jersey Dept. of Environmental
Protection
401 East State Street
P. 0. Box CN027
Trenton, NJ 08625
(609) 633-1142
Parnarouski, Dr. Michael
U. S. Coast Guard (G-MTH-1)
2100 2nd St., SW
Washington, DC 20593
(202) 267-1577
Patrinos, Dr. A. A. N.
Department of Energy, ER-74
Washington, DC 20545
(301) 353-3764
Pearson, Johnnie L., Chief
Model Application Section
USEPA (MD-14)
Research Triangle Park, NC 27211
(919) 541-5690
E-6
-------
Pendergrass, William R.
ATDD, P. 0. Box 2456
Oak Ridge, TN 37831
Perry, Dr. Steve
USEPA/NOAA (MD-80)
Research Triangle Park, NC 27711
(919) 541-1375
Peterson, W. B. -
USEPA (MD-80) f
Research TriangWPark, NC 27711
(919) 541-4564
Pierce, Thomas E.
NOAA Meteorologist
USEPA (MD-80)
Research Triangle Park, NC 27711
(919) 541-1375
FTS 629-1375
Pritchett, Thomas H.
USEPA-ERT
GSA Raritan Depot, Bldg. 10
Edison, NJ 08837
(201) 321-6738
Randerson, Dr. Darryl
National Weather Service
Nuclear Support Office, NOAA
P. 0. Box 14985
Las Vegas, NV 89114
(702) 598-3234
Rao, Dr. K. Shankar
ATDD, P. 0. Box 2456
Oak Ridge, TN 37831
(615) 576-1238
FTS 626-1238
Rao, S. Trivikr
Bureau of Air Research
New York State Dept. of
Environmental Conservation
Division of Air Resources
50 Wolf Road
Albany, NY 12233
(518) 457-3200
Rhodes, William J.
USEPA (MD-62B)
Research Triangle Park, NC 27711
(919) 541-2853
FTS 629-2853
Rodak, Victoria A.
ICF Technology
1850 K Street, NW, Suite 950
Washington, DC 20006
Rosensteel, Robert E.
USEPA (MD-13)
Research Triangle Park, NC 27711
(919) 541-5605
FTS 629-5605
Runkle, Gene E.
USDOE/AL, ESHD
P. 0. Box 5400
Albuquerque, NM 87115
(505) 846-2046
Sakenas, Cheryl
Office of Inspection & Enforcement
NRC, MS-MNBB-3302
Washington, DC 20555
Schiermeier, Francis A.
Meteorology Division (MD-80)
USEPA
Research Triangle Park, NC 27711
(919) 541-4541
Schroy, Jerry M.
Monsanto Company
Mail Code CS7H
800 N. Lindbergh
St. Louis, MO 63167
(314) 694-6174
Sedefian, Leon
New York Dept. of Environmental
Conservation
Division of Air
50 Wolf Road, Room 115
Albany, NY 12233
(518) 457-7605
E-7
-------
Shannon, Dr. Jack
Atmospheric Physics
ER Building 181
Argonne National Laboratory
Argonne, IL 60439
(312) 972-5807
Shephard, Cathleen
USEPA
OSWER/PS, WH-548 A
401 M Street, SW
Washington, DC 20460
(202) 475-8247
Snyder, Dr. William H.
USEPA (MD-81)
Research Triangle Park, NC 27711
(919) 541-2811
Steinberg, Ken W.
Exxon Research & Engineering Company
P. 0. Box 101
180 Park Avenue
Florham Park, NJ 07932
(201) 765-1209
Swank, Dr. Robert R., Jr.
Director of Research
Athens Environmental Research Lab
USEPA
College Station Road
Athens,.GA 30613
(404) 546-3128
FTS 250-3128
Topoleski, Gary
USEPA (TS-779)
401 M Street, SW
Washington, DC 20460
(202) 382-2251
Touma, Joe
USEPA, MD-14
Research Triangle Park, NC 27711
(919) 541-5681
Turner, D. Bruce
USEPA (MD-80)
Research Triangle Park, NC 27711
(919) 541-4564
Vaughan, Dr. William M.
Aero Viroment, Inc.
8515 Delmar Blvd., Suite 215
St. Louis, MO 63124
(314) 993-0543
Weber, Dr. Alan H.
Bldg. A773, Room A1012
E. I. Du Pont
Savannah River Laboratory
Aiken, SC 29808
(803) 725-3717
FTS 239-3717
Wiser, Dr. Herbert L.
Office of Air & Radiation
USEPA (ANR-443)
Washington, DC 20460
(202) 382-7747
Wratt, David S.
New Zealand Meteorology Service
P. 0. Box 722
Wellington, New Zealand
Yatnada, Tetsuji
MS F665
Los Alamos National Laboratory
P. 0. Box 1663
Los Alamos, NM 87545
(505) 667-8353
FTS 843-8353
E-8
------- |